THESIS Gour Bandyopadhyay 2016 PDF

An Analysis of the Non-Performing
Assets of Some Selected Public

Sector Banks in India
Thesis submitted to
Maulana Abul Kalam Azad University of Technology,
West Bengal
(Formerly known as West Bengal University of Technology)
In partial fulfillment of the Degree of
DOCTORATE OF PHILOSOPHY
2016
By
Gour Bandyopadhyay
M.Phil in Management,
The Heritage Academy
Maulana Abul Kalam Azad University of Technology,

West Bengal
Salt Lake, Kolkata-700064
West Bengal, India
Certificate from the Supervisor
This is to certify that the thesis entitled “An Analysis of the Non-
Performing Assets of Some Selected Public Sector Banks in India”
submitted by Gour Bandyopadhyay who got his name registered on 27th
April, 2009 to the West Bengal University of Technology for the award of
the degree of Doctorate of Philosophy, based on original research work
done by him under my supervision. He has completed the work truthfully
and successfully to the best of my knowledge. The contents of the thesis, in
full or in part, have not been submitted to any other Institute or University
for the award of any degree or diploma.
Prof. (Dr.) Malayendu Saha

Research Supervisor
Contents
Acknowledgement i
List of Abbreviations ii - iii
List of Charts iv - viii
List of Tables ix - xiii
List of Annexures xiv
Chapter Details
Chapter Chapter Name Page
Number Number
I Introduction 1 – 31
II Existing Prudential Norms on Asset Quality 32-58
III Practices of NPA Management in Selected PSBs in India 59-85
IV Dynamicity of NPA over Recent Years in Selected PSBs in 86-138

India
V Movement of NPA over Time in Selected PSBs in India 139-169
with Non Parametric Approach
VI Modeling NPA Time Series Data in Selected PSBs in India 170-193
with Semi Parametric Approach
VII Examination of Impact of NPAs on Strategic Banking 194-260
Variables in Selected PSBs in India
VIII Summary, Suggested Action Plan and Conclusion 261-274
Bibliography 275– 286

Annexure 287-299
Acknowledgement
First of all I would like to express my sincere gratitude to my Research Supervisor

Prof. (Dr.) Malayendu Saha who has given a lot of his valuable time and advice to me
throughout my journey in the world of ‘NPA in Public Sector Banks in India’, towards
successful completion of the Thesis.
I am also very grateful to the library staffs of MAKAUT for helping me to locate the
required books, journals and other materials without much inconvenience. Library
coordinators of Reserve Bank of India, Calcutta, State Bank of India Training College,
Calcutta, Allahabad Bank, Calcutta, Punjab National Bank Zonal Staff Training College,
Calcutta, National Library and ISI library have also extended their helping hands in
making my effort a success.
Last but not the least, I extend my heartiest thanks and gratitude to our most beloved and
respected Academic Advisor, Higher Education, Kalyan Bharti Trust, Prof. B. B. Paira,
but for whose continuous encouragement the task would remain very difficult. I am
equally indebted to Prof D. Das and Prof. P. Chakraborty for their sincerest support and
assistance, particularly during final stage of drafting the dissertation.
My sincere appreciation goes to my wife, Ratna and my children Tatain, Megha and
Neha, who have tolerated all sorts of my selfish behavior during preparation of this
dissertation.
(Gour Bandyopadhyay)
Date:
i
List of Abbreviations
Sl No. Abbreviations Expansions
1 AAGR Average Annual Growth Rate
2 ACB Audit Committee Board
3 ARC Asset Reconstruction Company
4 BBR Base Bank Rate
5 BPE Business Per Employee
6 CAR Capital Adequacy Ratio
7 CAGR Compound Annual Growth Rate
8 CBI Central Bank of India
9 CBSR Committee on Banking Sector Reform
10 CDR Credit Deposit Ratio
11 CFS Committee on the Financial System
12 CGTSI Credit Guarantee Trust for Small Industries
13 CRAR Capital Risk Adequacy Ratio
14 CRR Cash Reserve Ratio
15 CV Coefficient of Variation
16 CorDR Corporate Debt Restructuring
17 DCCO Date of Commencement of Commercial Operations
18 DICGC Deposit Insurance & Credit Guarantee Corporation
19 DOP Delegation of Powers
20 DRT Debt Recovery Tribunal
21 DWS Durbin Watson Statistics
22 DWT Durbin Watson test
23 EAS Early Alert System
24 ECGC Export Credit Guarantee Corporation
25 FIs Financial Institutions
26 FTNPA First time NPA
27 GA Gross Advance
28 GDP Gross Domestic Product
29 GNPA Gross Non-Performing Asset
30 GOI Government of India
31 HC Health Code
32 KVP Kishan Vikas Patra
33 LOESS Local Regression
34 L/C Letter of Credit
35 MAE Mean Absolute Error
36 NBFCs Non Banking Financial Companies
37 NBV Net Book Value
38 NIM Net Interest Margin
39 NNPA Net Non-Performing Asset
ii
Sl No. Abbreviations Expansions
40 NPAs Non-Performing Assets
41 NPL Non-Performing Loan
42 NPV Net Present Value
43 NS Negotiated Settlement
44 NSC National Savings Certificate
45 OLS Ordinary Least Squares
46 OPE Operating Profit per Employee
47 OTS One time settlement
48 PDC Post Dated Cheque
49 PNB Punjab National Bank
50 PrSBs Private Sector Banks
51 PSA Priority Sector Advance
52 PSBs Public Sector Banks
53 RBI Reserve Bank of India
54 RRB Regional Rural Bank
55 RSS Residual Sum of Squares
56 RVS Realizable Value of Security
57 SARFAESI Securitization and Reconstruction of Financial Assets and
Enforcement of Security Interest
58 SB Syndicate Bank
59 SBI State Bank of India
60 SBT State Bank of Travancore
61 SCB Schedule Commercial Bank
62 SLR Statutory Liquidity Ratio
63 SMA Special Mention Account
64 SME Small and Medium Enterprise
65 SW Shapiro - Wilk
66 TOL Tolerance Factor
67 UCO UCO Bank
68 VIF Variance Inflation Factor
69 VR Valuation Report
iii
List of Charts
Sl Chart Number & Title Page
No. No.
1 Chart-1.1.: GNPA of SCB and PSB 4
2 Chart-1.2.: GA of SCB and PSB 5
3 Chart-1.3.: GNPA Ratio of SCB and PSB 5
4 Chart-1.4.: Structure of Commercial Banks in India as on 31st March 2012 30
5 Chart- 4.1.: Box Plot of SBI and PSBs in Aggregate on GNPA Dataset (1996- 96
2012)
6 Chart- 4.2.: Box Plot of 5 selected banks on GNPA Dataset (1996-2012) 96
7 Chart- 4.3.: QQ Plot on GNPA – SBI (1996-2011) 99
8 Chart- 4.4.: QQ Plot on GNPA – SBT (1996-2011) 99
9 Chart- 4.5.: QQ Plot on GNPA – SB (1996-2011) 99
10 Chart- 4.6.: QQ Plot on GNPA – UCO (1996-2010) 99
11 Chart- 4.7.: QQ Plot on GNPA – CBI (1996-2011) 100
12 Chart- 4.8.: QQ Plot on GNPA – PNB (1996-2011) 100
13 Chart- 4.9.: QQ Plot on GNPA – PSB (1996-2011) 100
14 Chart- 4.10.: Best Model on GNPA – SBI 110
15 Chart- 4.11.: Best Model on GNPA – SBT 110
16 Chart- 4.12.: Best Competing Models on GNPA – SB 110
17 Chart- 4.13.: Best Model on GNPA – UCO 110
18 Chart- 4.14.: Best Model on GNPA – CBI 110
19 Chart- 4.15.: Best Model on GNPA – PNB 110
20 Chart- 4.16.: Best Model on GNPA – PSB 111
21 Chart- 4.17.: Box Plot of SBI and PSB in Aggregate on GA Dataset (1996-2012) 114
22 Chart- 4.18.: Box Plot of five selected banks on GA Dataset (1996-2012) 114
23 Chart- 4.19.: Q-Q Plot on GA - SBI (1996-2012) 117
24 Chart- 4.20.: Q-Q Plot on GA - SBT (1996-2012) 117
25 Chart- 4.21.: Q-Q Plot on GA - SB (1996-2012) 117
26 Chart- 4.22.: Q-Q Plot on GA - UCO (1996-2012) 117
27 Chart- 4.23.: Q-Q Plot on GA - CBI (1996-2012) 117
28 Chart- 4.24.: Q-Q Plot on GA - PNB (1996-2012) 117
29 Chart- 4.25.: Q-Q Plot on GA - PSB (1996-2012) 117
30 Chart- 4.26.: Best Competing Models on GA – SBI 126
31 Chart- 4.27.: Best Competing Models on GA– SBT 126
32 Chart- 4.28.: Best Competing Models on GA– SB 127
33 Chart- 4.29.: Best Competing Models on GA–UCO 127
34 Chart- 4.30.: Best Competing Models on GA– CBI 127
35 Chart- 4.31.: Best Competing Models on GA– PNB 127
36 Chart- 4.32.: Best Competing Models on GA– PSB 127
37 Chart- 5.1. Defining the window width on GNPA - SBI 148
38 Chart- 5.2. Defining the window width on GNPA - SBT 148
39 Chart- 5.3. Defining the window width on GNPA - SB 148
40 Chart- 5.4. Defining the window width on GNPA - UCO 148
41 Chart- 5.5.: Defining the window width on GNPA - CBI 149
42 Chart- 5.6.: Defining the window width on GNPA - PNB 149
43 Chart- 5.7.: Defining the window width on GNPA - PSB 149
44 Chart- 5.8.: Tricube weight function for parameter GNPA 149
45 Chart- 5.9.: The LOESS Fit on GNPA – SBI 150
iv
No. No.
46 Chart- 5.10.: The LOESS Fit on GNPA – SBT 150
47 Chart- 5.11.: The LOESS Fit on GNPA – SB 150
48 Chart- 5.12.: The LOESS Fit on GNPA – UCO 150
49 Chart- 5.13.: The LOESS Fit on GNPA – CBI 150
50 Chart- 5.14.: The LOESS Fit on GNPA – PNB 150
51 Chart- 5.15.: The LOESS Fit on GNPA – PSB 151
52 Chart- 5.16.: Robust LOESS Fit on GNPA – SBI 152
53 Chart- 5.17.: Robust LOESS Fit on GNPA –SBT 152
54 Chart- 5.18.: Robust LOESS Fit on GNPA –SB 152
55 Chart- 5.19.: Robust LOESS Fit on GNPA – UCO 152
56 Chart- 5.20.: Robust LOESS Fit on GNPA –CBI 152
57 Chart- 5.21.: Robust LOESS Fit on GNPA –PNB 152
58 Chart- 5.22.: Robust LOESS Fit on GNPA – PSB 152
59 Chart- 5.23.: Linear and LOESS Fits on GNPA - SBI 153
60 Chart- 5.24.: Linear and LOESS Fits on GNPA - SBT 153
61 Chart- 5.25.: Linear and LOESS Fits on GNPA - SB 153
62 Chart- 5.26.: Linear and LOESS Fits on GNPA - UCO 153
63 Chart- 5.27.: Linear and LOESS Fits on GNPA - CBI 153
64 Chart- 5.28.: Linear and LOESS Fits on GNPA - PNB 153
65 Chart- 5.29.: Linear and LOESS Fits on GNPA - PSB 154
66 Chart- 5.30.:Quantile Comparison Plot on GNPA – SBI 154
67 Chart- 5.31.: Quantile Comparison Plot on GNPA – SBT 154
68 Chart- 5.32.: Quantile Comparison Plot on GNPA – SB 154
69 Chart- 5.33.:Quantile Comparison Plot on GNPA - UCO 154
70 Chart- 5.34.: Quantile Comparison Plot on GNPA – CBI 154
71 Chart- 5.35.:Quantile Comparison Plot on GNPA – PNB 154
72 Chart- 5.36.:Quantile Comparison Plot on GNPA – PSB 155
73 Chart- 5.37.: Residual Plot on GNPA - SBI 155
74 Chart- 5.38.: Residual Plot on GNPA - SBT 155
75 Chart- 5.39.: Residual Plot on GNPA - SB 155
76 Chart- 5.40.: Residual Plot on GNPA - UCO 155
77 Chart- 5.41.: Residual Plot on GNPA - CBI 156
78 Chart- 5.42.: Residual Plot on GNPA - PNB 156
79 Chart- 5.43.: Residual Plot on GNPA - PSB 156
80 Chart- 5.44.: Abs Res vs Pred Plot on GNPA - SBI 156
81 Chart- 5.45.: Abs Res vs Pred Plot on GNPA - SBT 156
82 Chart- 5.46.: Abs Res vs Pred Plot on GNPA - SB 157
83 Chart- 5.47.: Abs Res vs Pred Plot on GNPA - UCO 157
84 Chart- 5.48.: Abs Res vs Pred Plot on GNPA - CBI 157
85 Chart- 5.49.: Abs Res vs Pred Plot on GNPA - PNB 157
86 Chart- 5.50.: Abs Res vs Pred Plot on GNPA - PSB 157
87 Chart- 5.51:Defining the window width on GA - SBI 158
88 Chart- 5.52:Defining the window width on GA - SBT 158
89 Chart- 5.53: Defining the window width on GA - SB 159
90 Chart- 5.54:Defining the window width on GA - UCO 159
91 Chart- 5.55: Defining the window width on GA - CBI 159
92 Chart- 5.56:Defining the window width on GA - PNB 159
93 Chart- 5.57: Defining the window width on GA - PSB 158
94 Chart- 5.58.: Tricube weight function for parameter GA 159
v
No. No.
95 Chart- 5.59.: The LOESS Fit on GA – SBI 160
96 Chart- 5.60.: The LOESS Fit on GA –SBT 160
97 Chart- 5.61.: The LOESS Fit on GA – SB 160
98 Chart- 5.62.: The LOESS Fit on GA – UCO 160
99 Chart- 5.63.: The LOESS Fit on GA – CBI 160
100 Chart- 5.64.: The LOESS Fit on GA – PNB 160
101 Chart- 5.65.: The LOESS Fit on GA – PSB 161
102 Chart- 5.66.: Robust LOESS Fit on GA – SBI 162
103 Chart- 5.67.: Robust LOESS Fit on GA – SBT 162
104 Chart- 5.68.: Robust LOESS Fit on GA – SB 162
105 Chart- 5.69.: Robust LOESS Fit on GA – UCO 162
106 Chart- 5.70.: Robust LOESS Fit on GA – CBI 162
107 Chart- 5.71.: Robust LOESS Fit on GA – PNB 162
108 Chart- 5.72.: Robust LOESS Fit on GA - PSB 163
109 Chart- 5.73.: Linear and LOESS Fits on GA -SBI 163
110 Chart- 5.74.: Linear and LOESS Fits on GA -SBT 163
111 Chart- 5.75.: Linear and LOESS Fits on GA -SB 163
112 Chart- 5.76.: Linear and LOESS Fits on GA -UCO 163
113 Chart- 5.77.: Linear and LOESS Fits on GA -CBI 164
114 Chart- 5.78.: Linear and LOESS Fits on GA -PNB 164
115 Chart- 5.79.: Linear and LOESS Fits on GA -PSB 164
116 Chart- 5.80.: Quantile Comparison Plot on GA - SBI 164
117 Chart- 5.81.: Quantile Comparison Plot on GA - SBT 164
118 Chart- 5.82.: Quantile Comparison Plot on GA - SB 164
119 Chart- 5.83.: Quantile Comparison Plot on GA - UCO 164
120 Chart- 5.84.: Quantile Comparison Plot on GA - CBI 165
121 Chart- 5.85.: Quantile Comparison Plot on GA - PNB 165
122 Chart- 5.86.: Quantile Comparison Plot on GA - PSB 165
123 Chart- 5.87.: Residual Plot on GA -SBI 165
124 Chart- 5.88.: Residual Plot on GA -SBT 165
125 Chart- 5.89.: Residual Plot on GA -SB 165
126 Chart- 5.90.: Residual Plot on GA -UCO 165
127 Chart- 5.91.: Residual Plot on GA -CBI 166
128 Chart- 5.92.: Residual Plot on GA -PNB 166
129 Chart- 5.93.: Residual Plot on GA -PSB 166
130 Chart- 5.94.: Abs Res vs Pred Plot on GA - SBI 166
131 Chart- 5.95.: Abs Res vs Pred Plot on GA – SBT 166
132 Chart- 5.96.: Abs Res vs Pred Plot on GA – SB 167
133 Chart- 5.97.: Abs Res vs Pred Plot on GA – UCO 167
134 Chart- 5.98.: Abs Res vs Pred Plot on GA – CBI 167
135 Chart- 5.99.: Abs Res vs Pred Plot on GA – PNB 167
136 Chart- 5.100.: Abs Res vs Pred Plot on GA – PSB 167
137 Chart- 6.1.: Penalized Spline on GNPA -SBI 178
138 Chart- 6.2.: Penalized Spline on GNPA -SBT 178
139 Chart- 6.3.: Penalized Spline on GNPA –SB 179
140 Chart- 6.4.: Penalized Spline on GNPA –UCO 179
141 Chart- 6.5.: Penalized Spline on GNPA –CBI 179
142 Chart- 6.6.: Penalized Spline on GNPA –PNB 179
143 Chart- 6.7.: Penalized Spline on GNPA –PSB 179
vi
No. No.
144 Chart- 6.8.: Penalized Spline, LOESS and Linear fit on GNPA – SBI 180
145 Chart- 6.9.: Penalized Spline, LOESS and Linear fit on GNPA – SBT 180
146 Chart- 6.10.: Penalized Spline, LOESS and Linear fit on GNPA – SB 180
147 Chart- 6.11.: Penalized Spline, LOESS and Linear fit on GNPA – UCO 180
148 Chart- 6.12.: Penalized Spline, LOESS and Linear fit on GNPA – CBI 181
149 Chart- 6.13.: Penalized Spline, LOESS and Linear fit on GNPA – PNB 181
150 Chart- 6.14.: Penalized Spline, LOESS and Linear fit on GNPA – PSB 181
151 Chart- 6.15.: Quantile Comparison Plot on GNPA -SBI 182
152 Chart- 6.16.: Quantile Comparison Plot on GNPA - SBT 182
153 Chart- 6.17.: Quantile Comparison Plot on GNPA -SB 182
154 Chart- 6.18.: Quantile Comparison Plot on GNPA -UCO 182
155 Chart- 6.19.: Quantile Comparison Plot on GNPA -CBI 182
156 Chart- 6.20.: Quantile Comparison Plot on GNPA -PNB 182
157 Chart- 6.21.: Quantile Comparison Plot on GNPA -PSBs 182
158 Chart- 6.22.: Residual vs Predictor Plot on GNPA – SBI 183
159 Chart- 6.23.: Residual vs Predictor Plot on GNPA – SBT 183
160 Chart- 6.24.: Residual vs Predictor Plot on GNPA – SB 183
161 Chart- 6.25.: Residual vs Predictor Plot on GNPA – UCO 183
162 Chart- 6.26.: Residual vs Predictor Plot on GNPA – CBI 183
163 Chart- 6.27.: Residual vs Predictor Plot on GNPA – PNB 183
164 Chart- 6.28.: Residual vs Predictor Plot on GNPA – PSB 183
165 Chart- 6.29.: Normal Q-Q Plot on GNPA – SBI 184
166 Chart- 6.30.: Normal Q-Q Plot on GNPA – SBT 184
167 Chart- 6.31.: Normal Q-Q Plot on GNPA – SB 184
168 Chart- 6.32.: Normal Q-Q Plot on GNPA – UCO 184
169 Chart- 6.33.: Normal Q-Q Plot on GNPA – CBI 184
170 Chart- 6.34.: Normal Q-Q Plot on GNPA – PNB 184
171 Chart- 6.35.: Normal Q-Q Plot on GNPA – PSB 184
172 Chart- 6.36.: Penalized Spline on GA – SBI 186
173 Chart- 6.37.: Penalized Spline on GA – SBT 186
174 Chart- 6.38.: Penalized Spline on GA – SB 186
175 Chart- 6.39.: Penalized Spline on GA – UCO 186
176 Chart- 6.40.: Penalized Spline on GA – CBI 186
177 Chart- 6.41.: Penalized Spline on GA – PNB 186
178 Chart- 6.42.: Penalized Spline on GA – PSB 187
179 Chart- 6.43.: Penalized Spline, LOESS and Linear fit on GA – SBI 187
180 Chart- 6.44.: Penalized Spline, LOESS and Linear fit on GA – SBT 187
181 Chart- 6.45.: Penalized Spline, LOESS and Linear fit on GA – SB 188
182 Chart- 6.46.: Penalized Spline, LOESS and Linear fit on GA – UCO 188
183 Chart- 6.47.: Penalized Spline, LOESS and Linear fit on GA – CBI 188
184 Chart- 6.48.: Penalized Spline, LOESS and Linear fit on GA – PNB 188
185 Chart- 6.49.: Penalized Spline, LOESS and Linear fit on GA – PSB 188
186 Chart- 6.50.: Quantile Comparison Plot on GA –SBI 189
187 Chart- 6.51.: Quantile Comparison Plot on GA - SBT 189
188 Chart- 6.52.: Quantile Comparison Plot on GA –SB 189
189 Chart- 6.53.: Quantile Comparison Plot on GA –UCO 189
190 Chart- 6.54.: Quantile Comparison Plot on GA –CBI 189
191 Chart- 6.55.: Quantile Comparison Plot on GA –PNB 189
192 Chart- 6.56.: Quantile Comparison Plot on GA -PSBs 189
vii
No. No.
193 Chart- 6.57.: Residual vs Predictor Plot on GA – SBI 190
194 Chart- 6.58.: Residual vs Predictor Plot on GA – SBT 190
195 Chart- 6.59.: Residual vs Predictor Plot on GA – SB 190
196 Chart- 6.60.: Residual vs Predictor Plot on GA – UCO 190
197 Chart- 6.61.: Residual vs Predictor Plot on GA – CBI 190
198 Chart- 6.62.: Residual vs Predictor Plot on GA – PNB 190
199 Chart- 6.63.: Residual vs Predictor Plot on GA – PSB 190
200 Chart- 6.64.: Normal Q-Q Plot on GA – SBI 191
201 Chart- 6.65.: Normal Q-Q Plot on GA – SBT 191
202 Chart- 6.66.: Normal Q-Q Plot on GA – SB 191
203 Chart- 6.67.: Normal Q-Q Plot on GA – UCO 191
204 Chart- 6.68.: Normal Q-Q Plot on GA – CBI 191
205 Chart- 6.69.: Normal Q-Q Plot on GA – PNB 191
206 Chart- 6.70.: Normal Q-Q Plot on GA – PSB 191
viii
List of Tables
Sl Table Number & Title Page No.
No.
1 Table- 2.1.: Summary of Progressive Changes in Income Recognition Norms 35
2 Table- 2.2.: Summary of Changes in NPA Norms 41
3 Table- 2.3.: Summary of Change in Asset Classification Norms 44
4 Table- 2.4.: Provisioning Norms for Doubtful Debts 45
5 Table- 2.5.: Summary of Changes in Provisioning Norms 46
6 Table- 2.6.: Provisioning Norms for Doubtful Debts (Infrastructure Projects) 48
7 Table- 2.7.: Provisioning Norms for Doubtful Debts (Non Infrastructure Projects) 48
8 Table- 2.8.: Methods of Valuation of Securities 50
9 Table- 3.1.: Calendar of Review of NPA accounts by the Board 61
10 Table- 3.2.: Calendar of Review of NPA accounts by the Executive Committee 61
11 Table- 3.3.: Calendar of Review of NPA accounts by the Audit Committee 61
12 Table- 4.1.: GNPA Dataset (Rs. in crores) 88
13 Table- 4.2.: GA Dataset (Rs. in crores) 88
14 Table- 4.3.: Values (with year) and ultimate sample size w.r.to the Parameter GNPA 97
15 Table- 4.4.: Ultimate Dataset on Banks with the corresponding years - GNPA 97
16 Table- 4.5.: Values of DW Statistic on observed data - GNPA 98
17 Table- 4.6.: Values of t-Statistics (Calculated and Tabulated) - GNPA 98
18 Table- 4.7.: Values of the SW statistic computed for Parameter GNPA 100
19 Table- 4.8.: Values of different statistical measures for Parameter GNPA 101
20 Table- 4.9.: Alternative Model Equations on GNPA - SBI 103
21 Table- 4.10.: Alternative Model Equations on GNPA - SBT 104
22 Table- 4.11.: Alternative Model Equations on GNPA - SB 104
23 Table- 4.12.: Alternative Model Equations on GNPA - UCO 104
24 Table- 4.13.: Alternative Model Equations on GNPA - CBI 105
25 Table- 4.14.: Alternative Model Equations on GNPA - PNB 105
26 Table- 4.15.: Alternative Model Equations on GNPA - PSB 106
27 Table- 4.16.: Values of precision criteria for Alternative Models on GNPA - SBI 106
28 Table- 4.17.: Values of precision criteria for Alternative Models on GNPA - SBT 107
29 Table- 4.18.: Values of precision criteria for Alternative Models on GNPA - SB 107
30 Table- 4.19.: Values of precision criteria for Alternative Models on GNPA - UCO 107
31 Table- 4.20.: Values of precision criteria for Alternative Models on GNPA - CBI 108
32 Table- 4.21.: Values of precision criteria for Alternative Models on GNPA - PNB 108
33 Table- 4.22.: Values of precision criteria for Alternative Models on GNPA - PSB 108
34 Table- 4.23.: Values of precision criteria for higher degree polynomials on GNPA - 109
PNB
35 Table- 4.24.: DW Values Calculated on Errors (using best model) – GNPA 111
36 Table- 4.25.: ‘t’ test Values Calculated on Errors (using best model) – GNPA 112
37 Table- 4.26.: Test of Normality on Errors (using best model) – Shapiro-Wilk Test – 113
GNPA
38 Table- 4.27.: Values (with year) and ultimate sample size w.r.to the Parameter GA 114
39 Table- 4.28.: Ultimate Dataset on Banks with the corresponding years - GA 115
40 Table- 4.29.: DW Statistics on observed data - GA 115
41 Table- 4.30.: Values of t-Statistics (Calculated and Tabulated) - GA 116
42 Table- 4.31.: Values of the SW test statistic computed for Parameter GA 118
43 Table- 4.32.: Values of the different statistical measures for Parameter GA 119
44 Table- 4.33.: Alternative Model Equations on GA - SBI 120
ix
No.
45 Table- 4.34.: Alternative Model Equations on GA - SBT 121
46 Table- 4.35.: Alternative Model Equations on GA - SB 121
47 Table- 4.36.: Alternative Model Equations on GA - UCO 122
48 Table- 4.37.: Alternative Model Equations on GA - CBI 122
49 Table- 4.38.: Alternative Model Equations on GA - PNB 122
50 Table- 4.39.: Alternative Model Equations on GA - PSB 123
51 Table- 4.40.: Values of precision criteria for Alternative Models on GA - SBI 123
52 Table- 4.41.: Values of precision criteria for Alternative Models on GA - SBT 124
53 Table- 4.42.: Values of precision criteria for Alternative Models on GA - SB 124
54 Table- 4.43.: Values of precision criteria for Alternative Models on GA - UCO 125
55 Table- 4.44.: Values of precision criteria for Alternative Models on GA - CBI 125
56 Table- 4.45.: Values of precision criteria for Alternative Models on GA - PNB 125
57 Table- 4.46.: Values of precision criteria for Alternative Models on GA – PSB 126
58 Table- 4.47.: DW Values Calculated on Errors (using best model) – GA 128
59 Table- 4.48.: ‘t’ test Values Calculated on Errors (using best model) – GA 128
60 Table- 4.49.: Summary Statistic Shapiro-Wilk Test on errors (using best model) – 129
GA
61 Table- 4.50.: Estimation of Accuracies of the Competing Models on GNPA – SBI 131
62 Table- 4.51.: Estimation of Accuracies of the Competing Models on GNPA – SBT 131
63 Table- 4.52.: Estimation of Accuracies of the Competing Models on GNPA – SB 131
64 Table- 4.53.: Estimation of Accuracies of the Competing Models on GNPA – UCO 132
65 Table- 4.54.: Estimation of Accuracies of the Competing Models on GNPA – CBI 132
66 Table- 4.55.: Estimation of Accuracies of the Competing Models on GNPA – PNB 132
67 Table- 4.56.: Estimation of Accuracies of the Competing Models on GNPA – PSB 132
68 Table- 4.57.: Best Forecasting Model Expressions in Respect of Parameter GNPA 133
69 Table- 4.58.: Forecast Values for Parameter GNPA 133
70 Table- 4.59.: Estimation of Accuracies of the Competing Models on GA – SBI 134
71 Table- 4.60.: Estimation of Accuracies of the Competing Models on GA – SBT 134
72 Table- 4.61.: Estimation of Accuracies of the Competing Models on GA – SB 134
73 Table- 4.62: Estimation of Accuracies of the Competing Models on GA – UCO 135
74 Table- 4.63.: Estimation of Accuracies of the Competing Models on GA – CBI 135
75 Table- 4.64.: Estimation of Accuracies of the Competing Models on GA – PNB 135
76 Table- 4.65.: Estimation of Accuracies of the Competing Models on GA – PSB 136
77 Table- 4.66.: Best Forecasting Model Expressions in Respect of Parameter GA 136
78 Table- 4.67.: Forecast Values for Parameter GA 136
79 Table- 4.68.: Annual Growth Rates for the Parameter GA in 6 PSBs and PSBs in 137
Aggregate
80 Table- 4.69.: Mean values calculated from Annual Growth Rate (Percentage) 138
81 Table- 5.1.: Summary Statistics of LOESS Model for selected PSBs on Parameter 150
GNPA
82 Table- 5.2.: Summary Statistics of precision criteria for selected PSBs on Parameter 157
GNPA
83 Table- 5.3.: Summary Statistics of LOESS Model for selected PSBs on Parameter 160
GA
GA
87 Table- 6.1.: Summary Statistics for Penalized Spline models on Parameter GNPA 178
x
No.
GNPA
89 Table- 6.3.: Summary Statistics for Penalized Spline models on Parameter GA 185
GA
93 Table- 7.1.: Gross NPA To Total Asset (GNPA/TA Ratio) Dataset 195
94 Table- 7.2.: Profit before Tax To Total Asset (PBT/TA Ratio) Dataset 196
95 Table- 7.3.: Operating Profit per Employee (OPE Ratio) Dataset 196
96 Table- 7.4.: Net Interest Margin to Total Asset (NIM/TA Ratio) Dataset 197
97 Table- 7.5.: Credit Deposit Ratio (CDR) Dataset 197
98 Table- 7.6.: Ratio of Credit Growth Dataset 198
99 Table- 7.7.: Capital Adequacy Ratio (CAR) Dataset 199
100 Table- 7.8.: (Net Interest Margin – Burden) to TA Ratio Dataset 199
101 Table- 7.9.: Priority Sector Advance To Total Advance Ratio (PSADV/TADV) 200
Dataset
102 Table- 7.10.: List of Statistically Significant Variables from Earlier Studies 203
103 Table- 7.11.: List of Statistically Significant Variables Identified for our Study 205
104 Table- 7.12.: Expected relationship between NPA and any of the strategic banking 207
variables
105 Table- 7.13.: Expected relationship between Profitability and any of the strategic 207
banking variables
106 Table- 7.14.: Correlation between NPA (GNPA/TA) and Selected Banking Variables 210
107 Table- 7.15.: Selection of Statistically Significant Strategic Banking Variables 212
having Association with NPA (GNPA/TA)
108 Table- 7.16.: Correlation between Profitability (PBT/TA) and Selected Banking 214
Variables
109 Table- 7.17.: Selection of Statistically Significant Strategic Banking Variables 216
having Association with Profitability (PBT/TA)
110 Table- 7.18.: DW and SW Statistics for Profitability (PBT/TA) and NPA 217
(GNPA/TA)
111 Table-7.19.: DW and SW Statistics for Profitability (PBT/TA) and NPA 218
(GNPA/TA) for UCO
112 Table- 7.20.: Regression Model Summary for Profitability (PBT/TA) and NPA 218
(GNPA/TA)
113 Table- 7.21.: Summary of the Coefficients of the Models Selected Banks 219
114 Table- 7.22.: Model Equations on Profitability (PBT/TA) and NPA (GNPA/TA) 220
115 Table- 7.23.: Regression Model Assumptions Verification for Profitability (PBT/TA) 221
and NPA (GNPA/TA) & Time (Year)
116 Table- 7.24.: Regression Model Collinearity Statistics for Profitability (PBT/TA) 221
(GNPA/TA) & Time (Year)
118 Table- 7.26.: Regression Model Coefficients, t stat for Profitability (PBT/TA) and 223
NPA (GNPA/TA) & Time (Year)
119 Table- 7.27.: Model Equations on Profitability (PBT/TA) and NPA (GNPA/TA) & 224
Time (Year)
120 Table- 7.28.: Regression Model Assumptions Verification Statistics for Profitability 224
(PBT/TA) and statistically significant variables including NPA (GNPA/TA)
xi
No.
121 Table- 7.29.: The Multiple Cubic Model for SB 225
122 Table- 7.30.: Multiple Regression Model Collinearity Statistics for Profitability 225
(PBT/TA) & NPA (GNPA/TA), other Statistically Significant Variables
123 Table- 7.31.: Regression Model Assumptions Verification for Profitability (PBT/TA) 226
and NPA (GNPA/TA) along with Other Statistically Significant Variables
124 Table- 7.32.: Regression Model Collinearity Statistics for Profitability (PBT/TA) and 227
NPA (GNPA/TA) & Other Statistically Significant Variables
(GNPA/TA) & Other Statistically Significant Variables
126 Table- 7.34.: Regression Model Coefficients, t stat for Profitability (PBT/TA) and 228
NPA (GNPA/TA) & Other Statistically Significant Variables
127 Table- 7.35.: Model Equations on Profitability (PBT/TA) and Strategic Banking 229
Variables including NPA (GNPA/TA)
128 Table- 7.36.: Regression Model Assumptions Verification Statistics for profitability 230
and statistically significant variables including NPA & Time
129 Table- 7.37.: Regression Model Collinearity Statistics for profitability (PBT/TA) and 231
statistically significant variables including NPA & Time
130 Table- 7.38.: Regression Model Summary for Profitability (PBT/TA) and 232
statistically significant variables including NPA & Time
131 Table- 7.39.: Regression Model Coefficients, t stat for profitability (PBT/TA) and 233
NPA (GNPA/TA). Time (Year) & other statistically significant variables
132 Table- 7.40.: Model Equations on Profitability (PBT/TA) and Strategic Banking 233
Variables including NPA (GNPA/TA) & Time (Year)
133 Table- 7.41.: DW and SW test Statistics for Productivity (OPE) and NPA 235
(GNPA/TA)
134 Table- 7.42.: Regression Model Assumptions Verification for Productivity (OPE) 236
135 Table- 7.43.: Assumptions Verification for Productivity (OPE) and NPA 236
(GNPA/TA) & Time (Year) with Multiple Quadratic, Multiple Cubic Models
136 Table- 7.44.: Regression Model Collinearity Statistics for Productivity (OPE) and 237
NPA (GNPA/TA) & Time (Year)
137 Table- 7.45.: Regression Model Summary for Productivity (OPE) and NPA 238
138 Table- 7.46.: Regression Model Coefficients, t stat for Productivity (OPE) and NPA 238
139 Table- 7.47.: Model Equations on Productivity (OPE) and Strategic Banking 239
Variables including NPA (GNPA/TA) & Time (Year)
140 Table- 7.48.: Regression Model Assumptions Verification for NIM/TA and NPA 240
(GNPA/TA)
141 Table- 7.49.: Regression Model Summary for NIM/TA and NPA (GNPA/TA) 241
142 Table- 7.50.: Regression Model Coefficients, t stat for NIM/TA and NPA 241
(GNPA/TA)
143 Table- 7.51.: Model Equations for NIM/TA and NPA (GNPA/TA) 242
144 Table- 7.52.: Regression Model Assumptions Verification for NIM/TA and 242
GNPA/TA & Year
145 Table- 7.53.: Regression Model Collinearity Statistics for NIM/TA & GNPA/TA and 243
Time (Year)
146 Table- 7.54.: Regression Model Summary for NIM/TA and GNPA/TA and Time 243
(Year)
147 Table- 7.55.: Regression Model Coefficients, t stat for NIM/TA & GNPA/TA and 244
xii
No.
Time (Year)
148 Table- 7.56.: Model Equations for NIM/TA and GNPA/TA & Time (Year) 244
149 Table- 7.57.: Regression Model Assumptions Verification for CDR and NPA 245
(GNPA/TA)
150 Table- 7.58.: Regression Model Assumptions Verification for CDR and NPA 246
151 Table- 7.59.: Regression Model Collinearity Statistics for CDR and NPA 246
152 Table- 7.60.: Regression Model Summary for CDR and NPA (GNPA/TA) & Time 247
(Year)
153 Table- 7.61.: Regression Model Coefficients, t stat for CDR and NPA (GNPA/TA) 247
& Time (Year)
154 Table- 7.62.: Model Equations on CDR and NPA (GNPA/TA) & Time (Year) 248
155 Table- 7.63.: DW and SW test Statistics for Credit Growth and NPA (GNPA/TA) 248
156 Table- 7.64.: Regression Model Summary for Credit Growth and NPA (GNPA/TA) 249
157 Table- 7.65.: Regression Model Coefficients, t stat for Credit Growth and NPA 249
(GNPA/TA)
158 Table- 7.66.: Model Equations on Credit Growth and NPA (GNPA/TA) 250
159 Table- 7.67.: Regression Model Assumption Verification for Credit Growth and 251
GNPA/TA & Time (Year)
160 Table- 7.68.: Regression Model Coefficients, t stat, Collinearity Statistics for Credit 251
Growth and NPA ( GNPA/TA) & Time (Year)
161 Table- 7.69.: Regression Model Summary for Credit Growth and NPA ( GNPA/TA) 252
& Time (Year)
162 Table- 7.70.: Regression Model Coefficients, t stat for Credit Growth and NPA 252
163 Table- 7.71.: Model Equations on Credit Growth and NPA (GNPA/TA) & Time 253
(Year)
164 Table- 7.72.: Regression Model Assumptions Verification for CAR and NPA 254
(GNPA/TA)
165 Table- 7.73.: Regression Model Summary for CAR and NPA (GNPA/TA) 254
166 Table- 7.74.: Regression Model Coefficients, t stat for CAR and NPA (GNPA/TA) 255
167 Table- 7.75.: Model Equations on CAR and NPA (GNPA/TA) 255
168 Table- 7.76.: Regression Model Assumptions Verification for CAR and GNPA/TA 255
& Time (Year)
169 Table- 7.77.: Regression Model Coefficients, t stat, Collinearity Statistics for CAR 256
and GNPA/TA & Time (Year)
170 Table- 7.78.: Regression Model Summary for CAR and GNPA/TA & Time (Year) 256
171 Table- 7.79.: Regression Model Coefficients, t stat for CAR and GNPA/TA & Time 257
(Year)
172 Table- 7.80.: Model Equations on CAR and NPA (GNPA/TA) & Time (Year) 257
xiii
List of Annexures:
Sl Annexure Annexure Title Page
No. No. No.
1 1A Status of GA, GNPAs, and GNPA/GA Ratio of all SCBs & PSBs 287
from 1993 – 2012
2 1B List of Scheduled Commercial Banks as on 31st March 2012 287
(excluding RRBs)
3 1C Bank wise GNPA/GA Ratio as on March 31st, 2008 of all PSBs. 288
4 7A Average Total Asset (TA) Dataset 289
5 7B Profit before Tax (PBT) Dataset 289
6 7C Aggregate Credit Dataset 289
7 7D Aggregate Deposit Dataset 290
8 7E Priority Sector Advance (PSADV) Dataset 290
9 7F Number of Employee 290
10 7G (1) Computation of Net Interest Margin (NIM) and Burden (SBI) 291
11 7G (2) Computation of Net Interest Margin (NIM) and Burden (SBT) 291
12 7G (3) Computation of Net Interest Margin (NIM) and Burden (SB) 291
13 7G (4) Computation of Net Interest Margin (NIM) and Burden (UCO) 292
14 7G (5) Computation of Net Interest Margin (NIM) and Burden (CBI) 292
15 7G (6) Computation of Net Interest Margin (NIM) and Burden (PNB) 293
16 7G (7) Computation of Net Interest Margin (NIM) and Burden (PSB) 293
17 7H (1) Computation of Total Overhead Expense (SBI) 293
18 7H (2) Computation of Total Overhead Expense (SBT) 294
19 7H (3) Computation of Total Overhead Expense (SB) 294
20 7H (4) Computation of Total Overhead Expense (UCO) 295
21 7H (5) Computation of Total Overhead Expense (CBI) 295
22 7H (6) Computation of Total Overhead Expense (PNB) 296
23 7H(7) Computation of Total Overhead Expense (PSB) 296
24 7I (1) Computation of Operating Profit (SBI) 296
25 7I (2) Computation of Operating Profit (SBT) 297
26 7I (3) Computation of Operating Profit (SB) 297
27 7I (4) Computation of Operating Profit (UCO) 298
28 7I (5) Computation of Operating Profit (CBI) 298
29 7I (6) Computation of Operating Profit (PNB) 298
30 7I (7) Computation of Operating Profit (PSB) 299
xiv
Chapter – I
Introduction
1.1 Introduction
1.2 Status of NPA in Indian Banking and the Impact of NPA on Indian
Economy
1.3 Review of Literatures
1.4 Identification of the Research Gap
1.5 Objective of the Study
1.6 Research Methodology
1.7 Scope of the Study
1.8 Scheme of Research Work
1
1.1. Introduction
Both in developed and developing countries, banking system has shown tremendous
dynamism and flexibility for adaptation towards adjusting itself to the sweeping changes in
the economy (Sharma, 1985). Banking system in India is no exception to the above
phenomena and has undergone major transformation twice, over the last five decades. Firstly,
nationalization of State Bank of India in 1955 followed by nationalization of 20 leading
commercial banks in 2 phases in 1969 and 1984, resulted in paradigm shift in the philosophy
and functioning of banking from traditional class banking to mass banking. The policy thrust
in those days was to spread banking service far and wide into the countryside. As a result,
banking in India witnessed unprecedented expansion in branch network, particularly in rural
and semi urban areas, tremendous increase in disbursement of credit, particularly in priority
and preferred sector of the economy and also in mobilization of deposit.
Secondly, the reform initiative started in 1992-93 in the banking industry with the primary
motive of making Indian banking system stable, competitive, transparent and vibrant which
brought a total shift in the mindset of bankers in India from mass banking to prudent banking.
In this direction, series of policy initiatives were introduced as per recommendations of
Narasimham committee in two phases. Firstly during 1992-93 to 1997-98 (first phase of
reforms), the primary objective was to stop erosion in the health of banking, bring uniformity
in banking practices and transparency in the balance sheet of banking companies in India.
Until 1992-93, the Indian Financial system was characterized by limited competition. Both
the products and players were circumscribed by heavy regulation in interest rates in credits,
government securities and capital markets. During these days commercial banks in India
were also subjected to administered structure with a highly segmented market with limited
competition and few distinct alternative products. The banking industry moved gradually
from a regulated environment to a deregulated market economy. The market developments
kindled by liberalization and globalization resulted in changes in the intermediation role of
banks. The pace of transformation was more significant with suitable support of
technological up-gradation. While the banking system had done fairly well in adjusting to the
new market dynamics, greater challenges were still there to face. Finally with the second
phase of reform, from 1998-99 onwards, the main focus was shifted towards raising Indian
banking to international standard. Accordingly, the financial sector was opened up for greater
international competition under WTO. Banks had to gear up to meet stringent prudential
2
capital adequacy norms under BASEL II and also had to cope with challenges posed by
technological innovations in banking. Commercial banks in India accepted the challenges in a
befitting manner to suitably adapt to such changes.
One of the major challenges that the banking industry is facing today is credit risks. Various
dimensions of credit risks expose this fast growing industry to mounting Non-Performing
Assets (NPAs). Though macro environmental factors including global recession, economic
slowdown, rising inflation and inadequate legal frame work have contributed heavily towards
growth of distress asset in the banking sector in India, micro banking factors, including poor
appraisal and follow-up, inappropriate mindset of the borrower and lender are no less
important. Pressure from regulators towards strict adherence to international norms on asset
quality has also contributed to rising bad loan figures in absolute terms in the Public Sector
Banks (PSBs) in India in the last two decades.
1.2. Status of NPA in Indian Banking and the Impact of NPA on Indian Economy
Banking is a financial intermediary which mobilizes savings of the surplus units in the form
of deposits and channelizes them in deficit units including agriculture, industry and services,
as loans and advances. Commercial banks in India particularly the PSBs are exposed to
numerous problems viz a dramatic hike in the bank total expenses, stiff competition with
highly technology driven new generation Private Sector Banks and Foreign Banks, enormous
task of changing the mindset of existing workforces, mobilizing low cost resources,
improving asset quality and improving on operating efficiency etc. However, the most
difficult problem is to address the issue of distress asset in the balance sheet and ensure
disbursement of funds in quality assets (loans and advances). As a lending institution, bad
debt in banking business is an eventuality and cannot be avoided. Bad loans or NPAs are loan
assets, which cease to generate income to the bank. NPAs have dampening effect on banking
system since long, though they were not in the public domain till early 90s (Khasnobis,
2006). By this time huge amount of advances involving uncertainty regarding repayment had
piled up, resulting in question about the health of Indian banking system and their ability to
honour their deposit commitments (Banerjee, Cole, & Duflo, 2004).
3
Norms on NPA was implemented for the first time by RBI in 1992-93, when Gross Non
Performing Asset (GNPA) of all public sector banks (PSBs) were Rs 39253 crores,
representing 23.18% of Gross Advances (GAs) and subsequently rose to Rs 1,12,489 crores,
representing 3.17% of the GAs as on March 2012. For all Schedule Commercial Banks
(SCBs), GNPA were Rs 47,300 crores representing 15.7% of GAs as on March 1997, which
rose to Rs 1,37,096 crores for SCBs representing 2.94% of GAs as on March 2012. This
indicates low quality of credit portfolio in absolute terms in the PSBs and SCBs in aggregate
in India. Such a high figure of loan defaults is directly eating away vitality of the banks in
India and compels to relook into their credit appraisal and follow up system, (Sharma, 2005).
The above estimates of NPAs in PSBs, is believed to significantly understate the actual
gravity of the situation (Basu, 2005). Such a mammoth NPA figure in PSBs is also a constant
worry for the regulators and the Ministry, as banking plays a significant role in a developing
country like India. Moreover, on account of the intricacies involved in handling the NPAs,
the task of managing asset quality in a bank has become quite challenging for the managers,
because a little indiscretion on any front may put a bank into serious trouble. GAs, GNPAs
and GNPA/GA ratio of all SCBs from March 1997 to March 2012 and of all PSBs from
March 1993 to March 2012 are given in Annexure 1A, a graphical presentation of which is
given in the following charts (chart nos 1.1 to 1.3).
Chart-1.1.: GNPA of SCB and PSB
4
Chart-1.2.: GA of SCB and PSB
Chart-1.3.: GNPA Ratio of SCB and PSB
Source: Statistical Tables Relating to Banks in India, RBI various issues (Annexure 1A)
A look at the GNPA curves for PSBs and SCBs exhibit upward trends in the last six years
(though prior to that period over a period of five years there was a declining trend in GNPA
due to number of factors including global recession, for which the GNPA figures increased
very sharply. The GA figures with respect SCBs as well as PSBs reveal existence of
continuous upward trends, which indicate increasing credit appetite in the various segments
of the society (thereby justifying the commonly held impression regarding ‘growth story’ in
the Indian economy). The GNPA ratio [(GNPA/GA) X 100] figures have registered decline
over a period of more than a decade till 2008-09 for PSB and 2009-10 for SCB indicating
apparent improvement over years, primarily because of the many-fold increase in the gross
advances and not due to the activities, like, better management of distress asset through
proper appraisal, follow up and recovery actions. The micro and macro impact of such deadly
virus are as under:
5
Micro or bank specific impact of NPA :
It is like a “double – edged knife” as it not only reduces net interest margin (as
interest cannot be charged on such accounts), but also erodes current profit generated
by other performing assets due to stringent provisioning norms on impaired loans
(Selvarajan & Vadivalagan, 2012).
Existence of distress assets in the balance sheet adversely affect bank’s profitability
due to funding costs and also puts deterrent effect on the solvency of the bank.
Moreover, unexpected increases in Non Performing Loans (NPLs) reduce the coverage
provided by the loan loss reserves, and lead to a deterioration of banks’ liquidity.
Asset quality of a bank is measured inter-alia, in terms of the percentage of NPA to total
advances and therefore a higher level of NPA leads to adverse comments and criticism
by all stake-holders which in-turn, adversely affects bank’s image.
It tends to de-motivate the staff and create investor apathy and shake the customer’s
loyalty. Public confidence depends on the intrinsic strength of banks, their sound
balance sheet and strong fundamentals, which is diluted with the existence of NPA level
of beyond tolerance.
Additional time and efforts required in handling NPA portfolio involve indirect cost
which a bank has to bear for no incremental gain (Srivastava, Bansal, & Sharma,
2013).
NPA creates a serious problem of asset- liability mismatch as the money lent for a
particular period, if not recovered in time, meeting the matching liability will be
extremely difficult and often becomes costly.
The cost of financial intermediation by banks is high because of the cross subsidization
of their NPAs.
Existence of large amount of NPAs in the balance sheet of a bank essentially leads to
bank failure (Hou, 2007).
6
There are evidences that even among the banks that do not fail, there is a negative
relationship between NPA and performance efficiency (Kwan & Elsenbeis, 1997).
Macroeconomic/ industry impact of NPA :
NPA implies bad loans and is a drag on bank’s resources. Scarce resources of the bank
get blocked, restricting the recycling to the productive sectors of the economy. Once the
credit to various sectors of the economy slows down, the economy is badly hit. There is
a slowdown in growth in GDP. NPA is one of the major causes of economic stagnation
problem (Hou, 2007).
Indirectly the burden of NPA is to be borne by the society as a whole. When

government comes to rescue by way of infusion of capital to a sick bank due to erosion
of capital on account of NPAs, the same comes out of government’s budgetary
resources which is contributed by the public in general, either in the form of tax
revenues or from the hard earned savings of the investing public. In any case, the
society is bearing the cost of these NPAs (Sharma, 2005).
NPA in bank is a serious problem and is plaguing the banking industry which is
reflected in decline in productivity and efficiency due to deterioration in the quality of
loan portfolio, restricting income generation and enhancement of capital funds,
accompanied by inadequate loan loss provisions (RBI, 1991).
It is a fact that at present, Indian banks are well-capitalised, with satisfactory capital adequacy
ratio. However, growing distress asset in stressed power and airline sectors is a matter of
great concern for all concerned. (RBI, 2012). Management of NPAs is as such sacrosanct to
ensure effective re-cycling of funds, and improvement of bottom lines. Therefore, reduction
in NPA should be treated as a national priority (GOI, 1998) and it is necessary that NPAs in
banks should be kept at the minimum possible level.
7
1.3. Review of Literature
The term NPA is not very old in Indian banking scenario. In fact, the term was coined for
the first time in 1991 as a part of prudential norms introduced by the committee on the
financial system (CFS) headed by Mr. M. Narsimham in November 1991. The review of
literature starts with careful study of Narsimham committee report on banking sector
reforms. The report reviews major problems of banking sector in India and recommends
suggestions towards raising it to international standard. Major recommendations of the
report are (i) gradual reduction of reserve ratios (SLR & CRR) to statutory minimum level,
(ii) de-regulation of interest rates on deposits and advances, (iii) introduction of prudential
norms on NPAs as also norms on capital adequacy ratio (CAR).
Subsequently, a different committee [Committee on Banking Sector Reforms (CBSR)] was

set up under the Chairmanship of Mr. M. Narsimham in December 1997 to review the
progress of reforms and to suggest remedial measures for strengthening banking system in
India. Major recommendations of the committee were, increase in minimum CAR, reduction
in NPA level and tightening the asset classification and provisioning norms, improvement in
the systems and procedures in the banking to promote efficiency in banking, convergence of
activities of Development Financial Institutions (DFIs) and banks over a period of time,
integration of financial markets and identification of separate specialized roles for different
players, strengthening the operation of rural financial institutions, amendments in legal and
legislative framework to give more autonomy and powers to banks particularly PSBs. Other
literatures on the subject of interest reviewed are:
Meeker and Gray (1987) in their research paper observe that information on bank asset
quality expressed in the form of NPA was made public in 1983 in US banking perspective.
Since then any stake holder could review the asset quality of a bank before taking a decision.
Basic objective of the study is to evaluate information on NPA as an indicator of asset
quality. A regression analysis comparing the NPA statistics with examiner classifications of
assets has been performed. The study suggests that the NPA information can be a useful aid
in analyzing the asset quality of banks, particularly when the information is timely.
8
Suneja (1992) outlines in brief the principal issues and operational aspects of credit
management. The author also prescribes norms and guidelines of sound lending and various
aspects in credit operations including appraisal of loan proposal, the sanctioning process,
disbursement and post disbursement follow up. Health Code (HC) System introduced by RBI
as a reporting and surveillance mechanism on the quality of loan portfolio for the banks in
India has been captured in a comprehensive manner.
Toor (1994) makes an attempt to capture various norms and guidelines issued by the RBI on
NPA and link it with the lending practices of banks. He stresses on the typical nature of
banking business, particularly in the changing scenario and rightly argued in favour of
servicing liabilities by creating assets in a business.
Shukla (1995) narrates in brief basic norms relating to income recognition, asset
classification and provisioning in the early years of introduction of prudential norms by RBI.
Mohan and Kapoor (1996) also contribute immensely through vivid description of frame
work of the prudential norms, followed by norms of income recognition, asset classification
and provisioning. The book also elaborately explains the procedures to be followed by the
banks in reporting NPA related figures.
Bernstein (1996) examines impact of NPAs on aggregate costs and on scale economy
represented by different asset size in banking sector. For the purpose of the study of impact of
NPA on bank cost, the researcher uses a translog cost function where dependent variable is
the sum of interest and non interest cost. The cost model examine NPA ratio, which has a
direct impact on bank cost. The study observes that NPAs have both a direct and indirect
effect on bank cost. Banks with high NPA, other factors remaining constant, will have higher
cost but the direct impact of NPA on bank cost appears to be small. Similarly scale economy
estimates for banks of different asset size, has been generated by translog cost function with
NPA ratio variable. The cost curve, estimated for banks with low NPA ratios, suggests that
scale economies may exist even for the large banks.
Berger and Young (1997) develop three econometric models to examine relationship among
loan quality, cost efficiency and bank capital. The paper sets four hypotheses, based on four
situations, namely ‘bad luck’, ‘bad management’, ‘skimping’ behavior and ‘moral hazard’
9
problems. The paper suggests that relationship between loan quality and cost efficiency move
in both direction. On the one hand, the data supports ‘bad luck’ hypothesis implying increase
in NPL is followed by decrease in measured cost efficiency. On the other hand, from the
industry perspective the data supports ‘bad management’ hypothesis over the ‘skimping’
hypothesis implying decrease in measured cost efficiency are followed by increase in NPL.
Finally the paper supports decrease in bank capital ratios generally precede increase in NPL
for banks with low capital ratio implying low capitalized banks may respond to moral hazard
problem by taking increased portfolio risk. Hence, cost efficiency may be an important
indicator of future problem loans and problem banks.
Joshi and Little (1996), describe the impact of financial sector reforms during 1991-2001.
With the help of statistical figure they make an attempt to explain that reforms in Indian
banking sector was not only the emerging solution to the crisis but also the only way for them
to come out of the shattered bottom lines and poor image.
Rangarajan (1997) argues that the broad objective of reform has been to create a viable and
efficient banking system. He further states that the four building blocks on which banking
sector reform stands include (i) modifying the policy framework, (ii) improving the financial
soundness and credibility of banks, (iii) creating a competitive environment and (iv)
strengthening of the institutional framework.
Tarapore (1998) presents a brief outline of backdrop of first phase of reform, progress made
in the banking sector during first phase of reform and advocates in favour of second phase of
banking sector reform with the following agenda: (i) re-organization of structure of banking
industry with the introduction of liberalized policy of opening of bank in private sector, (ii)
stricter prudential norms for capital adequacy as well as assets quality to take the banking
industry to international standard, (iii) improvement in the operating profit as an indicator of
success in the industry, (iv) further liberalisation on directed credit, (v) autonomy in banks,
(vi) merger or re-oganisation of weak banks, (vii) effective regulatory and supervisory
framework to handle increasing burden of NPAs.
Shajahan (1998), Naik (2006) and Poongavanam (2011) define NPA, discuss NPA norms
across the world, explain Health Code System prevalent in India before introduction of
Prudential Norms and basic Prudential Norms on income recognition, asset classification and
10
provisioning along-with factors responsible for growing NPA and effective mechanism
towards management of distress asset in commercial banks in India.
Ahluwalia (2001) observes that the experience of 1st generation reform is quite encouraging
as the performance of the banks has improved and level of NPA register significant decline.
However, he stresses that there is lot of scope of improvement as the asset quality
international bench mark still could not be reached.
Chansarkar (2001), Dayal (2001) and Govil (2001) present a brief progress and prospect
status of Indian banking in various areas including asset quality, profitability and
transparency and recommend in favour of implementation of second phase of reform
packages.
Misra and Dhal (2001) and Ranjan and Dhal (2003) examine the factors responsible for
increasing levels of NPAs and observe that NPLs are influenced by three major sets of
economic and financial factors, i.e., terms of credit, bank size induced risk preferences and
macroeconomic shocks. A panel regression model for forecasting GNPA ratio has been
developed. The empirical results from panel regression models suggest that terms of credit
variables have significant effect on the banks' NPL in the presence of bank size induced by
risk preferences and macroeconomic shocks. The study also observes that factors like
maturity of credit, better credit culture, and favorable macroeconomic and business conditions
lead to lowering of NPAs. Business cycle may have differential implications adducing to
differential response of borrowers and lenders.
Bhide, Prasad and Ghosh (2002) observe that the most critical issue in Indian banking is that
of its bad loans which has been understated over years due to lax regulatory provisions
towards asset classification and provisioning in banking in India.
Vasishtha and Rajaraman (2002) examine the relationship between NPA and operating
efficiency. With the help of panel regression on a time series data covering a period of 5
years i.e., 1995-96 to 1999-2000 of 27 PSBs. Two runs are attempted, one with capital
adequacy as the bank specific indicator of solvency and the second as operating profit as a
per cent of working funds as the indicator of operating efficiency. The estimated coefficient
of capital adequacy in the first run carries the expected sign but is found not statistically
11
significant. Thus, the fixed group effect in that run basically separate the banks into those
carrying statistically significant higher or lower NPAs than the mean for the pool. In the
second run, the coefficient for operating profits carries the expected negative sign and is
observed statistically significant. Two banks with higher than mean NPA carrying
insignificant intercepts after controlling for operating profit, clearly indicate that the high
NPAs in this case correlate low operating efficiency. But there are four banks with high
positive and significant intercepts, even after controlling for operating efficiency. In all cases,
the intercept is lower with operating efficiency thus showing that low operating efficiency
explains some, if not all of the NPA stock.
Machiraju (2003) and Shekhar & Shekhar (2005) develop a conceptual framework within
which the banks and other financial institutions operate. Attempt has been made to explain
the nature and role of financial system including commercial banks, management of funds,
advances and investment in addition to management of risks in banking with focus on credit
risk, asset quality and non performing assets. Donald and Koch (2006), focus on key issues
confronting decision making with particular reference to management of credit risk in the
bank.
Chaitanya (2004) in his paper on ‘Causes of NPA in Public Sector Banks’ examines with the
help of three year moving average the trends in NPA based on four ratios (GNPA to Gross
advance, GNPA to Total Assets, Net NPA to Net Advance and Net NPA to Total Assets) and
also attempts to forecast future value of NPA for next 3 years. Secondary data for the time
period 1992-93 to 2001-02 has been captured. The study examines various macro and micro
perspective of NPA in Indian banking system and recommends two prone strategies to deal
with this major problem as under: (i) reduction of existing NPAs and (ii) restriction of further
accumulation of NPAs.
Sharma (2005) in her research paper makes an attempt to study the problem of NPAs in PSBs
and also its impact on the performance of these banks. Impact of NPAs on the profitability of
the banks is analyzed by applying multiple regression. Impact of NPA on productivity,
represented by business per employee (BPE) and operating profit per employee (OPE) and
few other strategic banking variables like capital adequacy, credit deployment and deposit
mobilization along with some macroeconomic variable have also been examined with the
help of simple regression model. The issue of auto correlation and multi co-linearity has
12
been diagnosed with the help of Durbin Watson Statistics (DWS) and Variance Inflation
Factor (VIF) respectively. She observes that NPA not only affect performance of the banks,
but also cause tremendous damage to the entire economy by dampening the very foundation
of the credit system. The author suggests different preventive and curative measures to
control NPAs in PSBs and urges that efficient legal framework, improvement in credit
appraisal and monitoring skills of banks backed by strong political will may enable the Indian
banks to tackle the burning issue.
Gopalakrishnan (2006) examines the impact of NPAs on various micro and macro economic
variables, with the help of simple regression analysis. It is observed that there is significant
relation between NPAs and strategic variables stated above with very high value of R2.
Simple regression models generated are highly significant with F test result with significance
of less than .05 and moderate to very high R2 value ranging from 0.525 in case of net profit
to 0.920 in case of interest rate on advances. A multiple regression model has also been
employed to explain impact of NPA along with a few other strategic banking variables on net
profit of PSBs. The author has also developed a statistical model for forecasting future value
of gross advances, standard advances and NPAs with the help of regression analysis. Time
series data for nine years have been taken and regressed with a few strategic variables under
two scenarios to enable prediction of the above parameters for 11 years as a matter of
reference for the bankers in formulating their strategies.
Reddy, Babu, Mallikarjuna and Viswanath (2006) make a serious attempt to examine trend of
NPAs in PSBs in India with the help of variables like GNPA, NNPA, GNPA to GAs, GNPA
to total assets, NNPA to NAs, NNPA to total assets on time series data from 1994 to 2004.
With the help of statistical table, they observe that in ratio terms, NPAs have gone down
gradually. Though NNPA in actual term has also declined after a moderate rise, GNPA has
shown upward trends throughout. The researchers examine the level and trend of NPA in
individual PSB, with the help of Box Plot. GNPA ratios of individual banks have been
plotted on the Box Plot and it is observed from the chart that the boxes are skewed upward as
many banks have higher NPA ratio than median. Box Plot also helps to isolate the odd-men-
out (outliers) with staggering levels of NPA.
Bodla and Verma (2006) in their empirical research work, ‘Determinants of Profitability of
Banks in India : A Multivariate Analysis’, examine the impact of banking variables like
13
spread, net interest income, credit deposit ratio, NPAs as percentage of net advances,
provision and contingencies, operating expenses, business per employee on profitability in
PSBs in India. They use a time series data covering 1991-92 to 2003-04 and employ Multiple
Regression Technique. The study has brought out that the explanatory power of some
variables like, spread, net interest income, provision and contingencies and operating
expenses are significantly high while others like business per employee, credit deposit ratio,
NPAs as percentage of net advances are found with low explanatory power. Durbin Watson
Test (DWT) has been employed to examine the problem of auto-correlation in the time series
data employed herein. Moreover, to bring out the explanatory powers of each of the
independent variables under study, the square of partial correlation coefficient has been
worked out. The study concludes that variables such as provision and contingencies, non
interest income, spread and operating expenses have significant relationship with net profit
and therefore are major areas of concern in PSBs in India
Chipalkatti and Rishi (2007) examine whether weaker Indian banks have a tendency to under
provide for loan loss provision and understate gross NPA in order to boost their earnings and
capital adequacy ratio by examining bank data over 1996-2002, the period when banks in
India faced increasing competition and tight regulatory pressure. The study develops two
models to examine the issues and reveals that weak Indian banks identified by low
profitability and low capital ratio have understated their bad loans in the post 1999 period and
conclude that the magnitude of NPA in these banks may be more grave than is claimed by
authorities.
Pal and Malik (2007) make a serious effort to measure the performance of commercial banks
in terms of various financial characteristics established in financial ratios. Parameters of the
study primarily included profitability and efficiency ratios popularly used in the industry. To
investigate the factors of financial performance for making demarcation of commercial banks
on the basis of ownership, the most widely used statistical technique, ie Multinomial Logistic
Regression Method is employed on the pooled and individual years’ data. To make the
regression model free from multi co-linearity problems two models are selected which
include the independent variable having linear relationship. Both these models are estimated
by using the forward stepwise procedure for variable selection. It is found that foreign banks
came up with a dominating performance in India over PSBs and Private Sector Banks
(PrSBs) during the period under study. The foreign banks (FBs) are proved to be best
14
performers in generating business with a given level of resources implying they are better
equipped with banking skill, technological upgradation and managerial practices in
comparison to its competitor groups. They are also found to be more consistent with market
system as reflected in terms of net interest rate margin. The PSBs emerged as the next best
performer after foreign banks and are characterized by higher return on equity in comparison
to FBs and PrSBs and in terms of economizing their expenses as reflected in the expense ratio
and efficiency ratio. PrSBs have however emerged with a better performance in utilization of
resources as compared to PSBs.
Sethi and Bhatia (2007), introduce the readers to the banking business, with exclusive and
detailed discussion on banking sector reforms. Issues, dimensions and management of NPAs
in commercial banks have also been deliberated with reasonable precision.
Vallabh, Bhatia and Misra (2007) undertake an empirical approach to analyze the
movement of NPAs in SCBs, group-wise. The study establishes the results on the basis of the
impact of macroeconomic variables and bank specific variables on NPAs of a particular
group of banks. First of all, a few representative variables related to banks such as credit
deposit ratio and some macroeconomic variables such as GDP growth rate are chosen, and
movement in NPA is assumed to be explained by the movement of these representative
variables over the years. For the purpose of the study, regression analysis is undertaken to
examine sensitivity of NPAs with selected variables for different categories of SCBs on
dataset from 1996 to 2004. To check auto co-relation and co-linearity among independent
variables the study has used VIF statistics, while to confirm the mutual independence of
chosen independent variable DWT has been employed. The study has found that for PSBs R2
is 0.971 which implies the independent variables explain 97.1% of variation in NPA. In case
of PrSBs, R2 is 0.817 which implies the independent variables explain 81.7% of variation in
NPA. For FBs R2 is 0.640 which implies the independent variables explain 64.0% of variation
in NPA.
Vyas, Singh and Kashif (2007) in their paper make a serious attempt to establish relationship
between Capital to Risk Asset Ratio (CRAR) and NPAs and to study the impact of
determinants of credit on the credit growth in Indian Banks. Correlation analysis has been
used to examine the relationship CRAR and NPAs and regression analysis to identify
variables having significant impact on credit growth. The study observes that R comes to be
15
negative, which implies that there exist a negative relationship between CRAR and NPAs. It
is interesting to note that CRAR has negative impact on NPAs only for PSBs. The study also
examines, with the help of multiple regression analyses, that CRAR, NPAs along with a few
strategic banking variable have influence on credit growth (in %) and found that Indian
evidence makes capital requirements an attractive regulatory instrument since they serve to
reinforce the stability of banking system without apparently distorting the credit flows of the
banks.
Singla (2008) examines empirically the financial performance in terms of profitability of

sixteen selected banks for a period of five years (2000-01 to 2006-2007). The study reveals
that the profitability position was reasonably satisfactory during the period of study when
compared with the previous years. With the help of statistical measures like correlation
analysis and multiple regression analysis, the study examines the determinants of profitability
of selected banks. During the study period, it was observed that the return on net worth had a
negative correlation with the debt equity ratio. Interest income to working funds also had a
negative association with Interest coverage ratio and the NPA to net advances are negatively
correlated with interest coverage ratio.
Agnani (2010) examines relationship between NPAs and financial health of banks by
analyzing the impact of NPA on profitability of banks. Two PrSBs and three PSBs have been
identified for the study where structured questionnaires were used to find out the preference
of the respondent with respect to his choosing a particular bank for investment. Spearman’s
Rank correlation has been used for identification of the best bank. It is inferred in the study
that lower profitability or higher NPA taken in isolation do not reflect the investor’s
preference in deciding performance and future direction of success or failure of a bank in
real and absolute term. In fact, an investor gives higher weight to goodwill and customer
service of individual bank than NPAs or profitability.
Ahmed (2010) observes that deterioration in asset quality of the banks is primarily on account
of increasing level of overdue, which restricts recycling of funds, with negative impact on
socio-economic progress in the economy. He argues that improving the quality of loan asset
is the acid test of efficiency of a bank. The paper attempts to investigate empirically the asset
quality and loan recovery of Indian commercial banks and establishes a relationship between
growth in advance and growth in NPA of a bank. The paper also examines various factors
16
that affect NPAs of a commercial bank and presents a multi regression model in forecasting
NPA of a bank with a few strategic banking variables like priority sector advances (PSA),
credit deposit ratio (CDR), capital adequacy ratio (CAR), NPA to advance ratio as
independent variables. The regression equation has been fitted with dataset pertaining to the
period of eight years, ie, 2000 to 2007. It is observed that PSA, CDR, CAR and NPA to
advances ratio may not be considered as a very good determinant of dependent variable, ie,
NPAs. The researcher however concludes that NPA is an important parameter to assess the
financial health of a banking company as it reflects asset quality, credit risk, and efficiency in
the allocation of resources in deficit sectors and therefore various initiatives have been taken
to contain growth in NPAs.
Dash and Kabra (2010) examine determinants of NPLs in India considering both macro
economic and strategic banking variables. With the help of regression analysis an
econometric model is developed using a panel data set covering 10 years (1998-99 to 2008-
09) ) to examine the relationship between non-performing loans and several key
macroeconomic and bank specific variables, for predicting future value of NPA. The study
observes that the real effective exchange rate and the changes in real income as reflected by
growth in real GDP, have a significant positive and a significant negative relationship
respectively with NPAs. It also finds that commercial banks that are aggressive and charge
relatively higher interest rates incur greater NPLs.
Goyal (2010) in his analytical study strives to examine the status of NPAs in PSBs in India
for the period 2002-03 to 2008-09. Attempts were made to examine asset quality in PSBs in
general with sector wise performance in particular, with the help of statistical tools like
descriptive statistics, correlation, regression analysis etc. Major findings of the study include:
• Increase in gross and net advances over the period under study.
• Decline in GNPA ratio and NNPA ratio, which imply improvement in asset quality in
PSBs.
• All three categories of loan asset, ie, sub-standard asset, doubtful asset and loss asset have
registered declining trends in percentage terms which imply overall improvement in
recovery in PSBs.
17
• Though NPA in priority sector advance has registered downward trend, NPA in
agriculture sector has increased during the period under study.
The study concludes that prudential norms have helped banks to improve their performance
by trimming down the level of NPA.
Gupta and Jain (2010) examine factors like credit spreads, collateral, long-term structures and
commitments between borrowers and lenders over time, which are responsible for the
cropping up of distressed asset from the perspective of the financiers. An econometric model
has been captured to predict NPAs, taking into account bank level micro economic variables
in the selected banks. The study suggests a multi pronged approach for solution to the
problem of NPAs by addressing policy issues, strategic issues, and restructuring issues, legal
issues, reporting issues, supervisory issues, operational and procedural issues.
Rajeev and Mahesh (2010) in their exploratory paper examine the trends of NPAs in India
from various dimensions. The study of trend in this paper has been examined on both
GNPA, NNPA and sector wise distribution of NPA both on absolute figure of NPA as well as
on ratio data with a time horizon of 2002-2009. The study observes that decline in NNPA is
sharper than GNPA because of the increasing level of provisions. The study also observes
that mere recognition of the problem and self-monitoring has been able to reduce it to a great
extent. It also shows that PSBs in India, which function to some extent with welfare motives,
have performed rather well in reducing NPAs as their counterparts in the private sector.
Saluja and Lal (2010) in their comparative analysis of NPAs in different groups of
commercial banks in India, examine the status and trends of NPAs in commercial banks,
group wise as well as sector wise with a data set from 2004-05 to 2008-09. The study
observes that NPAs reduce the profitability of banks, weaken its financial health and erode its
solvency. For comparing the performance of public and private sector banks and foreign
banks in India with special reference to their NPAs, four banks each from public sector (State
Bank of India, Allahabad Bank, Bank of Baroda and United Bank of India), from private
sector (Axis Bank, HDFC Bank, ICICI Bank and IndusInd Bank) and from foreign banks
(Citibank, Deutsche Bank, HSBC Bank and Standard Chartered Bank) are selected. A
comparative analysis of all three categories is made on the basis of gross NPAs and net
NPAs. NPAs are also categorized into priority and non-priority sector for purpose of analysis.
18
The study concludes that though NPAs are present in every group of banks, PSBs are under
severe pressure of NPAs as compared to its counterparts i.e., PrSBs and FBs. Moreover,
greater quantum of NPAs is observed in non-priority sector than in priority sector.
Chaudhary and Sharma (2011) in an attempt to analyze efficiency of PSBs and PrSBs in
managing NPA, study trends and progress in NPA of PSBs and PrSBs, individual bank wise
as well as sector wise with the help of statistical table. The study observes that PSBs in India
are doing well in competing with PrSBs and FBs in the matter of management of NPA. The
study also examines various measure of management of NPA and recommends timely
implementation of suitable and stringent measures for efficient management of NPA in
banking sector.
Ghosh and Ghosh (2011) focus on composition, trend and management of NPA in PSB. The
study observes that NPA is an important parameter in the analysis of financial performance of
a bank as it results in decreasing margin and higher provisioning requirement for doubtful
debts. Therefore reduction of NPA is necessary to improve profitability of banks and comply
with the capital adequacy norms as per the Basel Accord. This study traces the movement of
the NPAs present in PSBs in India by analyzing the financial performance of the banks with
respect to key performance indicators and management of the NPAs under the purview of
new policy actions and regulatory compliance of the RBI. The study concludes that the
quality of advances in PSBs have deteriorated due to aggressive lending policy persuaded by
the banks. However, PSBs have been able to control NPAs by virtue of regulatory norms of
RBI and their close monitoring.
Hosmani and Hudagi (2011) in their study of magnitude and trend of NPA in PSBs in India,
observe that the level of NPA is very high and controlling the growing NPA is a challenging
task in Indian banking in general with PSBs in particular. The study examines association
between gross advances (GA) and GNPA though r = .838, there on statistically significant
relationship between GNPA and GA. The study also observes that GNPA and NNPA figure
in absolute term is on the rise though in ratio term (GNPA ratio and NNPA ratio), they are
gradually falling. As regards sector wise analysis of NPA, it is observed that NPA in priority
sector advances is still very high at 57.1% of the total NPA. The study recommends that
PSBs have to initiate timely remedial actions against degradation of performing assets.
19
Jayasree and Radhika (2011) in their paper examine relationship between net profit and
NPA’s and total advances. The impact of NPA on net profit and impact of total advances on
NPA are also examined with the help of simple regression analysis. For the purpose of
examining associations between two variables, correlation analysis is computed, while for
examining the relationship, simple regression is also computed. The study observes that there
exists negative relationship between NPA and net profit and NPA in SCBs, which implies
that with the increase in NPA, net profit will decline. The study also observes positive
correlation between NPA and Total advance in SCBs which means that with the increase in
total advance, NPA will increase. The regression analysis explains that NPA has negative
impact on Net profit whereas Total advance, has positive impact on NPA in SCBs. The study
concludes that banks are performing well in managing their NPA, mainly because of RBI
prescriptions and follow up of prudential norms, introduction of securitization in the country
and change in the legal framework.
Malyadri and Sirisha (2011) examine trend of NPAs in weaker sections in both PSBs and
PrSBs in India. The study is based on the secondary data for seven (7) years i.e. from 2004-
2010. The data has been analyzed by statistical tools such as percentages and compound
annual growth rate (CAGR). The study observes that the PSBs have achieved a greater
penetration compared to the PrSBs so far management of NPAs in the weaker sections is
concerned.
Nandy (2011) in his empirical research work observes that differential profit performance has
been attributed to various factors. The objective of the study is to identify these factors and
examine whether they have any significant influence on profitability of commercial banks in
India. For the purpose of the study a few variables, like interest income, other income,
interest expenses, operating expenses, net NPA and spread are identified. With a database
covering a period of 2004-07, the study employs multiple correlations and multiple
regression technique to examine the relation between above named variables with
profitability. The study concludes that interest expenses are the only good predictor for
profitability in a bank for the given dataset.
Rawlin and Sharan (2011) in their research paper, ‘Modeling the NPA of a Large Indian
Public Sector bank as a function of Total Assets’ make a sincere attempt to develop a
forecasting model for the NPA % both at the gross and net level of one of the India’s largest
20
PSB. A strong correlation has been observed between gross and net NPA% and the total
assets suggesting that the estimate of gross and net NPA can be made from total assets, by
fitting to linear and non linear models. A non linear curve estimation model is found to be
best fit by virtue of highest R2 value linking both gross and net NPA to advances, provide the
best curve fit and the least deviation from actual values. Thus by studying total assets an
overall picture of the banks, NPA level can be ascertained and effective strategy can be
formulated to address the most formidable problem in the banking industry.
Thiagarajan, Ayappan and Ramachandran (2011) in their empirical study make an attempt to
predict the determinants of the credit risk in the banking sector in India by using an
econometric model. The study identifies both macroeconomic as well as micro economic
bank specific variables and observes a high R2 for both PSBs and PrSBs which is a reflection
of the fitness of the model and its predictability. For the purpose of the study descriptive
statistics including mean and standard deviation of the selected variables have been captured.
Trend in the inflation and GDP growth for the period 2001 -2010 shows an exponential
increase and decline respectively. It is inferred by looking at the graphical representation of
nonperforming loans and the rate of GDP growth and inflation that there is an inverse
relationship between GDP growth and NPA while a positive co-relation between inflation
and NPA for the recent past 2 years.
Thiagarajan, Ayyappan, Ramachandran and Sakthivadivel (2011) in their paper empirically

evaluate the determinants of profitability in the public and private sector commercial banks in
India. Correlation analysis, multiple regression analysis and factor analysis have been used to
estimate the contribution of select bank specific variables towards profitability. The
correlation coefficient of the selected independent variables with the bank’s profitability has
been worked out in order to identify the most significant variable that has strong relationship
with the dependent variable. The study also examines the impact of several independent
variables on profitability for which multiple regression analysis has been undertaken. The
study also uses factor analysis to examine the value for the coefficient for regression when
the variables are regressed upon the factors. In factor analysis the varimax rotation has been
used. The study reveals that credit risk represented by NPA has a significant negative
influence on the profitability on both private and public sector banks. It also establishes that
NPA is positively influenced by GDP and negatively influenced by inflation. The study
concludes that with the recent trend of lower GDP and higher inflation, NPA is likely to rise
21
and as a consequence profitability is likely to decline. The multiple regression analysis
highlight that the return on investments and return on advances has a significant influence on
the profitability of private sector banks.
Yadav (2011) in his research paper examines the concept of NPA in PSBs, its magnitude and
impact. The primary objectives of the research are to analyze the impact of NPA on (i)
profitability, (ii) profitability with other variables and (iii) efficiency and productivity. For
the purpose of the study, secondary data for the period of 1994-95 to 2005-06 has been
collected from RBI Publication. Simple linear regression model is used to analyse the impact
on profitability represented by profit as a percentage of total asset and on productivity
represented by business per employee and profit per employee of public sector banks. It is
observed that NPAs in PSBs, in absolute terms, has shown increasing trend till 2001 and
declined later on, whereas its percentage has shown declining trends. The study also finds
that NPA has a negative impact on profitability and is statistically significant and also has
negative relationship with business per employee (a productivity variable) and magnitude of
relationship is statistically significant. The profitability of all PSBs is affected to a great
extent when NPAs working with other banking strategic variables and also affect productivity
and efficiency. High R2 value in the regression model explains variability of high order in the
productivity and efficiency parameter of PSBs.
Agarwal and Mittal (2012) evaluate the operational performance of the selected PSBs and
PrSBs in India and make a comparative study of status and progress of NPA in PSBs and
PrSBs with the help of time series data from 2001-02 to 2010-11. Two banks from each
group, namely SBI and PNB from PSB group and HDFC and ICICI from PrSB group have
been picked up for comparison. Various statistical tools like measures of central tendency,
frequency distribution, standard deviation, coefficient of variation and ANOVA test, along
with suitable ratio analysis have been employed to analyse and interpret the data. While
examining the issue of asset quality of PSBs and PrSBs, the study observes that the
magnitude of NPA is comparatively higher in PSBs as compared to PrSBs with the exception
of recent couple of years under study, when ICICI bank, a PrSB representative is exhibiting
higher NPA figure than it’s PSB counterparts.
22
Pradhan (2012), in his study of trends of NPA in PSBs during post reform period, examines
trends in GNPA, GNPA Ratio, PSB group-wise as well as individual PSB wise, and observes
that rate of decline in GNPA in absolute term has been extremely low for PSBs group-wise
during the last decade. However asset quality in PSBs in ratio terms like GNPA to gross
advances has reduced remarkably during post reform period. With regard to individual bank’s
performance in management of NPA, it is found that Andhra Bank and Indian Bank are the
leaders during the period under consideration.
Rajamohan (2012) makes an attempt to examine the status, growth and sector wise movement
of NPA in PSBs, group wise with dataset covering a period from 2001-02 to 2010-11. The
study observes that NPA figure in absolute term has registered consistent fall for six to seven
years for both Nationalised Bank (NB) Group and State Bank (SB) Group after which NPA
rose very sharply for the remaining three to four years. With regard to NPA in priority sector,
the study also observes that it remained moderately stable till 2009 beyond which NPA figure
rose quite sharply. The study examines whether there is significant difference in the growth
rate of NPA between NB Group and SB Group with the help of Mann Whitney U Test and
observes that there is significant difference between the growth rate of NPAs in these banks.
Rajput, Gupta and Chauhan (2012) provide an empirical approach to the analysis of
profitability indicators with a focal point on NPAs of commercial banks in the Indian context.
The paper examines factors contributing to NPA, its magnitude and consequences. The study
uses observation method as well as statistical tools like correlation analysis and regression
analysis. Two regression equations have been developed to explain the relationship. The
study identifies that there is a negative relationship between profitability and NPAs in PSBs
in India and concludes that NPA affects performance of the banks significantly. On the other
hand, factors like better credit culture, managing the risk and business conditions will
certainly lead to lowering of NPAs.
Rawlin, Sharan and Lakshmipathy (2012) in their study, ‘Modeling the NPA of a Midsized
Indian Nationalised Bank as a Function of Advances’, examine the relationship between
gross and net NPA% of a bank with its aggregate advances. A strong correlation is observed
between gross and net NPA% and the advances made. Based on the above observations
attempt is made to predict gross and net NPA% from advances by fitting to linear and non
linear models. A non linear curve estimation model linking both gross and net NPA to
23
advances provided the best curve fit and the least deviation from actual values. Thus by
simply looking at advances an overall picture of the banks NPA level can be ascertained.
Siraj and Pillai (2012) investigate the performance of Indian SCBs before and after global
financial crisis (2007-09). They examine various aspects of performance and asset quality of
those banks, group wise. For the purpose of analyzing the trends average annual growth rate
(AAGR) has been computed for selected variables and the output is presented graphically for
meaningful inference. They also examine movement of various NPA indicators, gross NPA,
net NPA, additions to NPA, reductions to NPA and provisions towards NPA and compare it
with total advances and total deposits of banks. The study examines bank-group wise
performance data for post-millennium period from 31st March 2001 to 31st March 2011. The
study also examines the impact of gross advance and total deposit on incremental NPA with
the help of simple regression analysis. A multiple regression model has also been developed
with NPA as a dependent variable with total deposits and total advances as independent
variable to examine the total effects of these variables in addition to NPAs of the bank. The
study observes that increased level of addition to NPA still remain a major concern for banks
in India. Though NPA shows improved recovery during first half of last decade, it remained
challenging during second half of the period. A notable result is the financial stability of
PSBs and increased susceptibility of PrSBs and FBs during financial crisis.
Veerakumar (2012) in an attempt to study NPA in priority sector advances (PSA) in a

commercial bank examines time series data from 2000-01 to 2009-10 of both GNPA and
NNPA (on absolute figure as well as ratio data) and presents trend analysis with the help of
curve estimation regression technique. It is observed that a polynomial regression model,
particularly a cubic curve is the best fit with very high R2 value, signifying a dependable
forecasting model. The researcher has also used multiple linear regression analysis to
examine the influence of different variables on gross NPA of SCBs.
Selvarajan and Vadivalagan (2013) examine the nature of the NPAs under the priority sector
advances (PSAs) in Indian Bank, a PSB in India and compare the same with PSBs in
aggregate and suggest measures to curve NPAs in Indian Bank. The PSAs have been
analyzed in detail under three major heads, viz., agriculture, small scale industries and other
priority sector along with weaker section advances, which forms part of PSAs. The study
observes that average percentage of PSAs by Indian Bank is 37.84% of its total advances,
24
while NPAs in PSAs accounts for 58.71% of total NPAs of Indian Bank. Further, the mean
percentage of GNPAs to total advances in Indian Bank is 2.73. The mean percentage of
NPAs in PSAs to total PSAs is arrived at as 4.11. Thus incidence of NPAs is more in PSAs.
The study suggests that Indian Bank has to restrict its advances to PSAs and boost up
recovery measures to improve its NPAs portfolio under PSAs.
1.4. Identification of Research Gap
A number of factors make the NPAs in PSBs in India an interesting subject for study.
 First, during the 1990s, India underwent liberalization of the banking sector with the
objective of enhancing efficiency, productivity, and profitability (RBI, 1991).
 Second, the banking sector passed through an important transformation, driven by the
need for creating a market-driven, productive, and competitive economy in order to
support higher investment levels and accentuate growth (GOI, 1998).
 Third, studies on NPAs in banking industry in emerging economies like India has great
relevance, as it dampens the bottom-lines and thereby poses a serious threat to the very
existence of the most important sector in propelling the desired growth and development
of the economy.
In view of the seriousness of the problem, numerous research studies have been conducted
on different issues concerning credit risks in banks including NPAs. However, empirical
works on NPA problems in PSBs are found inadequate. The exhaustive review of literature
on NPA demonstrates that majority of the research work has been undertaken on aggregate
PSBs data with primary focus on the areas which include
i) Trends of NPA in PSBs (Chaitanya (2004) and Reddy, Babu, Mallikarjuna &
Viswanath (2006)).
ii) Impact of NPA on micro as well as well macro economic variables (Sharma (2005),
Gopalakrishnan (2006), Bodla & Verma (2006), Singla (2008), Thiagarajan,
25
Ayyappan, Ramachandran and Sakthivadivel (2011), Nandy (2011), Yadav (2011),
Jayasree and Radhika (2011), Rajput, Gupta and Chauhan (2012)).
iii) Future value of NPAs with the help of simple/ multiple regression models (Berger &
Young (1997), Ranjan & Dhal, (2003), Misra & Dhal (2003) and Gopalakrishnan
(2006)). Forecasting NPA employing data of past loans (Graham & Humphrey (1978)).
However empirical work on individual bank-wise trend study and modeling for the purpose
of forecasting future value and conducting impact analysis on NPA in PSB has seldom been
attempted. It is against this backdrop that the present study is undertaken to fill up this gap
and make a modest contribution in the field of management of NPA in banks in India.
Accordingly, examination of movement of NPA over time and forecasting the same for
medium term with the help of univariate time series data by employing parametric Curve
Estimation Technique of Regression Analysis, non parametric LOESS Curve Fit Technique
and semi parametric Penalised Spline Curve Fit Technique for six selected PSBs and PSB in
aggregate has been attempted in the thesis. Similarly, for the purpose of examining the impact
of NPA on strategic banking variables, along with time variable in selected PSBs as well as
PSBs in aggregate, with the help of simple as well as multiple regression analysis have been
undertaken.
1.5. Objective of the Study
In view of the relative importance of NPAs in banking sector in India in general with PSBs in
particular, it is perceived that a comprehensive study in this area should be made. The present
study is a humble endeavour to examine various aspects of NPAs in selected PSBs and PSBs
in aggregate. The specific objectives embodied under the research are as follows:
i) to review the existing norms and concepts related to NPA,
ii) to review the mechanism for effective management of NPAs as initiated by PSBs
in India,
26
iii) to examine the overall trends of NPAs and to explore the dynamicity of NPA as the
variable under study over time in the PSBs in general with a few selected PSBs in
particular,
iv) to examine the status of inter relationship as is existed among the set of relevant
banking variables including NPA,
v) to examine the impact of NPAs on profitability and other strategic banking variables
along with time variables, and
vi) to suggest future action plan based on the findings emanated from the points iii, iv and v
as mentioned above.
1.6. Research Methodology
The study undertakes an empirical approach to analyse the movement of NPAs and impact of
NPA in six selected PSBs along with PSBs in aggregate in India during the last two decades,
based on secondary data related to the strategic banking variables including NPAs. The
secondary data have been collected from RBI publications, Prowess Database of Centre for
Monitoring Indian Economy and the annual reports of the selected PSBs.
The study includes examination of movement of NPA over time, developing forecasting
model for medium term with NPA as dependent variable and examining the impact of NPA
along with time variable on strategic banking variables. For the purpose of our analysis, a
time series data-set on parameters like GNPA and GA for 17 years i,e March 1996 to March
2012 for six selected PSBs along with PSBs in aggregate have been captured, screened and
collated. The data thus obtained has been tabulated, and analysed by using appropriate
financial ratios. Suitable statistical techniques (SPSS 19) and software packages (R 3.0.2)
have been extensively used in order to draw relevant inference.
With respect to the first objective, a detailed review of existing literature and RBI
circularized instruction has been studied and a comprehensive note on norms and concepts
of NPA alongwith has been delineated.
With respect to the second objective, a review of existing mechanism for effective
27
management of NPAs as initiated by PSBs in India has been delineation based on
researcher’s personal interaction with recovery cell officials of such banks and RBI
circularized instructions.
To examine dynamicity of NPA over time as stated in the third objective, parametric, non
parametric and semi parametric and nonlinear regression models have been invoked to
enhance the precision level of the fitted curves. The techniques of regression diagnostics
have been employed to judge the validity of the models. Towards this end, Box Plot has
been used to identify outliers. DW Test and t test have been employed to diagnose the
existence of independence in the data-sets in respect of the parameters GNPA and GA. To
examine the normality or non-normality of the distributions followed by the different data
sets on the above parameters, normal Q-Q plots have been constructed apart from finding
out the critical values (from table) of the SW Test Statistic. To determine the features of the
above data sets, values of the simple statistical measures, like, mean, range, standard
deviation, coefficient of variation along with measures of the properties, namely, skewness
and kurtosis have been computed. Parametric models, using time series regression analysis
(Curve Estimation Technique) have been employed to analyse the trends as are existed in
the above data-sets relating to the parameters, GNPA and GA respectively. Finally, the best
fitted model to represent the trend has been obtained in case of each data-set, and the
forecast values based on the said best fitted model for the respective data-set have been
generated. Non parametric and semi parametric models have also been developed by
employing LOESS Curve Fit and Penalised Spline technique respectively.
With respect to the mechanism for effective management of NPAs as initiated by PSBs in
India, pair wise correlation co-efficient have been computed and appropriate tests have
been employed to judge the significance status of these co-efficient. This has enabled us to
select the appropriate variables to be considered for further study.
To understand the impact of NPA as mentioned under objective number five, banking
variables selected by employing pair wise correlation under profitability and productivity in
addition to strategic banking variables like net interest margin, credit-deposit ratio, capital
adequacy ratio, credit growth and priority sector advances to total advances have been
regressed. Model validity has been studied to finally determine the most appropriate model.
28
To study the impact of NPA on the profitability of banks firstly simple regression have been
used to measure the relationship between two variables. But as there are number of factors
affecting the profitability, it is not appropriate to judge profitability with NPA alone, rather
few other strategic banking variables, along with NPAs have been considered through
multiple regression models which consist of determining a regression equation explaining the
relationship between dependent variables (profitability) with the independent ones. Similarly,
simple regression has been used to examine statistically significant relationship between
productivity related variable like OPE and other selected variables with NPAs. OPE or other
selected variables have been taken as dependent variable and NPA as independent and
models have been developed separately for individual banks and PSBs in aggregate for each
variable. For examining the impact of NPA on OPE and other selected variables as stated
above along with time variable, multiple regression analysis has been undertaken and model
developed separately under each case.
Assumptions of linear regression models, ie, independence of the errors and normality of the
error distribution have been verified with the help of DW statistics and SW statistics
respectively. Significance of the models has been checked at five percent level of F values.
Model coefficients have been checked at five percent level of t values. Also to diagnose the
issue of multi co-linearity in multiple regressions, the value of VIF and Tolerance Factor
(TOL) have also been computed and compared with standard.
1.7. Scope of the Study
SCBs in India comprise of the PSBs and the PrSBs. PSBs consists of the State Bank of India
(SBI) and its five subsidiaries, collectively termed as the SB Group and nineteen nationalized
banks, commonly referred to as NB Group and one other PSBs, ie, IDBI Bank. The twenty
PrSBs are further subdivided into old PrSB group and new PrSB group. In addition to these
PrSBs there exists thirty six foreign banks (FBs), privately owned with registrar office
located abroad. A detailed list of SCBs in India as on 31st March 2012 is given in Annexure-
1B, a summary of which is presented in the following chart which broadly exhibits the
structure of commercial banks in India, based on ownership.
29
Chart-1.4.: Structure of Commercial Banks in India as on 31st March 2012
Commercial Banks in India

(168)
Schedule Commercial Non Schedule Commercial

Banks (164) Banks (04)
Public Sector Banks (26) Private Sector Banks (20) Foreign Banks in India (36) Regional Rural Banks (82)
Nationalised Banks (19) Old Private Banks (13)
State Bank of India and its New Private Banks (7)

Associates (6)
Other Public Sector Bank

(1)
Figures in the bracket represent number of banks in the group.

Source: www.rbi.org.in
PrSBs and FBs are particularly excluded from the study, as they are not strictly comparable
with PSBs and they account for less than 16% and 5.30% of gross NPA’s of SCBs in India
respectively, as on March 2012. The Regional Rural Banks (RRBs) are also excluded from
the present study as their operations are confined to target groups in rural and semi urban
areas.
As the concept of NPA was first introduced in April 1992, therefore their post reform 1st
balance sheets became available one year later ie, in, 1992-93. However NPA data of all
PSBs are not available for 1992-93, 1993-94 and 1994-95 (only aggregate PSB figures are
available since 1992-93). This provides justification for 1995-96 as the starting year in the
present study. Further, the study covers the period up to 2011-12, i,e the latest year till the
dataset is available.
The study is primarily confined to six selected PSBs selected based on their GNPA to GA
ratio as on March 31st, 2008. First two ranks of banks have been selected from the SB Group,
ie, State Bank of India (SBI) and State Bank of Travancore (SBT) and first four ranks of
banks have been selected from NB Group, ie, Central Bank of India (CBI), UCO Bank
(UCO), Punjab National Bank (PNB), and Syndicate Bank (SB). Bank wise GNPA to GA
Ratio as on March 31st, 2008 of all PSBs are given in Annexure 1C.
30
1.8. Scheme of Research Work
The chapter 2 has elaborately described prudential norms on NPAs and a few other related
concepts. The concepts of loan assets and their various classifications have been spelt out. The
guidelines of RBI on NPAs, on income recognition, on asset classification and on
provisioning have been narrated in this chapter. A few operational concepts related to NPAs
have also been discussed in brief.
Chapter 3 delineates on the mechanism for effective management of NPAs as initiated by

PSBs in India.
In chapter 4, an attempt has been made to examine the dynamicity of NPA as the variable
under study over time, in six selected PSBs along with PSBs in aggregate. Parametric
models (using time series regression analysis) have been employed to examine the trends
existing in the above data-sets relating to the parameters, GNPA and GA respectively over
time. Finally, the best fitted model have been used to forecast values for three years.
Chapter 5 and 6 of the thesis attempts to employ non parametric [LOESS curve fit
(Quadratic)] model and semi parametric [Penalised Spline] model to examine the pattern
existed in the above data-sets relating to the parameters, GNPA and GA respectively over
time, followed by forecasting future values for three years.
Chapter 7 examines relationship between NPA and selected banking variables with the help
of pair-wise correlation analysis followed by regression analysis (simple as well as multiple)
to examine impact of NPA along with time variable on such variables and model developed
under each case.
Chapter 8 is having three sections. Section I is earmarked for summary of the thesis. Section
II narrates the future action plan based on the findings emanated from chapters 4, 5, 6 and 7.
The last Section (Section III) is dedicated to future application of the thesis, uniqueness of the
research work, social benefits of the research, scope of future research on the subject and
comprehensive conclusion.
31
Chapter II
Existing Prudential Norms on Asset Quality
2.1 Introduction
2.2 Review of literature
2.3 Norms Relating to Income Recognition.
2.4 Norms Relating to Asset Classification
2.5 Norms Relating to Provisioning
2.6 Treatment of Prudential norms on various types of Advances
2.7 Operational Concepts Related to NPAs
2.8 Conclusion
32
2.1. Introduction
In traditional banking business, lending to borrower (deficit units) are being financed by
deposits from customers (surplus units) and in the process, commercial banks are exposed to
the risk of default by the borrower in the payment of either principal or interest. This risk in
banking parlance is termed as "credit risk" and accounts where payment of interest and/or
repayment of principal are not forthcoming are treated as bad loans or distress assets NPAs.
Existence of NPA is an integral part of banking and therefore, all banks have NPAs in its
advance portfolio, may be in varied proportions.
Prudential Norms in India address asset quality and solvency in a banking business.
Accordingly, it has two separate set of norms for two specific banking issues. They are: (i)
Prudential Norms on Asset Quality, and (ii) Prudential Norms on Capital Adequacy. The
major issues arising in the case of Prudential Norms on Asset Quality essentially revolves
around three vital aspects of Income Recognition, Asset Classification and Provisioning. In
other words, such norms require that for ascertaining quality of assets in a bank, they are to
be classified based on recognition of income and be subjected to adequate provisioning.
Booking of unrealized interest on Non-Performing Assets to income account without
making adequate provisioning in respect of the debts doubtful of recovery will result in an
overstatement of the profit of the bank. Similarly, depicting such an advance as healthy
would not reflect the actual financial health of that bank. In its Annual Report, 1991-92, RBI
has stated that "if the balance sheet of a bank is to reflect actual financial health of that bank,
there has to be a proper system of recognition of income, classification of assets and
provisioning for bad debts on a prudential basis”.
A high-power Committee on the Financial System, with Shri M. Narasimham as the

Chairman, was constituted by RBI to make recommendations on the issue of financial sector
reforms with a view to ensuring that this sector operates on the basis of operational
flexibility and functional autonomy and thereby enhance efficiency, productivity and
profitability. The Committee studied the prevailing financial system, identified its short
comings and weaknesses, reflected on the decline in productivity and efficiency and erosion
of profitability due to deterioration in the quality of the loan portfolio, restricting income
generation and enhancement of capital funds, accompanied by inadequate loan loss
33
provisions. The Committee also observed that the accounting and disclosure practices, too,
did not always reflect the true state of affairs of banks and financial institutions. Against this
backdrop, the Committee made a wide range of suggestions and recommended inter-alia that
balance sheets of banks and financial institutions should be made in line with the
recommendations of the International Accounting Standards Committee. This implies that
assets of banks should be evaluated on the basis of their realisable values and that uniform
accounting policies should be adopted particularly with regard to income recognition and
provisioning. On the latter aspect, the Committee further observed that a proper system of
income recognition and provisioning is fundamental to the preservation of the strength and
stability of the banking system and to reflect its actual financial health, a proper asset
classification will, however, have to precede this exercise.
The objective of the Committee was to make banking sector more stable, transparent,
competitive and vibrant. One of the major recommendations of the Committee was with
regard to NPAs in banks and financial institutions, their identification, disclosure and
provisioning. Majority of the suggestions and recommendations of this committee were
accepted and resulted in evolvement of prudential norms on Income Recognition, Assets
Classification and Provisioning. Keeping intact the original viewpoint, RBI has been making
modifications to these norms taking into account the changes in the business environment
and need of the hour.
Before getting into a detailed discussion on the three aspects, let us review the earlier
accounting system prevalent in commercial banks in India prior to the introduction of the
Prudential Norms. Before 1990’s ‘security oriented approach’ was meticulously followed
for accessing recoverability of a particular debt. Interest booked in all loan accounts was
taken to total income of the bank irrespective of the fact whether interest debited earlier was
actually recovered or not. As these interest incomes were credited in the Profit and Loss
account, tax was paid and dividend was declared on the basis of profits so arrived at and
Reserves were erroneously inflated to the extent of the residual profits. When it was
ascertained that a particular debt was irrecoverable, it was too late. Moreover, this resulted
in overstatement of profits and net worth of banks. All these anomalies were rightly taken
care of by the Prudential Norms on asset quality implemented in Indian banking by RBI.
34
For understanding in letter and spirit the Prudential Norms in general with Non-Performing
Assets (NPAs) in particular, circularized instructions issued by Reserve Bank of India (RBI)
from time to time since 1992 have been examined, without going through the same it is not
possible to appreciate various norms and concepts relating to NPA and desired procedures being
followed by individual banks. The latest Master Circular covering all the amendments is given in
circular number DBOD.No.BP.BC.1/21.04.048/2013-14 dated 1st July 2013.
In addition to the above circularized instructions, number of books and published papers were
consulted for better understanding on the subject under reference. The detailed reviews are
included in Chapter-1. However, the references of the base literatures in connection with this
chapter are given below:
Suneja (1992), Shajahan (1998), Toor (1994), Shukla (1995), Mohan and Kapoor (1996), Naik
(2006), Shekhar and Shekhar ( 2005), Poongavanam (2011).
2.3. Norms Relating to Income Recognition
Recognition of income shall be objective and be based on record of recovery rather than any
subjective considerations like availability of security, net worth of borrower/ guarantor etc.
Income Recognition in the case of NPA account is only to the extent of realisation of
income. In other words, income (interest and other charges) cannot be booked by merely
charging to borrowal account unless it is realized. However, in case of performing asset,
income can be booked partly on actual basis and partly on accrual basis. Income
Recognition norms set in the Accounting Year 1992-93 was subjected to a number of
amendments towards raising the same to international standard, a brief account of which is given
in the following table.
Table- 2.1.: Summary of Progressive Changes in Income Recognition Norms

Effective Nature of Credit Facilities
Date of Term Loan Cash Credit Bill Purchased Agricultural Loan Other
Guidelines and Over and Discounted Accounts
( As on 31st Draft
March)
1993 4 quarters * 4 quarters $ 4 quarters # 4 quarters £
35
1998 2 quarters * 2 quarters $ 2 quarters # ** for 2 harvest seasons 2 quarters £
beyond due date but not
exceeding 2.5 years
2001 Over due, for Out of Order, Over due, 180 ** for 2 harvest seasons Over due, for
more than 180 days days beyond due date but not more than
180 days exceeding 2.5 years 180 days
more than 90 90 days days beyond due date but not more than 90
days exceeding 2.5 years days
more than 90 90 days days beyond due date ( in more than 90
days case of short duration days
crop) and for 1 harvest
seasons beyond due
date ( in case of long
duration crop)
* An account should be treated as NPA if interest remains ‘Past Due’ (Past Due means an amount
remaining outstanding or unpaid for 30 days beyond due date) for a period of
$ An account should be treated as NPA if the account remains ‘Out of Order’ for a period of
# An account should be treated as NPA if the Bill remains over due and unpaid for a period of
** An agricultural account should be treated as NPA if the account remains over due
£ Any other Credit Facility should be treated as NPA if any amount to be received in respect of that
facility remains Past Due for a period of
2.3.1. Reversal of Income
If any advance becomes NPA as at the end of the year, interest accrued and credited to
income account in the corresponding previous year should be reversed or provided for, if the
same is not realised. This will apply to Government Guaranteed Advances also. In case of
NPA accounts, fees, commission and similar income that have accrued should cease to
accrue in the current period and should be reversed and provided for with respect to the
previous year, if the same remains uncollected.
2.3.2. Appropriation of Recoveries
(i) Interest realised on NPAs may be taken in to income account provided the credits in the
accounts towards interest are not out of fresh/ additional credit facilities sanctioned to the
borrowers. In other words, interest realised in absolute terms in the form of fresh
remittances only should be accounted for as income.
36
(ii) Since RBI has not issued specific guidelines in the matter of appropriation of recovery
and has left it to the individual banks, banks are required to work out a prudent methodology
for appropriation of recoveries. It is, however, needless to stress that the methodology so
worked out shall be implemented in a uniform and consistent manner. In case a borrower
avails several facilities, appropriation of recoveries can be made in accordance with the
provisions of Indian Contract Act, 1872.
2.3.3. Interest Application
There is no objection to the banks using their own discretion in debiting interest to an NPA
account taking the same into Interest Suspense Account or maintaining only a record of such
interest in proforma accounts.
2.4. Norms Relating to Asset Classification
There are generally two ways of classification of assets, which are given as under:
• Asset classification under Health Code System.
• Asset classification under Prudential Norms.
2.4.1. Asset classification under Health Code (HC) System
With the growing volume of bank credit, the task of assessing the quality of individual
advances for the purpose of their effective monitoring and follow-up, as also for making
adequate provisions for bad and doubtful debts, was becoming increasingly difficult.
Consequently, RBI felt the need for a comprehensive and uniform grading and credit
monitoring system for aiding an on-going assessment of the quality (or health) of the
advance portfolio of banks which will, inter alia, provide them with the necessary
information regarding the health of the advances portfolio and the extent of the advances
causing concern in relation to total advances.
In the backdrop of the above, all the Indian banks were asked by the RBI to uniformly adopt
the HC System under which each bank was required to classify its advances at all branches,
including foreign branches, into eight categories with a HC ranging from 1 to 8 assigned to
each borrowal account.
37
The first such classification of advances was required to be done as on 31 December 1985,
and thereafter the information was to be updated on a quarterly basis. If it was not possible
to cover all accounts straightaway, a beginning was to be made by applying the
classification codes to all accounts with limit / outstanding balances of Rs.1 lakh and above.
Advances below the cut-off point (excluding the bad and doubtful debts which were to be
grouped separately as “other bad/doubtful debts”) were required to be grouped under ‘other
advances’. Importantly, the system should not result in dilution of control over the advances
below the cut-off point of Rs.1 lakh (Suneja, 1992). Features of each category of advance
under the system are given below:
Code 1: Satisfactory
This category covers all borrowers where

• Conduct of the account is satisfactory;
• All the terms and conditions of sanction are complied with;
• All accounts of the borrower are in order;
• The safety of advance is not in doubt.
Code 2: Irregular
This category of borrowers covers those accounts where
• The safety of the advance is not suspected, though there may be occasional irregularities
which may be considered as short- term phenomenon, e.g. the accounts are overdrawn beyond
the drawing power (based on the value of the security less margin) or the sanctioned limit for a
temporary period.
• Installments in respect of term loans overdue for less than six months or import bills under
Letter of Credit (L/C) and installments under deferred payment guarantees, if overdue for less
than three months.
• In respect of bills purchased/ discounted account, some of the bills (not exceeding 10% to
15% of the total outstanding in the bills purchased/discounted account of the borrowers) are
overdue for payment by less than 3 months and / or refund in respect of unpaid bills is not
forthcoming immediately.
38
Code 3: Sick: Viable / Under Nursing
Units in respect of which nursing / revival programmes are taken up should be included under
this category.
Code 4: Sick: Non-Viable / Sticky
Accounts of borrowers under this category are those where the irregularities mentioned above
persist for a period of six months and over and there are no immediate prospects of
regularization. Moreover, the accounts could throw up some of the usual signs of recipient
sickness such as:
• Apparent stagnation in the business as reflected by slow / negligible turnover in the
account;
• Frequent requests for overdrawing or issue of cheques without ensuring availability of
funds in the account;
• Bills purchased / discounted drawn by the borrower remaining overdue for 3 months and
more or recovery of such bills from the borrower poses difficulties;
• Unexpected delays in submission of stock statements / quarterly / half-yearly operating
statements / balance sheet or other information required by the bank.
Code 5: Advances Recalled
This category consists of those accounts, in which the repayment is doubtful and nursing is not
considered workable. If a decision has been taken or likely to be taken to recall the advance, such
borrowers are to be classified under this code.
Code 6: Suit-filed Accounts
This category consists of accounts where legal action or recovery proceedings under the Public
Debt Recovery Act, wherever applicable, have been initiated. Suit-filed accounts may further be
classified as (i) Advances where suits are pending for more than 5 years, (ii) Advances where
suits are pending for more than 2 years but less than 5 years, (iii) Advances where suits are
pending for 2 years or less.
39
Code 7: Decreed Accounts
The advances where suits have been filed and decrees obtained will come under this
category. This category may further be classified depending upon the pending period of the
decree.
Code 8: Debts classified by the bank as Bad and Doubtful
All advances appearing under HC numbers 3 to 7 and where the recoverability of bank dues
have become doubtful on account of shortfall in value of security, difficulty in enforcing and
realizing the securities or inability or unwillingness of the borrowers to repay the bank dues
will fall in this category.
The HC System was required to be monitored by the Board of Directors through half-yearly
returns and the same was also submitted to RBI. The latter, through Circular letter dated
March 19, 1994, considering the introduction of the Prudential Norms, decided that with
effect from April 1, 1994, "the HC classification shall cease to be a subject of supervisory
interests and reporting" (to the RBI and the Board of Directors). However, banks, if they so
desire, may continue to use the HC System as a Management Information System.
2.4.2. Asset classification under Prudential Norms
From perusal of balance sheet of any bank, it is observed that lion’s share of aggregate asset
are advances and investments. Loan Assets of a bank are broadly classified as :
A) Performing Assets or Standard Assets are those which do not disclose any problem or
weakness with regard to repayment of principal and interest. In other words such assets do
not carry more than normal risk attached to the business.
B) Non Performing Assets (NPA), on the other hand are loan assets, which cease to
generate income to the bank. It includes borrowers’ defaults or delays in interest or principal
repayment. A bank should classify an account as NPA only if the interest charged during
any quarter is not serviced fully within 90 days from the end of the quarter. These assets
have well defined credit weakness that jeopardize the liquidation of debts and may be
characterized by distinct possibilities that bank will sustain some losses. In other words, an
NPA may be defined as a credit facility in respect of which the interest and/or installment of
principal has remained unpaid for a specified period of time. With a view to moving towards
40
international best practices and to ensure greater transparency, it has been decided to adopt
the “90 days’ overdue” norm for identification of NPAs, from the year ended March 31,
2004. The specified period has been reduced in a phased manner from 4 quarters on March
1993 to 1 quarters on March 2004 as given in the following table.
Table- 2.2.: Summary of Changes in NPA Norms
Effective Date of Guidelines ( As Specified Period, beyond which the account will be
on 31st March) termed as NPA
1993 four quarters
1994 three quarters
1995 two quarters (i.e.180 days)
2004 one quarter (i.e.90 days)
Features of NPA for different categories of loans and advances are given below:
i) Term Loan : A Term Loan ( other than agricultural loan ) is treated as NPA if either
interest and/or installment of principal remains overdue for a period of more than 90 days.
ii) Cash Credit & Overdraft: In case of a running account like Cash Credit / Overdraft, it
is treated as NPA if the outstanding balances remains ‘Out of Order’ for a period of 90 days
or more. The account need not be treated as NPA merely because of some temporary
deficiencies. However, if there is a threat of loss or a doubt of its recoverability, then it
should be classified as non performing taking due note of the spirit behind the prudential
norms.
iii) Bills Purchased/ Discounted : In case of Bills Purchased/ Discounted ( Inland or

Foreign), it will be considered an NPA, if any amount of Bills Purchased/ Discounted
remains ‘Overdue” for a period of 90 days or more after its due date. The period of 90 days
is to be reckoned as from the date of presentation of the bill incase of a demand bill and
from the expiry of usance period of the bill in respect of usance bill. In case of bill
discounting facility is given in the form of a limit, if any bill remains overdue for 90 days or
more, the entire dues under the bill discounting limit should be treated as NPA.
41
iv) Agricultural Advances : In case of agricultural loan, it is treated as NPA if either
interest and/or installment of principal remains overdue for 2 harvest seasons for a short
duration crop (up to 12 months) and for 1 harvest season for long duration crop (beyond 12
months) beyond the due dates.
Since for all practical purposes, classification is to be done as on the date of balance sheet of
the banks, any classification done in between the two balance sheet dates for the purposes of
half yearly closing etc. will not have relevance so far final asset classification is concerned.
Asset classification for the purpose of balance sheet shall be, therefore, independent of such
interim classification.
2.4.3. Similarities between Asset Classification under HC System and under Prudential
Norm
Several "irregularities" included under the HC as stated below are similar to those
considered in the definition of "non-performing assets" under the new prudential norms.
• The features of "Out of Order" cash credit/ over draft accounts, ie, accounts overdrawn
beyond drawing power/ sanctioned limit, accounts with slow/ negligible turnover or
frequent over drawings included under clause (i) of HC 2 and clauses (i) and (ii) of HC 4,
are also one of the adverse features in determining NPAs.
• A NPA has been defined as a credit facility in respect of which an amount remains
unpaid for one quarter. Clauses (iii) and (iv) of HC 4 as given below are akin to this
definition. "(iii) in case of bills purchased/ discounted drawn by the borrower remaining
overdrawn for 3 months and more ; (iv) in. case of term loans, 6 or more monthly
installments/ two or more quarterly installments/ one or more half-yearly installments are
overdue"
In view of the above, it can be said that the concept of NPA was evolved a decade earlier,
though the term, as such, was not in vogue. The basic difference between the HC System
and the presently prescribed and used Prudential Norms on NPA is that provisioning under
the former is required to be made only with respect to the portion of the advance not covered
by realisable security whereas under the new norms, in addition to consideration of
availability of security, track record of payment (or non-payment) by the borrowers, is also
42
considered. Moreover, the earlier concept contained some amount of subjective
consideration while the present one is free from subjectivity.
2.4.4. Norms Relating to Classification of NPAs
Classification of NPAs are done, taking into account the degree of credit weaknesses, risks
and extent of dependence on collateral security for realization of dues. NPA accounts are
classified into following three categories.
2.4.4.1. Sub Standard Asset
A substandard asset is one, which has remained NPA for a period of less than or equal to 12
months. In such cases, the current net-worth of the borrower/ guarantor or the current
market value of the security charged is not enough to ensure recovery of the dues to the
banks in full.
2.4.4.2. Doubtful Asset
An asset would be classified as doubtful if it has remained in the sub standard category for a
period of 12 months. A loan classified as doubtful has all the weaknesses inherent in asset
that were classified as substandard, with the added characteristic that the weaknesses make
collection or liquidation in full, on the basis of currently known facts, conditions and values-
highly questionable and improbable.
2.4.4.3. Loss Asset
A Loss Asset is one where loss has been identified by the bank or internal or external
inspector or auditors or the RBI inspection as non recoverable and realizable value of
securities is less than or equal to 10% of the outstanding but the amount has not been written
off wholly. In other words, such an asset is considered uncollectable and of such little value
that its continuance as a bankable asset is not warranted, although there may be some
salvage or recovery value. However, NPA account where there is potential threats to
recovery on account of erosion in the value of security or non availability of security or
fraud by the borrower, it should be straight way classified as Doubtful or Loss asset as
appropriate. Depending upon period of remaining in overdue, norms relating to asset
classification has undergone several changes since it’s introduction in 1992 as is given in the
following table:
43
Table- 2.3.: Summary of Change in Asset Classification Norms
Effective Category of Asset

Date of Standard Sub Standard Doubtful Asset Loss Asset
Guidelines Asset Asset
31st March An asset which Identified NPA Identified NPA An asset where loss has
1993 does not disclose < 2 years > 2 years been identified by the
any problem bank or internal/ external
31st March Identified NPA < Identified NPA
and also does auditors or by RBI
2001 18 months > 18 months
not carry more inspection but the amount
31st March than normal Identified NPA < Identified NPA has not been written off,
2005 business .risk. 12 months > 12 months wholly or partly.
2.5. Norms Relating to Provisioning
Major source of loanable funds of a bank is deposits raised from the public and therefore
prudence demands that the credit risk and losses suffered should be adequately provided for
to take care of the depositors' money. It is therefore a sound policy to create adequate
provisions from the annual profits to take care of such losses/ contingencies. Such
provisioning would result in erosion of banks' net worth, in case there is no adequate profit
to cover the provisioning requirement.
2.5.1. Provisioning Norms for Standard Assets
Banks should make general provision for standard assets at the following rates :
a) direct advances to agricultural and Small and Medium Enterprises (SME) sectors at
0.25 per cent,
b) residential housing loans beyond Rs. 20 lakhs at 1 per cent,
c) advances to specific sectors, i.e., personal loans (including credit card receivables),
loans and advances qualifying as capital market exposures, commercial real estate loans,
and loans and advances to Non Banking Financial Companies (NBFCs) at 2 per cent,
d) all other advances not included in (a), (b) and (c) above, at 0.40 per cent.
(ii) The provisions on standard assets should not be reckoned for arriving at net NPAs.
44
(iii) The provisions towards Standard Assets need not be netted from gross advances but
shown separately as 'Contingent Provisions against Standard Assets' under 'Other Liabilities
and Provisions' in Schedule 5 of the balance sheet.
2.5.2. Provisioning Norms for Substandard Assets
A general provision of 15 per cent on total outstanding should be made without making any
allowance for Export Credit Guarantee Corporation (ECGC) guarantee cover and securities
available. The 'unsecured exposures' which are identified as 'substandard' would attract
additional provision of 10 per cent, i.e., a total of 25 per cent on the outstanding balance.
2.5.3. Provisioning Norms for Doubtful Assets
In case of advances classified as doubtful, provisioning is to be done separately for the

secured and unsecured portion, (Mohan, 1996, pp-39).
(i) 100 percent of the extent to which the advance is not covered by the realizable value of
the security to which the bank has a valid recourse and the realizable value is estimated on a
realistic basis.
(ii) With regard to the secured portion, provision may be made on the following basis, at
the rates ranging from 25 per cent to 100 per cent of the secured portion depending upon the
period for which the asset has remained doubtful.
Table- 2.4.: Provisioning Norms for Doubtful Debts
Period for which the advance has Percentage of Provision in Percentage of Provision in
been considered as Doubtful. Secured Advance Unsecured Advance
Upto one year 25% 100%
One to three years 40% 100%
More than three years 100% 100%
(iii) Banks are permitted to phase the additional provisioning consequent upon the reduction
in the transition period from substandard to doubtful asset from 18 to 12 months over a four
year period commencing from the year ending March 31, 2005, with a minimum of 20%
each year.
45
2.5.4. Provisioning Norms for Loss Assets
Loss Assets should be written-off. If loss assets are permitted to remain in the books for any
reason, 100 percent of the outstanding should be provided for. Provision norms set in the
Accounting Year 1992-93 were subjected to number of amendments, a brief account of which is
given in the following table.
Table- 2.5.: Summary of Changes in Provisioning Norms
Effective Category of Asset

Date of Guide- Standard Sub- Doubtful Asset Loss Asset
lines ( As on 31st Asset standard
March) Asset
1993 No Provision 10% a. 100% in respect of the portion The entire
is required which is not covered by realizable asset should
value of security be written
b. 20% to 50% of the covered off. If it is not
portion depending upon the period written off
for which the asset has remained for some
doubtful reason,
Period of Doubtful Percentage of provision
Status Provisioning should be
Upto 1 year 20 made for
1 to 3 years 30 100% of the
More than 3 yrs 50 balance
outstand-ing
2001 *
2002 $
2004 £
To start with in March 1993, facilities with an outstanding balance of Rs 25000/- and above were required
to be considered for provisioning, based on asset classification.
In respect of accounts with outstanding below Rs 25000/-, aggregate provisioning was to be made.
Aggregate provisioning of 2.5% of the total outstanding to be made on 31st March 1993 and the rate of
provision was raised to 5% for 31st March 1994 to 7.5% for 31st March 1995 to 10% for 31st March 1996
and 15% for 31st March 1997 (Toor, 1997, pp-33).
46
* Not less than 50% on the assets which have become doubtful on account of new norms (Doubtful
Norms made stringent from 2 year to 1.5 year ).
$ Balance of the provisions not made during the previous year. In addition to the provisions needed as on
March 31, 2002.
£ For outstanding stock of NPA in Doubtful category of more than 3 years with security backing
as on 31.03.2004,
1. 60% provision on secured portion as on 31.03.2005
ii.) on or after 01.04.2004,
1. 100% provision on secured portion
2.5.5. Provisioning Norms for Restructured Accounts (RBI Circular No 2012-13/309 dated
November 26, 2012
i. To enhance the provisioning requirement for restructured accounts classified as standard

advances from the existing 2.00 per cent to 2.75 per cent in the first two years from the date of
restructuring. In cases of moratorium on payment of interest/principal after restructuring, such
advances will attract a provision of 2.75 per cent for the period covering moratorium and two
years thereafter; and that
ii. Restructured accounts classified as non-performing advances, when upgraded to standard

category will attract a provision of 2.75 per cent in the first year from the date of up-gradation
instead of the existing 2.00 per cent.
When a loan to projects under implementation is restructured due to change in the Date of
Commencement of Commercial Operations (DCCO) beyond the original DCCO as envisaged at
the time of financial closure and classified as standard advances in terms of RBI guidelines, such
account would attract higher provisioning as per the details given below:
47
Table- 2.6.: Provisioning Norms for Doubtful Debts (Infrastructure Projects)
Particulars Provisioning Requirement

If the revised DCCO is within two years from the
original DCCO prescribed at the time of financial 0.40 per cent
closure
If the DCCO is extended beyond two years and 2.75 per cent – From the date of such restructuring till
upto four years or three years from the original the revised DCCO or 2 years from the date of
DCCO, as the case may be, depending upon the restructuring, whichever is later.
reasons for such delay
Table- 2.7.: Provisioning Norms for Doubtful Debts (Non Infrastructure Projects )
Particulars Provisioning Requirement
If the revised DCCO is within six months from 0.40 per cent
the original DCCO prescribed at the time of
financial closure
If the DCCO is extended beyond six months and 2.75 per cent – From the date of such restructuring
upto one year from the original DCCO for 2 years.
prescribed at the time of financial closure
2.5.6. Guidelines for Calculating Realisable Value of Security (RVS)

The maximum amount of RVS that can be taken for the purpose of provision is limited to
the extent of amount outstanding to be arrived at as under:-
a) Principal Security
b) Collateral Security
c) DICGC/ECGC Claim Received /Receivable
d) Sundry Credits/Amount kept in Suspense
e) Deposit as per Court Order remaining unencumbered.
f) Value of attached assets as per court order against which bank can have recourse to.
g) Self-liquidating securities such as National Savings Certificate (NSCs), Kishan Vikas
Patra (KVPs), Bank's term deposits, LIC policies eligible for surrender etc which can be
covered under Bankers Right of General Lien or Set-off.
h) Any other tangible asset belonging to the borrower over which bank can have recourse
under Bankers' Right of Setoff.
48
The total RVS is the summation of all the variables mentionted above viz, a+ b+ c+ d+ e+
f+ g+ h
2.5.7. Guidelines for Calculating the Amount of Provision

Amount of provision should be calculated on the amount outstanding arrived at as : Amount
Outstanding (Ledger outstanding + Amount kept in Suspense Debit, if any) - Unrealised
interest/ Other charges kept in sundry credits account.
a) In case of doubtful assets, besides making full provisions to the extent of shortfall in
security available in the account, banks are required to make an additional provision to the
extent of 20% to 100% of the secured portion of the balance outstanding in the account
depending upon the age of the doubtful asset. However, this additional provision need not
be made for that portion of book outstanding which is covered by securities of self-
liquidating nature and to the extent of money received in the account and held in sundry
credits account.
b) In case of suit filed account, in which Attachment Before Judgement is granted or a

Garnishee Order is served, value of such security can be taken into account for calculating
the RVS.
c) In case of accounts in which security documents have expired, the value of security
should not be taken for the purpose of arriving at the provision requirements. This is,
however, not applicable to pledge of goods covered by such documents.
d) In cases where PARI-PASSU charge on the assets/security is available, only concerned

bank's share of security should be taken while arriving at the RVS. Even where the second
charge on the assets is available in the account duly created or registered, bank should
properly assess the extent of security that could be considered as part of the RVS.
2.5.8. Methods of Valuation of Security

Though banks have discretion to set their policy for valuation of securities, the basic guiding
factor in assessing the value shall be that the valuation should be realistic, based on relevant
records and not on any extraneous factors. Methods practiced for arriving at the valuation of
49
certain prominent types of assets for the purpose of assessing the realisable value of specific
assets/securities are given in the following table.
Table- 2.8.: Methods of Valuation of Securities
Assets/ Securities Method/ Source of Valuation
Latest stock/debtors statement (not more than 3 months old) duly

certified by the borrower. In cases of account with balance of Rs.
(a) Stock/Debtors 5.00 crores and above, stock audit at annual intervals by external
agencies be appointed as per the guidelines approved by the Board in
order to enhance the reliability of stock valuation.
Realisable value (RV) as per Valuation Report (VR) not older than 3
(b) Land and Buildings
years
(c) Other Fixed Assets,

RV as per VR not older than 3 years. Where VR is not available,
(indigenous) like plant,
value considered may be at cost less depreciation @ 20% pa.
machinery etc.
(d) Other Fixed Assets RV as per VR not older than 3 years by an approved valuer.
(imported)
(e) Vehicle RV as per VR by an approved valuer or cost less depreciation @ 15%

pa, whichever is lower.
(f) LIC Policy Surrender value of LIC Policy
2.6. Treatment of Income Recognition, Asset Classification and Provisioning in respect

of Special Types of Advances
a) Advances against Bank Term Deposit, NSCs, KVPs, LIPs etc
Advances against Banks Term Deposits, NSCs, KVPs, Life Policies eligible for surrender
etc. are exempted from the compliance of prudential norms and therefore, these advances
are classified as Standard for all practical purposes. Accordingly, banks are not required to
treat such accounts as NPA although interest thereon remains unrealized for the specified
period. Interest on advances against such securities may be taken in to income account on
the due date provided, adequate margin is available.
50
b) Advances to Staff
In case of staff advances given for housing purpose or purchase of consumer articles or any
other similar advances, where interest and/ or installments are to be recovered from salary
and interest is payable after recovery of principal. In such cases, interest is not debited to the
principal account of the staff members, but is kept in separate account. Therefore, interest
does not become “overdue” and these advances are generally classified as Standard Assets.
As such banks are not required to treat such account as NPA and may recognize the interest
income as income.
c) Government Guaranteed Advances

RBI clarified vide circular dated 24.11.1993 that Government Guaranteed Advances are
exempted from the prudential norms for the purpose of asset classification and provisioning
only and not for the purpose of income recognition. Therefore, if any income with respect to
advances guaranteed by the Government remains overdue for specified period and thereby
advance becomes NPA, interest on such advances should not be taken in to income account,
unless the same is realised.
However, for the purpose of asset classification and provisioning, even if income with
respect to advances guaranteed by the Government remains overdue for specified period and
as such these advances are classified as Standard irrespective of the fact that interest and/ or
installments are over due for the specified period. With effect from 31st March, 2000, in
respect of advances sanctioned against State Government Guarantees, if the guarantee is
invoked and remains in default for more than 90 days, the Bank should make normal
provisions.
d) Advances against Gold Ornaments, Government Securities and all other kinds of
securities.
Advances against gold ornaments, Govt. Securities and all other kinds of securities are not
exempted from provisioning requirements and as such provision has to be made on such
advances depending on the respective asset classification and provisioning norms applicable
to them.
51
e) Advance with Moratorium or Gestation Period
In respect of finance for industrial projects, agricultural plantation etc where implementation
period is long, banks generally provide a moratorium (or gestation) period for payment of
interest and/ or installments. Hence, although the interest is debited to the borrowers account
on its accrual, it does not become due for payment but is added to the principal. Therefore,
such interest does not become over due with reference to the due date of charging interest
and be classified as Standard Asset. As such, interest accrued during the moratorium period
may be taken into income account.
f) Agricultural Advances
In case of agricultural advances, interest payment is on half yearly basis in line with the
harvest seasons, since the agriculturists do not have any other source of income than sale
proceeds of crops. Hence, agricultural season should be taken as the basis for determining
an advance as NPA or otherwise. As such if interest is not paid for two harvest seasons not
exceeding two half years after it has become due, then only such advance should be
considered as NPA.
g) Project under Implementation

(i) In case of project finance, the project should be followed up all through the
implementation stage to keep track of the progress and any possible variation in the
implementation period and cost/time overrun.
Category I: Projects where financial closure had been achieved and formally documented.
Category II: Projects sanctioned before 1997 with original project cost of Rs.100 crore or
more where financial closure was not formally documented.
Category III Projects sanctioned before 1997 with original project cost of less than Rs.100
crore where financial closure was not formally documented. In these cases the date of
completion of the project would be as originally envisaged at the time of sanction. In such
cases, the asset may be treated as standard asset only for a period not exceeding two years
beyond the date of completion of the project as originally envisaged at the time of sanction.
(ii) Banks may recognize income on accrual basis in respect of the above three categories of
projects under implementation, which are classified as a ‘standard' assets.
52
(iii) Banks should not recognise income on accrual basis in respect of the above three
categories of projects under implementation which are classified as a 'substandard' assets.
Banks may recognise income in such accounts only on realisation on cash basis.
Consequently, banks which have wrongly recognised income in the past should reverse the
interest if it was recognised as income during the current year or make a provision for an
equivalent amount if it was recognised as income in the previous year(s). As regards,
regulatory treatment of 'funded interest' recognised as income and 'conversion into equity,
debentures or any other instrument, banks should adopt the following:
(i) Funded Interest: Income recognition in respect of NPAs, regardless of whether these
are or are not subjected to restructuring/ rescheduling/ renegotiation of terms of the loan
agreement, should be done strictly on cash basis, only on realisation and not if the amount
of interest overdue has been funded. If, however, the amount of funded interest is
recognised as income, a provision for an equal amount should also be made simultaneously.
(ii) Conversion into equity, debentures or any other instrument: The amount
outstanding converted into other instruments would normally comprise principal and the
interest components. If the amount of interest due is converted into equity or any other
instrument, and income is recognised in consequence, full provision should be made for the
amount of income so recognised to offset the effect of such income recognition. Such
provision would be in addition to the amount of provision that may be necessary for the
depreciation in the value of the equity or other instruments, as per the investment valuation
norms. However, if the conversion of interest is into equity which is quoted, income can be
recognised at market value of equity, as on the date of conversion, not exceeding the amount
of interest converted into equity. Such equity must thereafter be classified in the “available
for sale" category and valued at lower of cost or market value. In case of conversion of
principal and/ or interest in respect of NPAs into debentures, such debentures should be
treated as NPA, ab initio, in the same asset classification as was applicable to loan just
before conversion and provision made as per norms. This norm would also apply to zero
coupon bonds or other instruments which seek to defer the liability of the issuer. On such
debentures income should be recognized only on realisation basis.
53
The income in respect of unrealised interest which is converted into debentures or any other
fixed maturity instrument should be recognised only on redemption of such instrument.
Subject to the above, the equity shares or other instruments arising from conversion of the
principal amount of loan would also be subject to the usual prudential valuation norms as
applicable to such instruments.
h) Renegotiated/ Rescheduled Advances

In case of Renegotiated / Rescheduled Advances, where terms of interest and repayment of
principal are renegotiated or rescheduled after the commencement of production, such
advance should be continued as NPA, if categorised as such before such renegotiation or
reschedulement and income accrued on such advances should not be taken as income.
However, fees and commission earned by the banks due to renegotiations or rescheduling of
outstanding advances should be recognised on accrual basis over the period of time covered
by the renegotiated or rescheduled extension of credit.
i) Advances granted under Rehabilitation Packages approved by Board for Industrial

and Financial Reconstruction (BIFR)/Term Lending Institutions
In respect of advances granted under rehabilitation packages approved by BIFR/ Term
Lending Institutions, the provision should continue to be made for the outstanding to the
Bank on the existing credit facilities as per the classification as substandard or doubtful
assets. In respect of additional facilities sanctioned, as per package finalised by BIFR and/or
Term Lending Institution, provision on additional facilities need not be made for the period
of one year from the date of disbursement.
j) Provision for Advance covered by Export Credit and Guarantee Corporation

(ECGC) / Credit Guarantee Trust for Small Industries (CGTSI) Guarantees
In case the advance covered by ECGC/ CGTSI guarantee becomes nonperforming, no
provision needs to be made towards the guaranteed portion. The amount outstanding in
excess of the guaranteed portion should be provided for as per the extant guidelines on
provisioning for nonperforming advances. Further, while arriving at the provision required
to be made for doubtful assets, realisable value of security should be first deducted from the
54
outstanding balance in respect of the amount guaranteed by these corporations and then the
provision should be made.
2.7. Operational Concepts Related to NPA

2.7.1. ‘Out of Order’ Status
An account should be treated as ‘out of order’ in any of the following situations
• If the outstanding balance remains continuously in excess of the sanctioned
limit/drawing power for a period of more than 90 days;
• If the outstanding balance in the principal operating account is less than the sanctioned
limit/drawing power, but there are no credits continuously for 90 days as on the date of
Balance Sheet;
• If there are credits in the account, but the credits are not enough to cover the interest
debited in the account during the same period.
2.7.2. Overdue Status

Any amount due to the bank under any credit facility is ‘overdue’ if it is not paid on the due
date fixed by the bank.
2.7.3. Take-Out Finance

Take-out finance is the product emerging in the context of the funding of long term
infrastructure projects. Under the arrangement, a bank financing infrastructure projects will
have an arrangement with any financial institution for transferring to the latter the
outstanding in respect of such financing in their books on a pre determined basis. In view of
the time lag involved in taking-over, the possibility of default in the meantime cannot be
ruled out. The norms of asset classification will have to be followed by the bank as the
account stands as balance sheet item as on the relevant date. If the asset has turned NPA on
the basis of the record of recovery, it should be classified accordingly. Income should not be
recognized on accrual basis and the same should be accounted for only when it is paid by
the takeover borrower/ taking over institution (if the arrangement so provides). Bank should
also make provisions against any asset turning into NPA, pending its takeover by taking
over institution. As and when the asset is taken over by the taking over institution, the
corresponding provisions could be reversed. However, on taking over such assets, the
55
concerned institution should make provisions from the actual date of becoming NPA, even
though the account was not in its books as on that date.
2.7.4. Early Alert System (EAS)

A system that captures early warning signals in respect of account’s showing first signs of
weaknesses. Under it, banks may designate a time limit for overdue accounts to determine
the thresholds for proactive interventions, well before the account becomes NPA. For
example, when there is a default in an account for 30 days, it may be shifted to a special
category named as Special Mention Account. Even when an account is otherwise regular, it
may be classified as potential NPA if there is
• Delay in submission of stock/ financial / other control statements
• Cheques issued by borrowers dishonored when presented for payment.
• Frequent Devolvement of Letter of Credits and /or Bank Guarantee
• Return of bills/ cheques discounted,
• Incomplete documentation and non-compliance of terms and conditions of sanction.
2.7.5. Special Mention Account (SMA)

SMA is a system of early recognition with timely and adequate interventions which may
form the focus of approach in dealing with slippage of NPAs. It is suggested that Banks may
introduce a new asset category, ie, potential NPA between ‘standard’ and ‘substandard’ for
their own internal monitoring and follow-up. The main characteristics of SMA are, the asset
deserves close management attention and if left uncorrected, the account may turn to NPA.
Apart from irregularity, SMA can also be on the basis of inadequate cash flows and such
accounts would require provisioning as per standard advances. It is the present practices
with PSBs that excepting serious impairment of credit and/or significant erosion in the value
of the security, no account can be classified as NPA, if it has not been figured as SMA
2.7.6. Floating Provision

Some banks make floating provision over and above the specific provisions made in respect
of Accounts identified as NPAs. Such provisions can be used only for contingencies under
extra-ordinary circumstances for making specific provisions impaired accounts after
obtaining board’s approval and prior permission of RBI. The board of the banks should lay
56
down approved policies as to what circumstances could be considered as extra-ordinary.
Such extra-ordinary circumstances could broadly fall under 3 categories, ie general, market
and credit.
General category may include situations where bank is put to unexpected loss due to events
such as civil unrest, collapse of currency in the country, natural calamities etc. Market
category would include events such as a general meltdown in the market, which affect the
entire financial system. Among the credit category only exceptional credit losses would be
considered as an extra-ordinary circumstance.
2.7.7. Potential NPA

This classification is more of practical and operational significance rather than theoretical
importance. These assets are still performing asset but exhibiting symptoms of irregularities.
A borrowal account may be termed as Potential NPA if it is out of order for a period of more
than 45 days but less than 90 days. With regard to Term Loan and other category of
advances, if they are overdue for more than 45 days but less than 90 days it will be termed
as Potential NPA. Letter of Credit Devolved account and Bank Guarantee Invoked account
may be termed as Potential NPA, if it remains irregular for a period of one month.
2.7.8. Gross Non Performing Asset (GNPA)

GNPA refers to the sum total of all loan assets that are classified as NPAs as per RBI
guidelines as on balance sheet date and reflect the asset quality of a bank. It consists of all the
nonstandard assets like as sub-standard, doubtful, and loss assets. In other words, GNPA is the
amount outstanding in the borrower’s accounts in the books of the banks other than the
interest which has been recorded and not debited to the borrower accounts.
2.7.9. Net Non Performing Asset (NNPA)

NNPAs refers to those type of NPAs in which the bank has deducted the provision regarding
NPAs. Net NPA shows the actual burden of banks. Since in India, bank balance sheets contain a
huge amount of NPAs and the process of recovery and write off of loans is very time consuming,
the provisions the banks have to make against the NPAs according to the central bank guidelines,
are quite significant. That is why the difference between gross and net NPA is quite high. It can
be calculated by following:
57
Net NPAs = Gross NPAs – (Suspended interest + Amount of provision + Amount of claims
received and not appropriated)
2.8. Conclusion
Before introduction of the Prudential Norms by RBI, there was no concept like ‘income
recognition’. All loan accounts were subjected to interest and other charges, irrespective of
the fact whether such interest or charges have been realised or remain overdue. Health Code
System was used during pre-reform era, classifying advance portfolio of a bank, primarily in
a subjective manner unlike present day classification of loan assets with the help of
objective criteria in a scientific manner. Prior to introduction of prudential norms of
accounting by RBI, all banks were following their own provisioning norms. But with the
introduction of prudential norms, RBI advised that the provisioning norms were to be
uniformly followed by all the banks in India. Provisioning for bad and doubtful debts were
made on the basis of classification of assets reckoning the period for which the asset
remained non-performing and the availability of security and the realisable value thereof.
All these norms and concepts, introduced by RBI in phased manner had compelled banks in
India to follow uniform accounting and disclosure practices and thereby brought paradigm
shift in the mind set of the bankers and other stake holders.
58
Chapter III
Practices of NPA Management in Selected PSBs in India
3.1 Introduction
3.2 Recovery Policy
3.3 Major Policy Measures
3.4 Post Recovery Processes
3.5 Writing off of the Outstanding Bad Loans
3.6 Conclusion
59
3.1. Introduction
Maintaining quality of the assets is the most important objective of a bank. NPAs have a
significant bearing on a bank’s overall financial health. This is a major determining factor of
profitability, productivity and other strategic banking variables. Therefore, all out efforts are
required to ensure that asset quality of a bank is maintained at an acceptable level. Banks try to
eliminate or reduce the NPAs through persuasion by way of compromises or negotiated
settlements with the borrowers or resorting to legal action. Keeping in view the need to have in
place a well planned approach to manage the problem of NPAs at the whole industry level, the
RBI had advised all banks to place before the Board of Directors, a Recovery Policy which sets
down the manner of recovery of dues, targeted level of reduction, norms for permitted sacrifice /
waiver, factors to be taken into account before considering waivers as per delegated authority,
and monitoring of write-off/ waiver cases etc. Accordingly, all PSBs have documented
individually their own Recovery Policy and placed the same for consideration by the Board.
In this chapter, a comprehensive review of existing mechanism for effective management of

NPAs as initiated by PSBs in India has been delineation based on researcher’s personal
interaction with recovery cell officials of such banks and RBI circularized instructions.
3.2. Recovery Policy
Recovery Policy of a PSB is framed with the primary objective of adopting aggressive recovery
measures and implementing case specific strategies, which may include persuasion,
restructuring, settlements, and legal methods, to ensure maximum and timely recovery of dues in
distress asset. To this end, all necessary measures including legal action/ sale of shares/ change
of management, etc. are taken. The emphasis of such policy is on timely action whether it is for
recovery of NPAs or legal action. It is recognized that any delay in decision-making reduces the
chances of recovery.
3.3. Major Policy Measures/Recovery Mechanism
The major policy stance of a bank is to contain further drift in the position of NPAs on the one
hand and simultaneously substantially reduce NPAs by adopting the following measures :
60
3.3.1. Review and Monitoring of Existing NPA Accounts
All existing NPA accounts deserve serious attention. Such accounts are monitored at different
levels of management as on the balance sheet date and are subjected to regular review. Calendar
of review of NPA accounts by the Board, Executive Committee, Audit Committee of Board as
followed by different PSBs are given in the following tables (Table .: 3.1 to Table .: 3.3)
Table- 3.1.: Calendar of Review of NPA accounts by the Board
Sr. No Review items Periodicity

1 Review of Recovery Policy Yearly.
2 Review of Willful Defaulters Half yearly
3 Review of slippages in asset classification in borrowal accounts with Every Board
outstanding of 5 crores and above and review of NPA accounts which Meeting
have registered recoveries of 1 crore and above
4 Review of NPAs including Recovery Status Every Board
Meeting
5 Review of Suit filed cases to include DRT as well as Civil Court Cases Half Yearly
6 Reporting of accounts with outstanding of 50 lakh and above that have Quarterly
become FTNPAs in the preceding quarter
7 Progress in cases where notices have been issued under SARFAESI Quarterly
Act (to be routed through ACB)
Table- 3.2.: Calendar of Review of NPA accounts by the Executive Committee

Sr. No. Review items Periodicity
1 Review of top 100 borrowal accounts of below `5 crores in each category Quarterly
of NPA i.e. substandard/doubtful/loss (75 in each quarter)
2 Review of Compromise Settlements/Bad Debts written-off under Quarterly
delegated authority.
Table- 3.3.: Calendar of Review of NPA accounts by the Audit Committee

Sr. No. Review items Periodicity
1 Review of First Time NPAs Every ACB
meeting
2 Progress in cases where notices have been issued under SARFAESI Act Quarterly
3 Review of loss assets (with outstanding balance Rs10 lakh and above for Half Yearly
more than two years where legal action has not been initiated)
61
3.3.1.1. Focused Recovery of NPAs
The cases that are classified as NPAs and have become First Time NPAs (FTNPA) are referred
to the Audit Committee Board (ACB), immediately after the preceding quarter to examine the
reasons for becoming NPA, the lessons learnt and staff accountability, if any, as per individual
bank's policy is fixed. After exhausting all normal avenues for recovery/resolution of NPAs, such
NPAs shall generally be transferred to Recovery Vertical, without delay, and in any case within a
period not exceeding one year from date of their being classified as NPA.
3.3.2. Close Monitoring and Follow-up of Existing Borrowal Accounts
Constant monitoring of an advance account as per terms and condition of sanction is a pre-
requisite for maintaining sound health of such account. Similarly, religious following up a
potential / First time NPA can ensure timely recovery. Depending on the type of exposure,
monitoring of accounts is done in PSBs as under:
3.3.2.1. Monitoring of Working Capital Exposures
Obtaining all periodical information such as stock/ book debt statements, information under QIS,
balance confirmation, annual financial statements for review of accounts and critically
examining them and scrutinizing the transactions in the account such as whether all sales
proceeds are routed through the account, whether withdrawals from the account (end use) are for
business purpose, etc.
(i) Sending written intimation on interest charged in the account and installment falling due.
(ii) Wherever working capital limits are Rs.5 crore or more and secured by current assets, stock
audit is conducted at least once in a year.
(iii) Periodic inspection of stock, which is used as an effective tool to keep the account regular
and to recover the bank’s dues. Any delay in submission of the stock statements beyond
10th of the following month is being followed-up.
(iv) Updating the net worth/ means of borrowers/ guarantors.
3.3.2.2. Monitoring of Term Loan Exposures
(i) Comparing of annual accounts to provisional quarterly results submitted by the company
and monitors major deviations and seeks necessary explanations from company.
62
(ii) Analysing budgeted cash flows to examine whether the projected cash flows are realistic
and sufficient to service interest and installments.
(iii) Frequency of plant visits based on the health of the company, availability and condition of
secured assets, prospects of resolution, etc. at least once a year.
3.3.2.3. Monitoring of Corporate Finance Sanctioned by Top Management (Board)
(i) In case of consortium/ multiple financing, a bank generally takes a pragmatic approach in
consultation with other banks/ financial institutions. Wherever units are working, banks
collect information on cash flow and monitor it to ensure that such flows are used for
business purposes.
(ii) Analysing Board agenda papers with respect to important items for discussion and their
impact on operations, long-term viability and repayment capacity of company.
(iii) Analysing Audit Reports with respect to end use of funds, diversion/ siphoning of funds,
status of implementation of projects/ systems.
(iv) Regular interaction with other secured creditors.
(v) Periodical review / rating of the accounts.
(vi) Feed-back from Nominee Director.
(vii) Exchange of information with other bankers.
(viii) Information in Public Domain
(ix) General conduct of the account
3.3.2.4. Insurance
Bankers always ensure that the assets charged to a bank are adequately insured. In the event the
borrower fails to take/ renew insurance cover, bank insures the assets adequately to safeguard its
interest. In case of closed units, the assets are covered under Master Insurance Plan to minimize
the cost to the Bank.
63
3.3.3. Turnaround through Restructuring
Based on promoter’s sincerity and capability to achieve a turnaround, a bank often examines
restructuring proposal and consider additional finance on a case-to-case basis as a part of
Corporate Debt Restructuring package/ BIFR rehabilitation scheme. Before committing
additional finance, a bank usually undertakes, a ‘Special Audit’ of all financial transactions and
books of accounts in order to ascertain real factors that contributed to the sickness of the
borrower. Potentially viable units are assisted by restructuring, after a techno-economic viability
study is made and after examining and confirming the quality of management. Financial viability
of the restructured unit is critically examined through appraisal of funds flow as projected by the
promoters.
3.3.3.1. Restructuring of Dues
Each PSB has designed broad guidelines for restructuring based on the guidelines issued by RBI,
from time to time. Such restructuring guidelines cover all business groups of the bank.
3.3.3.1.1. Corporate, Business, Financial and Operational Restructuring
A rehabilitation/ restructuring scheme is considered selectively, only after the viability of the
company is established beyond doubt on realistic projections, past track record, industry related
issues and subject to the fulfillment of following conditions:
(i) The unit becomes viable within 10 years, if it is engaged in infrastructure activities and 7
years in case of other units.
(ii) The repayment period of the restructured advances including moratorium, if any, shall not
exceed 15 years, in case of infrastructure advances and 10 years in case of other units.
(iii) The restructured dues of the bank are fully secured.
(iv) Promoters' sacrifice and additional funds brought in by them are minimum of 15% of the
bank's sacrifice. Relaxation on promoters' contribution is often considered on case-to-case
basis.
(v) Right of recompense is incorporated.
(vi) All possible measures are adopted/ explored towards credit enhancement.
64
(vii) Restructuring is explored as a first option in Sub Standard/ Doubtful accounts, if the unit is
found viable with suitable reliefs/ concessions. If long-term viability is doubtful,
prepayment/ settlement of dues is considered.
3.3.3.1.2. Concessions on Restructuring of Dues
In stressed standard cases as well as NPA accounts, overdue liquidated damages and compound
interest are waived (depending on cash flow), simple interest are funded so as to be payable
within 2 years, carrying interest rate linked to BBR (Bank Base Rate), interest rate on principal is
reduced from prospective date through linking it to BBR and payment of principal is rescheduled
based on cash flow projections. As far as possible, funded and deferred interest (both existing as
well as new) are recovered before the principal.
3.3.3.1.3. Security Issue on Restructuring of Dues
The financial restructuring package is usually accompanied by a suitable corporate, operational

and managerial restructuring to improve the gross profit levels. For restructuring package to be
successful, it is essential that majority (preferably all) secured creditors approve the package and
company has been able to tie up necessary working capital also.
(i) In case Funded Asset Coverage Ratio (FACR) is less than 1.25, possibility of obtaining
additional security is explored.
(ii) Conversion of a portion of the restructured dues into equity/ convertible bonds/ convertible
debentures are considered keeping in view statutory requirement under Section 19 of the
Banking Regulation Act, 1949 (in case of banks) and the relevant SEBI regulations.
(iii) Possibility of obtaining personal guarantees of promoters and pledge of shares with voting
rights is explored in all restructuring cases.
(iv) Consents of all the guarantors are obtained for any restructuring packages.
(v) All standard terms and conditions like right of recompense (based on certain performance
criteria decided by the bank), acceleration, review and revocation, promoters' contribution,
etc., are stipulated.
(vi) In respect of restructuring proposals through a common forum like BIFR, CorDR etc.,
equitable sacrifices by all stakeholders in the project are generally insisted upon.
65
3.3.3.1.4. General Guidelines for Restructuring
(i) The amount of economic sacrifice (calculated as per RBI guidelines), if any, in the element
of interest, measured in present value terms, is either written-off or provision is made.
(ii) The erosion in the fair value of the advance is computed as the difference between the fair
value of the loan before and after restructuring. Fair value of the loan before restructuring
is computed as the present value of cash flows representing the interest at the existing rate
charged on the advance before restructuring and the principal, discounted at a rate equal to
the bank’s BBR as on the date of restructuring plus the appropriate term premium and
credit risk premium for the borrower category as on the date of restructuring. Fair Value of
the loan after restructuring is computed as the present value of cash flows representing the
interest at the rate charged on the advance on restructuring and the principal discounted at a
rate equal to the bank’s BBR as on the date of restructuring plus the appropriate term
premium and credit risk premium for the borrower category on the date of restructuring.
(iii) Restructuring/ Rescheduling is applicable to all types of credit facilities including working
capital limits (except credit facilities extended to traders).
(iv) Restructuring is carried out such that the cut-off date for restructuring of dues is close to
the restructuring date.
(v) Restructuring/ Rescheduling is not allowed to be carried out repeatedly, unless there are
very strong and valid reasons which warrant such repeated restructuring/rescheduling.
(vi) During the pendency of the application for restructuring of the advance of a bank, the usual
asset classification norms would continue to apply. The process of reclassification of asset
shall not stop merely because the application is under consideration. However, as an
incentive for quick implementation of the package, if the approved package is implemented
by a bank as per the below mentioned time schedule, the asset classification status is
restored to the position which existed when the reference was made to the CorDR Cell in
respect of cases covered under the CorDR mechanism or when the restructuring application
was received by a bank in non- CorDR cases:
a) Within 120 days from the date of approval under the CorDR mechanism.
66
b) Within 90 days from the date of receipt of application by the bank in cases other than
those restructured under the CorDR mechanism.
3.3.3.2. Corporate Debt Restructuring (CorDR)
Objective of the CorDR framework is to ensure speedy and transparent mechanism for
restructuring of the corporate debts of viable entities facing problems, outside the purview of
BIFR, DRT and other legal proceedings, for the benefit of all concerned. The CorDR mechanism
covers only multiple banking accounts / syndication / consortium accounts with outstanding
exposure of Rs.10 Crore and above by banks and Financial Institutions (FIs). Reference to
CorDR System is triggered by –
(i) any or more of the secured creditor who have minimum 20% share in either working
capital or term finance, or
(ii) by the concerned corporate, if supported by a bank or FIs having stake as in (a) above.
3.3.3.2.1. Categories Corporate Borrowal Accounts for Restructuring under CorDR

mechanism
There are two categories corporate borrowal accounts for restructuring under CorDR mechanism
Category 1- Standard and Sub-Standard accounts
The Category 1 CorDR System is applicable only to accounts classified as 'standard / sub-
standard'. In a situation where a small portion (upto 10% by value) of debt by the bank might be
classified as doubtful, leaving at least 90% of creditors (by value), been classified as 'standard /
sub-standard' in the books, the entire account would be treated as Standard / Sub-standard only
for the purpose of judging the account as eligible for CorDR.
Additional finance, if any, is provided by all the creditors, of a 'standard / sub-standard account’
irrespective of whether they are working capital or term creditors, on a pro-rata basis. In case for
any internal reason, any creditor (outside the minimum 75% and 60%) does not wish to commit
additional financing, that creditor will have an exit option either by (a) arranging for its share of
additional finance to be provided by a new or existing creditor, or (b) agree to the deferment of
the first year's interest due to it, after CorDR package becomes effective.
67
Category 2- Doubtful Accounts
There are instances where the projects have been found to be viable by the creditors but the
accounts could not be taken up for restructuring under the CorDR system as they fall under
‘doubtful’ category. Hence, a second category of CorDR is introduced for cases where the
accounts have been classified as doubtful in the books of the creditors, and if a minimum of 75%
of creditors (by value) and 60 % creditors (by number) satisfy themselves of the viability of the
account and consent for such restructuring, subject to the following conditions:
(i) It is not binding on the creditors to take up additional financing worked out under the debt
restructuring package and the decision to lend or not to lend will depend on each creditor
Bank/FI separately. In other words, under the proposed second category of the CorDR
mechanism, the existing loans is restructured and it is up to the promoter to firm up additional
financing arrangement with new or existing creditors individually.
(ii) All other norms under the CorDR mechanism such as the standstill clause, asset classification
status during the pendency of restructuring under CorDR etc. are applicable to this category also.
3.3.3.2.2. Concession in Asset Classification and Provisioning Norms
The regulatory concession in asset classification and provisioning will be available if there is
compliance of four conditions as stipulated in RBI guidelines viz.,
(i) restructuring is done for first, time,
(ii) unit becomes viable in seven years and repayment of debt does not exceed 10 years,
(iii) promoter’s sacrifice is minimum 15% of creditor's sacrifice and
(iv) personal guarantee of promoters is stipulated, except in cases where the unit is affected by
external factors pertaining to the economy/ industry.
3.3.3.2.3. General guidelines being followed by PSBs under CorDR mechanism:
(i) Both debtor and creditor shall agree to a legally binding "stand-still" agreement whereby
both the parties commit themselves not to take recourse to any other legal action (except
criminal action) during the "stand-still" period.
68
(ii) Borrowers indulging in frauds and malfeasance will continue to remain ineligible for
restructuring under the CorDR mechanism. Inclusion of such borrowers will be considered
with the approval of the Core Group.
(iii) The accounts where recovery suits have been filed by the bank against the company, may be
eligible for consideration under CorDR system provided, the initiative to resolve the case
under the CorDR system is taken by at least 75% of the creditors (by value) and 60% of
creditors (by number). BIFR cases are not eligible for restructuring under the CorDR
system. However, large value BIFR cases may be eligible for restructuring under the
CorDR system, if specifically recommended by the CorDR Core Group. It shall be ensured
that the lending institutions complete all the formalities in seeking the approval from BIFR
before implementing the package.
(iv) As an incentive for quick implementation of the package, the asset classification status is
restored to the position where it existed when the reference to the CorDR was made if the
approved restructuring package is implemented within four months from the date of
approval. The additional provision made towards deterioration in the asset classification
during pendency of case with CorDR may be reversed.
(v) During the pendency of the application for restructuring of the advance of a bank, the usual
asset classification norms would continue to apply. The process of reclassification of asset
shall not stop merely because the application is under consideration.
(vi) The exit option is also available to all lenders within the minimum 75% by value and 60%
by number provided the purchaser agrees to abide by restructuring package approved by
CorDR.
(vii) One Time Settlement (OTS) is considered as an exit option wherever necessary as a part of
the restructuring package. If an account with any creditor is subject to OTS by a borrower
before its reference to the CorDR mechanism, any fulfilled commitments under such OTS
is not reversed under the restructured package. Further, payment commitments of the
borrower arising out of the OTS are factored into the restructured package.
(viii) The providers of additional finance, whether existing or new creditors, shall have a
preferential claim, being worked out under the restructuring package, over the providers of
69
existing finance with respect to the cash flows out of recoveries in respect of the additional
exposure.
(ix) The prudential norms, accounting issues and disclosure norms as per the guidelines issued
by regulatory authority are applicable to all accounts under CorDR mechanism.
3.3.4. One Time Settlement (OTS)/ Negotiated Settlement (NS)
3.3.4.1. Identification of an OTS/ NS Proposal
OTS or NS is explored/ encouraged based on the merits in respect of cases, where
(i) Restructuring is not likely to succeed and /or previous efforts for restructuring have failed.
(ii) The account has been identified as Focus account or as an NPA including doubtful and loss
asset account
(iii) Risks relating to promoters, management and industry are perceived to be relatively higher
and hence it would be advisable to exit from the company as early as possible.
(iv) Security and documentation formalities are not perfect and therefore recovery prospects
from borrower and guarantors are likely to be affected adversely and / or the Bank’s share
in estimated realization from sale of primary security and from guarantors after completion
of the due legal process will not be adequate to recover the bank’s dues in full.
It is however not generally applicable where the borrower is considered a willful defaulter by the
bank or where there is suspicion of fraud and/or an attempt to cheat the bank, except with the
approval of the Board/ EC.
3.3.4.2. Settlement Criteria in an OTS/ NS Proposal
(i) All dues including loans/ credit facilities, outstanding expenses (legal, insurance, valuation,
audit, etc.) are taken into account while working out the settlement proposal.
(ii) A large portion of settlement amount is normally arranged from external sources (including
sale of surplus assets) and the company shall not rely too much on the company's cash
flows or internal generation.
70
(iii) At least, 10 to 25% of the settlement amount is insisted upon as down payment in a non-
interest bearing ‘No-lien account’ and PDCs are obtained for the balance amount.
Exceptions to the same is considered subject to approval of the Delegated Authority.
(iv) In case of multiple lenders, while it is preferable to have a consensus among all secured
lenders, in the event there is no consensus or the settlement is getting delayed and the
company is prepared to make payment to the bank without approval of other lenders, the
same is accepted. If felt necessary, a condition is stipulated that in case company offers
better terms (on present value basis) to any other secured creditor, the bank is brought on
par with them.
(v) In case the borrower/ promoters are not able to pay the entire settlement amount in cash,
possibility of swapping the settlement amount with properties, shares of good companies
owned by the borrower / promoters, is explored, after carrying out valuation / due
diligence.
(vi) Right of recompense is not stipulated if the repayment period is upto one year and entire
funds are coming from external sources. In case part of funds are being arranged by sale of
assets charged to the bank or internal accruals, right of recompense is generally considered
to the extent of actual sale value of assets / internal accrual being higher than the projected
amount.
(vii) In respect of OTS and NS proposals through a common forum like BIFR, Corporate Debt
Restructuring, etc., equitable sacrifices by all stakeholders in the project is insisted upon.
Valuation of assets charged to the bank is generally insisted in all OTS/ NS cases and is
undertaken by an empanelled valuer with the bank.
(viii) In case the value of the property is Rs.50 crore and above, valuation is carried out by two
different independent valuers empanelled with the bank. The valuation shall also include
business valuation, if there are prospects.
(ix) While assessing the realizable value of the securities available to the bank, proper
weightage is given to the location, condition, marketable title and possession thereof.
(x) If an OTS/ NS arrangement envisages continuing exposure of the bank beyond a period of
one year from the date of agreement, additional security is insisted.
71
(xi) In the case of suit filed/ BIFR cases where an OTS/ NS is sanctioned, consent terms signed
by the bank and the borrower/ guarantor, filed with Court/ DRT and consent decree
obtained. Intimation is given to BIFR in respect of BIFR cases. Consent decree shall
contain the default clause to the effect that the debt to the bank would revive and become
payable under the decree with up-to-date interest at contracted rate, costs and expenses in
case the borrower does not pay the compromise amount as stipulated.
(xii) Necessary safeguards to protect the bank’s interests are stipulated in all OTS/ NS
approvals, including conditions like right of revocation/ review/ accelerated repayments,
etc.
(xiii) In cases, where the borrower defaults in payment of OTS/ NS amount, approval for
extension of time up to 6 months with payment of interest is referred to the respective
Delegated Authority for continuing/ revoking of the package explaining the reasons. Cases
where approval for extension of time for payment of OTS/ NS is beyond 6 months, the
same is referred to Recovery Committee. In case, where extension of time is considered
either without payment of interest or waiver of part interest, the same is referred to
Recovery Committee for approval.
(xiv) OTS/ NS sanctioned is conveyed in writing to the borrower(s) and his / their acceptance
along with consent of guarantors is obtained in writing immediately. The sanction letter
contains the terms and conditions governing the OTS. There is a default clause to the effect
that if the borrower does not pay the OTS/ NS amount (bullet payment/ installment), the
bank shall have the right to treat the sanction letter as withdrawn and in that event the
borrower will be liable to pay the entire gross dues to the bank. Endeavor is made to obtain
post-dated cheques at least for the installment (principal) amount, interest being payable
separately. In case of non-suit filed accounts, it is ensured that the documents are kept
alive by obtaining Letter of Acknowledgement of Debt from borrowers /guarantors.
(xv) In respect of fully provided/ written-off/ loss accounts where an OTS/ NS or a restructuring
package has been approved and is under implementation, pre-payments on PV basis is
accepted. Appropriate graded additional discount is considered to expedite recovery within
3 to 6 months with the approval of Delegated Authority as per existing Delegation of
Powers (DOP).
72
(xvi) In closed units the settlement amount is worked out on the basis of bank's pro-rata share in
the distress sale value of assets, while in units under operation, settlement amount shall be
based on bank's pro-rata share in realizable value of assets. Other collateral securities, if
available, are kept in view while arriving at the settlement amount. In case of constraints
which are financial, legal or of any other nature, which may hamper the recovery/
resolution, an appropriate discount of 10-30% over and above the applicable valuation is
offered with justifiable reasons, provided that the payments are made within 3 to 6 months.
The realisability of other collateral securities, if any, including personal guarantees and
corporate guarantees etc, are appropriately factored in the proposal.
(xvii)While working out the pro-rata share of the bank in the realizable / distress sale value of the
asset, care is being taken to consider first expenses incurred towards insurance, security as
well as expenses incurred towards sale of assets and secondly the statutory dues such as
sales tax and EPF. In respect of the companies under liquidation/ winding-up, the claims of
workmen and the claims of the secured creditors shall be considered on pari-passu basis
(xviii) All OTS proposals referred to the Delegated Authority are first referred to the Screening
Committee for its recommendation. Proposals involving modification in approved
settlement amount are also referred to Screening Committee.
3.3.4.3. Settlement Coverage in an OTS/ NS Proposal
Cases which are standard assets as on date but are facing difficulties and are likely to become
NPA, Substandard assets, Doubtful assets, Loss and Fully Written-Off Assets.
3.3.4.4. Settlement Amount in an OTS/ NS Proposal
Based on the general guidelines, the settlement amount is worked out as under:
(i) In case of Sub Standard assets, the crystallized amount could generally comprise at least
the outstanding principal and simple interest calculated at BBR plus spread from the date of
first default/ signs of weakness. However, the settlement at lower amount is considered
depending on the value of security, net loan outstanding and other related factors.
(ii) For Doubtful and Loss assets, the settlement below principal outstanding is considered on
merits of each case.
73
(iii) In Loss and Fully Written Off assets, if the unit is lying closed, settlement amount is
worked out on the basis of the bank's pro rata share in the distress sale value of assets. In
operating units, settlement amount is based on the bank's pro rata share in the realizable
value of assets.
(iv) Minimum settlement amount depends on the nature of the assets, security available, value
thereof etc.
(v) The OTS amount when payable in installments, the net present value (NPV) of the
settlement amount is calculated and this amount is generally not be less than the NPV of
the realiasable value of securities held with the bank.
3.3.4.5. Payment Terms in OTS/ NS Settlements
(i) The entire settlement amount is preferably paid within a period of 12 months with upfront
payment to the extent possible (at least 10 to 25%). Interest @ BBR plus spread is explored
on the outstanding crystallized amount from the date of LOI on merit basis.
(ii) Modifications in payment terms including rate of interest is considered on case-to-case

basis by the Delegated Authority.
(iii) If a part of the settlement amount is proposed to be recovered by way of conversion into
any instrument, the said conversion is made in accordance with the investment policy of the
bank/ other statutory rules, as applicable.
3.3.4.6. Relaxation of OTS/ NS Norms
In all NPA cases, exceptions/ deviations are considered, if any, if the following situations prevail
(i) Cases pending in the DRT/Courts/BIFR/AAIFR for more than 2 years.
(ii) Unit lying closed for more than 2 years.
(iii) Security has not been created / perfected.
(iv) Project remaining unimplemented and/or project has been abandoned.
(v) Companies which have continuously incurred gross loss for last three years or since the
year in which first default to the bank (in payment of interest) occurred.
(vi) Companies which have not achieved cash break even during the last 3 years.
74
(vii) Realizable value of mortgaged assets as assessed by an independent approved valuer being
lower than the loan outstanding.
Deviations if any can be considered on a case–to–case basis by the respective Delegated

Authority. In such cases, appropriate discounts below the principal outstanding are considered
subject to:
(i) Recovery of minimum of net loan outstanding, and
(ii) Entire payment is received within maximum period of one year.
(iii) Settlement is also considered below net loan outstanding in exceptional cases if available
realizable/ distress value of security is not expected to cover entire net loan outstanding and
the chances of deterioration in security are fast.
3.3.4.7. Authorization in OTS/ NS Settlements
OTS / NS proposals are approved by the delegated authority as per extant DOP. Quarterly
updates shall be placed before the delegated authority on progress in implementation / recoveries
in respect of OTS / Compromise proposals approved by the bank.
3.3.4.8. Release of Security Documents in OTS/ NS Settlements
Release of security/security document is completed normally within three months in accordance

with the existing DOP, if the terms and conditions of compromise are fulfilled and the entire
compromise amount has been paid by the borrower as per the terms of settlement and as per
situation.
3.3.5. Compromise Settlements in Suit Filed, BIFR and Lok Adalats Cases
3.3.5.1. Compromise Settlements in Suit Filed Accounts
In case of suit filed accounts where bank approves the compromise proposals, the compromised
amount is payable in one lump sum within a stipulated period or in installments. In such event
the bank invariably file consent terms before the Courts/ DRT for passing the decree/orders. The
consent terms shall include
(i) amount payable by the borrower calculated as on the date of filing the consent terms
75
(ii) payment of future interest at contractual rate. If the borrower/guarantor pays the agreed
amount as per the compromise, the decree shall stand satisfied.
(iii) in case of default in the payment as per the terms of compromise, the bank is entitled to
recover the entire decreed amount as awarded by the Court/ DRT, after adjusting the
amount, if any, paid by that time.
(iv) steps are taken immediately to enforce the decree/ recovery certificate.
3.3.5.2. Compromise Settlements in BIFR Accounts
In respect of BIFR cases, where the compromise settlement is entered into by the bank and forms
a part of the scheme sanctioned by BIFR, the settlement between borrower and the bank is
governed by the terms of the sanctioned scheme. In respect of other BIFR cases where
compromise settlement is not forming a part of BIFR sanctioned scheme, BIFR is kept informed
of the same. In the event of any breach of terms and conditions of the compromise settlement,
necessary action is taken after seeking the approval of BIFR.
3.3.5.3. Compromise Settlement through Lok Adalats
Under the Legal Services Authorities Act, Lok Adalat has been given the status of a Civil Court
and every award made by Lok Adalat is final and binding on all parties and no appeal lies to any
court against its award. There are certain advantages in using the forum of Lok Adalats by banks
and FIs in compromise settlement of their NPAs. There are no court fees involved when fresh
disputes are referred to it. It can take cognizance of any existing suit in the court as well as look
into and adjudicate upon fresh disputes. If no settlement is arrived at, the parties can continue
with court proceedings. Its decrees have legal status and are binding.
In order to make a definite impact on reduction of NPAs reference to Lok Adalat may include all
NPA accounts, both suit filed and non-suit filed accounts, which are in "doubtful" and "loss"
category, with outstanding balance up to Rs.`20 lakh.
3.3.6. Acquiring Real Estate Assets in satisfaction of Bank dues in NPA accounts
In an NPA account, wherever security by way of equitable mortgage of landed property is

available to the bank, such a security is sold only through a legal process and the proceeds
utilized to adjust the bank's dues. Such a procedure requires quite a long time to fructify, by
which time the value of the property declines or the property is encroached upon or even
76
clandestinely sold on the basis of duplicate documents etc. Moreover, during the intervening
period, the bank incurs interest on funding the NPA and does not get an opportunity to use the
property. In order to overcome the above said drawbacks, if the borrower / guarantor agrees, the
bank may acquire the property in partial or full satisfaction of the bank’s dues for the purpose of
bank's own use as office premises or for residential quarters for the employees. In certain cases
the property can be developed through property developers and sold to public to realize the
bank’s dues. While taking such decision, the Competent Authorities considers the following –
(i) That the land/ building is sold under auction is under physical possession of borrowers/
guarantors.
(ii) The bank after purchase is in a position to sell the property within a period of say 6 months
to 2 years and the progress in this regard is put up for information of the Recovery
Committee on a quarterly basis.
(iii) That there is no likelihood of the property being disposed off during the intervening period
i.e. from the date of acquisition to the date of sale of the property.
The Competent Authority for taking decision in the matter of acquiring the property /
participation in auction sale is held in execution of decree made in favour of the bank is the
Recovery Committee. In case the bank desires to develop the property and sell thereafter,
approval of Board/ EC is obtained.
3.3.7. Sale of Assets under SARFAESI Act, 2002
(i) In terms of Sec 13(2) and 13(4) of the Securitisation and Reconstruction of Financial
Assets and Enforcement of Security Interest (SARFAESI) Act, 2002, a bank can issue a
notice to the NPA borrower to discharge his liabilities within 60 days from the date of
notice failing which the bank can take possession of the secured assets with requisite
consent of secured creditors and sell the same for realizing the bank's dues.
(ii) Where the borrower / guarantors do not come forward with an acceptable settlement and
there are no other avenues for expeditious recovery of dues, it is imperative that the bank
makes swift and optimum usage of the SARFAESI Act 2002.
(iii) OTS is also considered in cases where SARFAESI action has already been initiated by the
bank with stipulation in the OTS terms that further action to be initiated within the
77
provisions of SARFAESI would be kept in abeyance and it would be reactivated in case of
failure of the OTS package.
(iv) A bank may explore the possibility of initiation of action under SARFAESI Act 2002 even
in cases where recovery application is already filed before DRT.
(v) Enforcement of security under the SARFAESI Act is completed as early as possible, except
in cases where borrower has obtained order of injunction restraining bank from proceeding
further under the Act. If the dues of the bank are not fully satisfied by appropriation of sale
proceeds of secured assets, bank takes steps for recovery of balance amount of outstanding
dues from the borrower(s) and guarantor(s) by filing application before the concerned
DRT.
(vi) Care is exercised while taking possession of the assets including proper identification of the
assets charged to the bank as also charged to the other lenders who have consented for
SARFAESI action.
(vii) The publication of Possession Notice is made in leading news papers (one in English and
one in vernacular language) as soon as possible but in any case not later than seven days
from the date of taking possession of the immovable property.
(viii) The valuer is appointed for valuation of assets taken over under the SARFAESI Act has to
be a person who is (i) registered under Section 34AB of the Wealth-tax Act, 1957, (ii) he or
his firm is empanelled as a valuer with the Bank as per the Board approved criteria.
3.3.7.1. Conditions for Initiation of Action under SARFAESI Act
(i) Borrowers identified for Initiation of action under the Act is evaluated based on following :
a. Type of security: generally saleable assets secured by exclusive charge are preferred for
faster resolution.
b. Ready availability of prospective buyer.
c. Nature and gravity of attendant problems viz. whether unit is in operation, labour
problems, pending statutory dues, other litigations, etc.
d. Chances and time frame for realization of the security and an estimate of watch and
ward expenses in the interim.
78
e. Number of co-lenders involved and likelihood of obtaining their consent, etc.
(ii) Initiation of action under the Act is approved only by the respective Delegated Authority.
(iii) After initiating action and in cases where possession of the secured assets has been taken by
the bank successfully, the next step is selling the assets to realize the bank's dues. For the
purpose, the bank would require to fix the Reserve Price (RP) before advertising for sale.
a. RP is fixed based on the valuation of the secured assets. For the purpose of fixing RP,
cases may broadly be divided into 2 categories:
i. Running Unit: The RP is fixed based on the realisable value of the secured assets.
ii. Closed Units: The RP is fixed based on the Distress Sale Value (DSV) of the
secured assets.
b. In case, first attempt for sale of assets fails, lowering of RP may be considered on a
case-to-case basis, with the approval of the Delegated Authority.
(iv) Earnest Money Deposit (EMD), which is normally stipulated shall be at least 5% of
the RP, and duly approved by the Delegated Authority.
(v) The bank appoints an "Authorized Officer" as per prevailing DOP for the purpose
of complying with formalities under the Act, for seizure of the assets / properties,
etc. The Authorized Officer so appointed shall not be a person who is, or has been
adjudicated insolvent, or has suspended payment or has compounded with his
creditors, or who is, or has been, convicted by a criminal court of an offence
involving moral turpitude.
(vi) After appropriation of sale proceeds of secured assets, recovery application is filed/
pursued with DRT against the guarantors for recovery of balance dues.
3.3.8. Sale of NPAs to Asset Reconstruction Companies (ARCs)/ Banks/ FIs/ NBFCs
A bank can sell the NPAs along with underlying securities to ARCs as well as to other banks/
FIs/ NBFCs, for recovering bank's dues from hard core NPAs under following conditions:
A. An asset is sold to ARC, provided it is
(i) a NPA including non performing Bond/ Debenture,
79
(ii) a standard asset, where
a. the asset is under consortium/ multiple banking arrangements
b. at least 75% by value of the asset is classified as NPA in the books of other banks
c. at least 75% (by value) of the banks/ FIs who are under the consortium /multiple
banking arrangements agree to the sale of the asset to ARC.
B. A NPA in the books of a bank is eligible for sale to other banks/ FIs/ NBFCs only, if it has
remained a NPA for at least 2 years in the books of the selling bank.
C. A bank is permitted to sell homogeneous pool of assets and also pool of NPA, on a portfolio
basis provided each of the NPA of the pool has remained NPA for at least 2 years in the books of
the selling bank. The pool of assets would be treated as a single asset in the books of the
purchasing bank.
3.3.8.1. Norms for Sale of NPAs
(i) A bank normally sells the NPAs on "As is where is basis and as is what is basis" without
any recourse to the bank i.e., the entire credit risk associated with the NPA is transferred to
the purchasing bank. However, a bank also sells NPAs to ARCs/ Banks/ FIs/ NBFCs on
"with recourse" basis i.e., subject to unrealised part of the asset reverting to the seller bank.
(ii) In case of such sale of NPAs, the entire sale consideration is received upfront on cash basis
without any recourse to the selling bank. On receipt of entire sale consideration, the asset is
taken out of the books of the selling bank.
(iii) In case of sale of NPAs to ARCs, the bonds/ debentures received as sale consideration
towards sale of financial assets to ARCs are classified as investments in the books of the
bank.
(iv) While selling NPAs, the Net Present Value (NPV) of the estimated cash flows associated
with the realizable value of the available securities net of the cost of realization is worked
out. The sale price is generally not lower than the NPV arrived at as indicated above.
(v) After such sale of NPAs, it is ensured that the asset is taken off the books of the bank and
after the sale there shall not be any known liability devolving on the selling bank.
80
(vi) Banks ensure that subsequent to such sale of the NPAs it shall not have any involvement
with reference to the assets sold and do not assume operational, legal or any other type of
risks relating to the financial assets sold.
(vii) The bidders eligible for purchase of NPAs are as per Bank's Policy i.e. any ARC registered
with RBI, Banks, FIs & NBFCs registered with RBI and also into the business of acquiring
NPAs from various Banks for at least last two years.
(viii) Banks are not eligible to participate in purchase of NPAs, which have been classified as
NPAs for a period less than two years.
(ix) Such sale does not ordinarily result in a debit to the profit and loss account. In exceptional
cases, sale at a loss may also be considered with proper justification for considering such
sale.
(x) The selling bank then pursue the staff accountability aspects
3.3.8.2. Valuation Procedure in Sale of NPAs
Valuation procedure ensures that the economic value of financial assets is reasonably estimated
based on the estimated cash flows arising out of repayments and recovery prospects. The
valuation procedures being adopted are as under:
(i) The NPAs proposed to be sold either on single basis or portfolio basis are valued by an
empanelled valuer.
(ii) The Replacement Value, Fair Market Value, Realizable Value and Distress Sale Value of all
underlying securities, collaterals as also the value of assets under personal guarantees
executed/ shares pledged with the bank are taken into consideration for fixation of RP.
(iii) As per RBI guidelines, two independent valuations is carried out in respect of NPAs where
the asset size is more than Rs.50 crore.
3.3.8.3. Delegation of Powers in Sale of NPAs
On finalization of highest bidder, the bank obtains necessary approvals as per extant DOP on sale
of NPAs to ARCs/ Banks/ FIs/ NBFCs.
(i) The single asset/ pool of assets (corporate/retail) for sale is referred to the approval of
Recovery Committee, if there is no loss to the Bank.
81
(ii) In case of sale of NPAs that would result into loss to the Bank, such transaction is effected
only with the approval of the Executive Committee (EC).
(iii) Ordinarily the Bank shall not purchase NPAs of other Banks/ FIs/ NBFCs. However, where
the Bank is likely to get majority control on the security created for the financial assistance
or where the acquisition will result in debt aggregation for quicker and better resolution of
the Bank's NPAs, they may be considered for purchase. Purchase of NPAs is undertaken
with the approval of the EC.
3.3.8.4. Accounting Policy in Sale of NPAs
(i) If the sale of assets to ARCs/Banks/ FIs/ NBFCs is at a price below the Net Book Value
(NBV) (i.e., book value less provisions held), the shortfall shall be debited to the profit and
loss account of that year.
(ii) In case of sale of assets to ARCs/Banks/FIs/NBFCs at an amount which is higher than NBV,
the excess provision shall not be reversed and the same shall be utilized to meet the
shortfall/loss on account of sale of other financial assets by the bank.
(iii) The usual accounting policy prevailed in the bank relating to receipts and payments of
money shall be applied for the sale of the NPAs also.
3.3.8.5. Concluding Legal formalities with the Purchaser
On receipt of the entire consideration money from the purchaser, the Deed of Assignment is
executed and also registered at the concerned Sub-Registrar Office where the immovable
property is situated. Accordingly the letter conveying approval for Assignment contains a clause
that such execution shall normally not exceed a time period of 6 months.
3.3.9. Change of Management or Induction of Strategic Investors
(i) The option of change of management is explored in the following specific circumstances,
whenever it is felt that the dues may not be recovered from the existing promoters even
with a restructuring package
(a) Cases involving fraud.
(b) Existing promoters do not have the resources or the willingness to turnaround the company
and repay the bank’s dues.
82
(c) Restructuring or settlement package failed even after 2 – 3 modifications.
(d) The promoters themselves indicate that change of management would be a better option.
(e) The institutions / banks lose faith in the management.
(f) In-fighting among the promoters, which could affect the bank’s recovery.
(ii) The proposals is examined based on financial soundness, capability, seriousness and
willingness to bring funds, minimum sacrifice from the secured creditors, satisfactory track
record with acceptability to institutions and banks, etc.
(iii) Other policy guidelines being followed by PSBs in respect of change of management are:
(a) Credentials and seriousness of the new promoters be checked. It would be preferable,
though not essential, to have some upfront refundable deposit from the new promoters.
(b) The level of sacrifices is decided taking into account the overall value of business and
assets of the company, size of liabilities, availability of other options, etc.
3.3.10. Legal Action
(i) Filing a suit against the borrower and the guarantors are considered as the last option, after
other feasible options of recovery of dues are exhausted, at an appropriate stage of the
account becoming NPA. Suits are filed before a bank’s claim gets time barred under Law
of Limitation of the country. In respect of NPA accounts, which are registered with BIFR,
prior permission of BIFR under Section 22 of SICA are obtained before filing recovery
application.
(ii) Various options available for recovery of bank's dues are explored simultaneously during
the process of filing the suit, at proceedings stage and even after the decree is obtained.
(iii) After being satisfied with enforceability of legal documents in possession, legal notice is
served against the borrower and the guarantors simultaneously. If there is no response to
the legal notices, banks usually proceed with the filing of a suit through their legal cell or
duly appointed advocate. Recovery officer in the branch co-ordinates with the Advocates /
Solicitors regularly to ensure that the cases are adjudicated expeditiously. Such officer also
ensures that bank’s advocates provide complete documents to Debt Recovery Tribunal
(DRT) to minimize loss of time and effort.
83
(iv) In case the borrower does not co-operate for a compromise settlement, vigorous steps are
taken for obtaining a decree from the Court / DRT.
3.3.10.1. Follow-up for Recovery from Decreed Accounts
(i) Once a decree/ Recovery Certificate is obtained, execution of the same is undertaken
immediately. Even when the decree is under execution, the borrower are personally
contacted and persuaded to pay off the dues. In case the decreed amount is not paid,
insolvency notice are served on the borrower through the court demanding payment failure
of which would be considered as an act of insolvency. Interim reliefs against the
defendant(s) are made by way of injunction, stay or attachment debarring him/ them from
transferring, or otherwise disposing of any property/ assets belonging to him without prior
permission of the Court /DRT. The interim order are made by way of injunction, stay or
attachment
(ii) On commencement of execution proceedings, the branch concerned play a major role by
identifying and furnishing particulars of properties for attachment and sale, substituting any
deceased judgment debtor, if any, by his/ her legal heirs, finding market for sale of
properties of the judgment debtors, etc.
3.3.11. Recovery of Bad Loans Secured by Equity
Credit, primarily secured by equity shares/units of mutual funds linked to equities listed on
recognized stock exchanges, and is monitored at regular interval to facilitate decision on disposal
for recovery of dues from defaulting borrowers. In case the capital market is volatile, the review
is done with very short interval. In addition, in the event of default, sale of equity shares of listed
companies pledged as collateral security are considered after invoking the pledge.
3.4. Post Recovery Processes
3.4.1 Appropriation of recovery in NPAs:
(i) In case of NPA accounts, the cash recovery are appropriated as per the order of
appropriation specified in the loan agreement / restructuring package, except in specific
cases, if stipulated otherwise.
84
(ii) Interest realized on NPAs are taken to income account provided the credits in the accounts
towards interest are not out of fresh / additional credit facilities sanctioned to the borrower
concerned.
3.4.2. Up-gradation of loan accounts classified as NPAs:
(i) If arrears of interest and principal are paid by the borrower in the case of loan accounts
classified as NPAs, such accounts are no longer treated as non-performing and are classified
as "standard' account.
(ii) The 'Sub-standard/Doubtful' accounts which have been subjected to restructuring etc.
whether in respect of principal installment or interest amount, by whatever modality,
becomes eligible to be upgraded to the 'Standard' category only after the specified period i.e,
a period of one year after the date when first payment of interest and of principal, whichever
is earlier, falls due under the rescheduled terms, subject to satisfactory performance during
the period.
(iii) Up-gradation of NPAs to Standard Assets shall be done in accordance with the existing
DOP.
3.5. Write-off the Outstanding
If all the efforts for recovery fail, PSBs resort to write-off the outstanding in distress
advances. When chances of recovery are negligible, some banks prefer to write-off an
advance to reduce its income and save tax. In such cases, banks should continue to make
efforts for recovery even after writing-off the advance.
3.6. Conclusion
Maintenance of loan loss to a minimum level is the top priority of the top management of all
PSBs, as with growing competition, interest spreads are becoming thin and a bank can hardly
afford to allow any slippage in quality of the assets and let the stressed assets portfolio drift
below the tolerance level. Moreover, accepting the challenges of the Income Recognition and
Asset Classification (IRAC) norms, PSBs are becoming increasingly sensitive to credit risks and
there is growing awareness to keep NPAs at a low level. Accordingly, they are forced to initiate
a series of measures as stated above to contain growth in NPA as per their Board approved
Recovery Policy.
85
Chapter IV
Dynamicity of NPA over Recent Years in Selected PSBs in India
4.1 Introduction
4.2 Review of Literature
4.3 Data Source
4.4 Methodological Aspects: Statistics and Diagnostics Procedures, Model

Dynamics over Time, Growth Rates
4.5 Results and Discussion
4.6 Aspects on Forecasting
4.7 Conclusion
86
4.1. Introduction
Banks in India, particularly public sector banks (PSBs), in spite of having completed almost two
decades of reforms era, are exposed to bad loans of high proportions, which is a matter of great
concern for the industry in general with its regulators in particular. There are various expressions
(measures) of such bad loans in banking industry, like, gross non- performing asset (GNPA), net
non-performing asset (NNPA), herein after referred to as economic parameters. Though NNPA
is recognized internationally as an indicator of distress assets in the banking industry, for the
purpose of our analysis on progress, and determination of trend over time, GNPA values,
represented in absolute terms, are considered. The rationality behind taking GNPA is that the
true picture of bad advance with respect to a bank is appropriately reflected in GNPA values,
while NNPA values determine the quantum of bad loans reduced by provision and adjustment
factors. Not only GNPA but also gross advances (GA) have been considered for our study as the
former represent a subset of the later. Accordingly, GNPA and GA, each of which, in effect,
have exposed the degree of distress assets in the banking sector over time (based on the data
sets). This has been studied in respect of six selected PSBs along with PSBs-aggregated. The
datasets (expressed in crores of rupees), on which analytical studies have been performed in this
thesis, are presented in tables (table no 4.1 and table no 4.2), relating to parameters, GNPA and
GA, respectively.
A time series is a sequence of data points, measured typically at successive points in time spaced
at uniform time intervals. Time series analysis comprises methods for analyzing time series data
in order to extract meaningful statistics and other characteristics of the data. Time series
forecasting is the use of a model to predict future values based on previously observed values.
Such ‘Time series analysis techniques’ may be divided into parametric, non-parametric and semi
parametric methods. A sincere attempt has been made to examine the movement GNPA and GA
over time and to develop forecasting models (parametric, non parametric and semi parametric)
for medium term for the two parameters, GNPA and GA, in our study. In this chapter, for the
purpose of examining the dynamicity of NPA over time in selected PSBs in India we have used
curve estimation techniques of regression analysis (parametric approach).
87
Table- 4.1.: GNPA Dataset (Rs. in crores)
Year SBI SBT SB UCO CBI PNB PSB
1996 10560.75 429.98 1223.38 1670.54 2289.30 2518.84 41661.00
1997 10968.77 586.60 1218.43 1685.14 2407.38 2424.99 43576.00
1998 11463.12 910.57 1128.64 1614.21 2333.48 2448.06 45653.00
1999 14064.75 885.62 1074.28 1716.18 2436.00 2832.19 51710.00
2000 15246.28 811.08 992.15 1651.41 2843.00 3126.77 53294.00
2001 15874.97 757.93 1074.60 1284.02 3253.00 3460.10 54774.00
2002 15485.85 727.61 1299.13 1332.65 3376.00 4139.86 56507.00
2003 13506.07 635.26 1420.17 1366.49 3244.00 4980.06 54089.00
2004 12667.21 661.76 1589.92 1479.12 3092.00 4670.13 51541.00
2005 12455.73 652.13 1432.78 1399.35 2621.00 3741.34 46597.00
2006 10375.76 609.95 1506.36 1234.74 2684.00 3138.29 41379.00
2007 9998.22 540.11 1559.81 1506.23 2572.00 3390.72 38425.00
2008 12837.34 570.83 1768.65 1651.95 2350.00 3319.30 39606.00
2009 15588.60 544.40 1594.54 1539.51 2316.00 2767.47 44032.00
2010 19534.89 641.98 2006.82 1666.43 2458.00 3214.41 57293.00
2011 25326.29 835.23 2598.97 3150.36 2394.00 4379.39 71080.00
2012 39676.50 1488.80 3182.70 4086.20 7273.00 8719.60 112489.00
Sources: (1) www.rbi.org.in and (2) Prowess Database
Table- 4.2.: GA Dataset (Rs. in crores)

1996 66170.09 3662.55 5834.00 6807.44 9574.65 13441.00 231321.00
1997 68469.22 4048.31 6306.57 5944.05 9629.51 14868.10 244213.00
1998 81068.71 4539.23 7371.91 6714.68 11399.50 16883.20 284971.00
1999 90390.42 4797.51 10021.30 7610.55 13992.00 20058.00 325328.00
2000 106991.40 5620.79 12818.50 8788.77 17095.60 23705.60 380077.00
2001 122776.30 6660.19 13654.40 10965.20 20255.30 29548.30 442134.00
2002 129588.70 7732.31 15558.40 13896.30 22966.00 36378.40 509369.00
2003 144604.60 9524.14 17028.40 16583.60 24839.20 43005.70 577813.00
2004 163447.90 11754.20 21690.60 21343.70 24637.50 49947.90 661976.00
2005 208988.80 15201.20 27713.40 28234.00 27593.70 62774.20 870850.00
2006 287416.90 19180.80 37659.00 37817.00 39185.00 76543.70 1070872.00
2007 342232.00 25005.10 52839.00 47471.00 53471.90 98281.70 1373777.00
2008 422331.00 28136.60 65197.00 55627.00 74287.00 119502.00 1696333.00
2009 548540.00 32710.90 82495.00 69669.00 86740.00 154703.00 2099938.00
2010 641480.00 38461.00 91450.00 83371.00 107118.00 186601.00 2512358.00
2011 771802.00 46044.00 108350.00 100561.00 131407.00 242107.00 3059870.00
2012 893613.00 55346.00 125617.00 117504.00 150725.00 293775.00 3550389.00
Sources: (1) www.rbi.org.in, (2) Prowess Database and (3) Annual Report of selected banks
88
The detailed reviews are included in Chapter-1. However, the references of the base papers in
connection with this chapter are given below:
Ahmed (2010), Berger and Young (1997), Dash and Kabra (2010), Gopalakrishnan (2006),
Gupta and Jain (2010), Ranjan and Dhal (2003), Rawlin and Sharan (2011), Rawlin, Sharan and
Lakshmipathy (2012), Reddy, Babu, Mallikarjuna and Viswanath (2006), Thiagarajan, Ayappan
and Ramachandran (2011), Siraj and Pillai (2012), Veerakumar (2012).
4.3. Data Source

The sources are as under:
(a) RBI Publications in their website (http://www.rbi.org.in,) and
(b) RBI Publications in hard copies (not available in Publications and Database in RBI website)
available in RBI library only. The titles of the items of information are given under (i) to (iv):
(i) Movement of Non Performing Assets (NPAs) of Scheduled Commercial Banks,
(ii) Non Performing Assets as percentage of Total Assets - Scheduled Commercial Banks
(iii) Non Performing Assets as percentage of Advances - Scheduled Commercial Banks
(iv) Bank Group-wise Classification of Loan Assets of Scheduled Commercial Banks
(c) Six selected PSBs websites and Annual Reports.
(d) Indian Bank’s Association website. http://www
(e) CMIE Prowess database,
4.4. Methodological Aspects: Statistics and Diagnostics Procedures, Model Dynamics over
Time, Growth Rates
Diagnostics procedures like Box and Whisker plots, DW tests, Q-Q plots, SW tests are
performed on the data included Table 4.1, and Table 4.2. Descriptive statistics are also calculated
on these dataset. Development of stochastic models to represent dynamicity over time by
parametric approach has also been undertaken. In the following, the different procedures related
to the above methods are explained at first.
89
4.4.1. Box and Whisker Plot
The Box Plot is a graphical representation of a sample data set and shows the lowest value,
highest value, median value, and positions of the first and third quartiles. It is quite useful in
analyzing small and medium data sets. A box plot is based on the Inter-Quartile Range (IQR).
IQR is the distance between the lower and upper quartiles. Thus, IQR = Q U - QL; QU and QL are
upper and lower Quartiles.
There are three elements of a box plot:
i. A rectangle (the box) is drawn with the ends (the hinges) drawn at the lower and upper
quartiles. The median of the data is shown in the box, typically by a line.
ii. The points at distances 1.5 (IQR) from each hinges mark are the inner fences of the data
set. Lines (the whiskers) are drawn from each hinge to the most extreme measurement
within the inner fence.
iii. A second pair of fences, the outer fences, appear at a distance of 3(IQR) from the hinges.
One symbol (usually "*") is used to represent measurements falling between the inner
and outer fences, and another (usually "0") is used to represent measurements that lie
beyond the outer fences. Outer fences are not shown unless one or more measurements
lie beyond them.
The length of the box in a box plot has special significance. The length represented by IQR is a
measure of the sample's variability and is especially useful for comparing two or more samples.
Visually comparing the lengths of the whiskers, if one is clearly longer, the distribution of the
data is probably skewed in the direction of the longer whisker.
A box plot is a good alternative or complement to a histogram and is usually better for showing
several simultaneous comparisons, particularly when the data set is small. Data-sets for 17 years
on GNPA and GA. (from March 1996 to March 2012) have been collected and collated with
respect to six selected PSBs, and PSB Group-wise, the sample sizes in case of two parameters,
however, are small. Box and Whisker Plot is used primarily to identify outliers, if any, in our
data set. BW plot improves the precision level of the model-analysis, particularly with respect to
the forecasting model.
90
For Box Plot, SPSS has a two stage flagging process.
I. Values which are more than three box lengths from either end of the box are denoted by
asterisks (extreme outliers).
II. Values which are between one and a half and three box lengths from either end of the box
are denoted by circle (mild outliers).
4.4.2. DW Test for testing auto correlation in a data series
Durbin–Watson statistic is a test statistic used to detect the presence of auto-correlation (a
relationship between values separated from each other by a given time lag in a given series of
data) in the residuals obtained from a regression analysis. Here, we have used DW (DW) statistic
to examine the mutual independence of the observations in a given series and specially to test for
independence of residuals after fitting a regression model. The formulation of DW test is as
follows. In polynomial models (which are found to have generated best fits) the errors are
uncorrelated (i.e., all order serial correlations, ρs are zero).
DW statistic is used the null hypothesis, H0 (ρs = 0) against the alternative hypothesis, H1 (ρs =
ρs, is the form of ρs in the first order auto-regressive process, et = ρ.et-1 + zt, where zt ~ N (0, σ2)
and is independent of et-1, et-2, et-3, ….., and of zt-1 , zt-2, …; t = 1, 2, ..n.), ρ=ρ1. The form of DW
statistic is: d = ( ei – ei-1)2 / ∑ e 2. The statistic, d is used only for a lower tailed test against
alternatives, ρ > 0. For testing ρ < 0 a lower-tailed test using the statistic 4 – d is used. In DW
Test, the calculated value of the test statistic d is compared to lower and upper critical
(theoretical) values (dL and dW). These critical values are available in standard books on
statistics.
The testing procedures are as follows:
1. One-sided test against alternative hypotheses, ρ > 0: (i) if d < dL, the conclusion is we
have to reject H0; which implies that the residuals are positively correlated, (ii) if d > du.
the conclusion is we have nothing to reject H0; which implies that the residuals are
independent, (iii) if dL < d < du, the test is inconclusive.
2. One-sided test against alternative hypotheses, ρ < 0: (i) if 4 - d < dL, the conclusion is we
have to reject H0; which implies that the residuals are negatively correlated, (ii) if 4 - d >
du. the conclusion is we have nothing to reject H0; which implies that the residuals are
independent, (iii) if dL < d < du, the test is inconclusive.
91
The size of the DW statistics depends upon the number of predictors in the model and the
number of observations.
4.4.3. Q-Q plot

A Q-Q plot ("Q" stands for quantile) is a probability plot, which is a graphical method for
comparing two probability distributions by plotting their quantiles against each other. First, the
set of intervals for the quantiles are chosen. A point (x, y) on the plot corresponds to one of the
quantiles of the second distribution (y-coordinate) plotted against the same quantile of the first
distribution (x-coordinate). Thus the plot obtained in this way represents a Q-Q Plot.
A Q-Q plot is used to compare the shapes of distributions, providing a graphical view of how
properties such as location, scale, and skewness are similar or different in the two distributions.
Q-Q plots can be used to compare collections of data, or theoretical distributions. The use of Q-Q
plots to compare two samples of data can be viewed as a non-parametric approach to comparing
their underlying distributions.
If the quantiles of the two distributions being compared are similar, the points in the Q-Q plot
will approximately lie on the line y = x. If the quantiles of the two distributions differ only in
their location or scale, the points in the Q-Q plot will fall on or near line, y = ax + b. The slope a
and intercept b are visual estimates of the scale and location parameters of the theoretical
distribution. The Q-Q plot can provide an graphical assessment (rather than reducing to a
numerical summary) of "goodness of fit" and hence can be used to compare two theoretical
distributions. A normal Q-Q plot compares a randomly generated, independent data-set plotted
on the vertical axis to a standard normal deviate population plotted on the horizontal axis. The
linearity of the points on the plot suggests that the randomly generated data are normally
distributed.
4.4.4. Shapiro-Wilk (SW) Test

The SW Test is appropriate for testing normality of a given set of observations for small sample
sizes (<50 samples), but can also be used for sample size as large as 2000.
The decision rule is:
I. If the significance-probability value of the SW Test statistic is greater than α (where, α
=0.05), the data can be assumed to be normal.
92
II. If the significance-probability value of the SW Test statistic is below α (where, α =0.05),
the data are assumed to significantly deviate from a normal distribution.
4.4.5. Descriptive Statistics

The primary purpose of statistical methods is to summarise the information contained in any
dataset. For this purpose, certain descriptive measures of statistics, which will describe the nature
of the data set in a general way, are used. The descriptive statistics used in this section to analyse
the datasets on Parameters GNPA and, GA with respect to aggregated PSBs in general with 6
selected PSBs, are mean, range, standard deviation, coefficient of variation, measures of
skewness and measures of kurtosis.
The above plots, and tests are performed on the data included Table 4.1 and Table 4.2. Statistical
measures, which describe the features of the data, are also calculated.
4.4.6. Development of Stochastic Models to represent Dynamicity over time and generating
Forecasts by Parametric Approach
4.4.6.1. Introduction
Modeling and Forecasting are the twin concepts obeying the eternal relationship of the mother
and the offspring in the arena of science, humanities and commerce across disciplines from time
immemorial. Such pursuits enable us to produce image-data which are as close as the
observational data, and the forecasts generated from such models portray the near-future scenario
with high perfection. To write a few words on the genesis of modeling, we observe that the
modeling endeavors put forth by humankind to achieve analytically a near accurate semblance of
the original observations (data) recall the deterministic and stochastic models. Though
deterministic models are used frequently, yet they fail to capture the uncertainties and variation
present in the real-life (social data), yet the stochastic models have the power to contain such
components representing real-life vagaries. As we visit the complexities of the stochastic models
developed till now, we find that there exist parametric, non-parametric and semi-parametric
models in the literature, which represent the dynamicity of the temporal data with considerable
accuracy.
93
In deterministic models two types (dependent (y) and independent (x)) of variables are
considered and to develop a model (an analytical expression or a mathematical equation
involving the two variables) different types of relationships are considered.
In stochastic models an error component is included in the expression of the relationship (viz.,
polynomial, exponential, power, logarithmic, logistic, inverse, harmonic, etc.) and the parameters
involved in the model-equation are estimated by using the method of least squares. A large
spectrum of relationships can be exploited and the model equations offer evaluation of predicted
values and estimation of forecast values. Any relationship (model) in which the independent
variable is restricted to the time variable only, is called a temporal model. Historical data on
variables collected over time at regular (non equi-spaced) intervals are called time series data and
such data are often used for forecasting the future values of the variable.
4.4.6.2. Galore of Parametric Model Equations

The set of variables are represented by the n pairs (yt, t), t being the time variable, and t = 1, 2,
3,…., n. The algebraic expression of the nth degree polynomial model with stochastic element,
et, is given below:
1. Polynomial Model: yt = a0 + a1.t + a2.t2 + a3.t3 + ………+.an-1tn-1 + an.tn + et, the estimates of
the model coefficients are obtained by using LST/MLE/MOM, the abbreviations have standard
connotations. et’s are assumed to be distributed as IID N (0, σ2), 0 and σ2 being the mean and
variance respectively. For n = 1, 2, 3, 4, 5,‘’’’’’’, linear, quadratic, cubic, biquadratic, fifth
degree,…. models are obtained.
2. Compound Model: yt = b0 * (b1t)
(
3. Growth Model: = + )
4. Logarithmic Model: = + ∗ ln (t)
5. Inverse Model: = + / (t)
( /( ))
6. S. Type Model: =
.
7. Exponential Model: = ∗
8. Logistic Model: yt = 1 / (1/u + (b0 * (b1t))) where u is the upper boundary value
94
9. Power Model: yt = b0 * (tb1)
The above models have been fitted to the GNPA and GA data sets mentioned above. In respect
of each data-set, the best model has been identified by employing the R2, Adjusted R2, F value
and its significance level, and lastly, Mean Absolute Error (MAE) criteria, where MAE = .
∑ ^ ^
| − |. are the observed and predicted values respectively.
With regard to forecasting of GNPA and GA variables beyond the years for which data have
been collected, the usual method is employed by using the closely competing rival models
selected from the galore. It may be mentioned at this point that it is also necessary to determine
the magnitude of errors created by the rival models and to compare their magnitudes.
In order to estimate and compare the accuracies of the competing models, the whole data set
(say, with n = 15) is partitioned into two groups. On the first group (with the first 13
observations) highly precise competing models are built up, and on the second set the (2 residual
observations), forecast values are obtained using the built up models. It is to note that for the last
two observations we are having the observed values as well as the corresponding forecast values.
Best Forecast model will be the one for which MAE is the least. The above procedure has been
applied on all available Banks, and on the two variables, GNPA and GA respectively.
4.4.7. Growth Rates
It is interesting to examine the annual growth rates, and, particularly if the best models (selected
using methods described in earlier sub-sections) are not linear. In such cases, the following
formulae may be used to describe annual growth rates.
For yt, t = 1, 2, …, n; annual growth rate = gi = (yi + 1 - yi)/yi, for i = 1, 2, 3, …..
The Average annual growth rate may be one of the following: A.M./G.M./H.M. of annual growth
rates with respect to a Bank.
4.5. Results and Discussion
This section contains the results obtained after having applied the methods discussed in section
4.4., and such are presented in sub-sections 4.5.1 and 4.5.2, for GNPA and GA respectively.
95
4.5.1. Results on GNPA Dataset
The results are presented in the order: Box and Whisker Plot, DW Test, Q-Q Plot, SW Test,
Descriptive Statistics, Model Development
4.5.1.1. Box and Whisker (BW) Plot
Chart- 4.1.: Box Plot of SBI and PSBs in Aggregate on GNPA Dataset (1996-2012)
Chart- 4.2.: Box Plot of 5 selected banks on GNPA Dataset (1996-2012)
The above two Charts present the extreme values, median, two Quartiles, Q1, Q3, and also
outliers, values of the above statistics are presented later (Section 4.5.1.5), however, the outliers
(magnitudes and respective years) are presented in Table 4.3.
96
Table- 4.3.: Values (with year) and ultimate sample size w.r.to the Parameter GNPA
Outlier PSB
SBI SBT SB UCO CBI PNB
Types (aggregate)
(n=15) (n=16) (n=15) (n=15) (n=16) (n=16)
(n=16)
Extreme 2012 2012 2012 2012 2012 2012 2012
outliers (39676.46) (1488.75) (3182.70) (4086.20) (7273.00) (8719.62) (112489.00)
(Rs. in 2011
crores) (3150.36)
Mild outliers 2011 2011
(Rs. in (25326.29) (2598.97)
crores)
It is observed from the Table 4.3, that in respect of the parameter GNPA, each of the banks
(including aggregate PSBs) has an extreme outlier datum corresponding to the year 2012. In
addition, in case of UCO there exists another extreme outlier datum (corresponding to the year
2011). Also that in respect of the parameter GNPA, mild outliers exists in case of each of SBI
and SB corresponding to the year 2011. In what follows, extreme outliers are omitted from the
purview of our study. The ultimate data-sets correspond to available years (with reduced
number) with respect to the different banks considered under the research work, and are
presented in table-4.4.
Table- 4.4.: Ultimate Dataset on Banks with the corresponding years - GNPA
PSB
Bank SBI SBT SB UCO CBI PNB (aggregate)
1996 to 1996 to 1996 to 1996 to 1996 to 1996 to
GNPA 1996 to 2011
2011 2011 2011 2010 2011 2011
4.5.1.2. DW Test
The values of DW statistic (1% level, k’=1) calculated on observed data (normally this test is
done on residuals obtained after fitting models) of six PSBs (and PSBs in aggregate) are given in
the following table (table 4.5).
97
Table- 4.5.: Values of DW Statistic on observed data - GNPA
Bank n DW Stat. dL du Interpretation

(1%) (1%)
SBI 16 0.0243 .844 1.086 Since .0243 is < .844, hence there is evidence that the
observations in the series are positively correlated
SBT 16 0.0269 .844 1.086 Since .0269 is < .844, hence there is evidence that the
SB 16 0.0202 .844 1.086 Since .0202 is < .844, hence there is evidence that the
UCO 15 0.0094 .811 1.070 Since .0094 is < .811, hence there is evidence that the
CBI 16 0.0063 .844 1.086 Since .0063 is < .844, hence there is evidence that the
PNB 16 0.0245 .844 1.086 Since .0245 is < .844, hence there is evidence that the
PSB 16 0.0127 .844 1.086 Since .0127 is < .844, hence there is evidence that the
The table values of t statistic (1% level, one sided) for testing the existence of auto correlation
(of first order) on the data-sets on six PSBs (and PSBs in aggregate) are given in the following
table. Also, values of the first order serial correlations (also the calculated values of t as
computed from serial correlations) with respect to the parameters GNPA, (t Test for Significance
of population correlation-one sided) are given in the last column as an alternative procedure (or,
just as a check).
Table- 4.6.: Values of t-Statistics (Calculated and Tabulated) - GNPA

Bank d.f. Calculated Table value Interpretation First order
value of t of t statistic auto-corr.
(1%-one on original
sided) data
SBI 13 5.7377 2.681 Since Calculated t is > Tabulated t, hence 0.8467
the variables are correlated in the
population
SBT 13 2.5774 2.65 Since Calculated t is < Tabulated t (very 0.5815
marginally less), hence the variables may
not be correlated.
SB 13 6.9458 2.681 Since Calculated t is > Tabulated t, hence 0.8875

population
UCO 12 2.1123 2.681 Since Calculated t is < Tabulated t, hence 0.5206
the variables are not correlated in the
population
98
CBI 13 5.1199 2.65 Since Calculated t is > Tabulated t, hence 0.8176
population
PNB 13 3.8276 2.65 Since Calculated t > Tabulated t, hence the 0.7279
variables are correlated in the population
PSB 13 4.0218 2.65 Since Calculated t is > Tabulated t, hence 0.7446
population
From the above table 4.5, as per DW Test of independence, it is observed that the data-sets for
all selected banks and aggregate PSB in respect of the parameter GNPA are correlated (not
independent). We also observe as per ‘t’ test for testing of independence is given in Table 4.6,
that the data-sets for UCO bank in respect of parameter GNPA, are independent, while that for
SBI, SBT, SB, CBI, PNB and PSB in aggregate are not independent.
The above two analyses performed by employing two methods (rather two methods of
calculation of correlation coefficient) produce almost similar inferences (on existence of
independence of observations) on data-sets on different banks in respect of the parameter,
GNPA.
4.5.1.3. Q-Q plot on GNPA Dataset

Chart- 4.3.: QQ Plot on GNPA – SBI (1996-2011) Chart- 4.4.: QQ Plot on GNPA – SBT (1996-2011)
Chart- 4.5.: QQ Plot on GNPA – SB (1996-2011) Chart- 4.6.: QQ Plot on GNPA – UCO (1996-2010)
99
Chart- 4.7.: QQ Plot on GNPA – CBI (1996-2011) Chart- 4.8.: QQ Plot on GNPA – PNB (1996-2011)
Chart- 4.9.: QQ Plot on GNPA – PSB (1996-2011)
Interpretation on the above seven Q-Q Plots:

For the above six selected PSBs, also PSB in aggregate, (with the exception SBI, SB and CBI, a
pattern is revealed) as the observations on each of the plots fall approximately on a straight line
with a positive slope, in respect of the parameter GNPA, it may be concluded that each of the
data-sets follows approximately a normal distribution (which is also confirmed by the SW test
applied on each of the above mentioned data-sets as evidenced below).
4.5.1.4. SW Test
The test outputs on SW Test are given in the following table:
Table- 4.7.: Values of the SW statistic computed for Parameter GNPA
SW
Bank Value – SW Statistic df Probability
SBI 0.844 16 0.011
SBT 0.963 16 0.724
SB 0.88 16 0.038
UCO 0.904 15 0.111
CBI 0.844 16 0.011
PNB 0.932 16 0.262
PSB (aggregate) 0.921 16 0.175
100
Interpretation of SW test outputs in respect of the parameter GNPA:
From the above table it is evident that the probability value of the SW statistic is greater than
0.05 (level of significance) for each of the banks (and also PSB in aggregate) with the exception
of CBI and hence each of the data-sets on selected PSBs (and also PSB in aggregate), with the
exception of CBI, can be assumed to be normal. Thus, the results obtained by application of the
two methods, Q-Q plot and SW test corroborate each other.
Normally, we may not proceed to build up parametric models (over time) on data sets in which
the conditions relating to (1) the independence of observations and (2) normality to be followed
by observations, are not satisfied, however, as our primary goal is to forecast GNPA values for
future years, we have also taken up the task of model building on such data-sets for which the
observations in the data sets are either not found to be independent or not to follow normality, or
both.
Summarily, the above diagnostic considerations suggest that further works on best model
identification on most of the above data sets (table-4.1 and table-4.2) can be carried out,
however, having taken due regard to our goal we have included data-sets on all banks (and on
aggregate over all banks) towards the task of model building, best model identification and
subsequently forecasting on all data-sets. In the following sub-section the values of different
statistical measures, obtained as SPSS outputs, are considered.
4.5.1.5. Important Statistical Measures
The values of the important statistical measures (mean, range, standard deviation, coefficient of
variation, measures of skewness and kurtosis) in respect of the parameter GNPA are given in
table 4.8. Such measures reveal the different characteristics inherited by a data-set.
Table- 4.8.: Values of different statistical measures for Parameter GNPA
Std. Coefficient
N Range Mean Deviation of Variation Skewness Kurtosis
Std. Std.
Bank S.S Statistic Statistic Statistic Statistic Statistic Error Statistic Error
SBI 16 15328.1 14122.2 3922.96 27.78 1.703 0.56 3.631 1.09
SBT 16 480.59 675.07 135.46 20.07 0.249 0.56 -0.593 1.09
SB 16 1606.82 1468.04 408.75 27.84 1.485 0.56 2.893 1.09
UCO 15 481.44 1519.86 161.67 10.64 -0.472 0.58 -1.262 1.12
CBI 16 1086.7 2666.92 375.76 14.09 0.871 0.56 -0.77 1.09
PNB 16 2555.07 3409.5 786.34 23.06 0.652 0.56 -0.449 1.09
PSB 16 32655 49451.1 8500.89 17.19 0.936 0.56 1.272 1.09
S.S: Sample Size
101
Interpretation on the values obtained for different statistical measures
At the very outset it is noted the important information as revealed from the above table are
presented in a little detail at first. From Table 4.8 it is seen that among the six selected PSBs, the
mean value with respect to parameter GNPA for SBI stands at Rs. 14122.16 crore (highest
value), while that for SBT stands at Rs.675.07 crore (lowest value). The standard deviation in
respect of the parameter GNPA for SBI is also the highest (Rs.3922.93 crore) while that for SBT
the standard deviation is also the lowest (Rs.135.46 crore). The mean value of bad loans for SB
and UCO are close to one another (Rs.1468.04 crore and Rs.1519.86 crore respectively), while
those for CBI, PNB and PSBs in aggregate are Rs.2666.92 crore, Rs 3409.50 crore and Rs.
49451.06 crore respectively. The standard deviation values for SB, UCO, CBI, PNB and PSBs in
aggregate are Rs. 408.75 crore, Rs 161.67 crore, Rs 375.76 crore, Rs 786.34 crore and Rs
8500.89 crore respectively, while in case of the range, for SBI it is highest (Rs 15328.07 crore)
and the same is the lowest for SBT (Rs 480.59 crore), with second lowest being for UCO (Rs.
481.44 crore). Ranges of bad loans for other banks vary in between these two extremes (lie
between Rs 1086.70 crore and Rs 2555.07 crore). The coefficient of variation (CV) in respect of
the parameter GNPA is the lowest (10.64%) for UCO (rank-1) while it is highest (27.78%) for
SBI (rank-6). As regards the remaining four banks, CV lies in between the above two extremes
and are 14.09% (CBI), 20.07% (SBT), 23.06% (PNB) and 27.84% (SB) respectively. The above
findings imply that there exist large amounts of variability with respect to bad debt (GNPA)
among the six selected banks and PSBs in aggregate during the period under consideration. Such
wide variations in mean, standard deviation and CV (%) justify our selection of sample banks in
terms of scales of operation and quantum of distress assets. From Table 4.8, it may be observed
that skewness coefficient of SBT is the smallest (0.249), which implies that over a period of 16
years, GNPA data for SBT are least asymmetric among the 6 selected banks. The values of the
skewness coefficient in respect of UCO, PNB, CBI, SB and SBI are - 0.472 (left skewed), and
for the rest banks the data-sets follow right skewedness, and the values are 0.652, 0.871, 1.485
and 1.703 respectively. Data on Aggregated PSBs are also skewed (0.936). Summarily, data on
all banks (excepting UCO) are right skewed - most values are concentrated on the left side of the
mean, while for UCO the data are left skewed - most values are concentrated on the right side of
the mean. From Table 4.8, it is seen that the values of the kurtosis coefficient are both positive
102
and negative (imply that the data are both leptokurtic, sharply peaked, and platykurtic, flatly
peaked). No one of the data-sets is mesokurtic. The value of the kurtosis coefficient is 3.631 (for
SBI) and 2.893 (for SB) hence the data are highly leptokurtic, while the values of the kurtosis
coefficients are negative for PNB (-0.449), SBT (-0.593), CBI (-0.770), and UCO (-1.262)
indicating that the status (flatness) of the respective data-sets vary between moderate to high
degrees of flatness.
4.5.1.6. Model Development on Seven Dataset by Parametric Approach
Development of parametric models in respect of the seven data-sets has been made with the help
of Regression Analysis using least squares estimation technique. Summary outputs (exhaustive
alternative model equations, values of different precision criteria, and graphs of best models for
seven data-sets) using SPSS package have been generated for selected PSBs.
Alternative Model Equations (11 equations) developed as stated above for six selected banks and
PSB in aggregate are given in the following tables.
Table- 4.9.: Alternative Model Equations on GNPA - SBI
Model Name Model expressions

Linear Y = 10459.481 + (430.904 * t)
Quadratic Y = 14135.342 + (-794.383 * t) + (72.076 * t2)
Compound Y = 11020.420 * (1.026t)
Growth Y = e (9.308 + (0.026 * t))
Logarithmic Y = 9778.371 + (2265.942 * ln(t))
Cubic Y = 3685.523 + (5629.361 * t) + (-844.575 * t2) + (35.947 * t3)
S-curve Y = e (9.613 + (-0.420 / t ))
Exponential Y = 11020.420 * (e (0.026 * t))
Inverse Y = 15421.386 + (-6148.845 / t)
Power Y = 10404.847 * (t0.143)
Logistic Y = 1 / ( 1/25330 + 0.0001 * 0.802t )
103
Table- 4.10.: Alternative Model Equations on GNPA - SBT
Linear Y = 711.301 + (-4.263 * t)
Quadratic Y = 671.921 + (8.864 * t) + (-0.772 * t2)
Compound Y = 684.981 * 0.996t
Growth Y = e( 6.529 + (-0.004 * t ))
Cubic Y = 231.097 + (279.849 * t) + (-39.441 * t2) + (1.516 * t3)
S-curve Y = e ( 6.564 + (-0.324 / t ))
Exponential Y = 684.981 * (e ( -0.004 * t ))
Inverse Y = 710.309 + (-166.798 / t)
Power Y = 628.753 * (t0.027 )
Logistic Y = 1 / ( 1/911 + (0.0001 * (1.131t )))
Table- 4.11.: Alternative Model Equations on GNPA - SB

Linear Y = 856.036 + (72 * t)
Quadratic Y = 1270.125 + (-66.029 * t) + (8.119 * t2)
Compound Y = 961.172 * (1.047t)
Growth Y = e ( 6.868 + (0.046 * t ))
Cubic Y = 1149.243 + (8.279 * t) + (-2.484 * t2) + (.416 * t3)
S-curve Y = e ( 7.356 + (-0.454 / t ))
Exponential Y = 961.172 * (e (0.046 * t ))
Inverse Y = 1616.671 + (-703.429 / t)
Power Y = 927.023 * (t0.223)
Logistic Y = 1 / ( 1/2600 + (0.002 * (0.784t )))
Table- 4.12.: Alternative Model Equations on GNPA - UCO

Linear Y = 1589.194 + (-8.666 * t)
Quadratic Y = 1883.370 + (-112.493 * t) + (6.489 * t2)
Compound Y = 1580.532 * (0.994t)
Growth Y = e ( 7.366 + (-0.006 * t ))
104
Logarithmic Y = 1672.123 + (-81.862 * ln(t))
Cubic Y = 1758.241 + (-31.425 * t) + (-5.778 * t2 ) + (0.511 * t3)
S-curve Y = e ( 7.281 + (0.182 / t ))
Exponential Y = 1580.532 * e ( (-0.006 * t ))
Inverse Y = 1459.066 + (274.841 / t)
Power Y = 1670.381 * (t-0.054)
Logistic Y = 1 / ( 1/1717 + (0.00001 * (1.142t )))
Table- 4.13.: Alternative Model Equations on GNPA - CBI

Linear Y = 2740.445 + (-8.650 * t )
Quadratic Y = 2011.604 + (234.297 * t) + (-14.291 * t2)
Compound Y = 2708.065 * (0.997t)
Growth Y = e ( 7.904 + (-0.003 * t ))
Logarithmic Y = 2539.100 + (66.677* ln(t))
Cubic Y = 1574.278 + (503.131 * t) + (-52.653 * t2) + (1.504 * t 3)
S-curve Y = e ( 7.918 + (-0.180 / t ))
Exponential Y = 2708.065 * (e ( -0.003 * t ))
Inverse Y = 2767.455 + (-475.811 / t)
Power Y = 2513.232 * (t0.026)
Logistic Y = 1 / ( 1/3377 + (0.00003 * (1.049t )))
Table- 4.14.: Alternative Model Equations on GNPA - PNB

Linear Y = 2849.479 + (65.884 * t)
Quadratic Y = 1872.904 + (391.409 * t) + (-19.149 * t2)
Compound Y = 2778.605 * (1.021t)
Growth Y = e ( 7.930 + (0.021 * t ))
Cubic Y = 1065.842 + (887.529 * t) + (-89.943 * t2) + (2.776 * t 3)
S-curve Y = e ( 8.221 + (-0.526 / t ))
Exponential Y = 2778.605 * (e ( 0.0213 * t ))
Inverse Y = 3762.453 + (-1670.449 / t)
Power Y = 2426.981 * (t0.165)
Logistic Y = 1 / ( 1/4981 + (0.0001 * 0.947t )))
105
Table- 4.15.: Alternative Model Equations on GNPA - PSB
Linear Y = 45832.5 + (425.713 * t)
Quadratic Y = 48824.438 + (-571.599 * t) + (58.665 * t2 )
Compound Y = (46164.024 * 1.007t )
Growth Y = e (10.740 +( 0.007 * t))
Cubic Y = 21846.269 + (16012.503 * t) + (-2307.841 * t2) + (92.804 * t3)
S-curve Y = e (( 10.838 + (-0.2012 / t ))
Exponential Y = 46164.024 * (e ( 0.007 * t ))
Inverse Y = 51675.856 + (-10529.298 / t)
Power Y = 44351.065 * t0.050
Logistic Y = 1 / ( 1/71081 + (0.00002 * (0.798t )))
Values of precision criteria for Alternative Models

Values of five (R, R2, Adjusted R2, α, and MAE) different precision criteria have been listed for
11 alternative models for each Bank; Pr. (F > F0) = α. Competing models are those for which the
values of R2, or Adjusted R2 are very close.
Table- 4.16.: Values of precision criteria for Alternative Models on GNPA - SBI
Model R R2 Adjusted R2 F Sig. MAE

Linear .523 .273 .222 5.270 .038 2492.281
Quadratic .634 .402 .310 4.370 .035 2523.287
Compound .493 .243 .189 4.498 .052 2420.318
Growth .493 .243 .189 4.498 .052 2420.318
Logarithmic .456 .208 .151 3.672 .076 2496.109
Cubic * .954 .910 .887 40.307 .000 899.368
S .409 .168 .108 2.817 .115 2497.508
Exponential .493 .243 .189 4.498 .052 2420.318
Inverse .377 .142 .081 2.326 .149 2545.47
Power .458 .210 .154 3.722 .074 2435.694
Logistic .475 .225 .170 4.072 .063 4495.578
* Represents best model.
106
Table- 4.17.: Values of precision criteria for Alternative Models on GNPA - SBT
Linear .150 .022 -.047 .322 .580 104.37735
Quadratic .187 .035 -.114 .235 .794 105.73243
Compound .093 .009 -.062 .122 .732 103.34912
Growth .093 .009 -.062 .122 .732 103.34912
Logarithmic .028 .001 -.071 .011 .918 110.24551
Cubic * .890 .793 .741 15.284 .000 48.35730
S .383 .147 .086 2.409 .143 105.07246
Exponential .093 .009 -.062 .122 .732 103.34912
Inverse .297 .088 .023 1.350 .265 107.36240
Power .105 .011 -.060 .156 .699 108.60331
Logistic .314 .098 .034 1.529 .237 117.68190
* Represents best model
Table- 4.18.: Values of precision criteria for Alternative Models on GNPA - SB

Linear .839 .703 .682 33.187 .000 164.7407
Quadratic * .924 .854 .831 37.887 .000 110.6492
Compound .866 .751 .733 42.139 .000 144.8594
Growth .866 .751 .733 42.139 .000 144.8594
Logarithmic .665 .442 .402 11.070 .005 224.0924
Cubic * .927 .860 .825 24.535 .000 115.9689
S .432 .187 .128 3.210 .095 247.6349
Exponential .866 .751 .733 42.139 .000 144.8594
Inverse .414 .172 .113 2.904 .110 257.602
Power .696 .484 .447 13.122 .003 201.3732
Logistic .586 .344 .297 7.335 .017 350.9934
* Represents competing models.
Table- 4.19.: Values of precision criteria for Alternative Models on GNPA - UCO
Linear .240 .057 -.015 .793 .389 132.90926
Quadratic .730 .532 .454 6.825 .010 83.14038
Compound .228 .052 -.021 .712 .414 134.28665
Growth .228 .052 -.021 .712 .414 134.28665
Logarithmic .396 .157 .092 2.417 .144 119.61302
Cubic * .757 .573 .457 4.922 .021 84.51743
S .408 .166 .102 2.595 .131 124.58851
Exponential .228 .052 -.021 .712 .414 134.28665
Inverse .418 .175 .111 2.752 .121 123.17639
107
Power .384 .147 .082 2.244 .158 120.12994
Logistic .349 .122 .054 1.802 .202 136.51686
Table- 4.20.: Values of precision criteria for Alternative Models on GNPA - CBI
Linear .110 .012 -.059 .170 .686 305.92585
Quadratic .750 .563 .496 8.368 .005 211.43323
Compound .100 .010 -.061 .141 .713 300.99481
Growth .100 .010 -.061 .141 .713 300.99481
Logarithmic .140 .020 -.050 .280 .605 308.85225
Cubic * .812 .660 .575 7.756 .004 182.28005
S .322 .103 .039 1.614 .225 284.01853
Exponential .100 .010 -.061 .141 .713 300.99481
Inverse .305 .093 .028 1.435 .251 290.68799
Power .154 .024 -.046 .340 .569 301.44410
Logistic .123 .015 -.055 .214 .651 392.49653
Table- 4.21.: Values of precision criteria for Alternative Models on GNPA - PNB

Linear .399 .159 .099 2.649 .126 573.06107
Quadratic .620 .385 .290 4.068 .042 488.99601
Compound .450 .203 .146 3.564 .080 563.77425
Growth .450 .203 .146 3.564 .080 563.77425
Logarithmic .522 .272 .220 5.235 .038 524.31242
Cubic .678 .460 .325 3.412 .053 472.64561
S .564 .318 .269 6.530 .023 530.92207
Exponential .450 .203 .146 3.564 .080 563.77425
Inverse .512 .262 .209 4.964 .043 553.73042
Power .579 .335 .288 7.063 .019 516.09971
Logistic .125 .016 -.055 .222 .644 761.34354
Table- 4.22.: Values of precision criteria for Alternative Models on GNPA - PSB

Linear .238 .057 -.011 .844 .374 6728.50919
Quadratic .274 .075 -.067 .527 .603 6948.50460
Compound .189 .036 -.033 .517 .484 6758.77934
Growth .189 .036 -.033 .517 .484 6758.77934
Logarithmic .267 .071 .005 1.077 .317 6419.29564
108
Cubic * .892 .796 .745 15.571 .000 3333.06914
S .293 .086 .021 1.316 .270 6185.75276
Exponential .189 .036 -.033 .517 .484 6758.77934
Inverse .298 .089 .024 1.367 .262 6154.87015
Power .239 .057 -.010 .845 .374 6503.90290
Logistic .409 .167 .107 2.806 .116 11406.12295
From the tables 4.16 to 4.22 (7 tables) the best models are selected as, Cubic, Cubic, Cubic
(though not a good fit), Cubic (though not a good fit), and Cubic with respect to SBI, SBT, UCO,
CBI, PSB (in aggregate) respectively based on the Adjusted R2 and MAE criteria. With regard to
SB there are two competing models (Cubic and Quadratic) of which Quadratic is the best model
based on the Adjusted R2 and MAE criteria. However, for PNB (the eleven models), the fits are
very poor. For PNB higher degree polynomials (of degrees higher than 3) are tried in order to
produce a closer fits (i.e., with higher values of Adjusted R2, lower value of MAE), and the
results of fit are given in table 4.23.
Table- 4.23.: Values of precision criteria for higher degree polynomials on GNPA - PNB
Adjusted R
Model R R Square F Sig. MAE
Square
Cubic 0.678 0.46 0.325 3.412 0.053 472.6456
Polynomial of
.710 .504 .380 4.059 .033 446.8576
degree 4
Polynomial of
.919 .845 .788 14.969 .000 204.8823
degree 5 *
* Represents competing model.
From the above 3 models, the model involving the polynomial of degree 5 has been selected
(based on the criteria, Adjusted R2, MAE).
109
The graphical presentations of the best/best competing models are given below.
Chart- 4.10.: Best Model on GNPA – SBI Chart- 4.11.: Best Model on GNPA – SBT
Chart- 4.12.: Best Competing Models on GNPA – SB Chart- 4.13.: Best Model on GNPA – UCO
Chart- 4.14.: Best Model on GNPA – CBI Chart- 4.15.: Best Model on GNPA – PNB
110
Chart- 4.16.: Best Model on GNPA – PSB
Diagnostics Checks on Best Model on Errors

Table- 4.24.: DW Values Calculated on Errors (using best model) – GNPA
Bank Name n DW dL du Interpretation
(Model Stat. (1%) (1%)
Name) (d)
SBI (Cubic) 16 1.133 0.844 1.086 Since 1.133 is > 1.086, hence there is evidence that the
observations in the series are not positively correlated, i.e.,
the values in the series are independent.
SBT 16 1.463 0.844 1.086 Since 1.463 is > 1.086, hence there is evidence that the
(Cubic) observations in the series are not positively correlated, i.e.,
SB 16 1.279 0.844 1.086 Since 1.279 is > 1.086, hence there is evidence that the
(Quadratic) observations in the series are not positively correlated, i.e.,
UCO 15 1.927 0.811 1.07 Since 1.927 is > 1.070, hence there is evidence that the
CBI (Cubic) 16 0.905 0.844 1.086 Since .844 is < .905 < 1.086, hence the test is inconclusive
PNB (5th 16 1.434 0.844 1.086 Since 1.434 is > 1.086, hence there is evidence that the
degree poly) observations in the series are not positively correlated, i.e.,
PSB 16 0.771 0.844 1.086 Since .771 is < .844, hence there is evidence that the
(Cubic) observations in the series are positively correlated
111
Table- 4.25.: ‘t’ test Values Calculated on Errors (using best model) – GNPA
Bank df Calculated Table Interpretation First order
(Model value of t value of t auto-corr.
Name) (1% - one on residuals
sided)
SBI (Cubic) 13 1.4177 2.65 Since Calculated t is < Tabulated t, hence 0.37
population
SBT (Cubic) 13 0.9549 2.65 Since Calculated t is < Tabulated t, hence 0.26
population
SB 13 1.0189 2.65 Since Calculated t is < Tabulated t, hence 0.27
(Quadratic) the variables are not correlated in the
population
UCO 12 0.0719 2.681 Since Calculated t is < Tabulated t, hence 0.02
(Cubic) the variables are not correlated in the
population
CBI (Cubic) 13 2.2078 2.65 Since Calculated t is < Tabulated t, hence 0.52
population
PNB (5th 13 1.056 2.65 Since Calculated t is < Tabulated t, hence 0.28
degree poly) the variables are not correlated in the
population
PSB (Cubic) 13 2.4592 2.65 Since Calculated t is < Tabulated t, hence 0.56
population
all selected banks ( with the exception of CBI and PSB in aggregate) in respect of the parameter
GNPA are not correlated (independent). We also observe as per ‘t’ test for testing of
independence is given in table 4.25, that the data-sets for all selected banks including PSB in
aggregate in respect of parameter GNPA, are independent. The above two analyses performed
by employing two methods (rather two methods of calculation of correlation coefficient) produce
almost similar inferences (on existence of independence of observations) on data-sets on
different banks in respect of the parameter, GNPA.
112
Table- 4.26.: Test of Normality on Errors (using best model) – Shapiro-Wilk Test – GNPA
Bank (Model Name) SW
Value of Statistic df Sig.
SBI (Cubic) 0.979 16 0.96
SBT (Cubic) 0.936 16 0.3
SB (Quadratic) 0.955 16 0.57
UCO (Cubic) 0.957 15 0.64
CBI (Cubic) 0.972 16 0.87
PNB (5th Degree Polynomial) 0.943 16 0.38
PSB in aggregate (Cubic) 0.936 16 0.3
From table-4.26, it is evident that the significance-probability value of the SW Test statistic is
greater than 0.05 in respect of the parameter GNPA for all banks studied and hence each of the
data-sets on selected PSBs (and also PSB in aggregate) can be assumed to be normal.
4.5.1.7. Important Findings Collated (As Emerged from Analysis)

1. The main objective of the research work undertaken is to obtain a quantitative assessment
of the movement of GNPA values over years with respect to six PSBs and PSBS
(aggregated), and to obtain the pattern of dynamicity different parametric models are
fitted.
2. The diagnostic methods applied on the GNPA values do suspect non-randomness and
existence of serial correlation in many data sets (as revealed from Tables given above).
3. 11 Parametric models have been tried and graph-plots of best (and competitive) models
(selected on the basis of standard Best Fit criteria) have been displayed in different
Charts, truly, the best identified models possess reasonable precision embedded in those
(details are delineated in the relevant parts) .
4. Residuals from best models are found to be random and they follow normality, thus
model post-requisites in regard to both randomness and normality have been satisfied,
and hence suggest future use of such models.
5. The underlying system of data generation possesses elements of stochasticity.
6. Such comprehensive studies are not available in the existing literature so far (as per the
review included in Chapter-1).
113
4.5.2. Results on GA Dataset
The results are presented in the order: Box and Whisker Plot, DW Test, Q-Q Plot, SW Test,
Descriptive Statistics, Model Development.
4.5.2.1. Box and Whisker Plot
Chart- 4.17.: Box Plot of SBI and PSB in Aggregate on GA Dataset (1996-2012)
Chart- 4.18.: Box Plot of five selected banks on GA Dataset (1996-2012)
The above two Charts present the extreme values, median, two Quartiles, Q1, Q3, and also
Outliers; values of the above statistics are presented later (Section 4.5.2.5), however, the Outliers
(magnitudes and respective years) are presented in table 4.27.
Table- 4.27.: Values (with year) and ultimate sample size w.r.to the Parameter GA
Outlier Type SBI SBT SB UCO CBI PNB PSB in
(n=17) (n=17) (n=17) (n=17) (n=17) (n=16) aggregate
(n=17)
Mild outliers 2012
(value in Rs. (293775)
Crores)
114
It is observed from the table 4.27, that in respect of the parameter GA, PNB has one mild outlier
corresponding to the year 2012. As extreme outliers are omitted from further analytical works
included in the study, the ultimate data-sets (on GA) correspond to available years (Table-4.28).
Table- 4.28.: Ultimate Dataset on Banks with the corresponding years - GA
Bank SBI SBT SB UCO CBI PNB PSB in aggregate
GA 1996 to 1996 to 1996 to 1996 to 1996 to 1996 to 1996 to 2012
2012 2012 2012 2012 2012 2012
4.5.2.2. DW Test
The values of DW statistic (1% level, k’=1) calculated on observed data (normally this test is
done on residuals obtained after fitting models) of six PSBs (and PSBs in aggregate) are given in
the following table (table 4.29).
Table- 4.29.: DW Statistics on observed data - GA
Bank N DW Stat. dL (1%) du (1%) Interpretation
SBI 17 0.0285 0.873 1.102 Since .0285 is < .873, hence there is evidence that the
SBT 17 0.0276 0.873 1.102 Since .0276 is < .873, hence there is evidence that the
SB 17 0.0285 0.873 1.102 Since .0285 is < .873, hence there is evidence that the
UCO 17 0.0294 0.873 1.102 Since .0294 is < .873, hence there is evidence that the
CBI 17 0.0397 0.873 1.102 Since .0397 is < .873, hence there is evidence that the
PNB 17 0.0386 0.873 1.102 Since .0386 is < .873, hence there is evidence that the
PSB 17 0.0287 0.873 1.102 Since .0287 is < .873, hence there is evidence that the
The table values of t statistic (1% level, one sided) for testing the existence of auto correlation
(of first order) on the data-sets (on GA) on six PSBs (and PSBs in aggregate) are given in the
table 4.30. Also, values of the first order serial correlations (also the calculated values of t as
computed from serial correlations) with respect to the parameters GA, (t Test for Significance of
115
population correlation-one sided) are given in the last column as an alternative procedure (or,
just as a check).
Table- 4.30.: Values of t-Statistics (Calculated and Tabulated) - GA
Bank df Calculated Table value Interpretation First order
value of t of t (1% - auto-corr. on
one sided) original
values
SBI 14 52.2772 2.624 Since Calculated t is > Tabulated t, hence the 0.9974
SBT 14 68.2895 2.624 Since Calculated t is > Tabulated t, hence the 0.9985
SB 14 43.7763 2.624 Since Calculated t is > Tabulated t, hence the 0.9964
UCO 14 72.8314 2.624 Since Calculated t is > Tabulated t, hence the 0.9987
CBI 14 38.2897 2.624 Since Calculated t is > Tabulated t, hence the 0.9953
PNB 14 76.3493 2.65 Since Calculated t is > Tabulated t, hence the 0.9988
PSB 14 75.1625 2.624 Since Calculated t is > Tabulated t, hence the 0.9988
all selected banks and aggregate PSB in respect of the parameter GA are correlated (not
independent). Similarly, we observe that in accordance with the ‘t’ test for testing of
independence given in table 4.27, the data-sets for all selected banks and aggregate PSBs are not
independent.
Thus the above two analyses performed by employing two methods (rather two methods of
calculation of correlation coefficient) produce similar inferences (on existence of independence
of observations) on data-sets on different banks in respect of the parameter, GA.
116
4.5.2.3. Q-Q plot on GA Dataset
Chart- 4.19.: Q-Q Plot on GA - SBI (1996-2012) Chart- 4.20.: Q-Q Plot on GA - SBT (1996-2012)
Chart- 4.21.: Q-Q Plot on GA - SB (1996-2012) Chart- 4.22.: Q-Q Plot on GA - UCO (1996-2012)
Chart- 4.23.: Q-Q Plot on GA - CBI (1996-2012) Chart- 4.24.: Q-Q Plot on GA - PNB (1996-2012)
Chart- 4.25.: Q-Q Plot on GA - PSB (1996-2012)
117
Interpretation on the above seven Q-Q Plots:
For the above six selected PSBs, also PSBs in aggregate, as the observations on each of the plots
does not fall approximately on a straight line with a positive slope in respect of the parameter
GA, it may be concluded that each of the data-sets does not follow approximately a normal
distribution (which is also confirmed by the SW test applied on each of the above mentioned
data-sets as evidenced below).
4.5.2.4. SW Test
The test outputs (SPSS package) on SW test are given in the following table:
Table- 4.31.: Values of the SW test statistic computed for Parameter GA
Bank SW
Statistic df Sig.
SBI 0.825 17 0.005
SBT 0.855 17 0.013
SB 0.833 17 0.006
UCO 0.837 17 0.007
CBI 0.809 17 0.003
PNB 0.822 17 0.004
PSB 0.832 17 0.006
From the above table it is evident that the probability value of the SW Test statistic is less than
0.05 (level of significance) for each of the banks (and also PSB in aggregate) in respect of the
parameter GA and hence each of the data-sets on selected PSBs (and also PSB in aggregate), can
be assumed to be non-normal. As mentioned above the results obtained by application of the two
methods, Q-Q plot and SW test corroborate each other.
Normally, we may not proceed to build up parametric models (over time) on data sets in which
the conditions relating to (1) the independence of observations or (2) normality to be followed by
observations, or both are not satisfied, however, if prediction and forecast accuracies of the
models built up on the data sets are high, the task of model building can be taken up.
In the following sub-section the values of different statistical measures, obtained as SPSS
outputs, are considered.
118
4.5.2.5. Important Statistical Measures
The values of the important statistical measures (mean, range, standard deviation, coefficient of
variation, measures of skewness and kurtosis) in respect of the parameter GA are given in table
4.32. Such measures reveal the different characteristics inherited by a dataset.
Table- 4.32.: Values of the different statistical measures for Parameter GA
Coefficient
Std. of
N Range Mean Deviation Variation Skewness Kurtosis
Std. Std.
S.S Statistic Statistic Statistic Statistic Statistic Error Statistic Error
SBI 17 827443 299407 264479 88.33 1.162 0.55 0.184 1.063
SBT 17 51683.5 18730.9 16244.3 86.72 1.029 0.55 0.011 1.063
SB 17 119783 41270.9 39165.3 94.9 1.027 0.55 -0.252 1.063
UCO 17 111560 37582.8 35875.8 95.46 1.087 0.55 0.053 1.063
CBI 17 141150 48524.5 45134.5 93.01 1.228 0.55 0.325 1.063
PNB 17 280334 87183.8 85288.8 97.83 1.336 0.55 0.919 1.063
PSB 17 3319068 1170093 1053457 90.03 1.161 0.55 0.234 1.063
S.S: Sample Size
Interpretation on the values obtained for different statistical measures

At the very outset it is noted the important information as revealed from the above table are
presented in a little detail at first. From table 4.29 it is seen that among the six selected PSBs, the
mean value with respect to parameter GA for SBI stands at Rs. 299406.53 crore (highest value),
while that for SBT stands at Rs.18730.87 crore (lowest value). The standard deviation in respect
of the parameter GA for SBI is also the highest (Rs.264478.60 crore) while that for SBT the
standard deviation is also the lowest (Rs.16244.32 crore). The mean value of aggregate advance
for SB and UCO are close to one another (Rs.41270.85 crore and Rs.37582.84 crore
respectively), while those for CBI, PNB and PSBs in aggregate are Rs.48524.51 crore, Rs
87183.75 crore and Rs. 1170093.41 crore respectively. The standard deviation values for SB,
UCO, CBI, PNB and PSBs in aggregate are Rs. 39165.32 crore, Rs 35875.82 crore, Rs 45134.52
crore, Rs 85288.76 crore and Rs 1053457.43 crore respectively, while in case of the range, for
SBI it is highest (Rs 827442.91 crore) and the same is the lowest for SBT (Rs 51683.45 crore),
with second lowest being for UCO (Rs. 111559.95 crore). Ranges of aggregate advance for other
banks vary in between these two extremes (lie between Rs 827442.91 crore and Rs 51683.45
crore). The coefficient of variation (CV) in respect of the parameter GA is the lowest (86.72%)
119
for SBT (rank-1) while it is highest (97.83%) for PNB (rank-6). As regards the remaining four
banks, CV lies in between the above two extremes and are 93.01% (CBI), 88.33% (SBI), 94.90%
(SB) and 95.46% (UCO) respectively. The above findings imply that there exist large amounts of
variability with respect to gross advance (GA) among the six selected banks and Public Sector
Banks in aggregate during the period under consideration. Such wide variations in mean,
standard deviation and CV (%) justify our selection of sample banks in terms of scales of
operation and quantum of gross advance. From Table 3.14, it may be observed that skewness
coefficient of SB is the smallest (1.027), which implies that over a period of 17 years, GA data
for SB are least asymmetric among the 6 selected banks. The values of the skewness coefficient
in respect of SBT, UCO, SBI, CBI and PNB are 1.029, 1.087, 1.162, 1.228 and 1.336
respectively. Data on Aggregated PSBs are also skewed (1.161). Thus for data on all banks are
right skewed (most values are concentrated on the left side of the mean). From Table 4.14, it is
seen that the values of the kurtosis coefficient are both positive and negative (imply that the data
are both leptokurtic, sharply peaked, and platykurtic, flatly peaked). No one of the data-sets is
mesokurtic. The values of the kurtosis coefficient are 0. 011 (for SBT), 0.53 (for UCO), 0.184
(for SBI), 0.325 (for CBI), 0.919 (for PNB) and hence the data are leptokurtic, while the value of
the kurtosis coefficients are negative for SB (-0.252) indicating that the status (flatness) of the
respective data are moderate.
4.5.2.6. Model Development on Seven Dataset on GA by Parametric Approach

Development of parametric models in respect of the seven data-sets on GA has been made with
the help of Regression Analysis using least squares estimation technique. Summary outputs
(exhaustive alternative model equations, values of different precision criteria, and the graphs of
best models for seven data-sets} using SPSS package have been generated for each of the seven
Banks, and are given in tables 4.33 to 4.39, charts 4.26. to 4.32.
Table- 4.33.: Alternative Model Equations on GA - SBI
Model Name Model expression

Linear Y = -130649.084 + (47783.957 * t)
Quadratic Y = 142162.735 + (-38367.144 * t) + (4786.172 * t2)
Compound Y = 44923.869 * (1.186t)
Growth Y = e (10.713 + (0.171 * t))
Logarithmic Y = -185924.729 + (246250.21 * ln(t))
120
Cubic Y = 62937.841 + (7963.204 * t) + (-1468.425 * t2) + (231.652 * t3)
S-curve Y = e (12.742 + (-2.428/t))
Exponential Y = 44923.869 * (e (0.171 * t))
Inverse Y = 411108.957 + (-552089.638 / t))
Power Y = 30841.044 * (t0.971)
Logistic Y = 1 / (1/893614 + (0.0001 * (0.602t)))
Table- 4.34.: Alternative Model Equations on GA - SBT

Linear Y = -8273.699 + (3000.508 * t)
Quadratic Y = 6668.049 + (-1717.939 * t) + (262.136 * t2)
Compound Y = 2515.010 * (1.198t)
Growth Y = e (7.830 + (0.181 * t))
Cubic Y = 4032.245 + (-176.533 * t) + (54.046 * t2) + (7.707 * t 3)
S-curve Y = e (9.990 + (-2.623/t))
Exponential Y = 2515.010 * (e (0.181 * t))
Inverse Y = 25909.654 + (-35481.155 / t)
Power Y = 1644.462 * (t1.042)
Logistic Y = 1 / (1/55347 + (0.002 * (0.631t)))
Table- 4.35.: Alternative Model Equations on GA - SB

Linear Y = -23170.899 + (7160.194 * t)
Quadratic Y = 14824.139 + (-4838.239 * t) + (666.580 * t2)
Compound Y = 4144.846 * (1.225t)
Growth Y = e (8.330 + (0.203 * t))
Cubic Y = 8625.290 + (-1213.181 * t) + (177.197 * t2) + (18.125 * t3)
S-curve Y = e (10775 + (-3.044/t))
Exponential Y = 4144.846 * (e (0.203 * t))
Inverse Y = 58202.773 + (-83686.096 / t)
Power Y = 2501.585 * (t1.184)
Logistic Y = 1 / (1/125618 + (0.001 * (0.602t)))
121
Table- 4.36.: Alternative Model Equations on GA - UCO
Linear Y = -21413.125 + (6555.107 * t)
Quadratic Y = 14058.960 + (-4646.604 * t) + (622.317 * t2)
Compound Y = 3744.206 * (1.226t)
Growth Y = e (8.228 + (0.204 * t))
Cubic Y = 7615.702 + (-878.617 * t) + (113.639 * t2) + (18.840 * t3)
S-curve Y = e (10.635 + (-2.837/t))
Exponential Y = 3744.206 * (e (0.204 * t))
Inverse Y = 52805.685 + (-75238.973 / t)
Power Y = 2382.165 * (t1.159)
Logistic Y = 1 / (1/117505 + (0.001 * (0.606t)))
Table- 4.37.: Alternative Model Equations on GA - CBI

Linear Y = -23901.313 + (8047.314 * t)
Quadratic Y = 24507.493 + (-7239.677 * t) + (849.277 * t2)
Compound Y = 6557.607 * (1.196t)
Growth Y = e (8.788 + (0.179 * t))
Cubic Y = 6578.870 + (3244.898 * t) + (-566.140 * t2) + (52.423 * t3)
S-curve Y = e (10.918 + (-2.581/t))
Exponential Y = 6557.607 * (e (0.179 * t))
Inverse Y = 67214.527 + (-92375.452 / t)
Power Y = 4366.527 * (t1.022)
Logistic Y = 1 / (1/150726 + (0.001 * (0.619t)))
Table- 4.38.: Alternative Model Equations on GA - PNB

Linear Y = -49897.740 + (15231.277 * t)
Quadratic Y = 41104.825 + (-13506.375 * t) + (1596.536 * t2)
Compound Y = 9435.922 * (1.219t)
122
Growth Y = e (9.139 + (0.198 * t))
Cubic Y = 1707.865 + (9532.783 * t) + (-1513.750 * t2) + (115.196 * t3)
S-curve Y = e (11.512 + (-2.921/t))
Exponential Y = 9316.007 * (e (0.198 * t))
Inverse Y = 122582.924 + (-174960.525 / t)
Power Y = 5779.361 * (t1.146)
Logistic Y = 1 / (1/293780 + (0.001 * (0.624t)))
Table- 4.39.: Alternative Model Equations on GA - PSB

Linear Y = -549954.559 + (191116.441 * t)
Quadratic Y = 517709.559 + (-146040.649 * t) + (18730.949 * t2)
Compound Y = 158633.036 * (1.197t)
Growth Y = e (11.974 + (0.179 * t))
Cubic Y = 200779.074 + (39298.816 * t) + (-6289.878 * t2) + (926.697 * t3)
S-curve Y = e (14.114 + (-2.592/t))
Exponential Y = 158633.036 * (e (0.179 * t))
Inverse Y = 1619170.666 + (2219565.849 / t)
Power Y = 105011.346 * (t1.029)
Logistic Y = 1 / (1/3550390 + (0.00005 * (0.581t)))
Values of precision criteria for Alternative Models

Values of five (R, R2, Adjusted R2, α, and MAE) different precision criteria have been listed for
11 alternative models for each Bank; Pr. (F > F0) = α. Competing models are those for which the
values of R2, or Adjusted R2 are very close.
Table- 4.40.: Values of precision criteria for Alternative Models on GA - SBI
Adjusted R α
Model R R Square F0 MAE
Square
.912 .832 .821 74.490 .000 90150.73
Linear
Quadratic * .996 .991 .990 775.234 .000 21209.42
Compound * .990 .979 .978 715.358 .000 26726.36
Growth * .990 .979 .978 715.358 .000 26726.36
123
Logarithmic .741 .549 .519 18.256 .001 145621.03
Cubic * .999 .998 .997 1914.541 .000 9935.71
S .657 .432 .394 11.416 .004 160201.77
Exponential * .990 .979 .978 715.358 .000 26726.36
Inverse .493 .243 .192 4.812 .044 190007.39
Power .886 .786 .771 54.941 .000 98946.84
Logistic .673 .453 .417 12.431 .003 151654.34
Table- 4.41.: Values of precision criteria for Alternative Models on GA - SBT

Adjusted R F0 α
Model R R Square Square MAE
Linear .933 .870 .861 100.398 .000 4798.24
Quadratic * .998 .996 .996 1826.256 .000 811.80
Compound * .995 .989 .988 1368.399 .000 869.59
Growth * .995 .989 .988 1368.399 .000 869.59
Logarithmic .768 .589 .562 21.523 .000 8328.60
Cubic * .999 .998 .998 2331.362 .000 497.40
S .674 .454 .418 12.480 .003 9883.34
Exponential * .995 .989 .988 1368.399 .000 869.59
Inverse .516 .266 .217 5.434 .034 11572.98
Power .902 .814 .802 65.727 .000 5645.93
Logistic .730 .533 .501 17.086 .001 7526.87
* Represents competing models
Table- 4.42.: Values of precision criteria for Alternative Models on GA - SB

Adjusted R F0 α
Linear .923 .852 .842 86.549 .000 12643.22
Quadratic * .996 .993 .992 943.076 .000 2638.48
Compound * .994 .989 .988 1314.445 .000 3101.03
Growth * .994 .989 .988 1314.445 .000 3101.03
Logarithmic .755 .570 .541 19.848 .000 21287.80
Cubic * .997 .994 .993 783.542 .000 2300.87
S .696 .485 .450 14.106 .002 23806.89
Exponential * .994 .989 .988 1314.445 .000 3101.03
Inverse .504 .255 .205 5.121 .039 28608.20
Power .913 .834 .823 75.154 .000 14033.64
Logistic .739 .547 .516 18.087 .001 17586.51
124
Table- 4.43.: Values of precision criteria for Alternative Models on GA - UCO
Adjusted R F0 α
Linear .923 .851 .841 85.892 .000 11553.93
Quadratic * .999 .997 .997 2416.563 .000 1641.76
Compound * .992 .983 .982 884.994 .000 2190.88
Growth * .992 .983 .982 884.994 .000 2190.88
Logarithmic .750 .563 .534 19.327 .001 19343.97
Cubic * 1.000 1.000 .999 8962.282 .000 654.86
S .646 .417 .379 10.745 .005 22015.12
Exponential * .992 .983 .982 884.994 .000 2190.88
Inverse .495 .245 .195 4.872 .043 25668.89
Power .890 .791 .778 56.932 .000 12754.85
Logistic .735 .540 .509 17.574 .001 16307.11
Table- 4.44.: Values of precision criteria for Alternative Models on GA - CBI

Adjusted R F0 α
Linear .900 .811 .798 64.211 .000 15902.47
Quadratic * .991 .982 .980 385.753 .000 5066.09
Compound * .987 .973 .972 549.024 .000 5548.53
Growth * .987 .973 .972 549.024 .000 5548.53
Logarithmic .728 .530 .499 16.912 .001 25021.45
Cubic * .997 .994 .993 710.982 .000 2823.49
S .667 .444 .407 11.999 .003 27245.65
Exponential * .987 .973 .972 549.024 .000 5548.53
Inverse .483 .234 .182 4.569 .049 32942.31
Power .890 .791 .777 56.889 .000 17418.99
Logistic .705 .497 .463 14.800 .002 23305.52
Table- 4.45.: Values of precision criteria for Alternative Models on GA - PNB

Adjusted R F0 α
Linear .902 .813 .801 65.325 .000 29835.7
Quadratic * .991 .983 .981 405.548 .000 9564.872
Compound * .997 .995 .995 2953.226 .000 5019.345
Growth * .997 .995 .995 2953.226 .000 5019.345
Logarithmic .730 .533 .502 17.106 .001 46637.29
Cubic * .999 .999 .999 4090.522 .000 2200.884
S .688 .473 .438 13.483 .002 48546.67
125
Exponential * .997 .995 .995 2953.226 .000 5019.345
Inverse .484 .235 .184 4.597 .049 59816.3
Power .910 .828 .817 72.354 .000 29065
Logistic .727 .528 .497 16.778 .001 41631.82
Table- 4.46.: Values of precision criteria for Alternative Models on GA – PSB

Adjusted R F0 α
Linear .916 .839 .829 78.325 .000 349809.33
Quadratic * .996 .992 .991 919.313 .000 81309.29
Compound * .994 .989 .988 1330.652 .000 80854.02
Growth * .994 .989 .988 1330.652 .000 80854.02
Logarithmic .746 .557 .527 18.827 .001 571126.60
Cubic * 1.000 .999 .999 5356.511 .000 23472.32
S .672 .451 .414 12.324 .003 627100.14
Exponential * .994 .989 .988 1330.652 .000 80854.02
Inverse .497 .247 .197 4.932 .042 753092.20
Power .898 .807 .794 62.749 .000 380543.41
Logistic .661 .437 .399 11.636 .004 628031.26
From the tables 4.40 to 4.46 (7 Tables) the best competing models are selected as, Quadratic or
Compound or Growth or Cubic or Exponential, with respect to SBI, SBT, SB, UCO, CBI, PNB
and PSB (in aggregate) respectively of which Cubic is the best model based on the Adjusted R2
and MAE criteria.
The graphical presentations of the best competing models are given below.
Chart- 4.26.: Best Competing Models on GA – SBI Chart- 4.27.: Best Competing Models on GA– SBT
126
Chart- 4.28.: Best Competing Models on GA– SB Chart- 4.29.: Best Competing Models on GA–UCO
Chart- 4.30.: Best Competing Models on GA– CBI Chart- 4.31.: Best Competing Models on GA– PNB
Chart- 4.32.: Best Competing Models on GA– PSB
127
Diagnostics Checks on Best Model on Errors
Table- 4.47.: DW Values Calculated on Errors (using best model) – GA

Bank N DW dL dU Interpretation
Name statistics
(Model
Name)
SBI 17 1.334 0.873 1.102 Since 1.334 is > 1.102, hence there is evidence that the
SBT 17 1.269 0.873 1.102 Since 1.269 is > 1.102, hence there is evidence that the
SB 17 0.837 0.873 1.102 Since .837 is < .873, hence there is evidence that the
UCO 17 1.906 0.873 1.102 Since 1.906 is > 1.102, hence there is evidence that the
CBI 17 0.825 0.873 1.102 Since .825 is < .873, hence there is evidence that the
PNB 17 1.795 0.873 1.102 Since 1.795 is > 1.102 hence there is evidence that the
PSB 17 1.409 0.873 1.102 Since 1.409 is > 1.102, hence there is evidence that the
Table- 4.48.: ‘t’ test Values Calculated on Errors (using best model) – GA
Bank df Calculated Table Interpretation First order
Name value of t value of t auto-corr. on
(Model (1% - one residuals
Name) sided)
SBI 14 1.1052 2.624 Since Calculated t< Tabulated t, hence the 0.28
(Cubic) variables are not correlated in the
population
SBT 14 1.338 2.624 Since Calculated t< Tabulated t, hence the 0.34
population
SB 14 2.5577 2.624 Since Calculated t< Tabulated t, hence the 0.56
population
UCO 14 0.1136 2.624 Since Calculated t< Tabulated t, hence the 0.03
population
128
CBI 14 1.8441 2.624 Since Calculated t< Tabulated t, hence the 0.44
population
PNB 14 0.1511 2.624 Since Calculated t< Tabulated t, hence the 0.04
population
PSB 14 0.8395 2.624 Since Calculated t< Tabulated t, hence the 0.22
population
all selected banks including PSB in aggregate (with the exception of SB and CBI) in respect of
the parameter GA are not correlated (independent). We also observe as per ‘t’ test for testing of
independence is given in table 4.48, that the data-sets for all selected banks including PSB in
aggregate in respect of parameter GA, are independent. The above two analyses performed by
employing two methods (rather two methods of calculation of correlation coefficient) produce
almost similar inferences (on existence of independence of observations) on data-sets on
different banks in respect of the parameter, GA.
Table- 4.49.: Summary Statistic Shapiro-Wilk Test on errors (using best model) – GA
Bank SW
(Model Name) Statistic df Sig.
SBI (Cubic) .969 17 .793
SBT (Cubic) .905 17 .076
SB (Cubic) .974 17 .888
UCO (Cubic) .966 17 .754
CBI (Cubic) .965 17 .729
PNB (Cubic) .944 17 .375
PSB (Cubic) .959 17 .617
Interpretation of SW Test output in respect of Parameter GA:

From the above table it is evident that the significance-probability value of the SW Test statistic
is greater than 0.05 irrespective of Banks on GA data-sets and hence each of the data-sets on
selected PSBs (and also PSB in aggregate) can be assumed to be normal.
129
1. The main objective of the research work undertaken is to obtain a quantitative assessment of
the movement of GA values over years with respect to six PSBs and PSBs (aggregated), and
to obtain the pattern of dynamicity different parametric models are fitted. Strikingly, the
values of GA follow smooth patterns over years irrespective of banks.
2. The diagnostic methods applied on the GA values do definitely suspect non-randomness and
existence of strong serial correlation in many data sets (as revealed from tables given above).
3. 11 Parametric models have been tried and graph-plots of best (and competitive) models
(selected on the basis of standard Best Fit criteria) have been displayed in different Charts,
truly, the best identified models possess reasonably high precision embedded in those (details
are delineated in the relevant portions of this chapter). Also, interestingly, a few GA data sets
are amenable to models, not of similar characters, with very high precision levels.
4. Residuals from best models are found to be random and they follow normality, thus model
post-requisites in regard to both randomness and normality have been very well satisfied. The
analytical framework and the results obtained indisputably uphold the validity of these
models recommending their future uses on GA data-sets on other banks also.
5. The underlying system of data generation reveals the impact of a pre-defined signal.
review included in chapter-1).
4.6. Aspects on Forecasting

In this section the prime goal is to compare the competing models (defined as models which
possess very close values of the R2 criterion} based on their powers to produce forecasts closest
to observed GNPA /GA values and the criterion to judge the above mentioned closeness is MAE
(defined earlier). In order to obtain the magnitudes of forecast errors, the relevant data-set is
divided into two partitions of which the first partition is used to generate the model while the
second partition is used to obtain forecasts using the model generated by using the data set
included under the first partition. As an example, for a Bank with GNPA data for 15 years (2
observations are excluded as being outliers), the whole data-set is divided into two sets of data,
the first set contains the first thirteen observations, and the second set contains the last two data.
As two forecasts are generated (corresponding to the data in the second set), the value of MAE
(defined earlier) is obtained for each of the competing models (found as the most precise) in
130
respect of SBI and the same procedure is continued for all other Banks, (and also for data-sets
belonging to the parameter GA). The particular model for which the value of MAE is least in
respect of a Bank (either in GNPA data set or in GA data set), the same may be regarded as the
most suitable one. In what follows, it is understood that the data included in different tables carry
their respective units after them.
4.6.1. Forecasting on GNPA Dataset

Table- 4.50.: Estimation of Accuracies of the Competing Models on GNPA – SBI
(First partition contains data for first 14 years; Second partition contains data for last 2 years)
Year Observed Values Predicted Values (Cubic Model)
2010 19534.89 17407.17636
2011 25326.29 21515.06611
MAE 2969.46877
Conclusion: In case of SBI, Cubic model is found as the best model and also best forecasting
model (based on the least value of MAE).
Table- 4.51.: Estimation of Accuracies of the Competing Models on GNPA – SBT

(First partition: Data for first 14 years; Second partition: Data for last 2 years)
2010 641.98000 695.70357
2011 835.23000 857.97196
MAE 38.23277
Conclusion: In case of SBT, Cubic model is found as the best model (also best forecasting
model, as judged on the basis of having least value of MAE).
Table- 4.52.: Estimation of Accuracies of the Competing Models on GNPA – SB

(First Partition: Data for first 14 years; Second Partition: Data for last 2 years)
Year Observed Values Predicted Value (Quadratic model) Predicted value (Cubic model)
2010 2006.82 1849.29885 1531.24287
2011 2598.97 1954.33179 1381.83103
MAE 401.07968 846.35805
Conclusion: In case of SB, Quadratic model is found as the best model and also best forecasting
model (based on the least value of MAE).
131
Table- 4.53.: Estimation of Accuracies of the Competing Models on GNPA – UCO
2009 1539.51000 1874.79517
2010 1666.43000 2188.47942
MAE 428.66730
Conclusion: In case of UCO, Cubic model is the best model (best forecasting model also).
Table- 4.54.: Estimation of Accuracies of the Competing Models on GNPA – CBI

2010 2457.89000 1848.38878
2011 2394.53000 1496.43688
MAE 753.79717
Conclusion: In case of CBI, Cubic model is the best model (also best forecasting model).
Table- 4.55.: Estimation of Accuracies of the Competing Models on GNPA – PNB

Year Observed values Predicted values (Fifth degree Polynomial model )

2010 3214.41 3989.33769
2011 4379.39 6419.48789
MAE 1407.51279
Conclusion: In case of PNB, Fifth degree Polynomial model is the best model (also best
forecasting model).
Table- 4.56.: Estimation of Accuracies of the Competing Models on GNPA – PSB

2010 57293.00000 41176.87912
2011 71080.00000 43340.41758
MAE 21927.85165
Conclusion: In case of PSB, Cubic model is the best model (also best forecasting model).
132
4.6.1.1. Summary of Findings at a Glance
Table- 4.57.: Best Forecasting Model Expressions in Respect of Parameter GNPA
Bank Model Name Expression
SBI Cubic Y = 3685.523 + (5629.361 * t) + (-844.575 * t2) + (35.947 * t3)
SBT Cubic Y = 231.097 + (279.849 * t) + (-39.441 * t2) + (1.516 * t3)
SB Quadratic Y = 1270.125 + (-66.029 * t) + (8.119 * t2)
UCO Cubic Y = 1758.241 + (-31.425 * t) + (-5.778 * t2 ) + (0.511 * t3)
CBI Cubic Y = 1574.278 + (503.131 * t) + (-52.653 * t2) + (1.504 * t 3)
PNB Polynomial of Y = 2923.189 + (-599.718 * t) + (167.342 * t2) + (-1.727 * t4) + (0.078 * t5)
degree 5
PSB Cubic Y = 21846.269 + (16012.503 * t) + (-2307.841 * t2) + (92.804 * t3)
(Y represents GNPA and t is time)
Using the above models the forecast values for the years, 2013, 2014 and 2015 with respect to
the parameter GNPA are given in the Table – 4.58.
Table- 4.58.: Forecast Values for Parameter GNPA
Forecasted value (Rs in Crores)

Bank Model 2013 2014 2015
SBI Cubic 41015.1 52312.9 66019.44
SBT Cubic 1333.28 1711.18 2183.07
SB Quadratic 2712.28 2946.67 3197.3
UCO Cubic 2301.41 2581.13 2907.57
CBI Cubic 2344.69 2444.67 2610.85
PNB Polynomial of degree 5 12439.83 20010.36 31145.63
PSB Cubic 103565 129497 161393.4
The movement of forecast values over the years under forecast (3 years beyond) follows the
model (the basic system from which the data are generated, has encountered stochasticity).
133
4.6.2. Forecasting on GA Dataset
Table- 4.59.: Estimation of Accuracies of the Competing Models on GA – SBI
Predicted Value
Year Observed Values Quadratic Compound Growth Cubic Exponential
2011 771802.00 718398.07 652059.91 652059.91 799929.18 652059.91
2012 893613.00 828121.86 769102.22 769102.22 970801.31 769102.22
MAE 59447.53 122126.44 122126.44 52657.75 122126.44
Conclusion: In case of SBI, Cubic mode is the best model (and best forecasting model).
Table- 4.60.: Estimation of Accuracies of the Competing Models on GA – SBT

Predicted Value
2011 46044.00 44336.96 45192.14 45192.14 45296.27 45192.14
2012 55346.00 50707.70 54119.10 54119.10 52386.48 54119.10
MAE 3172.67 1039.38 1039.38 1853.62 1039.38
Conclusion: In case of SBT, 3 models, namely, Compound, Growth and Exponential have least
MAE though Cubic is having highest values of R2, Adjusted R2 coefficients and least value of
MAE (calculated on 17 years data) coefficient the competing models. Hence, Cubic model (any
one of Compound, Growth and Exponential models) is recommended as the best model (best
forecasting model).
Table- 4.61.: Estimation of Accuracies of the Competing Models on GA – SB

Predicted Value
2011 108350.00 107419.04 108318.38 108318.38 117185.56 108318.38
2012 125617.00 124398.75 132878.73 132878.73 141490.14 132878.73
MAE 1074.60 3646.68 3646.68 12354.35 3646.68
134
Conclusion: In case of SB, Quadratic model has least MAE though Cubic model is having the
highest values of R2, Adjusted R2 coefficients and least value of MAE (calculated on 17 years
data) coefficient amongst the competing models. Hence, Quadratic model (Cubic model) is
recommended as the best forecast models (and best model).
Table- 4.62: Estimation of Accuracies of the Competing Models on GA – UCO

Predicted Value
2011 100561.00 95198.68 96912.99 96912.99 100090.45 96912.99
2012 117504.00 110010.20 118750.02 118750.02 118570.79 118750.02
MAE 6428.06 2447.02 2447.02 768.67 2447.02
Conclusion: In case of UCO, Cubic model is the best model (also the best forecasting model).
Cubic model is recommended as best on UCO data-set.
Table- 4.63.: Estimation of Accuracies of the Competing Models on GA – CBI

Predicted Value
2011 131407.00 117982.50 106178.22 106178.22 136627.74 106178.22
2012 150725.00 136510.33 126035.15 126035.15 169139.48 126035.15
MAE 13819.58 24959.31 24959.31 11817.61 24959.31
Conclusion: Cubic model is recommended as one best model and the best forecasting model. .
Table- 4.64.: Estimation of Accuracies of the Competing Models on GA – PNB

Predicted Value
2011 242107.0 177035.62 172904.59 172904.59 186710.29 172904.59
2012 293775.0 205995.56 209646.19 209646.19 227683.82 209646.19
MAE 76425.41 76665.61 76665.61 60743.94 76665.61
Conclusion: Cubic model is recommended as one best model and the best forecasting model.
135
Table- 4.65.: Estimation of Accuracies of the Competing Models on GA – PSB
Predicted Value
2010 3059870.00 2806759.92 2677554.78 2677554.78 3084837.10 2677554.78
2011 3550389.00 3231299.98 3189767.86 3189767.86 3717935.05 3189767.86
MAE 286099.55 371468.18 371468.18 96256.57 371468.18
Conclusion: The best Forecast model is the one for which the value of MAE coefficient is least.
On PSB data-set Cubic model is the best forecasting model, and is also the best model.
4.6.2.1. Summary of Findings at a Glance

Table- 4.66.: Best Forecasting Model Expressions in Respect of Parameter GA
Bank Model Name Expression
SBI Cubic Y = 62937.841 + (7963.204 * t) + (-1468.425 * t2) + (231.652 * t3)
SBT Compound Y = 2515.010 * (1.198t)
Growth Y = e(7.830 + (0.181*t))
Exponential Y = 2515.010 * (e(0.181 * t))
SB Quadratic Y = 14824.139 + (-4838.239 * t) + (666.580 * t2)
UCO Cubic Y = 7615.702 + (-878.617 * t) + (113.639 * t2) + (18.840 * t3)
CBI Cubic Y = 6578.870 + (3244.898 * t) + (-566.140 * t2) + (52.423 * t3)
PNB Cubic Y = 1707.865 + (9532.783 * t) + (-1513.750 * t2) + (115.196 * t3)
PSB Cubic Y = 200779.074 + (39298.816 * t) + (-6289.878 * t2) + (926.697 * t3)
Y represents GA and t is time.
Using the above models the forecast values for 2013, 2014 and 2015 are given in Table – 4.67.
Table- 4.67.: Forecast Values for Parameter GA

Bank Model Forecasted value (Rs in Crores)
2013 2014 2015
SBI Cubic 1081498.85 1273036.68 1488045.96
SBT Compound 65412.14 78392.78 93949.35
Growth 65412.14 78392.78 93949.35
Exponential 65412.14 78392.78 93949.35
SB Quadratic 143707.63 163532.84 184691.21
136
UCO Cubic 138494.14 161168.78 186218.45
CBI Cubic 187287.8 223423.8 264403.7
PNB Cubic 354664.77 426494.88 508429.80
PSB Cubic 4274735.99 5033027.46 5884382.66
The movement of forecast values over the years under forecast (3 years beyond) follows the
model (the basic system from which the data are generated possesses an inkling of an in-built
artificial work-out embedded in its frame).
A BREATHER from deep analytics

The above rigorous analytics portray comprehensibly the GNPA and GA scenarios prevailing in
the public sector banks in India, however, at the end of this Chapter results obtained by adopting
a simple, easy and crude way of understanding the annual movement of GA values by simple
and well known measures (described in Section 4.4.7), are given in the following Tables.
Table- 4.68.: Annual Growth Rates for the Parameter GA in 6 PSBs and PSBs in Aggregate
1997 0.0347 0.1053 0.0810 -0.1268 0.0057 0.1062 0.0557
1998 0.1840 0.1213 0.1689 0.1296 0.1838 0.1355 0.1669
1999 0.1150 0.0569 0.3594 0.1334 0.2274 0.1880 0.1416
2000 0.1837 0.1716 0.2791 0.1548 0.2218 0.1819 0.1683
2001 0.1475 0.1849 0.0652 0.2476 0.1848 0.2465 0.1633
2002 0.0555 0.1610 0.1394 0.2673 0.1338 0.2312 0.1521
2003 0.1159 0.2317 0.0945 0.1934 0.0816 0.1822 0.1344
2004 0.1303 0.2341 0.2738 0.2870 -0.0081 0.1614 0.1457
2005 0.2786 0.2933 0.2777 0.3228 0.1200 0.2568 0.3155
2006 0.3753 0.2618 0.3589 0.3394 0.4201 0.2193 0.2297
2007 0.1907 0.3037 0.4031 0.2553 0.3646 0.2840 0.2829
2008 0.2340 0.1252 0.2339 0.1718 0.3893 0.2159 0.2348
2009 0.2988 0.1626 0.2653 0.2524 0.1676 0.2946 0.2379
2010 0.1694 0.1758 0.1086 0.1967 0.2349 0.2062 0.1964
2011 0.2032 0.1972 0.1848 0.2062 0.2267 0.2975 0.2179
2012 0.1578 0.2020 0.1594 0.1685 0.1470 0.2134 0.1603
137
Table- 4.69.: Mean values calculated from Annual Growth Rate (Percentage)
SBI SBT SB UCO CBI PNB PSB
Arithmetic Mean 17.9658 18.6772 21.5807 19.9970 19.3820 21.3781 18.7708
Geometric Mean 15.6269 17.3423 18.8647 NA* NA* 20.6555 17.5985
Harmonic Mean 12.7249 15.7297 16.1315 NA** NA** 19.8563 16.0992
* Data-sets contain negative entries: GM values are incalculable as real values
** Data-sets contain both negative and positive values, and calculation of HM values is meaningless
Findings:
The average annual growth rates (in percentage) of a bank are calculated by using the familiar
and simple statistics, like, AM/GM/HM of growth rates mentioned above in case of each Bank
under the purview of the study. The AM of annual growth rate is highest for SB (21.58%) and is
lowest for SBI (17.97%). PNB attains the position of having the second highest annual growth
rate (21.38%), very close to that of SB.
The GM of annual growth rate is highest for PNB (20.66%) and lowest for SBI (15.63%). GM of
annual growth rates in case of other banks lie in between these two extreme values.
The HM of annual growth rates is highest for PNB (19.86%) and lowest for SBI (12.72%). The
HM of annual growth rates in case of other banks lie in between these two values.
The AM, GM and HM of annual growth rates is lowest for SBI. The AM, GM and HM of annual
growth rates is always on higher side in case of PNB.
4.7. Conclusion
The table 4.1 and table 4.2 present the data-sets relating to parameters, GNPA and GA,
respectively. A look at the GA values with respect to most PSB’s reveals existence of continuous
upward trends over the last seventeen years, which indicates increasing credit appetite in the
various segments of the society and thereby justifying the commonly held impression (not fully
correct) regarding “growth story’ in the Indian economic sector. The GNPA values for most
PSB’s exhibit upward trends in the last three years (though prior to that period there was a
declining trend in the GNPA values over a period of six years in most PSB’s, excepting a few)
due to a number of factors, which included the global recession also, for which the GNPA
figures increased very sharply. The dataset in respect parameters GNPA and GA of all the banks
including PSBs in aggregate have exhibited similar features like curve linearity and upward
trends in the recent past and foreseeable future.
138
Chapter V
Movement of NPA over Time in Selected PSBs in India with

Non Parametric Approach
5.1 Introduction
5.3 Data Source
5.4 Methodological Aspects: Diagnostics Procedures and Goodness of Fit

Statistics, LOESS Curve Fitting Model
5.7 Conclusion
139
5.1. Introduction
A model is a representation of an observed dataset of a real life phenomenon presented as per

standard analytical practices and expressed by means of equation. It is well known that the
linear model is most attractive because of its simplicity to fit, easily understandable results
and availability of a wide range of techniques for testing the assumptions. However, it should
also be noted that, in many cases because of an intrinsic nonlinearity in the data, linear model
becomes inappropriate. Moreover, real life data situations are not easy to model because of
the uncertainty and variation prevailing in the natural system and also due to presence of the
interaction of unknown factors interplaying and governing the entire real life phenomenon. In
respect of real life data, parametric non linear models and related methods may be too
restrictive and as a result various non parametric techniques (that do not assume the data have
any characteristic structure or parameters) for modeling non linear time series data have
become popular amongst the researchers, to examine the underlying structure of the data.
While analyzing a non linear time series data either parametric non linear or non parametric
time series model is used. In the later method, according to Casas and Gao (2007), the data
may be allowed to speak for themselves in determining mathematical relationship between
variables. This makes the non parametric techniques more feasible and flexible. For fitting
smooth curves to time series non linear data, LOESS is considered as a powerful and simple
technique. LOESS is the acronym for “local regression”.
In this Chapter, the data-sets relating to parameters GNPA and GA in respect of six selected
PSBs along with PSBs-aggregated, given in the Table 4.1 and Table 4.2 respectively in
Chapter 4, have been analysed with the help of LOESS procedure and forecasting future
value for medium term have been attempted.
LOESS methodology was first proposed by Cleveland (1979, as cited in Jacoby, 2000) for
modeling and smoothing bivariate data and further developed by Cleveland and Devlin
(1988, as cited in Jacoby, 2000). According to McLeod (n.d.), “the technique provides a
general and flexible non linear family of models to fit bivariate data” (p. 1). LOESS does not
require any priori specification of the relationship between the dependent and independent
variables and is often used as a scatter plot smoother. The technique also contains inferential
procedures for confidence intervals and other statistical tests and can also be applied to
multivariate data. For all of these reasons, LOESS is a useful tool for data exploration and
140
analysis in the social sciences (Jacoby, 2000). Dias (n.d.) states “depicting the ‘local’
relationship between a response variable and a predictor variable over parts of their ranges,
which may differ from a ‘global’ relationship determined using the whole data set toward
smoothing a scatter plot is the essence of the LOESS procedure” (p. 1).
Therefore LOESS procedure can be used to model sets of real data in many disciplines
including banking and finance. The detailed reviews are included in Chapter-1. However, the
references of the base papers in connection modeling and forecasting have been cited in
Section 4.2 (Chapter 4).
5.3. Data Source
Details of data source are given in Section 4.3 (Chapter 4).
5.4. Methodological Aspects: Diagnostics Procedures and Goodness of Fit Statistics,

LOESS Curve Fitting Model
Quantile Comparison plots of the Residuals, Residuals vs Predictor plot, Absolute Residual
vs Predictor plot are performed on the data included in Table 4.1 and Table 4.2 (given in
Chapter 4). Development of non parametric curve fitting (LOESS Fit) and goodness of fit
statistics like R2LOESS, significance of LOESS F statistics have also been computed. In the
following sub sections different procedures related to the above are explained at first.
5.4.1. Diagnostics Procedures
Best fit in nonparametric regression is found with proper variable selection as well as the
choice of smoothing parameter. Like parametric regression, residuals are the most important
diagnostic component in case of nonparametric regression also. Several plots are used as
diagnostic tool as described in subsection 5.4.1.1 to 5.4.1.3.
5.4.1.1. Quantile Comparison Plots of the Residuals
In non parametric regression, residuals are defined as the difference between the observed
values of the dependent (Y) variable, and the corresponding fitted values for the respective
occurrences of the independent (X) variable values. This plot shows the residual plot from the
original LOESS curve that are fitted to the data. The points in this figure are obtained by
plotting the LOESS residual values (on the vertical axis) against normal norm quantiles plot
(on the horizontal axis) to check the presence of outliers in the data variables. The plot is also
used to detect departures from normality in the residual distribution.
141
5.4.1.2. Residuals vs Predictor Plot
The LOESS residuals are defined as: Ei=yi- ĝ(xi), where yi is the observed value and ĝ(xi) is
the fitted value. It is to be noted that, since the m evaluation points have been used to find the
LOESS curve (the vjs) are imaginary values which are usually different from the n observed
values of the independent variable hence interpolation between the two closest occurrences of
the equally-spaced evaluation points are used to find the fitted values for the empirical
observations, ĝ(xi).
The LOESS residuals are plotted against either the corresponding fitted values or the values
of the original independent variable (X) and a LOESS curve is fitted to the points within the
residual plot. A flat line located at the zero value on the vertical axis in the residual plot is
obtained as a result of this new application of the LOESS smoother.
Jacoby (2000) states that:
The reasoning is as follows. The LOESS residuals measure the variability in Y that remains
after the dispersion of the fitted values (and hence, the smooth curve) is taken into account.
Any systematic functional dependencies between X and Y should be picked up by the
original smooth curve fitted to the bivariate data. To the extent that the LOESS fitting process
does so successfully, there should be no discernible patterns of any kind among the residuals;
this, in turn, would produce a horizontal line when a smooth curve is fitted to the residual plot
(Cleveland, 1993). (p. 590)
If the plot does not reveal any kind of systematic relationship (whether linear, cubic,
quadratic, etc.) between the residuals and the predicted values, we may say LOESS model is
appropriate for our data.
Jacoby (2000) argues that:
Residual plots can provide the analyst with useful guidance for controlling the LOESS fitting
process (Cleveland, 1993, 1994) (p.591). “This is particularly important for selecting the
proper value of the smoothing parameter α, as the appropriate λ value and the need for the
robustness weights can often be determined through visual inspection of the original scatter
plot” (Jacoby, 2000, p. 591).
5.4.1.3. Absolute Residuals vs Predictors
The LOESS absolute residuals are plotted against the corresponding predicted values of the
original independent variable (X). The points in this figure are obtained by plotting the
142
LOESS absolute residual values (on the vertical axis) against predictors (on the horizontal
axis). More effective is to plot residuals (or their absolute or squared values) against the
predictor variable or, equivalently, against the fitted values. This plot is primarily used to
detect dependence of the residuals on predictors. The lack of fit would result in a graph
showing the residuals departing from zeros in a systematic fashion likely a “megaphone”
shape.
5.4.2. Goodness of Fit Statistics
5.4.2.1. R2LOESS
The LOESS smooth curve discloses the structure within the data. By examining the
‘goodness of fit’ the researchers can understand how well the smooth curve characterizes the
data values. In case of nonparametric methods like LOESS the ‘goodness of fit’ is partially
appropriate. The ratio of the sum of squares in the LOESS fitted values to the total sum of
squares in the dependent variable gives us a summary fit statistics similar to R2.
Where, ĝ(xi) is the LOESS fitted value for observation i, is the mean of the LOESS
fitted values, yi is the dependent variable value for observation i, is the sample mean for
the dependent variable.
There are three major limitations in case of interpretation of R2LOESS value.
i. The R2LOESS cannot be interpreted as variance explained in true sense.
ii. When robustness weights are used in the fitting process, misleading or even
meaningless results can be produced by R2LOESS statistic.
iii. A large R2LOESS value should not always be expected.
The LOESS smoother does not fit a particular, narrowly defined model to data (Weisberg
(1996). Therefore the ‘goodness of fit’ concept becomes tricky. However, the statistics can be
used to give limited interpretation. It conveys the size of the fitted value variance, expressed
as a ratio of the total variance in Y and provide an effective summary of the degree to which
the LOESS fitted values track the empirical data points in the scatterplot. (Jacoby, 2000)
143
Decision Rule: A relatively high R2LOESS value would lead to the conclusion that the smooth
curve summarizes nearly all of the total dispersion in the dependent variable.
5.4.2.2. LOESS F statistics
The researchers use the approximate degrees of freedom to carry out F tests. The degrees of
freedom is also used to construct the confidence envelopes around the fitted curve.
The approximate degrees of freedom is given by, dfMOD=trace(S). There are several other
equivalent ways to define the degrees of freedom also. The residual degrees of freedom is
given by,dfRES=n - dfMOD. The estimated error variance is given by, S2=Σe2i/dfRES.
It is to be noted that in case of nonparametric regression the degrees of freedom are not
necessarily whole numbers. The researchers generally plot the 95% confidence band or
envelope for the regression function by joining the confidence intervals for each of the X-
values together.
The researcher use F tests to compare the residual sums of squares for alternative nested
models. Since the degrees of freedom are approximate hence these tests are approximate
only. It is worthwhile to note that the test of nonlinearity by constructing it with a linear
regression model is permitted by the nonparametric regression. Here, we have nested models
because a linear model is a special case of a general nonlinear relationship.
The F test takes the usual form:
Where, RSS0 is the residual sum of squares from the linear model, RSS1 is the residual sum
of squares from the nonparametric model and trace(S) is the df for the nonparametric model.
Decision rule: If p-value of F test is less than 0.05 then we conclude that there is no linear
relationship between response and predictor variables.
5.4.3. Non Parametric Curve Fitting (LOESS Fit)
There are different types of nonparametric regression in use. One of the most commonly used
method is LOESS or LOWESS procedure. The short form of locally weighted scatterplot
smoother is LOWESS and local regression is LOESS. LOWESS and LOESS is essentially
144
same thing. “LOESS has some highly desirable statistical properties, (it) is easy to compute,
and … (it) is easy to use” (Cleveland, 1993, p. 94). These models are used to fit local
polynomial regressions and join them together.
5.4.3.2. Steps in Developing LOESS Smooth Fit
The procedure for developing a LOESS smooth fit as described by Jacoby ( 2000), in his
published paper ‘LOESS: a non parametric graphical tool for depicting relationships between
variables’ and by Jacoby (n.d.), in his Lecture note ‘Lecture 14: Non-normal Error
Distribution and Nonlinear Functional Forms’ on ‘Regression III : Advanced Methods’ in
Michigan State University is stated below:
5.4.3.2.1. Defining the window width
The first step is to define the window width (q) that encloses the closest neighbours to each
data observation (the window half-width is labeled h). Here, for each data point the
researcher selects the ‘q’ nearest neighbours in terms of their X-value. This q was chosen to
represent a particular percentage of the data. The researcher typically chooses the window
width by trial and error. The researchers start with the first observation and move through the
data, finding the ‘q’ closest observations to each case. In each case, the observation is called
the focal X. Initially, m equally-spaced locations across the range of X values have been
defined calling these vj, where j = 1 to m. At each of these vj the LOESS curve will be
evaluated. We choose ĝ(vj) as the LOESS fitted value at vj.
After this the 2 parameters α (between 0 and 1) and λ (either 1 or 2) are supplied. The number
of observations used in each local regression (called ‘the window’) is found by using the
value of α. The window width is defined as, q= αn (where n is the number of empirical data
points, and q is truncated to an integer value). The degree of polynomial that is fitted to the
data is given by the value of λ.
5.4.3.2.2. Weighting the data
To give the greatest weight to the observations that are closest to the focal X observation, a
weight function is chosen by the researchers. In most of the cases the tricube weight function
is used. The distance from the point of evaluation (vj) to the ith observation xi is given by
Δ[i](vj) = |xi - vj|. The distances are sorted from smallest to largest. The same distance,
expressed as a proportion of the distance from vj to the farthest data point within the window
is given by, Δ[i](vj)* = Δ[i](vj)/Δ[q](vj). The neighborhood weight are calculated for all
observations, from i=1 to n. using the tricube weight function as follows:
145
5.4.3.2.3. Locally weighted least squares
The researchers then apply a polynomial regression to the focal X observation, using only the
nearest neighbour observations. This polynomial regression uses weighted least squares
(using the tricube weights). As a result the weighted residual sum of squares is minimized.
Normally either local linear regression or local quadratic regression is used but the use of
higher order polynomials is also allowed. The fitted value for the focal X value is then
calculated from this regression. The fitted value is then plotted on the scatterplot. First of all,
the coefficient bk is to be found that minimizes the following:
The value of λ can be set as 1 or 2, (λ=1 represent a linear equation, while λ= 2 represent a
quadratic equation) is fitted to the weighted data using above equation. The fitted value is
obtained by taking the predicted value of Y at vj, after the coefficients of the preceding
equation are found.
5.4.3.2.4. The Nonparametric Curve
For each observation in the data, defining the window width, weighting the data and locally
weighted least squares steps are carried out. For each value of X, there is a separate local
regression. For each focal X, a fitted value from these regressions is obtained. In other words,
all the points (vj ,ĝ(vj)) are calculated using the above regression method. These points are
plotted and connected as the solid lines. As a result, the local polynomial nonparametric
regression curve is obtained. A smoother curve is obtained by increasing the window width.
It is important to note that in this method, coefficient estimates are absent. Here, the
relationship between X and Y is graphed.
146
5.4.3.2.5. Robustness Weights
In the LOESS fitting procedure the robustness step is optional. Belsley (1980, as cited in
Jacoby, 2000) observes that the unusual observations can adversely and strongly influence
the LOESS like other least square methods.
Such unusual observations (outliers) are down weighted by the robustness step of the LOESS
procedure and as a result, the more concentrated areas of data points are more likely to be
followed by a smooth curve. The robustness weights in the LOESS fitting procedure is
routinely included to produce a graphical summery of the bi-variate data. The residuals may
not be normally distributed with a mean of zero and constant variance, which is considered to
be the only potential disadvantage of robust LOESS estimation. Thus it can be said, a no
robust LOESS fit can produce a misleading representation of the predominant structure
within the data (Jacoby, 2000).
5.4.4. Comparing the OLS and LOESS Fits
If there is a certain degree of correlation between the observations, ordinary least square
(OLS) method can be misleading. Hence, comparison between OLS and LOESS fits are
performed to find whether there is any departure from linearity in these data or not.
5.4.5. Software Used:
In this case we have used R software.
This section contains the results obtained after having applied the methods discussed in
section 5.4 and such are presented in sub section 5.5.1 and 5.5.2 for parameters GNPA and
GA respectively. The results are presented in the order: Developing LOESS smooth fit
(Defining the window width, Weighting the data, LOESS Fit, Robustness Weights),
Comparing the OLS and LOESS fits, Diagnostics (Quantile Comparison Plots of the
Residuals, Residuals vs Predictor Plot and Absolute Residuals vs Predictors Plot) and
Goodness of fit statistics (R2LOESS and LOESS F statistics)
147
5.5.1. Results on GNPA Dataset
5.5.1.1. Developing LOESS Smooth Fit
In GNPA dataset, there are 17 vj values (that is, n=17), uniformly spaced in the closed
interval from 1 to 17 (i.e., year 1996-2012); hence, v1=1, v2=2, and so on, up to v17=17. The
locations of the various vjs are shown as tick marks on the horizontal scale of all the figures.
Note that the vjs would usually range only across the observed values of the X variable. In
this case, α is set as 0.6 (i.e., 60% of the data), so αn=10. Thus, the window will always
enclose the 10 empirical data points that fall closest to the current vj on the horizontal axis.
Figures below show the window for vj=8th year (Year 2003) (an arbitrarily-selected value,
located by the asterisk along the horizontal axis in the figure). Note that the physical width of
this window will change at different values of vj, as the distance to the 10th closest data point
changes. Also, evaluation points close to either the maximum or minimum X values will be
asymmetric; to pick some obvious examples, v1 will have all 10 points in its window arrayed
to the right, while v17 will have all 10 points arrayed to its left. In any event, the window will
always contain the 10 closest empirical data points, regardless of their direction and/or
distance from vj.
Chart-5.1. Defining the window width on GNPA - SBI Chart- 5.2. Defining the window width on GNPA - SBT
Chart-5.3. Defining the window width on GNPA - SB Chart-5.4. Defining the window width on GNPA - UCO
148
Chart- 5.5.: Defining the window width on GNPA - CBI Chart-5.6.: Defining the window width on GNPA - PNB
Chart- 5.7.: Defining the window width on GNPA - PSB

For each vj, neighborhood weights are calculated. Neighborhood weights are calculated for
all observations, from i=1 to n for all GNPA values. The shape of the tricube weight function,
as well as the specific weights assigned to the 11 observations, is shown in figure below. It is
noted that the observations with X values close to 8th (year 2003) have large weights, near
1.0. The weights fall off fairly quickly for observations with X values substantially different
(in either direction) from the current evaluation point of vj=8th (year 2003).
Chart- 5.8.: Tricube weight function for parameter GNPA
5.5.1.1.3. Summary Statistics of the Non parametric (LOESS) Fit (Parameter GNPA)
The summary output of LOESS fit models (GNPA dataset) for six selected PSBs along with
PSB in aggregate, generated by R Software are given in the following tables:
149
Table- 5.1.: Summary Statistics of LOESS Model for selected PSBs on Parameter GNPA
No. of Observation 17
Equivalent Number of Parameters 5.36
Trace of smoother matrix 5.91
Normalize: TRUE, Span: 0.6, Degree: 2, Family: Gaussian,
Control settings
Surface: interpolate, Cell = 0.2
5.5.1.1.4. LOESS Fit (Quadratic) Model

The GNPAs of the selected banks plotted (on the vertical axis) against the Time (Years) from
1996-2012 (on the horizontal axis) are shown below.
Chart- 5.9.: The LOESS Fit on GNPA – SBI Chart- 5.10.: The LOESS Fit on GNPA – SBT
Chart- 5.11.: The LOESS Fit on GNPA – SB Chart- 5.12.: The LOESS Fit on GNPA – UCO
Chart- 5.13.: The LOESS Fit on GNPA – CBI Chart- 5.14.: The LOESS Fit on GNPA – PNB
150
Chart- 5.15.: The LOESS Fit on GNPA – PSB
Observations on the above non parametric models (smoothed by LOESS Fit) given in chart
number 5.9 to 5.15 for six selected PSBs along with PSB in aggregate are given below:
1. The general diagonal orientations of the points suggest that GNPA values are highly
correlated to Year for all the banks.
2. The curves also follow the central tendency of the Y variable’s values across the range of
the X variable which reveals the curvilinear nature of the relationship between GNPAs
and Time (Year).
3. These curves were obtained without any prior specification about the functional form of
the relationship. The sigmoid (i.e. S-like’) shapes of all the curves are produced by the
LOESS procedure. The slopes of the fitted curves are negative at some Years on the
horizontal axis. In contrast, the slopes become very shallow near the right and left sides
of the curve.
4. These curves clearly show that a linear model would provide a misleading depiction of
the relationship between GNPA values and Time (Year).

The effect of robust fitting option is given below. The solid black lines show the robust
LOESS (LOESS.smooth) curves. The red lines show the LOESS curves obtained when the
robustness weights are omitted from the fitting procedure. The deep blue lines show the
robust lowess curves. The light blue lines show the lowess curves obtained when the
robustness weights are omitted from the fitting procedure. All the robust LOESS and LOESS
curves are fit with α set at 60% and λ=2.
151
Chart- 5.16.: Robust LOESS Fit on GNPA – SBI Chart- 5.17.: Robust LOESS Fit on GNPA –SBT
Chart- 5.18.: Robust LOESS Fit on GNPA –SB Chart- 5.19.:Robust LOESS Fit on GNPA – UCO
Chart- 5.20.: Robust LOESS Fit on GNPA –CBI Chart- 5.21.: Robust LOESS Fit on GNPA –PNB
Chart- 5.22.: Robust LOESS Fit on GNPA – PSB
152
5.5.1.2. Comparing the OLS and LOESS fits
With given GNPA dataset, ordinary least square (OLS) method are misleading. The
following figures support our statement. The straight line is the OLS fit; the curve line is the
LOESS smooth from a local linear regression with a span (α) =0.6 and degree (λ) = 2. There
is a clear departure from linearity in these data and the linearity does not at all fit well.
Chart- 5.23.: Linear and LOESS Fits on GNPA - SBI Chart- 5.24.: Linear and LOESS Fits on GNPA - SBT
Chart- 5.25.: Linear and LOESS Fits on GNPA - SB Chart- 5.26.: Linear and LOESS Fits on GNPA - UCO
Chart- 5.27.: Linear and LOESS Fits on GNPA - CBI Chart- 5.28.: Linear and LOESS Fits on GNPA - PNB
153
Chart- 5.29.: Linear and LOESS Fits on GNPA - PSB
5.5.1.3.Diagonstics
5.5.1.3.1. Quantile Comparison Plot of Residuals
Chart- 5.30.:Quantile Comparison Plot on GNPA – SBI Chart-5.31.: Quantile Comparison Plot on GNPA – SBT
Chart- 5.32.: Quantile Comparison Plot on GNPA – SB Chart-5.33.:Quantile Comparison Plot on GNPA - UCO
Chart-5.34.: Quantile Comparison Plot on GNPA – CBI Chart-5.35.:Quantile Comparison Plot on GNPA – PNB
154
Chart- 5.36.:Quantile Comparison Plot on GNPA – PSB
From the above figures it is evident that in respect of residuals of GNPA dataset, outliers are
present in SBT, CBI and PSB in varying proportions (marginal for SBT and moderately high
for CBI and PSB). Rests of the banks (SBI, SB, UCO and PNB) are free from outliers. From
the above it may be concluded that the residuals of GNPA dataset follow approximately a
normal distribution.
5.5.1.3.2. Residual vs Predictor Plot

Chart- 5.37.: Residual Plot on GNPA - SBI Chart- 5.38.: Residual Plot on GNPA - SBT
Chart- 5.39.: Residual Plot on GNPA - SB Chart- 5.40.: Residual Plot on GNPA - UCO
155
Chart- 5.41.: Residual Plot on GNPA - CBI Chart- 5.42.: Residual Plot on GNPA - PNB
Chart- 5.43.: Residual Plot on GNPA - PSB
The dotted horizontal line is a visual baseline corresponding to residual value of zero. The
above ‘Residual vs Predictor Plot’ (chart number 5.37 to 5.43) shows that the residuals
appear randomly scattered around zero indicating that the LOESS model describes the data
well. It can also be stated that the model provide adequate representation of the GNPA bi-
variate dataset. Looking at the plots above we can say that for our GNPA dataset the plots
exhibit the absence of relationship between residual and predicted value and therefore can be
concluded that LOESS model exhibiting simple curvelinear relationship, is appropriate for
our GNPA dataset.
5.5.1.3.3. Absolute Residual vs Predictor (Abs Res vs Pred) Plot

Chart- 5.44.: Abs Res vs Pred Plot on GNPA - SBI Chart- 5.45.: Abs Res vs Pred Plot on GNPA - SBT
156
Chart- 5.46.: Abs Res vs Pred Plot on GNPA - SB Chart- 5.47.: Abs Res vs Pred Plot on GNPA - UCO
Chart- 5.48.: Abs Res vs Pred Plot on GNPA - CBI Chart- 5.49.: Abs Res vs Pred Plot on GNPA - PNB
Chart- 5.50.: Abs Res vs Pred Plot on GNPA - PSB
Above figures (chart No 5.44 to 5.50) exhibit that none of the graphs, excepting for the
models of SB, looks like megaphone shape. Hence the graphs show the justification of our
Loess models for the given GNPA dataset.
5.5.1.4. Goodness of fit statistics for LOESS smooth curve

A summary of R2LOESS, Residual standard error, LOESS F-statistic and P-value for all the
selected PSBs along with PSB in aggregate are given in the following table:
Table- 5.2.: Summary Statistics of precision criteria for selected PSBs on Parameter GNPA
Bank R2LOESS Residual Standard Error LOESS F-statistic P-value
SBI 0.97929 1296 84.26919 0.00000
SBT 0.91502 86.8 27.03522 0.00001
SB 0.97668 108.2 36.23729 0.00000
UCO 0.93869 223.8 31.79255 0.00001
CBI 0.78025 694.2 7.93439 0.00294
PNB 0.90009 595.1 16.55596 0.00013
PSB 0.97686 3298 91.06756 0.00000
157
The R2LOESS for the smoothed relationship between GNPAs and Time (Year) are given in the
above table. This indicates that such a relatively high R2LOESS value would lead to the
conclusion that the smooth curve summarizes nearly all of the total dispersion in the
dependent variable. It may also be noted that all the p-values < 0.05, which supports the
nonlinear relationships between GNPA of the selected banks and Time (Year).

1. The main objective of the research work undertaken is to examine underlying pattern of
the GNPA data set (linear or non linear) with respect to time through graphical
presentation and to develop a non parametric model for the purpose forecasting future
value of parameter GNPA for medium term (3 years).
2. The GNPA data set exhibits curve linearity for all the banks including aggregate PSB.
3. LOESS Curve Fit has been tried on GNPA dataset and the same has been displayed in
different charts for different banks, truly, such non parametric model of possess higher
degree of precision in comparison to best parametric model (captured in chapter 3) in
terms of explanatory power of the model ( R2), details are delineated in the relevant
parts.
review included in chapter-1).
5.5.2. Results on GA Dataset

5.5.2.1. Developing LOESS Smooth Fit
As we have 17 observations for both GNPA and GA, the same procedures as described in
previous subsection 5.5.1.1.1 has been followed.
Chart-5.51:Defining the window width on GA - SBI Chart-5.52:Defining the window width on GA – SBT
158
Chart-5.53: Defining the window width on GA - SB Chart-5.54:Defining the window width on GA – UCO
Chart-5.55: Defining the window width on GA – CBI Chart-5.56:Defining the window width on GA – PNB
Chart-5.57: Defining the window width on GA – PSB

As we have 17 observations for both GNPA and GA, the same procedures as described in
previous subsection 5.5.1.1.2 has been followed.
Chart- 5.58.: Tricube weight function for parameter GA
159
5.5.2.1.3. Summary Statistics of the Non parametric (LOESS) Fit (Parameter GA)
To fit a non parametric (LOESS) model the command : fit <- loess(GASBI~f(Year)) is used,
The summary output of LOESS fit models (GA Dataset) for six selected PSBs along with
PSB in aggregate, generated by R Software are given in the following tables:
Table- 5.3.: Summary Statistics of LOESS Model for selected PSBs on Parameter GA
No. of Observation 17
Equivalent Number of Parameters 5.36
Trace of smoother matrix 5.91
Normalize: TRUE, Span: 0.6, Degree: 2, Family: Gaussian,
Control settings
Surface: interpolate, Cell = 0.2
5.5.2.1.3. LOESS Fit (Quadratic) Model

The GAs of the selected banks plotted (on the vertical axis) against the Time (Years) from
1996-2012 (on the horizontal axis) are shown below.
Chart- 5.59.: The LOESS Fit on GA – SBI Chart- 5.60.: The LOESS Fit on GA –SBT
Chart- 5.61.: The LOESS Fit on GA – SB Chart- 5.62.: The LOESS Fit on GA – UCO
Chart- 5.63.: The LOESS Fit on GA – CBI Chart- 5.64.: The LOESS Fit on GA – PNB
160
Chart- 5.65.: The LOESS Fit on GA – PSB
Observations on the above non parametric models (smoothed by LOESS Fit) given in chart
number 5.59 to 5.65 for six selected PSBs along with PSB in aggregate are given below :
1. The general diagonal orientations of the points suggest that GA values are highly
correlated to Year for all the banks.
2. The curves also follow the central tendency of the Y variable’s values across the range of
the X variable which reveals the curvilinear nature of the relationship between GAs and
Time (Year).
3. These curves are obtained without any prior specification about the functional form of the
relationship. The Parabolic shape of the growth curve is produced by the LOESS Curves.
The slopes of the fitted curves are always positive at all Years on the horizontal axis. This
phenomena indicates that all the banks found Gross Advances are good way to create
assets and they are strongly emphasizing that over the years.
the relationship between GA values and Time (Year).
Thus, the nonparametric smooth curve provides useful information that could be incorporated
into a parametric model specification for these (or other) data. Despite its potential
importance, the curve linearity in the relationship between GAs and Year would have been
completely invisible in a standard, linear regression analysis — at least until an examination
of the residuals. Unfortunately, the latter step is often omitted from empirical research efforts.
And even when residual plots are produced, it is easy to overlook patterns that may exist
within them. The LOESS procedure helps the researcher avoid the intermediate step of fitting
a model that turns out to be an inaccurate representation of the data. Instead, the evidence
from the LOESS curve can be used to formulate a more accurate description of the data in the
first place.
161
The effect of robust fitting option is given below. The solid black lines show the robust
LOESS (LOESS.smooth) curves. The red lines show the LOESS curves obtained when the
robustness weights are omitted from the fitting procedure. The deep blue lines show the
robust lowess curves. The light blue lines show the lowess curves obtained when the
robustness weights are omitted from the fitting procedure.
All the robust LOESS and LOESS curves are fit with α set at 60% and λ=2.
Chart- 5.66.: Robust LOESS Fit on GA – SBI Chart- 5.67.: Robust LOESS Fit on GA – SBT
Chart- 5.68.: Robust LOESS Fit on GA – SB Chart- 5.69.: Robust LOESS Fit on GA – UCO
Chart- 5.70.: Robust LOESS Fit on GA – CBI Chart- 5.71.: Robust LOESS Fit on GA – PNB
162
Chart- 5.72.: Robust LOESS Fit on GA - PSB
5.5.2.2. Comparing the OLS and LOESS fits

As stated earlier, with given GA dataset, ordinary least square (OLS) method is misleading.
The following figures support our statement. The straight line is the OLS fit; the curve line is
the LOESS smooth from a local linear regression with a span (α) =0.6 and degree (λ) = 2.
There is a clear departure from linearity in these data - the linearity does not seem to fit well.
Chart- 5.73.: Linear and LOESS Fits on GA –SBI Chart- 5.74.: Linear and LOESS Fits on GA -SBT
Chart- 5.75.: Linear and LOESS Fits on GA –SB Chart- 5.76.: Linear and LOESS Fits on GA -UCO
163
Chart- 5.77.: Linear and LOESS Fits on GA –CBI Chart- 5.78.: Linear and LOESS Fits on GA -PNB
Chart- 5.79.: Linear and LOESS Fits on GA -PSB
5.5.2.3. Diagonstics
5.5.2.3.1. Quantile Comparison Plot of Residuals
Chart-5.80.: Quantile Comparison Plot on GA - SBI Chart- 5.81.: Quantile Comparison Plot on GA - SBT
Chart- 5.82.: Quantile Comparison Plot on GA - SB Chart- 5.83.: Quantile Comparison Plot on GA - UCO
164
Chart- 5.84.: Quantile Comparison Plot on GA - CBI Chart- 5.85.: Quantile Comparison Plot on GA - PNB
Chart- 5.86.: Quantile Comparison Plot on GA - PSB
From the above figures it is evident that in respect of residuals of GA dataset, outliers are
present in SBI, SBT, SB, CBI and PNB in varying proportions (marginal for SBI, SB and
CBI moderately high for SBT and PNB). Rests of the banks (UCO and PNB) are free from
outliers. From the above it may be concluded that the residuals of GNPA dataset follow
approximately a normal distribution.
5.5.2.3.2. Residual vs Predictor Plot

Chart- 5.87.: Residual Plot on GA -SBI Chart- 5.88.: Residual Plot on GA -SBT
Chart- 5.89.: Residual Plot on GA -SB Chart- 5.90.: Residual Plot on GA -UCO
165
Chart- 5.91.: Residual Plot on GA -CBI Chart- 5.92.: Residual Plot on GA -PNB
Chart- 5.93.: Residual Plot on GA -PSB
The dotted horizontal line is a visual baseline corresponding to residual value of zero. The
above ‘Residual vs Predictor Plot’ (chart number 5.87 to 5.93) shows that the residuals
appear randomly scattered around zero indicating that the LOESS model describes the data
well. It can also be stated that the model provide adequate representation of the GA dataset.
Looking at the plots above we can say that the plots exhibit the absence of relationship
between residual and predicted value and therefore can be concluded that LOESS model
exhibiting simple curve linear relationship, is appropriate for our GA dataset.
5.5.2.3.3. Absolute Residual vs Predictor (Abs Res vs Pred) Plot

Chart-5.94.: Abs Res vs Pred Plot on GA - SBI Chart-5.95.: Abs Res vs Pred Plot on GA – SBT
166
Chart- 5.96.: Abs Res vs Pred Plot on GA – SB Chart-5.97.: Abs Res vs Pred Plot on GA – UCO
Chart-5.98.: Abs Res vs Pred Plot on GA – CBI Chart-5.99.: Abs Res vs Pred Plot on GA – PNB
Chart-5.100.: Abs Res vs Pred Plot on GA – PSB
Above figures (chart No 5.94 to 5.100) exhibit that none of the graphs, excepting for the
models of CBI, looks like megaphone shape. Hence the graphs show the justification of our
Loess models for the given dataset.
5.5.2.4. Goodness of fit statistics for LOESS smooth curve

A summary of R2LOESS, Residual standard error, LOESS F-statistic and P-value for all the
selected PSBs along with PSB in aggregate is given in the following table:
167
Table- 5.4.: Summary Statistics of precision criteria for selected PSBs on Parameter GA
Bank R2LOESS Residual Standard Error LOESS F-statistic P-value
SBI 0.99922 9090 608.57270 0.00000
SBT 0.99896 650.1 346.89700 0.00000
SB 0.99902 1511 424.74550 0.00000
UCO 0.99969 775.9 1366.80600 0.00000
CBI 0.99869 2016 406.27770 0.00000
PNB 0.99953 2288 1115.72300 0.00000
PSB 0.99980 18340 2282.69800 0.00000
The R2LOESS for the smoothed relationship between GAs and Time (Year) are given in the
above table. This indicates that such a very high R2LOESS value would lead to the conclusion
that the smooth curve summarizes all most all of the total dispersion in the dependent
variable. It may also be noted that all the p-values < 0.05, which supports the nonlinear
relationships between GA of the selected banks and Time(Year).

1. The main objective of the research work undertaken is to examine underlying pattern of
the GA data set (linear or non linear) with respect to time through graphical presentation
and to develop a non parametric model for the purpose forecasting future value of
parameter GA for medium term (3 years).
2. The GA data set exhibits curve linearity for all the banks including aggregate PSB.
3. LOESS Curve Fit has been tried on GA dataset and the same has been displayed in
different charts for different banks, truly, such non parametric model possess still higher
degree of precision in comparison with best parametric model in terms of explanatory
power of the model ( R2), details are delineated in the relevant parts .
review included in Chapter-1).
5.6. Aspect of Forecasting

The main purpose of constructing a time series model is to forecast. Forecasting is a
quantitative estimate about the likelihood of future events which is developed on the basis of
current and past information. This information is embodied in the form of a model. By
extrapolating models beyond the period over which they were estimated, one can make
forecast about the future events. The non parametric models (smoothed by LOESS Fit)
obtained based on our dataset from1996 to 2012, have been extrapolated beyond the sample
168
period to get the forecasted values for the years, 2013, 2014 and 2015 with respect to the
parameters GNPA and GA are given in the Table – 5.5 and 5.6.
5.6.1. GNPA Dataset

Table- 5.5.: Forecast Values (Rs in Crores) for Parameter GNPA
Bank df Residual Scale 2013 2014 2015
SBI 11.73986 1460.24100 47769.89 60512.54 75269.91
SBT 11.73986 107.46330 1688.78 2163.76 2730.39
SB 11.73986 121.97670 3704.51 4436.93 5283.92
UCO 11.73986 250.99010 4960.88 6260.83 7780.24
CBI 11.73986 766.04620 7613.80 10026.34 12954.47
PNB 11.73986 696.48160 9864.07 12887.13 16519.65
PSB 11.73986 4113.17400 134061.40 170382.70 212990.00
5.6.2. GA Dataset
Table- 5.6.: Forecast Values (Rs in Crores) for Parameter GA
Bank df Residual Scale 2013 2014 2015
SBI 11.73986 9141.67400 1045556.00 1202892.00 1371598.00
SBT 11.73986 755.09740 62739.48 72068.65 82270.67
SB 11.73986 1653.61700 143014.70 161321.50 180477.10
UCO 11.73986 846.66840 136584.70 157407.60 179905.70
CBI 11.73986 2295.45500 178669.80 206381.90 235954.40
PNB 11.73986 2621.52100 351173.90 417222.00 490197.00
PSB 11.73986 19059.29000 4165363.00 4813296.00 5512412.00
5.7. Conclusion
The study on the aspects of movement of NPA offers a perpetual challenge to the bankers and
researchers in general as forecasting future NPA is essential to monitor, manage and mitigate
the deadly virus in banking system. In this chapter, LOESS, a very popular non parametric
technique, (which do not possess nice parametric interpretation, but offer more precise
representation, when evaluated against different measure of closeness) has been fitted to
GNPA and GA data sets for six selected PSBs along with PSB in aggregate and it has been
observed that particularly in respect of GNPA dataset, these non parametric models are very
much relevant and useful as it enhances the precision level of the predicted values, in
comparison to parametric models as observed in chapter 4.
169
Chapter VI
Modeling NPA Time Series Data in Selected PSBs in India with

Semi Parametric Approach
6.1 Introduction
6.3 Data Source
6.4 Methodological Aspects: Diagnostics Procedures and Goodness of Fit

Statistics, Curve Fitting using Penalized Spline Method
6.7 Conclusion
170
6.1. Introduction
Semi parametric regression is concerned with the flexible incorporation of non-linear functional
relationships in regression analyses. Any application area that benefits from regression analysis
can also benefit from semi parametric regression. Since a semi parametric regression model
contains both parametric and non parametric parts, it offers a good compromise between
parametric and non parametric regression models (Wu, 2010).
Researchers devoted serious attention on semi parametric modeling of either independent or

dependent time series data. Their focus on research interest has been primarily on estimation and
testing of both parametric and non parametric components in semi parametric models. Interest of
researchers also includes use of semi parametric methods in model estimation, specification
testing and selection of non linear time series data. The semi parametric modeling technique
comprises the two aims, flexibility and simplicity of statistical procedures by introducing partial
parametric components. The flexibility of semi parametric modeling has made it a widely
accepted statistical technique.
Semi parametric regression can also be of substantial value in complex real life problem. Such
models reduce the complexity in the data set to summarize that we can understand. Appropriate
application of such models ensure retaining essential features of the data set while discarding
unimportant details and hence they aid sound decision making. Indeed, these models are found to
have more representative power many times, when compared against different performance
criteria, to explain and model the real-life data situations. To capture the data characteristics,
emanated from a real life data situation, in this chapter, we have employed semi parametric
techniques. The table 4.1 and table 4.2 in chapter 4 present the data-sets relating to parameters,
GNPA and GA, respectively.
Yao & Lee (2008) states that:
Due to its simplicity and effectiveness for handling different semi parametric smoothing
problems, penalized spline regression has recently become a popular tool for solving various
estimation problems, ranging from environmental modeling (Wood and Augustin, 2002) to
longitudinal and functional data analysis (Harezlak, Ryan, Giedd, & Lange, 2005; Yao & Lee,
171
2006) to remote sensing imaging (Clarke et al., 2006). When comparing to smoothing splines, an
attractive property of penalized spline regression is the ease for conducting statistical inference
(Ruppert et al., 2003). (p 259)
Salkowski (2008) in his work has given examples of using spm function of the SemiPar package
to fit the semiparametric model where temperature deviation is a function of the year.
The detailed reviews are included in chapter-1. However, the references of the base papers in
connection have been cited in section 4.2 (chapter 4).
6.3. Data Source
Details of data source are given in section 4.3 (chapter 4).
6.4. Methodological Aspects: Diagnostics Procedures and Goodness of Fit Statistics, Curve
Fitting using Penalized Spline Method
Quantile Comparison plots of the Residuals, Residuals vs Predictor plot, Normal Q-Q plot are
performed on the data included in Table 4.1 and Table 4.2 (given in chapter 4). Development of
semi parametric curve fitting (Penalized spline) and goodness of fit statistics like R2, significance
of F statistics have also been computed. In the following sub section, the different procedures
related to the above are explained at first.
6.4.1. Diagnostics Procedures
Best fit in semi parametric regression is found with proper variable selection as well as the
choice of smoothing parameter. The residuals from a semi parametric regression fit can be
employed as a useful diagnostic tool as described in subsection 6.4.1.1 to 6.4.1.3. Such residuals
are used in order to determine whether the smooth curve adequately incorporates all of the
interesting structure in the data. The residuals are scrutinized for systematic patterns that may
remain after a hypothesized structural representation has been fitted to the empirical data.
6.4.1.1. Quantile Comparison Plots of the Residuals
In semi parametric regression, residuals are defined as the difference between the observed
values of the Y variable, and the corresponding fitted values for the respective occurrences of the
X variable values. This plot shows the residual plot from the original penalized spline curve that
is fitted to the data. The points in this figure are obtained by plotting the penalized spline residual
172
values (on the vertical axis) against normal norm quantiles plot (on the horizontal axis) to check
the presence of outliers in the data variables. The plot is also used to detect departures from
normality in the residual distribution.
6.4.1.2. Residuals vs Predictor Plot
As defined in the above subsection (6.4.1.1) the error is given by, Ei = yi-fˆ(xi)
The equation above is very similar to the familiar formula for calculating residual values in
regression analysis. However, there is one important difference. The K knots used to find the
semi parametric regression curve are imaginary values which are usually different from the n
observed values of the independent variable, X. Therefore, the fitted values for the empirical
observations, fˆ(xi) are typically obtained from the model.
Once the semi parametric regression residuals are calculated, they are plotted against either the
corresponding fitted values or (more commonly) the values of the original X variable. Then, a
semi parametric regression curve is fitted to the points within the residual plot. This new
application of the semi parametric regression smoother should produce a flat line located at the
zero value on the vertical axis in the residual plot. The reasons are as follows. The semi
parametric regression residuals measure the variability in Y that remains after the dispersion of
the fitted values (and hence, the smooth curve) is taken into account. Any systematic functional
dependencies between X and Y should be picked up by the original smooth curve fitted to the
bivariate data. To the extent that the semi parametric regression fitting process does so
successfully, there should be no discernible patterns of any kind among the residuals; this, in
turn, would produce a horizontal line when a smooth curve is fitted to the residual plot. If the
plot does not reveal any kind of systematic relationship (whether linear, cubic, quadratic, etc.)
between the residuals and the predicted values, we may say penalized spline model is appropriate
for our data.
6.4.1.3. Normal Q-Q Plot
Narrated in section 4.4.3 (chapter 4)
173
6.4.2. Goodness of Fit Statistics
6.4.2.1. R2 value
When a semi parametric regression curve is fitted to data, attention is usually focused on the
shape of the resultant curve because that feature is most revealing of the structure within the
data. However, it is also useful to consider how well the smooth curve characterizes the
empirical data values. This latter phenomenon is usually called ‘goodness of fit’, although that
term is only partially appropriate in the case of semi parametric regression. A summary fit
statistic similar to an R2 value can be obtained by taking the ratio of the sum of squares in the
semi parametric regression fitted values to the total sum of squares in the dependent variable.
Decision Rule: A relatively high R2 value would lead to the conclusion that the smooth curve
summarizes nearly all of the total dispersion in the dependent variable.
6.4.2.2. F statistics
The concept of degrees of freedom for semi parametric regression models is not as intuitive as
for linear models since there are no parameters estimated. Nonetheless, the degrees of freedom
for a semi parametric regression model are a generalization of the number of parameters in a
parametric model. The analogy to the linear model is not perfect, but approximate. Using the
approximate degrees of freedom, we can carry out F-tests to compare different estimates applied
to the same dataset; Compare different levels of polynomial fits; Compare the smoothed model
to a linear model etc. Determining the degrees of freedom is also necessary for constructing
confidence envelopes around the fitted curve. Unlike in linear model, the degrees of freedom for
semi parametric regression model are not necessarily whole numbers.
F-tests comparing the residual sums of squares for alternative nested models can be carried out in
exactly the same way as for linear models—these tests are only approximate because of the
approximation of the degree of freedom.
Perhaps the most useful aspect of semi parametric regression is that it allows us to test for
nonlinearity by contrasting it with a linear regression model. In general for semi parametric
regression, if we are testing two nested models, we can construct the usual F-ratio as
(( − )/( , − , ))⁄( / ( − , ))
174
where RSS0 is the residual sum of squares for the smaller model (linear model in our case), and
RSS1 is that for the larger model (semi parametric univariate penalized spline smoothing). Also
dfres, 0 is the degree of freedom for the RSS0 and dfres,1.is the degree of freedom for the RSS1.
Decision Rule: If p-value of F test is less than 0.05 then we conclude that there is no linear
relationship between response and predictor variables.
6.4.3. Semi Parametric Curve Fitting (Penalized spline)
Penalized splines have gained much popularity as a flexible tool for smoothing and semi-
parametric modeling. Penalized splines are low-rank smoothers, i.e. amount of knots used for
estimation is far less than the number of observations, which significantly reduces the numerical
effort. According to Yao & Lee (2008) there are two important components in fitting penalized
splines – a) selection of the smoothing parameter and b) choice of the number of knots and their
location.
6.4.3.2. Penalized spline procedure
The theoretical framework for developing a Penalized Spline fit as described by Nettleton (n.d.),
in his Handout is briefly given below.
Let us consider the model y = f(x ) + e where i=1,2….n
If f(x) is linear, then f(x) = β + β x
The linear model tries to approximate f(x) as a linear combination of two basis functions: where
bo(x) =1 and b1(x) = x, then
f(x) = β b (x) + β b (x)
Similarly, for a quadratic model i.e. f(x) = β + β x + β x
we can write, f(x) = β b (x) + β b (x) + β b (x)
Let us consider, S (x) = (x − k )+
Where, k1 is a specified real value.
Therefore f(x) can be approximated by β b (x) + β b (x) + u S (x)
175
where u1 is an unknown parameter (like β0 and β1).
It may be noted that,
β b (x) + β b (x) + u S (x) = β + β x + u (x − k )
This function is a continuous function since it is a linear combination of continuous function. At

the same time it is piecewise linear. It is simple example of linear spline. In this case k1 is known
as knot. In case of polynomial regression, we consider a linear spline model as a special case.
The linear spline function can become more flexible by adding more knots k1 … kk when
f(x) is approximated by, β + β x + ∑ u s (x)
= β + β x + ∑ u (x − k )+
Thus the model becomes, f(x) = β + β x + ∑ u (x − k )+
In many cases f(x) becomes non-smooth. Hence a less flexible (more smooth) estimate of f(x) is
usually preferred. This can be obtained using penalized least squares.
For this two things are important
(i) Smoothing parameter and (ii) Penalty for roughness. It is for the researcher to choose proper
smoothing parameter and the knots (k1, k2 … kn). If smoothing parameter is small we get a non
smooth fit.
Now, there are few strategies for choosing the smoothing parameter.
A) Cross Validation–It is a general strategy for choosing smoothing parameter.
B) Generalized Cross Validation– It is an approximation to cross validation. The amount of

smoothing can be selected by using the generalized cross validation function (Kageyama, Pal,
Pal, Medda & Basu, 2004).
Next the number of knots and their locations has to be decided. Ruppert (2002) states that “there
must be enough knots to fit features in the data, but after this minimum necessary number of
knots has been reached, further increases in K often have little effect on the fit” (p. 740). Here, K
means number of knots. Eilers and Marx (2004) stress that “equally- spaced knots are always to
be preferred”, while Ruppert (2002) emphasize utilization of quantile-based knots. In general,
both approaches do the work equally well for most of the cases. Wu (2010) points out that both
176
the location and number of knots are equally important for splines as they determine the degree
of smoothing. Knots can either be placed uniformly or all the distinct time point can be used as
knots. The percentile based knot placing rule can also be used. Thus it can be said that a good fit
can be obtained if there are enough knots. Penalization prevents a fit that is too rough even when
there are many knots.
In our case the semi-parametric model is fitted using the SemiPar package from the R statistical
system. SemiPar is a simple tool to construct a nonlinear regression. This package provides a
convenient way to fit splines to data using R. In the SemiPar package, f is estimated using
penalized spline smoothing.
For univariate smoothing, SemiPar’s default basis function is radial: | x− κk |p, where κk is the
location of the kth knot, and p is the degree. The default degree for the radial basis is 3, so the
default function is:
f(x) = β + β x + ∑u |x − K | where, k=1,…… K
6.4.4. Software Used
In this chapter, we have used SemiPar package of R software for developing forecasting model.
This section contains the results obtained after having applied the methods discussed in section
6.4 and such are presented in sub section 6.5.1 and 6.5.2 for parameters GNPA and GA
respectively.
6.5.1. Result on GNPA Dataset
The SemiPar package has algorithms for selecting knots, if knots are not provided by the user.
The smoothness of the fit can also be left to SemiPar, but users can control smoothness in two
ways. In a univariate fit, the smoothing parameter is the ratio of the smoothing variance to the
error variance. Large values of the smoothing parameter (σ2ε /σ2u) produce smoother functions.
Alternatively, users can specify the degrees of freedom for the fit. The more degree of freedom
in the fit, the less smooth the function is.
177
Now the GNPA dataset as dependent variable with time (year as 1996,1997,1998….2012 taken
as 1, 2, 3, ….17 respectively) as independent variable are processed through the SemiPar
package of R software and summary statistics for non linear components are given below.
Table- 6.1.: Summary Statistics for Penalized Spline models on Parameter GNPA
Bank df spar knots

SBI 4.955 1.707 3
SBT 4.774 3.039 3
SB 4.767 3.079 3
UCO 4.674 3.515 3
CBI 4.69 3.441 3
PNB 4.911 2.169 3
PSB 4.974 1.417 3
Note this includes 1 df for the intercept
6.5.1.1. The plots of the semi parametric models (smoothing by penalized spline)
The GNPA and time (year) dataset are plotted on vertical axis and horizontal axis respectively.
The plots of the semi parametric models (smoothing by penalized spline) for six selected PSBs
along with PSB in aggregate are given below:
Chart- 6.1.: Penalized Spline on GNPA –SBI Chart- 6.2.: Penalized Spline on GNPA -SBT
178
Chart- 6.3.: Penalized Spline on GNPA –SB Chart- 6.4.: Penalized Spline on GNPA –UCO
Chart- 6.5.: Penalized Spline on GNPA –CBI Chart- 6.6.: Penalized Spline on GNPA –PNB
Chart- 6.7.: Penalized Spline on GNPA –PSB
The findings from the above figures are detailed below:

1. The general diagonal orientations of the points suggest that GNPAs are highly correlated
to Year for all the banks.
2. The curves follow the central tendency of the Y variable’s (GNPA) values across the
range of the X variable (Time). In doing so, the curvilinear nature of the relationship
between GNPAs and Time is revealed immediately.
179
relationship. Instead, the sigmoid (i.e. ‘S-like’) shape of the curve is produced by the
semi parametric penalized spline regression procedure. The slopes of the fitted curves are
negative at some Years on the horizontal axis.
the relationship between GNPA values and time (year).
6.5.1.2. Comparing the Penalized spline, LOESS and OLS

Chart- 6.8.: Penalized Spline, LOESS and Linear fit Chart- 6.9.: Penalized Spline, LOESS and Linear fit
on GNPA – SBI on GNPA – SBT
on GNPA – SB on GNPA – UCO
180
on GNPA – CBI on GNPA – PNB
Chart- 6.14.: Penalized Spline, LOESS and Linear fit

on GNPA – PSB
As noted earlier, the ordinary least square (OLS) method sometime gives misleading
information. The above figures support our statement. The straight line (blue doted) is the OLS
fit; the curve lines are the loess line (red dashed) from a local linear regression with a span (α)
=0.6 and degree (λ) = 2 and a smooth line (black solid) from semi parametric regression. There is
a clear departure from linearity in these data—the linearity does not seem to fit well and both the
non parametric and semi parametric regression support nonlinear data modeling.
181
6.5.1.3. Diagnostics
6.5.1.3.1. Quantile Comparison plot of residuals
Chart- 6.15.: Quantile Comparison Plot on GNPA -SBI Chart- 6.16.: Quantile Comparison Plot on GNPA - SBT
Chart- 6.17.: Quantile Comparison Plot on GNPA -SB Chart- 6.18.: Quantile Comparison Plot on GNPA -UCO
Chart- 6.19.: Quantile Comparison Plot on GNPA -CBI Chart- 6.20.: Quantile Comparison Plot on GNPA -PNB
Chart- 6.21.: Quantile Comparison Plot on GNPA -PSBs
From the above figures it is evident that in case of CBI, PNB and PSB outliers are present. Rests
of the banks (SBI, SBT, SB and UCO) are free from outliers.
182
6.5.1.3.2. Residual vs Predictor plot
Chart- 6.22.: Residual vs Predictor Plot on GNPA – SBI Chart- 6.23.: Residual vs Predictor Plot on GNPA – SBT
Chart- 6.24.: Residual vs Predictor Plot on GNPA – SB Chart-6.25.: Residual vs Predictor Plot on GNPA – UCO
Chart- 6.26.: Residual vs Predictor Plot on GNPA – CBI Chart-6.27.: Residual vs Predictor Plot on GNPA – PNB
Chart- 6.28.: Residual vs Predictor Plot on GNPA – PSB
The dotted horizontal line is a visual baseline corresponding to residual value of zero. The above
‘Residual vs Predictor Plots’ (chart number 5.22 to 5.28) show that the residuals appear
randomly scattered around zero indicating that the penalized spline model describes the data
well. It can also be stated that the model provide adequate representation of the GNPA bi-variate
183
dataset. Looking at the plots above we can say that the plots exhibit the absence of relationship
between residual and predicted value and therefore it can be concluded that penalized spline
model exhibiting simple curve linear relationship is appropriate for our GNPA dataset.
6.5.1.3.3. Normal QQ plot

Chart- 6.29.: Normal Q-Q Plot on GNPA – SBI Chart- 6.30.: Normal Q-Q Plot on GNPA – SBT
Chart- 6.31.: Normal Q-Q Plot on GNPA – SB Chart- 6.32.: Normal Q-Q Plot on GNPA – UCO
Chart- 6.33.: Normal Q-Q Plot on GNPA – CBI Chart- 6.34.: Normal Q-Q Plot on GNPA – PNB
Chart- 6.35.: Normal Q-Q Plot on GNPA – PSB
184
The q-q plot has been done to test the normal distribution of the residuals for selected PSBs
along with PSB in aggregate. It may be observed that the residuals in respect of GNPA dataset of
all banks with the exception of SB follow approximately normal distribution.
6.5.1.4. Goodness of fit statistics

R2 value and significance of F value on GNPA dataset generated by R software are given below.
Table- 6.2.: Summary Statistics of precision criteria for selected PSBs on Parameter GNPA
Bank R2 F sig
SBI 0.98133 222.85120 0.00000
SBT 0.82629 31.98005 0.00001
SB 0.97226 79.26349 0.00000
UCO 0.91856 65.12458 0.00000
CBI 0.74170 18.18644 0.00024
PNB 0.90485 44.06718 0.00000
PSB 0.98208 282.53940 0.00000
From the above table it is apparent that R2 value is very high (above 0.90) for PSB in aggregate,
SBI, SB, UCO and PNB while it is moderately high (above 0.70 but less than 0.90)for SBT and
CBI. Hence we may say that the smooth curve summarizes nearly all of the total dispersion in
the dependent variable. The p-value of F test is less than 0.05 for all banks including PSB in
aggregate and therefore, we may conclude that there is no linear relationship between response
and predictor variables.
6.5.2. Result on GA Dataset

Now the GA dataset as dependent variable with time (year as 1996,1997, 1998….2012 taken as
1, 2, 3, ….17 respectively) as independent variable are processed through the SemiPar package
of R software and summary statistics for non linear components are given below.
Table- 6.3.: Summary Statistics for Penalized Spline models on Parameter GA
Bank df spar knots
SBI 4.979 1.326 3
SBT 4.896 2.291 3
SB 4.972 1.45 3
UCO 4.984 1.214 3
CBI 4.936 1.927 3
PNB 4.984 1.196 3
PSB 4.991 0.982 3
Note this includes 1 df for the intercept
185
6.5.2.1. The plots of the semi parametric models (smoothing by penalized spline)
The plots of the semi parametric models (smoothing by penalized spline) for six selected PSBs
along with PSB in aggregate are given below:
Chart- 6.36.: Penalized Spline on GA – SBI Chart- 6.37.: Penalized Spline on GA – SBT
Chart- 6.38.: Penalized Spline on GA – SB Chart- 6.39.: Penalized Spline on GA – UCO
Chart- 6.40.: Penalized Spline on GA – CBI Chart- 6.41.: Penalized Spline on GA – PNB
186
Chart- 6.42.: Penalized Spline on GA – PSB
The GA and time (year) dataset are plotted on vertical axis and horizontal axis respectively. The
findings from the above figures are detailed below:
1. The general diagonal orientations of the points suggest that GAs are highly correlated to
Year for all the banks.
2. The curves follow the central tendency of the Y variable’s (GA) values across the range
of the X variable (Time). In doing so, the curvilinear nature of the relationship between
GAs and Time is revealed immediately.
relationship. The parabolic shape of the growth curve is produced by the penalized spline
method. The slopes of the fitted curves are positive at some Years on the horizontal axis.
the relationship between GA values and time (year).
6.5.2.2. Comparing the Penalized spline, LOESS and OLS

Chart- 6.43.: Penalized Spline, LOESS and Linear fit on Chart- 6.44.: Penalized Spline, LOESS and Linear fit on
GA – SBI GA – SBT
187
GA – SB GA – UCO
GA – CBI GA – PNB
Chart- 6.49.: Penalized Spline, LOESS and Linear fit on

GA – PSB
As noted earlier, the ordinary least square (OLS) method sometime gives misleading
information. The above figures support our statement. The straight line (blue doted) is the OLS
fit; the curve lines are the loess line (red dashed) from a local linear regression with a span (α)
=0.6 and degree (λ) = 2 and a smooth line (black solid) from semi parametric regression. There is
a clear departure from linearity in these data—the linearity does not seem to fit well and both the
non parametric and semi parametric regression support nonlinear data modeling.
188
6.5.2.3. Diagnostics
6.5.2.3.1. Quantile Comparison plot of residuals
Chart- 6.50.: Quantile Comparison Plot on GA –SBI Chart- 6.51.: Quantile Comparison Plot on GA - SBT
Chart- 6.52.: Quantile Comparison Plot on GA – SB Chart- 6.53.: Quantile Comparison Plot on GA –UCO
Chart- 6.54.: Quantile Comparison Plot on GA –CBI Chart- 6.55.: Quantile Comparison Plot on GA –PNB
Chart- 6.56.: Quantile Comparison Plot on GA -PSBs
From the above figures it is evident that in case of SBT, UCO and PNB outliers are present.
Rests of the banks (SBI, SB, CBI and PSB) are free from outliers.
189
6.5.2.3.2. Residual vs Predictor plot
Chart- 6.57.: Residual vs Predictor Plot on GA – SBI Chart- 6.58.: Residual vs Predictor Plot on GA – SBT
Chart- 6.59.: Residual vs Predictor Plot on GA – SB Chart- 6.60.: Residual vs Predictor Plot on GA – UCO
Chart- 6.61.: Residual vs Predictor Plot on GA – CBI Chart- 6.62.: Residual vs Predictor Plot on GA – PNB
Chart- 6.63.: Residual vs Predictor Plot on GA – PSB
The dotted horizontal line is a visual baseline corresponding to residual value of zero. The above
‘Residual vs Predictor Plot’ (chart number 5.57 to 5.63) show that the residuals appear randomly
scattered around zero indicating that the penalized spline model describes the data well. It can
also be stated that the model provide adequate representation of the GA bi-variate dataset.
190
Looking at the plots above we can say that for our GA dataset, the plots exhibit the absence of
relationship between residual and predicted value and therefore it can be concluded that
penalized spline model exhibiting simple curvelinear relationship, is appropriate for our GA
dataset.
6.5.2.3.3. Normal QQ plot:

Chart- 6.64.: Normal Q-Q Plot on GA – SBI Chart- 6.65.: Normal Q-Q Plot on GA – SBT
Chart- 6.66.: Normal Q-Q Plot on GA – SB Chart- 6.67.: Normal Q-Q Plot on GA – UCO
Chart- 6.68.: Normal Q-Q Plot on GA – CBI Chart- 6.69.: Normal Q-Q Plot on GA – PNB
Chart- 6.70.: Normal Q-Q Plot on GA – PSB
191
The q-q plot has been done to test the normal distribution of the residuals for all the banks. It
may be observed that the residuals in respect of GA dataset of all banks follow approximately
normal distribution.
6.5.2.4. Goodness of fit statistics

R2 value and significance of F value on GA dataset generated by R software are given below.
Table- 6.4.: Summary Statistics of precision criteria for selected PSBs on Parameter GA
Bank R2 F sig
SBI 0.99909 1211.25800 0.00000
SBT 0.99816 480.63910 0.00000
SB 0.99857 675.70690 0.00000
UCO 0.99962 2552.73500 0.00000
CBI 0.99644 351.74390 0.00000
PNB 0.99953 2617.78500 0.00000
PSB 0.99970 3530.95000 0.00000
From the above table it is apparent that R2 value is high and hence we may say that the smooth
curve summarizes nearly all of the total dispersion in the dependent variable. The p-value of F
test is less than 0.05 and therefore we may conclude that there is no linear relationship between
response and predictor variables.
6.6. Aspects on Forecasting

The main purpose of constructing a time series model is to forecast. In this case forecasting has
been done by extrapolating models beyond the period over which they are estimated. The semi
parametric models (Penalised Spline Curve Fit) obtained based on our dataset from1996 to
2012, have been extrapolated beyond the sample period to get the forecasted values for the
years, 2013, 2014 and 2015 with respect to the parameters GNPA and GA are given below.
6.6.1. Parameter GNPA

Table- 6.5.: Forecast Values (Rs in Crores) for Parameter GNPA
Bank 2013 2014 2015
SBI 51323.02 67531.33 86941.44
SBT 1702.003 2186.049 2761.025
SB 3896.222 4847.518 6005.538
UCO 5121.764 6618.705 8420.167
192
CBI 8200.298 11178.383 14854.072
PNB 11234.27 15598.59 21036.09
PSB 146360.1 194670.2 253413.6
6.6.2. Parameter GA
Table- 6.6.: Forecast Values (Rs in Crores) for Parameter GA
Bank 2013 2014 2015
SBI 1050842 1217456 1401010
SBT 63477.19 73640.7 85067.55
SB 141952.8 159559.2 177969.6
UCO 137953.4 160665.5 186069.1
CBI 180639.1 211084 244680.5
PNB 360837.3 437842.7 526481.4
PSB 4205179 4909130 5693669
Forecasted Values of GNPA for three years (2013, 2014 and 2015) of all the selected PSBs
exhibit an alarming phenomenon of continuous upward rise, which puts question mark on the
wisdom and integrity of the top management in PSBs in India in handling credit portfolio. Such a
situation undoubtedly deserves immediate and serious attention on the part of the regulators to
relook into the practices of credit appraisal and monitoring of credit in PSBs in India.
6.7. Conclusion
In an effort to examine trends in NPA and develop a forecasting model for the purpose of
analysis and control of distress asset in banks various approaches and techniques are used by the
researchers and experts. In this chapter, penalized spline, a very popular semi parametric
technique has been fitted to GNPA and GA data sets for six selected PSBs along with PSB in
aggregate and it has been observed that these semi parametric models are very much relevant and
useful as it enhances the precision level of the models in respect of parameter GNPA in
comparison to parametric models as observed in chapter 4.
193
Chapter VII
Examination of Impact of NPAs on Strategic Banking Variables

in Selected PSBs in India
7.1. Introduction
7.3. Data source
7.4 Diagnostics Procedures - Durbin Watson Test, Shapiro-Wilk Test,

Tolerance & Variance Inflation Factor and Selection of Statistically
Significant Variables & Impact Analysis
7.6 Conclusion
194
7.1. Introduction
Existence of large amount of distressed assets in the balance sheet of a bank not only deprives it
of interest income but also exposes the organization to high rate of provisions out of income
generated by other performing assets and thus adversely affects bank’s profitability.
Performance of a bank receives a big blow, image get shattered because of NPA ratio of above
tolerance which leads to adverse comments by investors, auditors and inspectors. It de-motivates
the staff and creates investor apathy and shakes the customers’ loyalty. As a result, productivity
and other strategic banking variables also receive serious setback.
For the purpose of examining degree of association and relationship between variables including
NPA, statistical tool like simple correlation analysis has been used while regression analysis both
simple as well as multiple has been employed to examine the impact of NPA on strategic
banking variables including profitability and productivity. We have initially identified and
captured ratio dataset for variables like NPA Ratio, Profitability, Productivity along with few
other strategic banking variables namely Net Interest Margin, Credit Deposit Ratio, Capital
Adequacy Ratio and Ratio of Credit Growth etc in respect of six selected PSBs and PSBs in
aggregate. The data-sets, on which analytical studies have been performed in this chapter, are
presented in tables below:
Table- 7.1.: Gross NPA To Total Asset (GNPA/TA Ratio) Dataset
YEAR SBI SBT SB UCO CBI PNB PSB

1996 0.079265 0.129399 0.088178 0.127554 0.110491 0.083852 0.093596
1997 0.072896 0.081269 0.078552 0.117112 0.1021 0.072901 0.08575
1998 0.068203 0.107612 0.064415 0.102313 0.085306 0.065413 0.07688
1999 0.069942 0.088659 0.053146 0.09588 0.076183 0.065766 0.073708
2000 0.062999 0.069681 0.040449 0.081131 0.07572 0.062254 0.064752
2001 0.055012 0.056316 0.03879 0.054272 0.074592 0.058821 0.05748
2002 0.046631 0.046913 0.043305 0.048296 0.06807 0.060658 0.052045
2003 0.037274 0.035688 0.042911 0.042343 0.05896 0.06254 0.044503
2004 0.032305 0.030721 0.038941 0.037583 0.051062 0.04951 0.037499
2005 0.028701 0.024648 0.028843 0.028446 0.039551 0.032727 0.029526
2006 0.021751 0.020068 0.02661 0.02121 0.037294 0.023108 0.021854
2007 0.018847 0.015443 0.020747 0.022036 0.030544 0.022026 0.017239
2008 0.019919 0.013894 0.01801 0.020065 0.021579 0.01836 0.014492
2009 0.018479 0.011649 0.013434 0.015258 0.017003 0.012411 0.013006
195
2010 0.019351 0.0118 0.014904 0.013342 0.014846 0.011826 0.014071
2011 0.022229 0.012806 0.017585 0.020884 0.012179 0.012976 0.014646
2012 0.030972 0.018957 0.018777 0.023691 0.033045 0.020846 0.019838
Sources: GNPA/TA Ratio Dataset has been computed with the help of (1) GNPA dataset and (2)
Average Total Asset (TA) Dataset as given in Table- 4.1(chapter 4) and in Annexure 7A respectively.
Table- 7.2.: Profit before Tax To Total Asset (PBT/TA Ratio) Dataset

1996 0.006242 0.007888 0.001454 -0.01807 -0.00355 -0.00319 -0.00023
1997 0.016474 0.010287 0.004317 -0.01225 0.006651 0.010839 0.009755
1998 0.017784 0.014347 0.005174 -0.0061 0.00797 0.020774 0.012729
1999 0.007711 0.004892 0.007717 -0.00379 0.004945 0.01233 0.006553
2000 0.013188 0.008208 0.009852 0.002555 0.005301 0.010595 0.009366
2001 0.008925 0.009109 0.009282 0.001509 0.001185 0.009776 0.006966
2002 0.01114 0.011599 0.009102 0.006146 0.003618 0.011157 0.010897
2003 0.012612 0.015299 0.013797 0.007188 0.008092 0.014864 0.014424
2004 0.012677 0.017956 0.017259 0.011904 0.014437 0.018759 0.017348
2005 0.015028 0.014279 0.008725 0.00759 0.008896 0.017906 0.012688
2006 0.014478 0.011153 0.010285 0.003615 0.005223 0.014983 0.011856
2007 0.014374 0.013385 0.010454 0.005245 0.008257 0.014091 0.012858
2008 0.016198 0.012845 0.010239 0.005709 0.008019 0.018231 0.013794
2009 0.01681 0.019074 0.008731 0.00559 0.006789 0.020994 0.014689
2010 0.013795 0.017334 0.008719 0.007779 0.009359 0.021335 0.013612
2011 0.013126 0.013524 0.008697 0.006059 0.008439 0.01958 0.01351
2012 0.014445 0.010156 0.008421 0.006211 0.003447 0.016842 0.012026
Sources: PBT/TA Ratio Dataset has been computed with the help of (1) Profit before Tax (PBT) dataset
and (2) Average Total Asset (TA) Dataset as given in Annexure 7B and Annexure 7A respectively.
Table- 7.3.: Operating Profit per Employee (OPE Ratio) Dataset

1996 0.013145 0.011251 0.002667 -0.00064 0.004143 0.005729 0.0089
1997 0.014384 0.011592 0.003291 -0.00065 0.007514 0.009252 0.010659
1998 0.018489 0.020588 0.004941 0.000764 0.007279 0.017348 0.015849
1999 0.014459 0.013667 0.006955 0.002226 0.008053 0.013486 0.01347
2000 0.019592 0.015495 0.008591 0.006487 0.00899 0.013427 0.019217
2001 0.019139 0.018874 0.011573 0.007754 0.010072 0.017073 0.018651
2002 0.029352 0.02769 0.015542 0.019489 0.018949 0.027184 0.030309
2003 0.040818 0.038062 0.025239 0.025611 0.024691 0.042325 0.042332
2004 0.046692 0.058375 0.042387 0.032494 0.039944 0.056262 0.05814
196
2005 0.053643 0.068528 0.041085 0.034032 0.042509 0.049148 0.0546
2006 0.057312 0.057759 0.043407 0.034747 0.040107 0.060834 0.062942
2007 0.058949 0.061092 0.060514 0.044687 0.038805 0.066643 0.068497
2008 0.077563 0.061656 0.057012 0.046043 0.035245 0.072024 0.077242
2009 0.087028 0.099264 0.065874 0.050189 0.040432 0.104022 0.095589
2010 0.096453 0.087539 0.072159 0.067581 0.059118 0.129685 0.112652
2011 0.113904 0.130973 0.098601 0.122961 0.084871 0.172606 0.153405
2012 0.151644 0.116358 0.12721 0.128954 0.089778 0.183582 0.225001
Sources: OPE Ratio Dataset has been computed with the help of (1) Operating Profit Dataset and (2)
Number of Employee Dataset as given in Annexure 7I (1 to 7) and Annexure 7F respectively.
Table- 7.4.: Net Interest Margin to Total Asset (NIM to TA Ratio) Dataset

1996 0.035522 -0.00125 0.035846 0.025633 0.035228 0.035911 0.030206
1997 -0.00108 -0.00304 -0.01592 -0.01957 0.035446 -0.00932 0.011561
1998 -0.00587 -0.01222 -0.01876 0.022275 0.034735 -0.01409 0.007236
1999 0.030151 -0.02322 -0.01214 0.024944 0.0328 0.0365 0.018059
2000 0.028629 0.024221 0.033646 0.027146 -0.01683 0.031846 0.023731
2001 0.029048 0.029358 0.039477 0.027983 0.03323 0.034653 0.029095
2002 0.027345 0.027365 0.036919 0.026446 0.03095 0.033631 0.026391
2003 0.027536 0.029373 0.036552 0.027331 -0.01166 0.039228 0.02471
2004 0.028528 0.031722 0.035005 0.030354 -0.00884 0.038419 0.026284
2005 0.030415 0.033872 0.034098 0.028626 0.035832 0.035048 0.031396
2006 0.03268 0.031424 0.033225 0.026896 0.033069 0.034363 0.030412
2007 0.028885 0.02837 0.028597 0.024794 0.029385 0.033865 0.028154
2008 0.026411 0.023301 0.021106 0.018615 0.019391 0.030611 0.02356
2009 0.024743 0.027444 0.021868 0.017514 0.016357 0.030637 0.023631
2010 0.023449 0.025737 0.020948 0.018608 0.015374 0.031192 0.022982
2011 0.028549 0.026005 0.030066 0.025488 0.027085 0.034984 0.027862
2012 0.033793 0.023307 0.03 0.022624 0.023484 0.03207 0.027573
Sources: NIM to TA Ratio Dataset has been computed with the help of (1) Net Interest Margin (NIM)
Dataset (2) Average Total Asset (TA) Dataset as given in 7G (1 to 7) and Annexure 7A respectively.
Table- 7.5.: Credit Deposit Ratio (CDR) Ratio Dataset

1996 0.620627 0.58226 0.385927 0.359116 0.450727 0.467497 0.524618
1997 0.502043 0.528693 0.366387 0.315847 0.381342 0.411878 0.461725
1998 0.505784 0.487825 0.394762 0.387937 0.406224 0.417326 0.457427
1999 0.487215 0.441672 0.447755 0.382877 0.417622 0.43058 0.449619
197
2000 0.498432 0.503919 0.516005 0.415593 0.41609 0.435285 0.47387
2001 0.46778 0.552805 0.522663 0.468313 0.453621 0.499601 0.480735
2002 0.446505 0.552428 0.521385 0.476944 0.451606 0.535988 0.492157
2003 0.465206 0.575819 0.531802 0.508021 0.414514 0.530619 0.503886
2004 0.495682 0.564486 0.484842 0.525591 0.392789 0.537155 0.511211
2005 0.551358 0.615271 0.577372 0.559037 0.448997 0.585583 0.567395
2006 0.688866 0.725728 0.680031 0.685277 0.563808 0.623532 0.681849
2007 0.774558 0.794923 0.657104 0.724467 0.625728 0.690667 0.722368
2008 0.775521 0.795856 0.673011 0.689308 0.66169 0.717912 0.732477
2009 0.731064 0.775466 0.703561 0.686518 0.651192 0.737522 0.725758
2010 0.785849 0.755871 0.772534 0.673971 0.650084 0.748411 0.731214
2011 0.81025 0.79171 0.7875 0.681941 0.723284 0.773754 0.755738
2012 0.831295 0.774396 0.782698 0.750243 0.751952 0.77393 0.77535
Sources: CDR Ratio Dataset has been computed with the help of (1) Aggregate Credit Dataset and (2)
Aggregate Deposit Dataset as given in Annexure 7C is given in Annexure 7D respectively.
Table- 7.6.: Ratio of Credit Growth Dataset

1996 0.23275 0.0102 0.223569 -0.15724 0.139953 0.091498 0.180007
1997 -0.07102 0.08202 0.115685 -0.03049 -0.01261 0.000678 0.036716
1998 0.193013 0.066078 0.212239 0.408223 0.218791 0.156863 0.172328
1999 0.242158 0.049661 0.34321 0.109024 0.194735 0.196133 0.177299
2000 0.191138 0.341833 0.368921 0.226296 0.166097 0.17718 0.220106
2001 0.15788 0.246782 0.074538 0.321767 0.261795 0.356792 0.182608
2002 0.063528 0.162252 0.134835 0.269689 0.130307 0.22559 0.15379
2003 0.140324 0.23336 0.095447 0.243471 -0.0037 0.170463 0.140676
2004 0.146453 0.213918 0.266267 0.295378 0.03544 0.173923 0.155653
2005 0.28139 0.333791 0.294585 0.340789 0.24212 0.279261 0.283409
2006 0.293646 0.270607 0.364285 0.351532 0.374163 0.235292 0.371685
2007 0.288523 0.30549 0.416939 0.257142 0.381821 0.294385 0.302145
2008 0.235467 0.142377 0.239606 0.172232 0.40934 0.237121 0.247721
2009 0.301691 0.158683 0.272927 0.249119 0.171044 0.294569 0.25687
2010 0.164812 0.179742 0.108841 0.199127 0.232798 0.20619 0.185101
2011 0.197504 0.197158 0.181133 0.200792 0.230984 0.297455 0.234256
2012 0.1465 0.202017 0.157688 0.166237 0.137116 0.21341 0.173671
Sources: Ratio of Credit Growth Dataset has been computed with the help of (1) Incremental Credit and
(2) Aggregate Credit Dataset as given in Annexure 7C.
198
Table- 7.7.: Capital Adequacy Ratio Dataset

1996 0.116 0.094 0.0842 0.0783 0.0263 0.0823 0.0872
1997 0.1217 0.0817 0.088 0.0316 0.0941 0.0915 0.1
1998 0.1458 0.1148 0.105 0.0907 0.104 0.0881 0.1153
1999 0.1251 0.1027 0.0957 0.0963 0.1188 0.1079 0.112
2000 0.1149 0.1109 0 0.0915 0.1118 0.1031 0.1066
2001 0.1279 0.1179 0.1172 0.0905 0.1002 0.1024 0.1118
2002 0.1335 0.1254 0.1212 0.0964 0.0958 0.107 0.118
2003 0.135 0.113 0.1103 0.1004 0.1051 0.1202 0.127
2004 0.1353 0.1136 0.1149 0.1188 0.1243 0.131 0.132
2005 0.1245 0.1105 0.107 0.1126 0.1215 0.1478 0.129
2006 0.1188 0.1115 0.1173 0.1112 0.1103 0.1195 0.122
2007 0.1234 0.1168 0.1174 0.1156 0.104 0.1229 0.124
2008 0.1354 0.1268 0.1122 0.1009 0.1042 0.1296 0.125
2009 0.1297 0.1213 0.1137 0.0975 0.1175 0.1259 0.123
2010 0.12 0.1189 0.112 0.1135 0.1081 0.1297 0.121
2011 0.1069 0.1082 0.112 0.1187 0.1074 0.1176 0.1178
2012 0.1205 0.1118 0.1081 0.1103 0.1196 0.1159 0.118
Sources: (1) www.rbi.org.in, (2) Prowess Database, and (3) Annual Reports of selected banks
Table- 7.8.: (Net Interest Margin – Burden) to TA Ratio Dataset

1996 0.006242 0.007888 0.001454 -0.01807 -0.00355 -0.00319 -0.00043
1997 0.008967 0.005576 0.004317 -0.01225 0.006397 0.007146 0.006115
1998 0.011074 0.007481 0.004718 -0.0061 0.006394 0.013394 0.00856
1999 0.005111 0.004332 0.007054 -0.00379 0.00394 0.008739 0.004324
2000 0.008477 0.005708 0.008792 0.0018 0.004013 0.008126 0.006276
2001 0.005559 0.007244 0.008481 0.001395 0.001065 0.007882 0.004525
2002 0.007322 0.007797 0.008352 0.005962 0.003293 0.00824 0.007489
2003 0.008569 0.009609 0.010398 0.006429 0.005553 0.010576 0.010095
2004 0.009388 0.011355 0.010633 0.011064 0.010208 0.011754 0.012019
2005 0.009919 0.009341 0.008111 0.007026 0.005392 0.012335 0.00959
2006 0.009238 0.008511 0.009477 0.003378 0.003577 0.010598 0.008734
2007 0.008561 0.009329 0.009524 0.004625 0.005914 0.010004 0.009013
2008 0.010441 0.009398 0.008636 0.005006 0.005052 0.011332 0.009729
2009 0.010812 0.013007 0.007691 0.005528 0.004193 0.013861 0.010152
2010 0.00908 0.012578 0.00604 0.008104 0.006392 0.014369 0.009565
2011 0.007254 0.011158 0.007091 0.00601 0.00637 0.013136 0.00925
199
2012 0.009139 0.0065 0.007748 0.006428 0.002422 0.011695 0.008731
Sources: (NIM – Burden) to TA Ratio Dataset has been computed with the help of (1) NIM and Burden
Dataset and (2) Average Total Asset (TA) Dataset as given in Annexure 7G (1to7) and Annexure 7A
respectively.
Table- 7.9.: Priority Sector Advance To Total Advance (PSADV/TADV) Ratio Dataset

1996 0.249814 0.357336 0.274477 0.323734 0.312421 0.32269 0.29939
1997 0.304907 0.353914 0.275238 0.399476 0.349768 0.345188 0.332042
1998 0.294445 0.357508 0.271227 0.333525 0.34117 0.359385 0.331219
1999 0.28036 0.368342 0.315214 0.292448 0.324432 0.375619 0.327257
2000 0.263782 0.332331 0.256578 0.287426 0.389554 0.402697 0.310798
2001 0.265457 0.315045 0.290511 0.272502 0.380711 0.387144 0.306048
2002 0.261505 0.304003 0.275542 0.310453 0.385356 0.391083 0.308077
2003 0.254879 0.335999 0.309194 0.304219 0.413208 0.398568 0.323179
2004 0.270404 0.372192 0.325733 0.317531 0.451967 0.439061 0.345945
2005 0.285929 0.383566 0.362672 0.362092 0.447828 0.467922 0.359466
2006 0.305645 0.411025 0.369887 0.34988 0.434847 0.456849 0.353282
2007 0.302416 0.395392 0.328109 0.346311 0.391207 0.378146 0.340212
2008 0.286084 0.417313 0.321312 0.330255 0.328758 0.386747 0.32385
2009 0.264768 0.399204 0.323715 0.312859 0.313865 0.318109 0.301594
2010 0.269923 0.36638 0.343843 0.29525 0.321427 0.356994 0.307951
2011 0.306055 0.374729 0.301323 0.243156 0.312271 0.324803 0.305545
2012 0.288362 0.362129 0.294621 0.250342 0.261145 0.313277 0.28816
Sources: (PSADV/TADV) Ratio Dataset has been computed with the help of (1) Priority Sector Advance
Dataset and (2) Aggregate Credit (Advances) Dataset as given in Annexure 7E and Annexure 7C
respectively.
The detailed reviews are included in Chapter-1. However, the references of the base papers in
connection with impact analysis are given below:
Sharma (2005), Gopalakrishnan (2006), Bodla and Verma (2006), Vyas, Singh and Kashif
(2007), Pal and Malik (2007), Singla (2008), Agnani (2010), Thiagarajan, Ayyappan,
Ramachandran and Sakthivadivel (2011), Nandy (2011), Yadav (2011), Jayasree and Radhika
(2011), Rajput, Gupta and Chauhan (2012).
200
7.3. Data Source
Narrated in Section 4.3 (Chapter 4)
7.4. Diagnostics Procedures - Durbin Watson Test, Shapiro and Wilk Test, Tolerance &
Variance Inflation Factor and Selection of Statistically Significant Variables & Impact
Analysis
There are two principal assumptions which justify the use of linear regression models for
purposes of prediction and decision making:
(i) Independence of the errors, ie, there is no serial correlation between the errors of the
variables. Here Durbin Watson (DW) test statistics is employed
(ii) Normality of the error distribution; The assumption of normality is tested by Shapiro-Wilk
(SW) test.
If any of these assumptions is violated (i.e., if there is non normality and/or, serial correlation),
then the forecasts, confidence intervals, and economic insights yielded by a regression model
may be (at best) inefficient or (at worst) seriously biased or misleading. In that case the best
option is to redefine the models.
7.4.1. Durbin Watson Test
Narrated in Section 4.4.2 (Chapter 4)
Durbin Watson (DW) tests whether adjacent residuals are correlated ie, whether there is any
serial correlation between the errors of the variables. The test statistics can vary between 0 and 4.
Applying thumb rule a value of 2 implies that the residuals are uncorrelated. A value greater than
2 indicates a negative correlation between adjacent residuals whereas a value below 2 indicates a
positive correlation. As a very conservative rule of thumb, Field (2009) suggests that values less
than 1 or greater than 3 are definitely cause for concern, which has been applied in this chapter.
7.4.2. Shapiro and Wilk Test
Narrated in Section 4.4.4 (Chapter 4)
201
7.4.3. Tolerance and Variance Inflation Factor
Multi-co linearity is a high degree of correlation (linear dependency) among several independent
variables. It commonly occurs when a large number of independent variables are incorporated in
a regression model. It is because some of them may measure the same concepts or phenomena.
Only existence of multi-co linearity is not a violation of the OLS assumption. However, a perfect
multi-co linearity violates the assumption that X matrix is full ranked, making OLS impossible.
When a model is not full ranked, that is, the inverse of X cannot be defined, there can be an
infinite number of least squares solutions. For examining multi-co linearity, a judgment by
checking related statistics, such as tolerance value or variance inflation factor (VIF) is generally
attempted.
The tolerance and VIF is calculated using the formula (1 - R2k) and 1/ (1 - R2k), R2k is the
coefficient of determination for regression of the ith independent variable on all the other
independent variables. In case of linear relationship, all the variables involved should have a
small tolerance. Multi-co linearity becomes an issue in case the tolerance value is low but the
standard errors are large. Since, VIF = 1/ Tolerance, its value is always 1 or more. For
determining the presence of multi- co linearity the VIF value is examined though there is no set
norm as regards acceptable VIF value to examine multi-co linearity. Some argue that a tolerance
value less than .1 or VIF value more than 10 often indicates multi-co linearity. Others insist that
magnitude of model’s R2 be considered determining significance of multi-co linearity. We will
consider the first one only. In case of high multi- co linearity the VIF value will be high and
there will be instability of the b and beta coefficients.
7.4.4. Selection of Statistically Significant Variables
For the purpose of identifying strategic banking variables, earlier works which have used
different banking variables, provide a useful basis of our identification with respect to the totality
of such banking variables, a brief outline of which is given below.
202
Table- 7.10.: List of Statistically Significant Variables from Earlier Studies
Sl. Researcher/ Objective of the Variables considered in the Study

No. Author Study
1 Singla (2008), To analyze the 1. Interest Coverage Ratio
financial 2. Net Profit Ratio
performance of 3. Ratio of Debt Equity
banks in India
4. Return on Net Worth
5. Capital Adequacy Ratio
6. Interest Income in Proportion to Working fund
7. Operating Profit Ratio
8. NPA to Net Advances Ratio
9. Return on Investment
2 Pal and Malik To measure the 1. Return on Equity
(2007) performance of 2. Return on Asset
commercial 3. Asset Utilization
banks in India
4. Expense Ratio
5. Net Interest Margin
6. Efficiency Ratio
7. Earning Assets to Total Loan
8. Business per Employee
9. Net NPA to Net Advances
3 Vallabh, Bhatia To analyse the 1. Priority Sector Loans to Total Advances
and Mishra impact of NPAs 2. Credit Deposit Ratio
(2007) on the 3. Capital Adequacy Ratio
performance of
4. Liquidity Risk
financial health
5. Gross NPA to Gross Advances
of a bank.
2. Operating Expenses per Employee
3. Profit per Branch
4. Business per Branch
To evaluate 5. Establishment expenses per Employee
performance of
4. Rao (2007) 6. Profit per Employee
Commercial
7. Return on Assets
Banks
8. Return on Equity
9. Net Interest Margin as a percentage of Working fund
10. Non Interest Income as percentage of Total Income
11. Credit Deposit Ratio
203
No. Author Study
5. Bodla and others To study 1. Spread
(2006) determinants of 2. Net Interest Income
profitability of 3. Credit Deposit Ratio
banks 4. Net NPA to Net Advances
5. Provisions and Contingencies
6. Establishment Expenditure
8. Net Profit per Employee
6. Sharma (2005) To examine 1. GNPA as % of Total Advances
impact of NPA 2. Difference between Spread and Burden
on strategic 3. Priority Sector Advances as % of Total Advances
banking variables 4. Credit Deposit Ratio
5. Establishment Expenses as % of Total Expenses
6. Fixed Deposit as a % of Total Deposit
7. Capital Adequacy Ratio
9. Operating Profit per Employee
7. ICRA (2004) To examine the A. Operational Parameters
bench mark of 1. Return on Assets
performance of 2. Net Interest Margin
Indian Banks
3. Operating Expense Ratio
4. Asset Quality (GNPA % & NNPA%)
B. Efficiency Parameters
2. Business per Branch
3. Expenses per Employee
4. Expenses per Branch
5. Profitability per Employee
6. Profitability per Branch
8. Kannan, and To examine 1. Net Profit to Total Assets
Swamy (2001) profitability of 2. Log of Assets,
Indian Banks 3. Wages as percentage of Total Assets
4. Net Non Performing Assets (NNPA)
5. Net Interest Income as percentage of Total Assets
6. Share of contingent liability to Total Liability
7. Capital to Risk weighted Asset Ratio (CRAR)
204
No. Author Study
9. Bhatia and To study factors 1. Priority Sector Advances
Verma (1998-99) determining the 2. Credit Deposit Ratio
profitability of 3. Composition of Bank Deposits
PSBs in India
4. Establishment Expenses
5. Spread
6. Burden
In addition to the variables stated above, based on bankers’ perception, gathered from personal
visit to bank officials about significance of a variable in our study, a few new variables have also
been picked up. The variables identified for pair-wise correlation is given below along with their
significance.
Table- 7.11.: List of Statistically Significant Variables Identified for our Study
Sl Variable Name Significance of the Variable
No
1 GNPA to Average GNPA is the aggregate loan assets that are classified as NPA as on the
Total Asset balance sheet date. The ratio of GNPA to average total assets of a bank
(Representative of indicates the quality of credit portfolio. High ratio indicates low quality
NPA Parameter) of credit portfolio of the bank and vice versa.
2 Profit before Tax to One of the most significant ratio to demonstrate profitability of an
Average Total Asset organization, it measures how well bank resources are being used to
(Representative of generate income. It is often calculated as Net profit before tax divided by
Profitability average total assets. Higher the ratio, better is the company’s status in
Parameter) terms of profitability index and vice versa.
3 Operating Profit Per Operating profit is calculated as difference between operating income
Employee and operating expenses. Operating profit divided by the number of
(Representative of employee gives us the desired ratio. Higher the ratio for a company in
Productivity comparison to its competitors gives an indication of better utilization of
Parameter) labour force, ie, better productivity in the organization and viceversa.
5 Net Interest Margin Net interest margin (NIM) refers to is the difference between the interest
(NIM) to Average income and interest expenses. The ratio shows the management’s ability
Total Asset to manage the interest revenue and interest costs judiciously and
(Representative of therefore, the interest risk aspect in banking. As interest rate and asset
Interest Margin quality changes, bank’s net interest margin or spread will also change.
Parameter) Higher the margin, greater the profitability for a banking company and
vice versa.
6 Credit Deposit Ratio Aggregate credit divided by aggregate deposit gives us the desired ratio.
The ratio indicates extension of a credit by a bank per unit of deposit.
This ratio depends heavily on prevailing economic conditions,
government policy and market interest rate along with number of bank
specific factors. Higher the ratio implies greater amount of credit is
pumped in the economy and vice versa.
205
Sl Variable Name Significance of the Variable
No
7 Ratio of Credit Increase / decrease in credit or advance has a direct bearing with NPA in
Growth the banking industry. This ratio is calculated as change in total credit to
total credit. Higher the ratio implies greater amount of credit is released
by the banking system and increase in the credit appetite in the economy
and vice versa.
8 Capital Adequacy Capital adequacy reflects the overall financial condition of a bank in
Ratio terms of capital contributed by the owners. In other words, it indicates
whether the bank has enough capital to absorb unexpected losses. Capital
adequacy ratio is calculated as Capital (Tier1+Tier2+Tier3) / risk
weighted assets. The higher the CAR, the stronger the overall financial
condition of the bank as it ensures higher safety against bankruptcy and
vice versa..
9 Priority Sector The ratio refers to advances in selected sectors of the economy where
Advances to Total some benefits like interest subsidy is available to the borrowers and
Advances therefore greater the allocation of credit in priority sector, lesser will be
the profitability of the bank.
10 Difference between Burden refers to difference between Non-interest operating expenditure
NIM and Burden to and Non-interest operating income. Difference between NIM and Burden
Average Total Asset implies net earning before tax and provision. The ratio signifies net
income generating potential of the assets of a bank. Higher the ratio
implies better management of the NIM and Burden and vice versa.
11 Time 1,2,3 ……. 17 years
For the purpose of examining degree of association and relationship between variables identified,
statistical tool like simple pair wise correlation co-efficient has been computed and appropriate
tests have been employed to judge the significance status of these co-efficient. This has enabled
us to select the appropriate variables to be considered for further impact study. In this we have
computed Pearson Correlation Coefficient and its corresponding p value (Sig. 2 tailed).
7.4.4.1. Correlation Analysis

Pearson Correlation Coefficient measures the strength and direction of the linear relationship
between two variables. The correlation coefficient can range from -1 to +1. -1 indicates a perfect
negative correlation, +1 indicates a perfect positive correlation, and 0 indicates no correlation at
all. By definition we know that any variable correlated with itself has a correlation of 1. If the
Sig (significance) value of the correlation coefficient is less than .05, then it can be said that the
correlation is statistically significant. However, we have also selected a few variables for our
further study which have Sig value of marginally higher than .05. Before going to see the
outcome of the correlation analysis, let us also explore the nature of relationship as perceived in
the industry between (i) the dependent variable (i.e. NPA) and (ii) the dependent variable (i.e.
206
Profitability) and any one of the independent variables including NPA considered for the present
study as under :
Table- 7.12.: Expected relationship between NPA and selected strategic banking variables
Name of the Banking Variable Abbreviation for the variable Expected
used Relation
Profit before Tax to Average Total Asset PBT/TA (-)ve
Operating Profit Per Employee OPE (-)ve
Credit Deposit Ratio CDR (-)ve
Credit Growth CDRGROWTH (-)ve
Net Interest Margin to Average Total Asset NIM/TA (-)ve
Capital Adequacy Ratio CAR (-)ve
Table- 7.13.: Expected relationship between Profitability and selected strategic banking variables
Name of the Banking Variable Abbreviation for the variable Expected
used Relation
Gross Non Performing Asset to Average Total GNPA/TA (-)ve
Asset
Operating Profit Per Employee OPE (+)ve
Credit Deposit Ratio CDR (+)ve
Credit Growth CDRGROWTH (+)ve
Net of NIM and Burden to Average Total Asset (NIM-BURDEN)/TA (+)ve
Priority Sector Advance to Average Total Asset PSADV/TADV (+)ve
7.4.5. Impact Analysis of Statistically Significant Variables
For the purpose of examining impact of NPA (GNPA/TA) over the statistically significant
banking variables like Profitability (PBT/TA) and Productivity (OPE) in addition to variables
like NIM/TA Ratio, Credit-Deposit Ratio, Capital Adequacy Ratio, Ratio of Credit Growth etc
have been regressed.
(ii) There are two principal assumptions which justify the use of linear regression models for the
purpose of examining the impact of independent variable(s) on dependent variable. They are
independence of the errors and normality of the error distribution for which Durbin Watson
(DW) test and Shapiro-Wilk (SW) test have been employed.
First of all, impact of NPAs on the profitability of banks has been attempted with the help of
simple linear regression. For few banks, which could not qualify linearity requirement as stated
207
above, we have examined alternate conventional models like Quadratic, Cubic, Logarithmic,
Exponential and Growth and applied the model which suits the most with the given dataset. To
examine the impact of time along with NPA, multiple regression model is invoked and
diagnosed for auto correlation and multi co linearity. Same process has been followed for other
variables as stated above.
However, apart from NPA, as there are number of variables affect profitability of a bank, it is not
appropriate to judge profitability with NPA alone, rather few other strategic banking along with
NPA have been considered through multiple regression model which consist of determining a
regression equation explaining relationship between dependent variables (profitability) with the
independent ones. Equation has been estimated through ‘ENTER’ method of SPSS. This simply
means that we conducted a full entry regression in which all predictors were entered into the
model and got the summary of the models. Significance of the values has been checked at five
percent level. Different statistics of the model (e.g. R2, F statistics and Sig of F value) have been
studied to finally determine the most appropriate model. The p-value (Sig) of the F-stat is
compared to the alpha level (typically 0.05) and, if smaller, it can be concluded that the
independent variables reliably predict the dependent variable. To analyze the model coefficients
we examine t-stat and its sig for each independent variable. Coefficients having p-values less
than alpha (typically 0.05) are considered as statistically significant. To diagnose the problem of
the multi co-linearity in multiple regressions, the value of tolerance factor (TOL) along with
variance inflating factor (VIF) have also been computed and compared with standard.
7.5.1. Selection of Statistically Significant Strategic Banking Variables
To examine the status of inter relationship existing among the set of relevant financial variables
including NPA, bivariate correlation analysis have been employed to study whether the
relationship between two variables (NPA and any of the strategic banking variables) of our data
is linear. In the process we will finally select statistically significant strategic banking variables
for our further study.
208
7.5.1.1. Selection of statistically significant strategic banking variables having association
with NPA
7.5.1.1.1. Hypotheses
We know that if NPA (GNPA/TA) increases, profitability (PBT/TA), productivity (OPE),

NIM/TA, CDR, CDRGROWTH and CAR decreases. The statement indicates that existence of a
negative correlation between NPA and strategic banking variable as stated above (pair wise).
Therefore our hypotheses become:
H10: No linear correlation exists between the variables (GNPA/TA and PBT/TA).
H11: A negative linear correlation exists between the variables (GNPA/TA and PBT/TA).
H20: No linear correlation exists between the variables (GNPA/TA and OPE).
H21: A negative linear correlation exists between the variables (GNPA/TA and OPE).
H30: No linear correlation exists between the variables (GNPA/TA and NIM/TA).
H31: A negative linear correlation exists between the variables (GNPA/TA and NIM/TA).
H40: No linear correlation exists between the variables (GNPA/TA and CDR).
H41: A negative linear correlation exists between the variables (GNPA/TA and CDR).
H50: No linear correlation exists between the variables (GNPA/TA and CDRGROWTH).
H51: A negative linear correlation exists between the variables (GNPA/TA and CDRGROWTH).
H60: No linear correlation exists between the variables (GNPA/TA and CAR).
H61: A negative linear correlation exists between the variables (GNPA/TA and CAR).
A pair wise correlation between NPA (GNPA/TA) and profitability (PBT/TA), productivity
(OPE) others (NIM/TA, CDR, CDRGROWTH, CAR) for 6 selected PSBs along with PSBs in
aggregate have been calculated and presented in a table format as under:
209
Table- 7.14.: Correlation between NPA (GNPA/TA) and Selected Banking Variables

GNPA/T GNPA/T GNPA/T GNPA/T GNPA/T GNPA/T GNPA/T
A A A A A A A
PBT/TA Pearson -0.447 -.582* -.553* -.890** -.512* -.695** -.705**
Correlation
Sig. (2- 0.072 0.014 0.021 0.000 0.036 0.002 0.002
tailed)
N 17 17 17 17 17 17 17
** ** ** ** ** **
OPE Pearson -.751 -.786 -.776 -.691 -.816 -.826 -.738**
Correlation
Sig. (2- 0.001 0.000 0.000 0.002 0.000 0.0000 0.001
tailed)
N 17 17 17 17 17 17 17
**
NIM/TA Pearson -0.368 -.796 -0.415 -0.348 0.14 -0.302 -0.469
Correlation
Sig. (2- 0.146 0.000 0.097 0.171 0.591 0.239 0.058
tailed)
N 17 17 17 17 17 17 17
** ** ** ** ** **
CDR Pearson -.668 -.775 -.914 -.918 -.810 -.940 -.864**
Correlation
Sig. (2- 0.003 0.000 0.000 0.000 0.000 0.000 0.000
tailed)
N 17 17 17 17 17 17 17
** * **
CDRGR Pearson -0.413 -.610 -0.159 -.537 -0.475 -.611 -.597*
OWTH Correlation
Sig. (2- 0.099 0.009 0.541 0.026 0.054 0.009 0.011
tailed)
N 17 17 17 17 17 17 17
* ** * **
CAR Pearson 0.055 -.577 -0.306 -.760 -.563 -.734 -.766**
Correlation
Sig. (2- 0.835 0.015 0.232 0.000 0.019 0.001 0.000
tailed)
N 17 17 17 17 17 17 17
*. Correlation is significant at the 0.05 level (2-tailed).
**. Correlation is significant at the 0.01 level (2-tailed).
7.5.1.1.2. Statistical Interpretation

1. From the above table it is clear that the profitability (PBT/TA) of the banks SBT, SB,
UCO, CBI, PNB and PSB have strong negative linear correlations over NPA
(GNPA/TA). We also get a support in favour of our statement by a 2-tailed significance
test at 95% confidence level, because all p values<0.05. The profitability (PBT/TA) of
210
the other bank like SBI also has negative linear correlations over NPA (GNPA/TA). But,
we find statistical significance at 90% confidence level to nullify null hypothesis. So, we
can say that the relationship between the two variables of the bank SBI is may not be due
to by chance at 90% confidence level.
2. From the above table it is clear that the productivity (OPE) of the banks SBI, SBT, SB,
UCO, CBI, PNB and PSB have strong negative linear correlations over NPA
(GNPA/TA). We also get a support in favour of our statement by a 2-tailed significance
test at 95% confidence level, because all p values<0.05, sufficient to reject the null
hypothesis.
3. From the above table it is clear that the NIM/TA of the bank SBT has a strong negative
linear correlation over NPA (GNPA/TA). We also get a support in favour of our
statement by a 2-tailed significance test at 95% confidence level for SBT, because p
values<0.05. NIM/TA of the bank SB and PSB have negative linear correlation over
GNPAs. We also get a support in favour of our statement by a 2-tailed significance test at
90% confidence level for SB and PSB because p values<0.10. So, we can say that the
relationship between the two variables of the two banks SB and PSB are may not be due
to by chance at 90% confidence level. The NIM/TA of the other banks like SBI, UCO
and PNB also have negative linear correlations over NPA (GNPA/TA), while NIM/TA of
CBI has positive linear correlations over NPA (GNPA/TA). But, we found no statistical
significance even at 90% confidence level to nullify null hypothesis. So, we can say that
the relationship between these two variables of the four banks is may be due to by chance
(valid for this sample only).
4. From the above table it is clear that the CDR of the banks SBI, SBT, SB, UCO, CBI,
PNB and PSB have strong negative linear correlations over NPA (GNPA/TA). We also
get a support in favour of our statement by a 2-tailed significance test at 95% confidence
level, because all p values<0.05, sufficient to reject the null hypothesis.
5. From the above table it is clear that the CREGROWTH of the banks SBT, UCO, PNB
and PSB has strong negative linear correlations over GNPAs. We also get a support in
favour of our statement by a 2-tailed significance test at 95% confidence level, because
all p values<0.05. The CREGROWTH of the bank SBI and CBI also have negative linear
correlation over NPA (GNPA/TA). We also get a support in favour of our statement by a
211
2-tailed significance test at 90% confidence level for SBI and CBI because p
values<0.10. But, we found no statistical significance at 95% confidence level to nullify
null hypothesis. So, we can say that the relationship between the two variables of the two
banks SBI and CBI are may not be due to by chance at 90% confidence level. But the
CREGROWTH of the bank SB also has negative linear correlation over NPA
(GNPA/TA). However, we do not get a support in favour of our statement by a 2-tailed
significance test at 90% confidence level because p values>0.10. So, we can say that the
relationship between the two variables of the bank SB is may be due to by chance at 90%
confidence level (valid for this sample only).
6. From the above table it is clear that the CAR of the banks SBT, UCO, CBI, PNB and
PSB has strong negative linear correlations over NPA (GNPA/TA). We also get a
support in favour of our statement by a 2-tailed significance test at 95% confidence level,
because all p values<0.05. The CAR of the other banks like SB also has negative linear
correlations over NPA (GNPA/TA) and SBI has positive linear correlations over NPA
(GNPA/TA). But, we do not find any statistical significance at 95% confidence level to
nullify null hypothesis. So, we can say that the relationship between these two variables
of the two banks is may be due to by chance (valid for this sample only).
From the above correlation analysis, following statistically significant strategic banking variables
having association with NPA (GNPA/TA) with level of sig of 5% or marginally higher than 5%
have been finally selected.
Table- 7.15.: Selection of Statistically Significant Strategic Banking Variables having Association with
NPA (GNPA/TA)
Bank Statistically significant banking variable
SBI PBT/TA, OPE, CDR
SBT PBT/TA, OPE, NIM/TA, CDR, CDRGROWTH, CAR
SB PBT/TA, OPE, CDR
UCO PBT/TA, OPE, CDR, CDRGROWTH, CAR
CBI PBT/TA, OPE, CDR, CDRGROWTH , CAR
PNB PBT/TA, OPE, CDR, CDRGROWTH, CAR
PSB PBT/TA, OPE, NIM/TA, CDR, CDRGROWTH, CAR
212
7.5.1.2. Selection of statistically significant strategic banking variables having association
with Profitability
7.5.1.2.1. Hypotheses
We know that if NPA (GNPA/TA) decreases, profitability (PBT/TA) increases. The statement
indicates existence of a negative correlation between profitability (PBT/TA) and NPA
(GNPA/TA). Again, if productivity (OPE), (NIM–BURDEN)/TA, CDR, CDRGROWTH and
PSADV/TADV increases, profitability (PBT/TA) increases. The statement indicates that
existence of a positive correlation between strategic banking variable as stated above (pair wise).
Therefore our hypotheses become:
H70: No linear correlation exists between the variables (PBT/TA and GNPA/TA).
H71: A negative linear correlation exists between the variables (PBT/TA and GNPA/TA).
H80: No linear correlation exists between the variables (PBT/TA and OPE).
H81: A positive linear correlation exists between the variables (PBT/TA and OPE).
H90: No linear correlation exists between the variables (PBT/TA and (NIM-BURDEN)/TA).
H91:A positive linear correlation exists between the variables (PBT/TA and (NIM-
BURDEN)/TA).
H100: No linear correlation exists between the variables (PBT/TA and CDR).
H101: A positive linear correlation exists between the variables (PBT/TA and CDR)
H110: No linear correlation exists between the variables (PBT/TA and CDRGROWTH).
H111: A positive linear correlation exists between the variables (PBT/TA and CDRGROWTH)
H120: No linear correlation exists between the variables (PBT/TA and PSADV/TADV).
H121: A positive linear correlation exists between the variables (PBT/TA and PSADV/TADV)
Pair wise correlation between profitability (PBT/TA) and NPA (GNPA/TA), productivity (OPE),
and others [(NIM - BURDEN)/TA, CDR, CDRGROWTH, PSADV/TADV)] for 6 selected banks
and PSB in aggregate are calculated and presented in a table format as under:
213
Table- 7.16.: Correlation between Profitability (PBT/TA) and Selected Banking Variables

PBT/T PBT/T PBT/T PBT/T PBT/T PBT/T PBT/T
A A A A A A A
GNPA/TA Pearson -.447 -.582* -.553* -.890** -.512* -.695** -.705**

Correlation
Sig. (2-tailed) .072 .014 .021 .000 .036 .002 .002
N 17 17 17 17 17 17 17
OPE Pearson .370 .531* .254 .528* .337 .570* .437
Correlation
Sig. (2-tailed) .143 .028 .326 .029 .185 .017 .079
N 17 17 17 17 17 17 17
(NIM- Pearson .913 .891 .920 .999 .973 .986 .994
BURDEN)/TA Correlation
Sig. (2-tailed) .000 .000 .000 .000 .000 .000 .000
N 17 17 17 17 17 17 17
CDR Pearson .319 .447 .298 .691** .055 .537 .403
Correlation
Sig. (2-tailed) .212 .071 .245 .002 .834 .026 .108
N 17 17 17 17 17 17 17
CDRGROWT Pearson -.019 .216 .126 .675** -.036 .374 .175
H Correlation
Sig. (2-tailed) .944 .404 .630 .003 .892 .139 .502
N 17 17 17 17 17 17 17
PSADV/TAD Pearson .535* .315 .410 -.331** .385 .110 .318
V Correlation
Sig. (2-tailed) .027 .218 .103 .195 .126 .676 .214
N 17 17 17 17 17 17 17
*. Correlation is significant at the 0.05 level (2-tailed).
**. Correlation is significant at the 0.01 level (2-tailed).
7.5.1.2.2. Statistical Interpretation

1. From the above table it is clear that the NPA (GNPA/TA) of the banks SBT, SB, UCO,
CBI, PNB and PSB has strong negative linear correlations over profitability (PBT/TA).
We also get a support in favour of our statement by a 2-tailed significance test at 95%
confidence level, because all p values<0.05. The profitability (PBT/TA) of the other bank
like SBI also has negative linear correlations over NPA (GNPA/TA). But, we find
statistical significance at 90% confidence level to nullify null hypothesis. So, we can say
214
that the relationship between the two variables of the bank SBI is may not be due to by
chance at 90% confidence level.
2. From the above table it is clear that the productivity (OPE) of the banks SBT, UCO and
PNB has positive linear correlations over profitability (PBT/TA). We also get a support
in favour of our statement by a 2-tailed significance test at 95% confidence level, because
all p values<0.05. The productivity (OPE) of the other bank like PSB also has positive
linear correlations over profitability (PBT/TA). But, we find statistical significance at
90% confidence level to nullify null hypothesis. So, we can say that the relationship
between the two variables of PSB may not be due to by chance at 90% confidence level.
The productivity (OPE) of the other banks like SBI, SB and CBI also has positive linear
correlations over profitability (PBT/TA). But, we find no statistical significance even at
90% confidence level to nullify null hypothesis. So, we can say that the relationship
between the two variables of the three banks are may be due to by chance at 90%
confidence level (valid for this sample only).
3. From the above table it is clear that the (NIM - BURDEN)/TA of all the banks have
strong positive linear correlations over profitability (PBT/TA). We also get a support in
favour of our statement by a 2-tailed significance test at 95% confidence level, because
all p values<0.05.
4. From the above table it is clear that the CDR of the banks UCO and PNB has strong
positive linear correlations over profitability (PBT/TA). We also get a support in favour
of our statement by a 2-tailed significance test at 95% confidence level, because all p
values<0.05. The CDR of the other bank like SBT also has positive linear correlations
over profitability (PBT/TA). But, we find statistical significance at 90% confidence level
to nullify null hypothesis. So, we can say that the relationship between the two variables
of the bank SBT is may not be due to by chance at 90% confidence level. The CDR of the
other banks like SBI, SB, CBI and PSB has positive linear correlations over profitability
(PBT/TA). But, we find no statistical significance at 95% confidence level to nullify null
hypothesis. So, we can say that the relationship between these two variables of four banks
may be due to by chance (valid for this sample only).
215
5. From the above table it is clear that the CDRGROWTH of the bank UCO has strong
values<0.05. The CDRGROWTH of the other banks excepting SBI and CBI also have
positive linear correlations over profitability (PBT/TA). But, we find no statistical
significance at 95% confidence level to nullify null hypothesis. So, we can say that the
relationship between these two variables of the two banks may be due to by chance (valid
for this sample only).
6. From the above table it is clear that the PSADV/TADV of the banks SBI has strong
values<0.05. The PSADV/TADV of the other banks are not statistically significant at 5%
level.
From the above correlation analysis, following statistically significant strategic banking variables
having association with Profitability (PBT/TA) with level of sig of 5% or marginally higher than
5% have been finally selected.
Table- 7.17.: Selection of Statistically Significant Strategic Banking Variables having Association with
Profitability (PBT/TA)
Bank Statistically significant banking variable
SBI GNPA/TA, (NIM-BURDEN)/TA, PSADV/TADV
SBT GNPA/TA, OPE, (NIM-BURDEN)/TA, CDR
SB GNPA /TA, (NIM-BURDEN)/TA
UCO GNPA /TA, OPE, (NIM-BURDEN)/TA, CDR, CDRGROWTH
CBI GNPA /TA, (NIM-BURDEN)/TA
PNB GNPA /TA, OPE, (NIM-BURDEN)/TA, CDR
PSB GNPA /TA, OPE, (NIM-BURDEN)/TA
7.5.2. Examination of Impact of NPA on Strategic Banking Variables

We analyze the data for a simple linear regression models to examine impact of NPA on strategic
banking variables
216
7.5.2.1. Impact of NPA on Profitability
To examine the impact of NPA (GNPA/TA) over profitability (PBT/TA), simple linear
regression is tried as under:
7.5.2.1.1. Simple Regression between Profitability (PBT/TA) and NPA (GNPA/TA).
7.5.2.1.1.1. Evaluating Model Fit and Verifying Assumptions
Summary of DW and SW test statistics are given in the following table:
Table- 7.18.: DW and SW Statistics for Profitability (PBT/TA) and NPA (GNPA/TA)
Bank Durbin Watson Shapiro-Wilk Statistic Sig.
SBI 2.063 0.949 0.44
SBT 1.694 0.954 0.531
SB 1 0.797 0.002
UCO 0.864 0.972 0.852
CBI 1.461 0.974 0.881
PNB 1.256 0.928 0.202
PSB 1.547 0.939 0.301
Dependent Variable : Profitability ( PBT/TA)
DW statistics value for models of all banks is higher than 1.2 except for the models SB and
UCO. This suggests that the assumption of independent (no serious serial correlation) errors has
been met for models of the other banks, but not for the SB and UCO. The p values of the SW
Statistic for models of all banks are higher than 0.05 except for the model of SB. This suggests
that the assumption of normality of the error distribution has been met for models of the other
banks, but for the model of SB the assumption is doubtful.
7.5.2.1.1.2. Redefining Model for SB and UCO

Since the linear model is not suitable for SB and UCO we look for an alternate model. We tried
few conventional alternative models like Quadratic, Cubic, Logarithmic, Exponential and
Growth. For UCO we found model in the form of z = a + bx + cy, where x = GNPA/TA and y=
(GNPA/TA)2 and z = PBT/TA somehow suitable. But, do not find such model suitable for SB
(i.e, not matching of both Independence and Normality simultaneously). Therefore, we have
dropped SB from the modeling frame work. Summary of DW and SW test statistics for quadratic
model for UCO is given below:
217
Table-7.19.: DW and SW Statistics for Profitability (PBT/TA) and NPA (GNPA/TA) for UCO
UCO 1.986 0.896 0.059
Dependent Variable: PBT/TA (Profitability)
DW statistics value of 1.986 for quadratic model for UCO suggests that the assumption of
independent errors (no serious serial correlation) has been met for the model UCO. The p value
of the SW Statistic is marginally higher than 0.05 for the quadratic model of UCO. This suggests
that the assumption of normality of the error distribution has also been met for this model.
7.5.2.1.1.3. Model Summary for Profitability (PBT/TA) and NPA (GNPA/TA)

In the table below we presented the summaries of the models:
Table- 7.20.: Regression Model Summary for Profitability (PBT/TA) and NPA (GNPA/TA)
Bank R R2 df F Sig. of F
SBI .447 0.2 1 3.739 .072
SBT .582 0.339 1 7.702 .014
UCO .968 0.937 2 103.436 .000
CBI .512 0.262 1 5.332 .036
PNB .695 0.483 1 14.02 .002
PSB .705 0.497 1 14.813 .002
From the above table, we see that R is generally higher than 0.5 except SBI (.447a). R2 in the
model summary for all banks are higher than 0.3 except SBI (.200) and maximum .937 for UCO.
In other words we can say that the proportion of variance the profitability (PBT/TA) of SBI that
is “explained” or “accounted for” by knowledge of the NPA of SBI is approximately 20% and
93.7% (quadratic model) for UCO and rest falling in between. The F-ratio ranges from 3 to 104.
The test of significance reveals that the probability of obtaining an F-stat as the one we’ve
obtained on the model degrees of freedom are very smalls, all less than .05 for all models
excepting SBI. Hence, we can conclude the goodness of fit for all models are well established at
95% confidence interval and at 90% confidence interval for the model of SBI.
The above facts indicate not a very good fit based on our data for the models of all banks except
UCO. It is also clear from the above facts that the profitability (PBT/TA) is not only dependent
218
on parameter NPA (GNPA/TA) but to other parameters also. This gives us the basic foundation
for multiple regressions in the later segments.
7.5.2.1.1.4. Model Co-efficient Analysis

A table containing coefficients statistics, t-test and p-values of the models are given below.
Table- 7.21.: Summary of the Coefficients of the Models Selected Banks
Bank Unstandardized Coefficients Standardized T Sig.
Coefficients
B Std. error Beta
SBI (Constant) 0.016 0.002 10.17 0.000
GNPA/TA -0.065 0.034 -0.447 -1.934 0.07
SBT (Constant) 0.015 0.001 12.06 0.000
GNPA/TA -0.06 0.022 -0.582 -2.775 0.01
UCO (Constant) 0.003 0.002 1.946 0.07
GNPA/TA 0.205 0.07 1.004 2.929 0.01
2
(GNPA/TA) -2.901 0.515 -1.932 -5.637 0.000
CBI (Constant) 0.01 0.002 5.693 0.000
GNPA/TA -0.065 0.028 -0.512 -2.309 0.04
PNB (Constant) 0.022 0.002 9.829 0.000
GNPA/TA -0.17 0.045 -0.695 -3.744 0.000
PSB (Constant) 0.016 0.001 11.66 0.000
GNPA/TA -0.102 0.027 -0.705 -3.849 0.002
For SBI the NPA (b=-0.065) is not significant (p=0.07), but only just so (as it is marginally
above .05), and the coefficient is negative which would indicate that larger NPA is related to
lower profitability, which is what we would expect. For all other banks’ the effect of NPA is
significant (p < .05) and its coefficient is negative indicating that larger NPA is related to lower
profitability.
7.5.2.1.1.5. Impact Analysis of the Independent Variable

Our main objective of developing regression models is to examine the impact of NPA
(GNPA/TA) over profitability (PBT/TA). From the above output table we find our regression
equations given below to get the impact of NPAs in the models.
219
Table- 7.22.: Model Equations on Profitability (PBT/TA) and NPA (GNPA/TA)
Bank Equation
SBI PBT/TA (Profitability) = .016-.065*( GNPA/TA)
SBT PBT/TA (Profitability) = .015-.060*( GNPA/TA)
UCO PBT/TA (Profitability) = .003+.205*( GNPA/TA) -2.901*(GNPA/TA)2
CBI PBT/TA (Profitability) = .010-.065*( GNPA/TA)
PNB PBT/TA (Profitability) = .022-.170*( GNPA/TA)
PSB PBT/TA (Profitability) = .016-.102*( GNPA/TA)
1. SBI: for one unit of change in GNPA/TA, PBT/TA (Profitability) is effected by -.065 units.
Clearly the impact on PBT/TA (Profitability) is negative and nearly 6.5% of GNPA/TA.
2. SBT: for one unit of change in GNPA/TA, PBT/TA (Profitability) is effected by -.060 units.
3. UCO: for one unit of change in GNPA/TA, PBT/TA (Profitability) is effected by 0.205 units
{the impact on PBT/TA (Profitability) is positive and nearly 20.5% of GNPA/TA} and for
one unit of change in (GNPA/TA)2, PBT/TA (Profitability) is effected -2.901 units, {the
impact on PBT/TA (Profitability) is negative and nearly 290.1% of (GNPA/TA)2}
4. CBI: for one unit of change in GNPA/TA, PBT/TA (Profitability) is effected by -.065 units.
5. PNB: for one unit of change in GNPA/TA, PBT/TA (Profitability) is effected by -.170 units.
6. PSB: for one unit of change in GNPA/TA, PBT/TA (Profitability) is effected by -.102 units.
7.5.2.1.2. Multiple Regression between Profitability (PBT/TA) and NPA (GNPA/TA) &
Time (Year)
Summaries of the assumptions verification statistics of the models is given below:
220
Table- 7.23.: Regression Model Assumptions Verification for Profitability (PBT/TA) and NPA
SBI 2.054 .949 .434
SBT 1.726 .953 .512
SB 1.242 .661 .000
UCO 1.199 .965 .721
CBI 1.689 .954 .530
PNB 1.255 .928 .204
PSB 1.593 .949 .447
The values of DW statistics computed are all higher than 1.19 and less than 2.06. This suggests
that the assumption of independent (no serious serial correlation) errors has been met for all the
models. The p values of the SW statistic of all banks are higher than 0.05 except for the model of
SB. This suggests that the assumption of normality of the error distribution has been met for the
other models, but for the model SB the assumption is doubtful. Since the linear model is not
suitable for SB we look for conventional alternate models like Quadratic, Multiple Quadratic,
Cubic, Logarithmic, Exponential and Growth. For SB we finally found no suitable model which
can incorporate Time (Year).
7.5.2.1.2.2. Co linearity Analysis of the Models

A summary table of the co linearity statistics of the models is given below.
Table- 7.24.: Regression Model co linearity statistics for Profitability (PBT/TA) and NPA (GNPA/TA) &
Time (Year)
Bank Co linearity Statistics
Tolerance VIF
SBI (Constant)
GNPA/TA 0.158 6.315
YEAR 0.158 6.315
SBT (Constant)
GNPA/TA 0.171 5.836
YEAR 0.171 5.836
UCO (Constant)
GNPA/TA 0.156 6.406
221
YEAR 0.156 6.406
CBI (Constant)
GNPA/TA 0.078 12.82
YEAR 0.078 12.82
PNB (Constant)
GNPA/TA 0.093 10.71
YEAR 0.093 10.71
PSB (Constant)
GNPA/TA 0.078 12.89
YEAR 0.078 12.89
From the above table we may observe that the tolerance for NPA and Time for the models CBI,
PNB and PSB are less than 0.1, but VIF is >10. We found the basis of redefining of these three
models by elimination of variable, which makes very little sense in these cases. So, we conclude
models incorporating Time are not possible for CBI, PNB and PSB. The VIF for the other
models are quite satisfactory.
7.5.2.1.2.3. Model Summary

Table- 7.25.: Regression Model Summary for Profitability (PBT/TA) and NPA (GNPA/TA) & Time
(Year)
Bank R R2 df F Sig of F
SBI .449 .202 2 1.767 .207
SBT .584 .341 2 3.618 .05
UCO .910 .828 2 33.688 .000
From the above table it is found that R is generally higher than 0.44. R2 is all higher than 0.20
and maximum .828 for UCO. In other words we can say that the proportion of variance of the
profitability (PBT/TA) of SBI that is “explained” or “accounted for” by knowledge of the NPA
and Time of SBI is approximately 20.2% and 82.8 % for UCO and rest falling in between. The
F-ratio ranges from 1.7 to 33.7. The test of significance reveals that the probability of obtaining
an F-stat as the one we’ve obtained on the model degrees of freedom are small for SBT and
222
UCO, both less than .05 but not for SBI (higher than .05). Hence, we may say that the goodness
of fit for SBT and UCO are well established at 95% confidence level.
We may therefore, conclude that only for UCO a high quality of model fit is exhibited here and
decision of multiple correlation analysis including variable Time for UCO is justified here.

A summary table containing coefficients statistics, t-stat with sig. of the models are given below.
Table- 7.26.: Regression Model Coefficients, t stat for Profitability (PBT/TA) and NPA (GNPA/TA) &
Time (Year)
Bank Unstandardized Standardized T Sig.
Coefficients Coefficients
B Std. error Beta
SBI (Constant) 0.017 0.007 2.483 0.026
GNPA/TA -0.08 0.087 -0.551 -0.919 0.374
YEAR -7.23E-05 .000 -0.114 -0.19 0.852
SBT (Constant) 0.016 0.006 2.718 0.017
GNPA/TA -0.069 0.054 -0.668 -1.274 0.223
YEAR -7.13E-05 .000 -0.094 -0.179 0.86
UCO (Constant) 0.023 0.007 3.35 0.005
GNPA/TA -0.271 0.057 -1.327 -4.728 .000
YEAR -0.001 .000 -0.475 -1.692 0.113
We notice that one p-value (marked) is very small, less than 0.01, and hence, we can conclude
that the co-efficient is not zero for this predictor only and this coefficient is not by chance at 99%
confidence level and the other coefficients could be due to by chance (valid only for this
sample).

(GNPA/TA) and Time (Year) over profitability (PBT/TA). From the above output table we
found our regression equations given below to get the impact of predictors in the models.
223
Table- 7.27.: Model Equations on Profitability (PBT/TA) and NPA (GNPA/TA) & Time (Year)
Bank Equation
SBI PBT/TA (Profitability)= 0.017-0.080*( GNPA/TA)-0.00007228*( YEAR)
SBT PBT/TA (Profitability)= 0.016-0.069*( GNPA/TA)-0.00007125*( YEAR)
UCO PBT/TA (Profitability)= 0.023-0.271*( GNPA/TA)-0.001*( YEAR)
1. SBI: -0.080 times of GNPA/TA and -0.00007228 times of (YEAR) are responsible to
predict PBT/TA (Profitability) of SBI. Clearly the above quantities are the impacts of the
predictors on PBT/TA (Profitability) of SBI.
2. SBT: -0.069 times of GNPA/TA and -0.00007125 times of (YEAR) are responsible to
predict PBT/TA (Profitability) of SBT. Clearly the above quantities are the impacts of the
predictors on PBT/TA (Profitability) of SBT.
3. UCO: -0.271 times of GNPA/TA and -0.001 times of (YEAR) are responsible to predict
PBT/TA (Profitability) of UCO. Clearly the above quantity is the impact of the predictor
on PBT/TA (Profitability) of UCO.
7.5.2.1.3. Multiple Regression between: Profitability (PBT/TA) and other statistically

significant variable (derived from correlation matrix) including NPA (GNPA/TA)
Summaries of the assumptions verification statistics of the models is given below :
Table- 7.28.: Regression Model Assumptions Verification Statistics for Profitability (PBT/TA) and
statistically significant variables including NPA (GNPA/TA)
SBI 1.682 .942 .343
SBT 2.031 .955 .533
SB 1.691 .806 .002
UCO 2.253 .931 .227
CBI 1.513 .947 .408
PNB 1.284 .950 .459
PSB 1.684 .928 .199
224
DW statistics for all banks are all higher than 1.28 and less than 2.26. This suggests that the
assumption of independent (no serious serial correlation) errors has been met for all the models.
The p values of the SW Statistic of models of all banks, except for the model SB are higher than
0.05 This suggests that the assumption of normality of the error distribution has been met for the
other models, but for the model of SB, the assumption is doubtful.
7.5.2.1.3.2. Redefining Model for SB

Since the linear model is not suitable for SB we are in search of an alternate. We tried all the
conventional alternative models like Quadratic, Multiple Quadratic, Cubic, Multiple Cubic,
Logarithmic, Exponential and Growth. For SB we found model in the form of z = a+bx + cy,
where x = GNPA/TA, y= ((NIM - BURDEN)/TA)3 and z = PBT/TA somehow suitable.
Summary of DW and SW test statistics for quadratic model for SB is given below:
Table- 7.29.: The Multiple Cubic Model for SB
SB 1.553 .941 .335

The DW statistics is 1.553. This suggests that the assumption of independence (no serious serial
correlation) errors has been met for the model SB. The p value of the S-W Statistic is higher than
0.05 for the model SB. This suggests that the assumption of Normality of the error distribution
has been met for this model at 95% confidence.

A summary table of the co linearity statistics of the models is given below:
Table- 7.30.: Multiple Regression Model Co linearity Statistics for Profitability (PBT/TA) & NPA
(GNPA/TA), other Statistically Significant Variables
Bank Co linearity Stat
Tolerance VIF
SBI (Constant)
(NIM_BURDEN)/TA .765 1.306
PSADV/TADV .905 1.106
GNPA/TA .776 1.288
SBT (Constant)
GNPA/TA .319 3.139
OPE .247 4.052
CDR .262 3.813
225
SB (Constant)
GNPA/TA .819 1.221
{(NIM - BURDEN)/TA}3 .819 1.221
UCO (Constant)
GNPA/TA .026 38.144
OPE .252 3.976
CRD_GROWTH .432 2.313
CDR .048 20.923
CBI (Constant)
GNPA/TA .841 1.189
PNB (Constant)
GNPA/TA .048 20.644
OPE .133 7.511
CDR .036 27.649
PSB (Constant)
GNPA/TA .249 4.015
OPE .439 2.276
We interpret the tolerance for most of the variables for the models of UCO and PNB are less than
0.1, with VIF is >10. Accordingly, we examine the basis of redefining of these two models once
again. The VIF for the other models are quite satisfactory.
7.5.2.1.3.4. Redefining Model for UCO and PNB

Since the multiple linear models with the existing set of variables are not suitable for UCO and
PNB we are in search of an alternate by reducing few variables from the models. The decision of
elimination of variables is taken through correlation analysis. We find CDR is highly correlated
with other variables. So, we remove CDR and rerun the multiple regression analysis and model
fit summary statistic is given below.
Table- 7.31.: Regression Model Assumptions Verification for Profitability (PBT/TA) and NPA
(GNPA/TA) along with Other Statistically Significant Variables
UCO 2.030 .940 .323
PNB 1.289 .943 .353
226
DW statistics for both banks are higher than 1.28 and less than 2.1. This suggests that the
assumption of independent (no serious serial correlation) errors has been met for both the
models. The p value of the SW Statistic for all models are greater than 0.05. This suggests that
the assumption of normality of the error distribution has been met at 95% confidence level for
both the models.

Table- 7.32.: Regression Model Co linearity Statistics for Profitability (PBT/TA) and NPA (GNPA/TA)
& Other Statistically Significant Variables for UCO and PNB
Tolerance VIF
UCO (Constant)
GNPA/TA .138 7.261
OPE .415 2.409
CRD_GROWTH .439 2.279
PNB (Constant)
GNPA/TA .241 4.149
OPE .318 3.148
We can say that tolerance for NPA and other variables for all the models are above 0.1 and VIF
is <10. This implies that the multi co linearity is not cause of concern for these models.

Table- 7.33.: Regression Model Summary for Profitability (PBT/TA) and NPA (GNPA/TA) & Other
Statistically Significant Variables
SBI .960 .921 3 50.615 .000
SBT .914 .835 4 15.193 .000
SB .944 .892 2 57.517 .000
UCO .999 .999 4 2354.747 .000
CBI .983 .966 2 198.449 .000
PNB .986 .973 3 153.791 .000
PSB .996 .992 3 536.866 .000
227
From the above table, we see that R is generally higher than 0.91 and R2 value for models of all
banks all higher than 0.83 and maximum .999 for UCO. In other words we can say that the
proportion of variance the profitability (PBT/TA) of PSB that is “explained” or “accounted for”
by knowledge of the NPA and other variables of SBT is approximately 83.5% and 99.9 % for
UCO and rest falling in between. The F-ratio ranges from 15 to 2355. The test of significance
reveals that the probability of obtaining an F-stat as the one we’ve obtained on the model degrees
of freedom are very smalls, all less than .01 for all models. Hence, we can say that the goodness
of fit for all models are well established at 99% confidence interval. We can safely conclude that
a high quality of models fit is exhibited. Therefore, the decision of multiple regression analysis is
justified here.

Table- 7.34.: Regression Model coefficients, t stat for Profitability (PBT/TA) and NPA (GNPA/TA) &
Other Statistically Significant Variables
B Std. error Beta
SBI (Constant) -.015 .004 -3.573 .003
(NIM_BURDEN)/TA 1.534 .164 .832 9.351 .000
PSADV_TADV .053 .014 .308 3.762 .002
GNPA/TA .001 .013 .009 .100 .922
SBT (Constant) .007 .005 1.568 .143
GNPA/TA -.023 .021 -.219 -1.053 .313
OPE .005 .024 .048 .205 .841
(NIM_BURDEN)/TA 1.470 .251 .943 5.867 .000
CDR -.011 .007 -.354 -1.546 .148
SB (Constant) .006 .001 5.627 .000
GNPA/TA -.030 .015 -.193 -1.981 .068
{(NIM - BURDEN)/TA}3 8547.669 983.037 .846 8.695 .000
UCO (Constant) .001 .000 1.126 .282
GNPA/TA .001 .006 .004 .128 .900
OPE -.007 .003 -.033 -2.086 .059
(NIM_BURDEN)/TA 1.047 .028 1.029 37.781 .000
CRD_GROWTH -.001 .001 -.012 -.787 .446
CBI (Constant) .002 .001 2.919 .011
GNPA/TA -.019 .007 -.148 -2.743 .016
(NIM - BURDEN)/TA 1.226 .072 .915 17.004 .000
PNB (Constant) .000 .002 .091 .929
228
GNPA/TA -.001 .023 -.004 -.043 .966
OPE .003 .009 .028 .346 .735
(NIM_BURDEN)/TA 1.437 .094 .967 15.237 .000
PSB (Constant) -.001 .001 .902 .383
GNPA/TA .013 .007 .093 1.871 .084
OPE .001 .003 .009 .243 .812
(NIM_BURDEN)/TA 1.449 .052 1.060 27.911 .000
We notice that some p-values (marked) are small, less than 0.05, and hence, we can conclude
that these co-efficients are not zero for the related predictors only and these coefficients are not
by chance at 95% confidence level and the other coefficients could be due to by chance (valid
only for this sample).

(GNPA/TA) and other variables over profitability (PBT/TA). From the above output table we
found our regression equations given below to get the impact of predictors in the models.
Table- 7.35.: Model Equations on Profitability (PBT/TA) and Strategic Banking Variables including NPA
(GNPA/TA)
Bank Equation
SBI PBT/TA (Profitability)= -0.015+0.001*( GNPA/TA)+ 0.053*( PSADV/TADV)+ 1.534*(NIM -
BURDEN)/TA
SBT PBT/TA (Profitability)= 0.007-0.023*( GNPA/TA)+0.005*(OPE)+ 1.470*(NIM -
BURDEN)/TA-0.011*(CDR)
SB PBT/TA (Profitability)= 0.006-0.030*( GNPA/TA)+ 8547.669*((NIM - BURDEN)/TA)3
UCO PBT/TA (Profitability)= 0.001+0.001*( GNPA/TA)-0.007*( OPE)+ 1.047*(NIM -

BURDEN)/TA – 0.001*(CRD_GROWTH)
CBI PBT/TA (Profitability)= 0.002-0.019*( GNPA/TA)+ 1.226*(NIM - BURDEN)/TA
PNB PBT/TA (Profitability)= 0.000-0.001*( GNPA/TA)+ 0.003*(OPE)+ 1.437*(NIM -
BURDEN)/TA
PSB PBT/TA (Profitability)= -0.001+0.013*( GNPA/TA)+ 0.001*(OPE) + 1.449*(NIM -
BURDEN)/TA
1. SBI: 0.001 times of GNPA/TA, 0.053 times of (PSADV/TADV) and 1.534 times of
(NIM - BURDEN)/TA are responsible to predict PBT/TA (Profitability) of SBI. Clearly
the above quantities are the impacts of the predictors on PBT/TA (Profitability) of SBI.
2. SBT: -0.023 times of GNPA/TA, 0.005 times of (OPE) and 1.470 times of (NIM -
BURDEN)/TA and -0.011 times of CAR are responsible to predict PBT/TA
229
(Profitability) of SBT. Clearly the above quantities are the impacts of the predictors on
PBT/TA (Profitability) of SBT.
3. SB: -0.030 times of GNPA/TA and 8547.669 times of ((NIM - BURDEN)/TA)3 are
responsible to predict PBT/TA (Profitability) of SB. Clearly the above quantities are the
impacts of the predictors on PBT/TA (Profitability) of SB.
4. UCO: 0.001 times of GNPA/TA, -0.007 times of (OPE), 1.047 times of (NIM -
BURDEN)/TA and -0.001 times CRD_GROWTH are responsible to predict PBT/TA
(Profitability) of UCO. Clearly the above quantities are the impacts of the predictors on
PBT/TA (Profitability) of UCO.
5. CBI: -0.019 times of GNPA/TA and 1.226 times of (NIM - BURDEN)/TA are
responsible to predict PBT/TA (Profitability) of CBI. Clearly the above quantities are the
impacts of the predictors on PBT/TA (Profitability) of CBI.
6. PNB: -0.001 times GNPA/TA, 0.003 times of (OPE) and 1.437 times of (NIM -
BURDEN)/TA are responsible to predict PBT/TA (Profitability) of PNB. Clearly the
above quantities are the impacts of the predictors on PBT/TA (Profitability) of PNB.
7. PSB: 0.013 times GNPA/TA, 0.001 times OPE and 1.449 times of (NIM -
BURDEN)/TA are responsible to predict PBT/TA (Profitability) of PSB. Clearly the
above quantities are the impacts of the predictors on PBT/TA (Profitability) of PSB.
7.5.2.1.4. Multiple Regression between: Profitability (PBT/TA) and statistically significant

variable including NPA (GNPA/TA) & Time (Year)
Summaries of the assumptions verification statistics of the models are given below:
Table- 7.36.: Regression Model Assumptions Verification Statistics for profitability and statistically
significant variables including NPA & Time
SBI 1.798 .956 .562
SBT 2.098 .952 .490
SB 1.543 .946 .397
UCO 2.048 .966 .753
CBI 1.596 .931 .223
PNB 1.480 .942 .341
PSB 2.424 .934 .252
230
DW statistics of models of all banks are higher than 1.4 and less than 2.5. This suggests that the
assumption of independent (no serious serial correlation) errors has been met for all the models.
The p values of the SW statistic are all higher than 0.05. This suggests that the assumption of
Normality of the error distribution has been met for all the models.

Table- 7.37.: Regression Model co linearity statistics for profitability (PBT/TA) and statistically
significant variables including NPA & Time
Tolerance VIF
SBI (Constant)
PSADV/TADV .885 1.130
GNPA/TA .139 7.218
YEAR .150 6.662
SBT (Constant)
GNPA/TA .101 9.938
OPE .069 14.565
CDR .195 5.123
YEAR .026 39.136
SB (Constant)
GNPA/TA .110 9.066
{(NIM - BURDEN)/TA}3 .707 1.414
YEAR .125 7.991
UCO (Constant)
GNPA/TA .028 36.179
OPE .071 14.122
CRD/GROWTH .419 2.385
YEAR .018 54.427
CBI (Constant)
GNPA/TA .062 16.098
NIM_BURDEN_TA .730 1.370
YEAR .068 14.780
PNB (Constant)
GNPA/TA .075 13.415
OPE .127 7.849
YEAR .037 26.730
PSB (Constant)
GNPA/TA .009 106.552
OPE .029 34.210
231
YEAR .005 208.757
We interpret the tolerance for most of the variables of all models except for SBI and SB models
are less than 0.1, and VIF is also >10. So, it is clear from the fact that for SBI and SB regression
model including Year is justified and the model for SBI takes the form of z = a + bx + cy + dL +
kT, where x = GNPA/TA, y= PSADV/TADV, L= (NIM - BURDEN)/TA, T= YEAR and z =
PBT/TA and for SB takes the form of z = a + bx + cy + kT, where x = GNPA/TA, y= ((NIM -
BURDEN)/TA)3 , T=YEAR and z = PBT/TA somehow suitable. So, we are moving to present
the model summaries of SBI and SB in the next segment.

Table- 7.38.: Regression Model Summary for Profitability (PBT/TA) and statistically significant variables
including NPA & Time
SBI .961 .924 4 36.460 .000
SB .944 .892 3 35.722 .000
From the above table, we see that R is 0.961 and R2 is 0.924 for SBI and .944 and .892 for SB. In
other words we can say that the proportion of variance the profitability (PBT/TA) of SBI that is
“explained” or “accounted for” by knowledge of the NPA (GNPA/TA), Time (YEAR) and other
variables of SBI is approximately 92.4% and that for SB is approximately 89.2%.
The F-ratio for our data 36.46 for SBI and 35.722 for SB. The test of significance reveals that the
probability of obtaining an F-stat as the one we’ve obtained on the model degrees of freedom are
very smalls, all less than .01 for the model. Hence, we can conclude the goodness of fit for
models of SBI and SB are well established at 99% confidence interval and a high quality of
models fit exhibit here. Therefore, the decision of multiple regression analysis including Time
(Year) is justified for SBI as it improves the precision criterion, but not for SB as it does not
improve the precision criterion.

232
Table- 7.39.: Regression Model Coefficients, t stat for profitability (PBT/TA) and NPA (GNPA/TA).
Time (Year) & other statistically significant variables
B Std. error Beta
SBI (Constant) -.016 .005 -3.438 .005
GNPA/TA .020 .031 .139 .648 .529
(NIM_BURDEN)/TA 1.557 .171 .845 9.098 .000
PSADV/TADV .051 .014 .300 3.542 .004
YEAR .000 .000 .137 .669 .516
SB (Constant) .005 .003 1.397 .186
GNPA/TA -.022 .043 -.143 -.521 .611
{(NIM - BURDEN)/TA}3 8626.283 1095.912 .854 7.871 .000
YEAR .000 .000 .050 .195 .849
by chance at 95% confidence level and the other coefficients could be due to by chance (valid
only for this sample).

Our main objective of developing regression models is to examine the impact of NPA and other
variables including time (Year) over profitability (PBT/TA). From the above output table we
develop regression equations as under to show the impact of predictors in the models.
Table- 7.40.: Model Equations on Profitability (PBT/TA) and Strategic Banking Variables including NPA
Bank Equation
SBI PBT/TA (Profitability)= -0.016+0.020*( GNPA/TA)+ 0.051*( PSADV/TADV)+ 1.557*(NIM
- BURDEN)/TA +0.000087*(YEAR)
SB PBT/TA (Profitability)= 0.005-0.022*( GNPA/TA)+ 8626.283*{(NIM -
BURDEN)/TA}3+0.000035*(YEAR)
1. SBI: 0.020 times of GNPA/TA, 0.051 times of (PSADV/TADV), 1.557 times of (NIM -
BURDEN)/TA and 0.000087 times of YEAR are responsible to predict PBT/TA (Profitability) of
SBI. Clearly the above quantities are the impacts of the predictors on PBT/TA (Profitability) of SBI.
2. SB: -0.022 times of GNPA/TA, 8626.283 times of (NIM - BURDEN)/TA3 and 0.000035 times of
YEAR are responsible to predict PBT/TA (Profitability) of SB. Clearly the above quantities are the
impacts of the predictors on PBT/TA (Profitability) of SB.
233
7.5.2.1.5. Summary of findings on Impact of NPA on Profitability
(i) NPA (GNPA/TA) of all PSBs, with the exception of SBI and SB have been found to have
statistically significant (p value less than 0.05) negative impact on Profitability
represented by PBT/TA, though at varying degree. NPA of SBI also has negative impact
on Profitability with p value of coefficient marginally higher than 0.05, while for SB such
model seems infeasible. Precision of a model represented by R2 is very high for UCO
(0.937), moderate for PNB (0.483) and PSB (0.497) and low for SBI, SBT and CBI.
(ii) With the inclusion of Time (Year) in the predictor variable, multiple regression modeling
finally restricted to SBI, SBT and UCO. Other banks have been dropped due to presence
of multi co linearity beyond tolerance level. NPA of SBI, SBT and UCO have negative
impact on profitability with p value of coefficient for most of the predictors are higher
than 0.05, putting question mark on the statistical validity of such model. R2 in the model
is still very high for UCO (0.828), while it has marginally increased for SBI (0.202) and
SBT (0.341) with the inclusion of Time (Year) in the predictor variable.
(iii)With the introduction of selected statistically significant banking variables (as given in
Table 6.17) without Time (Year) in the predictor variable with NPA, feasible multiple
regression modeling for all selected PSBs could be done with statistical significance (as
multi co linearity is not a cause of concern for any bank). NPA has negative impact on
profitability of SBI, SBT and UCO with p value of coefficient for most of the predictors
are less than 0.05. R2 in the models for all banks, including UCO are very high, while the
same have increased considerably with minimum .835 for SBT and maximum .999 for
UCO, while other banks are in between these two.
(iv) With the introduction of selected statistically significant banking variables (as given in
Table 6.17) along with Time (Year) variable in the predictor variable with NPA, multiple
regression modeling finally restricted to SBI and SB. Other banks have been dropped due
to presence of multi co linearity beyond tolerance level. NPA of SB has negative impact
on profitability with p value of coefficient for most of the predictors are higher than 0.05,
putting question mark on the statistical validity of the such model. With the introduction
Time (Year) in the predictor variable, R2 in the models are still very high for SBI (0.924)
and SB (0.892).
234
7.5.2.2. Impact of NPA on Productivity
To examine the impact of NPA over productivity, simple linear regression is tried.
7.5.2.2.1. Simple Regression between: Productivity (OPE) and NPA (GNPA/TA)
Table- 7.41.: DW and SW test statistics for Productivity (OPE) and NPA (GNPA/TA)
SBI .334 .783 .001

SBT .798 .885 .039
SB .391 .886 .039
UCO .367 .754 .001
CBI .467 .870 .022
PNB .396 .899 .066
PSB .384 .751 .000
Dependent Variable: Productivity (OPE)
DW statistics are all less than 0.8. This suggests that the assumption of independent (no serious
serial correlation) errors has been violated seriously and models are highly affected by serial
correlations. The p values of the SW statistic of only one model (PNB) are higher than 0.05
while for others they are not. This suggests that the assumption of normality of the error
distribution has been met only for PNB.
Though a high negative linear correlation is present between the Productivity (OPE) and NPA
(GNPA/TA), but due to the violation of linear model assumptions, we feel developing a linear
model concept would be nonsense idea. Since the linear models are not suitable for all the banks,
we are in search of an alternate. We have tried for the conventional alternate models. But, we
found no suitable model combination for all the banks in our limited study area. So, we are
dropping these models.
7.5.2.2.2. Multiple Regression between Productivity (OPE) and NPA (GNPA/TA) & Time
(Year)
In the table below we presented the summaries of the assumptions verification of the models:
235
Table- 7.42.: Regression Model Assumptions Verification for Productivity (OPE) and NPA (GNPA/TA)
& Time (Year)
SBI 0.855 .954 0.516
SBT 2.136 .935 0.260
SB 1.264 .958 0.601
UCO 0.926 .979 0.943
CBI 1.259 .962 0.662
PNB 0.399 .917 0.130
PSB 1.172 .980 0.961
DW statistics of all models are higher than 1.1 and less than 2.14, except for SBI, UCO and
PNB, where DW is less than 0.93. This suggests that the assumption of independent (no serious
serial correlation) errors has been met for the models other than SBI, UCO and PNB. The p
values of the SW Statistic for models of all banks are higher than 0.05. This suggests that the
assumption of normality of the error distribution has been met for all the models.
7.5.2.2.2.2. Model Re-defined

Since the linear model is not suitable for SBI, UCO and PNB we are in search of an alternate.
We tried all the conventional alternate models like Quadratic, Multiple Quadratic, Cubic,
Multiple Cubic, Logarithmic, Exponential and Growth. Finally we find few suitable models of
the form z = a + bx + cy, where x = (Linear or higher powers of GNPA/TA), y= (Linear or
higher powers of YEAR) and z = Productivity (OPE), which incorporate time (Year).

Table- 7.43.: Assumptions Verification for Productivity (OPE) and NPA (GNPA/TA) & Time (Year)
with Multiple Quadratic, Multiple Cubic Models
SBI 1.300 .959 .614
UCO 1.265 .941 .336
PNB 1.684 .972 .859
DW statistic values ranges between 1.265 and 1.684. These suggest that the assumption of
independent (no serious serial correlation) errors has been met for the models SBI, UCO and
PNB. The p value of the SW statistic is higher than 0.05 for the models SBI, UCO and PNB.
236
This suggests that the assumption of normality of the error distribution has been met for this
model at 95% confidence level.

Table- 7.44.: Regression Model Co linearity statistics for Productivity (OPE) and NPA (GNPA/TA) &
Time (Year)
Tolerance VIF
SBI (Constant)
GNPA/TA .334 2.990
(YEAR)2 .334 2.990
SBT (Constant)
GNPA/TA .171 5.836
YEAR .171 5.836
SB (Constant)
(GNPA/TA) .145 6.905
YEAR .145 6.905
UCO (Constant)
(GNPA/TA)3 .708 1.412
(YEAR)3 .708 1.412
CBI (Constant)
GNPA/TA .078 12.824
YEAR .078 12.824
PNB (Constant)
GNPA/TA .165 6.066
(YEAR)2 .165 6.066
PSB (Constant)
GNPA/TA .078 12.893
YEAR .078 12.893
We interpret the tolerance for GNPA and Year for the models CBI and PSB are less than 0.1, but
VIF is >10. We find the basis of redefining of these two models once again. The facts in this case
are that elimination of variable is absurd in these cases. So, we conclude models incorporating
time (Year) are not possible for CBI and PSB. Therefore, we are dropping these two models. The
VIF for the other models are quite satisfactory.
237
7.5.2.2.2.5. Model Summary for Productivity (OPE) and NPA (GNPA/TA) & Time (Year)
Table- 7.45.: Regression Model Summary for Productivity (OPE) and NPA (GNPA/TA) & Time (Year)
SBI .989 .978 2 315.753 .000
SBT .958 .917 2 77.736 .000
SB .978 .957 2 155.727 .000
UCO .967 .936 2 102.445 .000
PNB .989 .978 2 304.841 .000
From the above table, we see that R is generally higher than 0.95 and R2 is all higher than 0.91
and maximum .978 (SBI and PNB). In other words we can say that the proportion of variance of
the productivity (OPE) of SBI & PNB that is “explained” or “accounted for” by knowledge of
the NPA (GNPA/TA) and Time (Year) of SBI & PNB is approximately 97.8% and 91.7 % for
SBT and rest falling in between. We can say a high quality of models fit is exhibited here. So,
the decision of multiple regression analysis included time (Year) is justified here. The F-ratio
ranges from 77.736 to 315.753. The test of significance reveals that the probability of obtaining
an F-stat as the one we’ve obtained on the model degrees of freedom are very smalls, all less
than .05. Hence, we can conclude the goodness of fit for all models are well established.
7.5.2.2.2.6. Model Co-efficient Analysis for Productivity (OPE) and NPA (GNPA/TA) &
Time (Year)
Table- 7.46.: Regression Model Coefficients, t stat for Productivity (OPE) and NPA (GNPA/TA) & Time
(Year)
B Std. error Beta
SBI (Constant) -.008 .008 -1.053 .310
GNPA/TA .286 .124 .157 2.299 .037
(YEAR)2 .000 .000 1.113 16.347 .000
SBT (Constant) -.057 .021 -2.687 .018
GNPA/TA .428 .191 .416 2.243 .042
YEAR .010 .001 1.321 7.119 .000
SB (Constant) -.103 .018 -5.630 .000
GNPA/TA 1.092 .237 .671 4.607 .000
YEAR .011 .001 1.565 10.744 .000
238
UCO (Constant) .006 .005 1.201 .250
3
(GNPA/TA) -3.811 4.935 -.062 -.772 .453
(YEAR)3 0.00002 .000 .933 11.612 .000
PNB (Constant) -.062 .016 -3.940 .001
GNPA/TA .899 .223 .398 4.034 .001
(YEAR)2 .001 .000 1.339 13.577 .000
We notice that p-values (marked) of all predictors {(except predictor (GNPA/TA)3} are small,
less than 0.05, and hence, we can conclude that these co-efficients are not zero for the related
predictors only and these coefficients are not by chance at 95% confidence level and the other
coefficients of predictor (GNPA/TA)3 could be due to by chance (valid only for this sample).

Our main objective of developing regression models is to find the impact of NPA (GNPA/TA)
and other variables over Productivity (OPE). From the above output table we find our regression
equations given below to get the impact of predictors in the models.
Table- 7.47.: Model Equations on Productivity (OPE) and Strategic Banking Variables including NPA
Bank Equation
SBI Productivity (OPE)= -0.008+0.286*( GNPA/TA)+0.000*(YEAR)2
SBT Productivity (OPE)= -0.057+0.428*( GNPA/TA)+ 0.010*(YEAR)
SB Productivity (OPE)= -0.103+1.092*( GNPA/TA)+ 0.011*(YEAR)
UCO Productivity (OPE)= 0.006-3.811*( GNPA/TA)3+0.00002*(YEAR)3
PNB Productivity (OPE)= -0.062+0.899*( GNPA/TA) + 0.001*( YEAR)2
1. SBI: 0.286 times of GNPA/TA and 0.000 times YEAR2 are responsible to predict
Productivity (OPE) of SBI. Clearly the above quantity is the impact of the predictor on
Productivity (OPE) of SBI.
2. SBT: 0.428 times of GNPA/TA and 0.010 times of (YEAR) are responsible to predict
Productivity (OPE) of SBT. Clearly the above quantities are the impacts of the predictors on
Productivity (OPE) of SBT.
3. SB: 1.092 times of GNPA/TA and 0.011 times of (YEAR) are responsible to predict
Productivity (OPE) of SB. Clearly the above quantities are the impacts of the predictors on
Productivity (OPE) of SB.
239
4. UCO: -3.811 times of (GNPA/TA)3 and 0.00002 times of (YEAR)3 are responsible to
predict Productivity (OPE) of UCO. Clearly the above quantity is the impact of the predictor
on Productivity (OPE) of UCO.
5. PNB: 0.899 times of GNPA/TA and 0.001 times of (YEAR)2 are responsible to predict
Productivity (OPE) of PNB. Clearly the above quantities are the impact of the predictor on
Productivity (OPE) of PNB.
7.5.2.2.3. Summary of findings of Impact of NPA on Productivity
(i) As simple linear as well as conventional model are not feasible with the dataset, though a
high negative linear correlation is present between the Productivity (OPE) and NPA
(GNPA/TA), no impact analysis could be performed.
finally restricted to SBI, SBT, SB, UCO and PNB. Other banks have been dropped due to
presence of multi co linearity beyond tolerance level. Surprisingly, NPA of SBI, SBT, SB
and PNB have positive impact on productivity, when time is added in the predictor
variable, with p value of coefficient for most of the predictors are less than 0.05. R2 in the
models are very high, with maximum for SBI & PNB (.978) and minimum for SBT
(.917) and others are in between.
7.5.2.3. Impact of NPA on NIM/TA
To examine the impact of NPA (GNPA/TA) over NIM/TA, simple linear regression is tried.
7.5.2.3.1. Simple Regression between NIM/TA and NPA (GNPA/TA)
7.5.2.3.1.1. Evaluating Model Fit and Verifying Assumptions:
In the table below we presented the summaries of the Assumptions Verification of the models:
Table- 7.48.: Regression Model Assumptions Verification for NIM/TA and NPA (GNPA/TA)
SBT 1.401 .922 .157
PSB 1.223 .970 .823
Dependent Variable: NIM/TA
DW statistic values ranges between 1.223 to 1.401. These suggest that the assumption of
independent (no serious serial correlation) errors has been met for the models SBT and PSB. The
p value of the SW statistic is higher than 0.05 for the both the models. This suggests that the
assumption of normality of the error distribution has been met for this model at 95% confidence
level.
240
7.5.2.3.1.2. Model Summary for NIM/TA and NPA (GNPA/TA)
Table- 7.49.: Regression Model Summary for NIM/TA and NPA (GNPA/TA)
SBT .796 .633 1 25.872 .000
PSB .469 .220 1 4.220 .058
From the above table, we see that R is .796 and .469 R2 is .633 and .220 for SBT and PSB
respectively. In other words we can say that the proportion of variance of the NIM/TA of SBT
& PSB that is “explained” or “accounted for” by knowledge of the NPA (GNPA/TA) of SBT &
PSB is approximately 63.3% and 22.0% respectively. We can say a high quality of model fit is
exhibited for model of SBT while not a good fit of the model for PSB based on our data. The F-
ratio ranges from 4.220 to 25.872. The test of significance reveals that the probability of
obtaining an F-stat as the one we’ve obtained on the model degrees of freedom are very smalls,
for SBT is less than .05, while for PSB it marginally higher than .05. Hence, we can conclude the
goodness of fit for all models are well established (SBT at 95% and PSB at 90% confidence
level).
7.5.2.3.1.3. Model Co-efficient Analysis:

Table- 7.50.: Regression Model Coefficients, t stat for NIM/TA and NPA (GNPA/TA)
B Std. error Beta
SBT (Constant) .036 .004 8.459 .000
GNPA/TA -.371 .073 -.796 -5.086 .000
PSB (Constant) .029 .003 10.657 .000
GNPA/TA -.110 .054 -.469 -2.054 .058
For PSB the NPA (b=-0.110) is not significant (p=0.058), but only just so (as it is marginally
above .05), and the coefficient is negative which would indicate that larger NPA is related to
lower profitability, which is what we would expect. For SBT the effect of NPA is significant (p <
.05) and its coefficient is negative indicating that larger NPA is related to lower profitability.
Hence, we can conclude that these co-efficients are not zero for the related predictors only.
241
7.5.2.3.1.4. Impact Analysis of the Independent Variable:
Our main objective of developing regression models is to find the impact of NPAs over
NIM/TA. From the above output table we found our regression equations given below to get the
impact of GNPAs in the models.
Table- 7.51.: Model Equations for NIM/TA and NPA (GNPA/TA)
Bank Equation
SBT NIM/TA = 0.036-0.371*( GNPA/TA)
PSB NIM/TA = 0.029-0.110*( GNPA/TA)
1. SBT: for one unit of change in GNPA/TA, NIM/TA is effected by -0.371 units. Clearly
the impact on NIM/TA is negative and nearly 37.1% of GNPA/TA.
2. PSB: for one unit of change in GNPA/TA, NIM/TA is effected by -0.110 units. Clearly
the impact on NIM/TA is negative and nearly 11% of GNPA/TA.
7.5.2.3.2. Multiple Regression between NIM/TA and NPA (GNPA/TA) & Time (Year)
Table- 7.52.: Regression Model Assumptions Verification for NIM/TA and GNPA/TA & Year
SBT 1.873 .937 .285
PSB 1.253 .965 .733
DW statistics are both all higher than 1.253. This suggests that the assumption of independent
(no serious serial correlation) errors has been met for both the models. The p values of the SW
statistic of both the models are more than 0.05. This suggests that the assumption of normality of
the error distribution has been met for both the models.
7.5.2.3.2.2. Co linearity Analysis of the Models:

242
Table- 7.53.: Regression Model Co linearity Statistics for NIM/TA & GNPA/TA and Time (Year)
Tolerance VIF
SBT (Constant)
YEAR .171 5.836
GNPA/TA .171 5.836
PSB (Constant)
YEAR .078 12.893
GNPA/TA .078 12.893
We interpret the tolerance for NPA and Year for the model PSB is below 0.1, but VIF is >10. We
found the basis of redefining of these two models once again. The facts indicate the elimination
of variable, which is absurd in these cases. So, we conclude linear model incorporating Year is
not possible for PSB. So, we are dropping it. The VIF for the other model (SBT) is quite
satisfactory.
7.5.2.3.2.3.1 Model Summary:

Table- 7.54.: Regression Model Summary for NIM/TA and GNPA/TA and Time (Year)
SBT .825 0.68 2 14.902 .000
We see that R is very high (0.825) and R2 is 0.680 for SBT. In other words we can say that the
proportion of variance of the NIM/TA of SBT that is “explained” or “accounted for” by
knowledge of the NPA and Year of SBT is approximately 68%.
We can say a high quality of model fit is exhibited for model of SBT based on our data. The F-
ratio of SBT is 14.902. The test of significance reveals that the probability of obtaining an F-stat
as the one we’ve obtained on the model degrees of freedom are very smalls, for SBT is less than
.01. Hence, we can conclude the goodness of fit for all models are well established for SBT at
99% confidence level). So, the decision of multiple regression analysis incorporating Time
(Year) only for SBT is justified here as inclusion of time has improved the precision level of the
model.
243
7.5.2.3.2.4. Model Co-efficient Analysis:
A summary table of the coefficients of the models is given below.
Table- 7.55.: Regression Model Coefficients, t stat for NIM/TA & GNPA/TA and Time (Year)
B Std. error Beta
SBT (Constant) 0.062 .019 3.315 .005
YEAR -0.002 .001 -.526 -1.441 .172
GNPA/TA -0.593 .170 -1.274 -3.491 .004
We notice that p-value of the predictor GNPA/TA (marked) is small, less than 0.05, and hence,
we can conclude that this co-efficient is not zero for the related predictors only and this
coefficient is not by chance at 95% confidence level and the other coefficient could be due to by
chance .
7.5.2.3.2.5. Impact Analysis of the Independent Variable:

Our main objective of developing regression models is to find the impact of NPAs and Time
(Year) over NIM/TA. From the above output table we found our regression equations given
below to get the impact of predictors in the models.
Table- 7.56.: Model Equations for NIM/TA and GNPA/TA & Time (Year)
Bank Equation
SBT NIM/TA = 0.062-0.593*(GNPA/TA)-0.002*(YEAR)
SBT: -0.593 times of GNPA/TA and -0.002 times of (YEAR) are responsible to predict
NIM/TA of SBT. Clearly the above quantities are the negative impacts of the predictors on
NIM/TA of SBT.
7.5.2.3.3. Summary of findings of Impact of NPA on NIM/TA

(i) NPA (GNPA/TA) of SBT has been found to have statistically significant (p value less
than 0.05) negative impact on NIM/TA and R2 is moderately high at .633. NPA of PSB
also has negative impact on NIM/TA with p value marginally higher than 0.05 and R2is
low at .220.
(ii) With the inclusion of Time (Year) in the predictor variable, multiple regressions
modeling finally restricted to SBT. The other bank in the earlier modeling of the
dependent variable NIM/TA has been dropped due to presence of multi co linearity
244
beyond tolerance level. NPA of SBT have negative impact on NIM/TA with p value of
coefficient for the predictor NPA is less than 0.05, while predictor Time (Year) is more
than 0.05. With the introduction Time (Year) in the predictor variable, R2 in the model is
still very high for SBT and has improved from .633 in the earlier model to .680.
7.5.2.4 Impact of NPA on CDR

To examine the impact of NPA (GNPA/TA) over CDR, simple linear regression is tried.
7.5.2.4.1 Simple Regression between CDR and NPA (GNPA/TA)
7.5.2.4.1.1 Evaluating Model Fit and Verifying Assumptions
Table- 7.57.: Regression Model Assumptions Verification for CDR and NPA (GNPA/TA)
SBI .327 .948 .426
SBT .604 .927 .191
SB .639 .894 .054
UCO .593 .936 .271
CBI .526 .952 .489
PNB .891 .944 .371
PSB .355 .970 .816
Dependent Variable: CDR
DW statistics are all less than 0.9. This suggests that the assumption of independent (no serious
serial correlation) errors has been violated seriously and models are highly affected by serial
correlations. All the p values of the SW statistic are higher than 0.05. This suggests that the
assumption of normality of the error distribution has been met for all the models.
Though a high negative linear correlation presents between the CDR and GNPA/TA, but due to
the violation of linear model assumptions we feel developing a linear model concept would be
nonsense idea. Since the linear models are not suitable for all the banks, we are in search of an
alternate. We tried for the conventional alternate models. But, we found no suitable model
combination for the selected PSBs in our limited study area. So, we are dropping these models.
7.5.2.4.2. Multiple Regression between CDR and NPA (GNPA/TA) & Time (Year)
245
Table- 7.58.: Regression Model Assumptions Verification for CDR and NPA (GNPA/TA) & Time (Year)
SBI .510 .962 .670
SBT .657 .929 .210
SB 1.473 .900 .067
UCO 1.102 .862 .017
CBI .683 .953 .510
PNB .868 .905 .081
PSB .518 .964 .698
DW statistics of few (SB & UCO) are higher than 1.1, rest are below .87. This suggests that the
assumption of independent (no serious serial correlation) errors has not been met for all the
models other than SB and UCO. The assumption of normality was tested by SW test and the p
values of the SW Statistic all higher than 0.05, with the exception of UCO. This suggests that the
assumption of normality of the error distribution has been met for all the models (at 95%
confidence level) except UCO. Thus, the model SB is justified by independence and normality of
the error distribution. Since the linear models are not suitable for the other banks, we are in
search of an alternate. We have tried for the conventional alternate models. But, we found no
suitable model combination for the other banks in our limited study area. So, we are keeping
only SB model and analyze it further and drop the other models.

Table- 7.59.: Regression Model Co linearity Statistics for CDR and NPA (GNPA/TA) & Time (Year)
Tolerance VIF
SB (Constant)
GNPA/TA .145 6.905
YEAR .145 6.905
We interpret the tolerance for NPA (GNPA/TA) and Time (Year) for the model SB is above 0.1
and VIF is <10. Hence co linearity is not an issue in model SB.

246
Table- 7.60.: Regression Model Summary for CDR and NPA (GNPA/TA) & Time (Year)
SB .972 0.944 2 118.57 .000
From the above table, we see that R is generally higher than 0.97 and R2 is higher than 0.94. In
other words we can say that the proportion of variance of the CDR of SB is “explained” or
“accounted for” by knowledge of the NPA (GNPA/TA) & Time (Year) of SB is approximately
94.4 %. We can say a high quality of models fit is exhibited here. So, the decision of multiple
regression analysis, including time is justified for this model.
The F-ratio is 118.570 and its sig is less than .05 for SB. Hence, we can conclude the goodness
of fit for SB model is well established.

Table- 7.61.: Regression Model Coefficients, t stat for CDR and NPA (GNPA/TA) & Time (Year)
B Std. error Beta
SB (Constant) .385 .080 4.781 .000
YEAR .024 .005 .869 5.243 .000
GNPA/TA -.693 1.047 -.110 -.662 .519
We notice that p-values (marked) in respect of predictor Time (Year) are small, less than 0.05,
and hence, we can conclude that these co-efficients are not zero for the related predictors only
and these coefficients are not by chance at 95% confidence level and the other coefficients could
be due to by chance (valid only for this sample).

Our main objective of developing regression models is to examine the impact of CDR and NPA
(GNPA/TA) & Time (Year). From the above output table we found our regression equations
given below to get the impact of predictors in the models.
247
Table- 7.62.: Model Equations on CDR and NPA (GNPA/TA) & Time (Year)
Bank Equation
SB CDR= 0.385-0.693*( GNPA/TA)+ 0.024*(YEAR)
1. SB: -0.693 times of GNPA/TA and 0.024 times of YEAR are responsible to predict CDR
of SB. Clearly the above quantity is the impact of the predictor on CDR of SB.
7.5.2.4.3. Summary of findings of Impact of NPA on CDR

(i) As simple linear as well as conventional model are not feasible with the dataset, though a
high negative linear correlation is present between the CDR and NPA (GNPA/TA), no
impact analysis could be performed.
(ii) With the inclusion of Time (Year) in the predictor variable, multiple regressions
modeling finally restricted to SB. As simple linear as well as conventional non linear
model are not feasible with the dataset for the rest of the banks, idea of modeling on such
banks have dropped. NPA of SBT has negative impact on CDR. With the introduction
Time (Year) in the predictor variable, R2 in the model is still very high for SBT at .944.
7.5.2.5. Impact of NPA on Credit Growth (CDRGROWTH)

To examine the impact of NPA (GNPA/TA) over CDRGROWTH, simple linear regression is
tried.
7.5.2.5.1. Simple Regression between: CDRGROWTH and NPA (GNPA/TA)
Table- 7.63.: DW and SW test statistics for CDRGROWTH and NPA (GNPA/TA)
SBT 1.512 .915 .120
UCO 1.442 .956 .555
CBI 1.284 .955 .538
PNB 1.727 .920 .148
PSB 1.418 .961 .642
Dependent Variable: CDRGROWTH
All the DW statistics are higher than 1.28 and less than 1.73. This suggests that the assumption
of independent (no serious serial correlation) errors has been met for all the models. The p values
of the SW statistic are all higher than 0.05. This suggests that the assumption of normality of the
248
error distribution has been met for all the models. So, we keep all the models and analyze them
further.

Table- 7.64.: Regression Model Summary for CDRGROWTH and NPA (GNPA/TA)
SBT .610 .372 1 8.872 0.009
UCO .537 .288 1 6.074 0.026
CBI .475 .225 1 4.367 0.054
PNB .611 .374 1 8.956 0.009
PSB .597 .357 1 8.314 0.011
From the above table, we see that R is generally higher than 0.47 and less than 0.62 and R2 is all
higher than 0.22 and maximum .374. In other words we can say that the proportion of variance of
the CDRGROWTH of CBI that is “explained” or “accounted for” by knowledge of the NPA of
CBI is approximately 22.5% and 37.4 % for PNB. We can say an average quality of model fit for
all banks are exhibited here. So, the CDRGROWTH is not explained properly by NPA. Hence,
multiple regression analysis would be other alternative to get better models.
The F-ratio ranges from 4.367 to 8.956. The test of significance reveals that the probability of
obtaining an F-stat as the one we’ve obtained on the model degrees of freedom are very smalls,
all less than .05 except models of CBI (marginally higher). Hence, we can conclude the goodness
of fit for all models are well established at 95% confidence level but CBI at 90% confidence
level.

Table- 7.65.: Regression Model Coefficients, t stat for CDRGROWTH and NPA (GNPA/TA)
B Std. error Beta
SBT (Constant) .261 .031 8.349 .000
GNPA/TA -1.602 .538 -.610 -2.979 .009
UCO (Constant) .313 .050 6.245 .000
GNPA/TA -1.950 .791 -.537 -2.465 .026
249
CBI (Constant) .296 .056 5.330 .000
GNPA/TA -1.894 .906 -.475 -2.090 .054
PNB (Constant) .303 .035 8.747 .000
GNPA/TA -2.093 .699 -.611 -2.993 .009
PSB (Constant) .274 .028 9.633 .000
GNPA/TA -1.612 .559 -.597 -2.883 .011
by chance at 95% confidence level and the other coefficients could be due to by chance (for this
sample only).

Our main objective of developing regression models is to find the impact of NPA (GNPA/TA)
and other variables over CDRGROWTH. From the above output table we found our regression
equations given below to get the impact of predictors in the models.
Table- 7.66.: Model Equations on CDRGROWTH and NPA (GNPA/TA)

Bank Equation
SBT CDRGROWTH = 0.261-1.602*(GNPA/TA)
UCO CDRGROWTH = 0.313-1.950*(GNPA/TA)
CBI CDRGROWTH = 0.296-1.894*(GNPA/TA)
PNB CDRGROWTH = 0.303-2.093*(GNPA/TA)
PSB CDRGROWTH = 0.274-1.612*(GNPA/TA)
1. SBT: for one unit of change in GNPA/TA, CDRGROWTH is affected by -.060 units.
Clearly the impact on CDRGROWTH is negative and nearly 6.0% of GNPA/TA.
2. UCO: for one unit of change in GNPA/TA, CDRGROWTH is affected -1.950 units.
Clearly the impact on Credit Growth is negative and nearly 195.0% of GNPA/TA.
3. CBI: for one unit of change in GNPA/TA, CDRGROWTH is affected by -1.894 units.
4. PNB: for one unit of change in GNPA/TA, CDRGROWTH is affected by -2.093 units.
250
5. PSB: for one unit of change in GNPA/TA, CDRGROWTH is affected by -1.612 units.
7.5.2.5.2. Multiple Regression between CDRGROWTH and NPA (GNPA/TA) & Time (Year)
In the table below we presented the summaries of the regression model assumptions verification
of the models:
Table- 7.67.: Regression Model Assumption Verification for CDRGROWTH and GNPA/TA & Time
(Year)
SBT 1.976 .947 .409
UCO 1.827 .850 .011
CBI 1.401 .941 .332
PNB 1.736 .922 .157
PSB 1.677 .962 .667
DW statistics are higher than 1.40 and lower than 1.98. This suggests that the assumption of
independent (no serious serial correlation) errors has been met for all the models. The p values of
the SW statistic of the all models with the exception of UCO are higher than 0.05. This suggests
that the assumption of normality of the error distribution has been met for all the models except
UCO (at 95% confidence level). So, we keep all the models except UCO and analyze them
further.

A summary table of the Co linearity stats of the models is given below.
Table- 7.68.: Regression Model Coefficients, t stat, Co linearity Statistics for CDRGROWTH and NPA
( GNPA/TA) & Time (Year)
Tolerance VIF
SBT (Constant)
Year .171 5.836
GNPA/TA .171 5.836
CBI (Constant)
Year .078 12.824
GNPA/TA .078 12.824
PNB (Constant)
Year .093 10.708
GNPA/TA .093 10.708
PSB (Constant)
251
Year .078 12.893
GNPA/TA .078 12.893
We interpret the tolerance for GNPA and Year for the models CBI, PNB and PSB are below 0.1,
but VIF is >10. But, removing the variable from the models would change the models into a
single predictor model (which we have already developed earlier). So, we are dropping the above
models, but are keeping SBT for further study.

Table- 7.69.: Regression Model Summary for CDRGROWTH and NPA ( GNPA/TA) & Time (Year)
SBT .699 .488 2 6.676 .009

From the above table, we see that R is generally higher than 0.69, while R2 is higher than 0.48. In
other words we can say that the proportion of variance of the CDRGROWTH of SBT that is
“explained” or “accounted for” by knowledge of the NPA and Year of SBT is approximately
48.8%. So, the decision of multiple regressions incorporating time for models of SBT is justified.
The F-ratio of SBT is 6.676 and its sig is less than .05. Hence, we can conclude the goodness of
fit for SBT model is well established.

Table- 7.70.: Regression Model Coefficients, t stat for CDRGROWTH and NPA (GNPA/TA) & Time
(Year)
B Std. error Beta
SBT (Constant) .495 .134 3.691 .002
GNPA/TA -3.574 1.214 -1.360 -2.945 .011
Year -.016 .009 -.825 -1.785 .096
252
by chance at 95% confidence level and the other coefficients could be due to by chance (for this
sample only).

Our main objective of developing regression models is to find the impact of NPA (GNPA/TA) &
Time (Year) over CDRGROWTH. From the above output table we find our regression equations
given below to get the impact of predictors in the models.
Table- 7.71.: Model Equations on CDRGROWTH and NPA (GNPA/TA) & Time (Year)
Bank Equation
SBT CDRGROWTH = 0.495-3.574*( GNPA/TA)-0.016*(YEAR)
1. SBT: -3.574 times of GNPA/TA and -0.016 times of YEAR are responsible to predict
CDRGROWTH of SBT. Clearly the quantities are the mathematical measures of the
impacts of the predictors on CDRGROWTH of SBT and it is negative.
7.5.2.5.3. Summary of findings of Impact of NPA on CDRGROWTH

(i) NPA of all selected PSBs, with the exception of CBI have been found to have statistically
significant (p value less than 0.05) negative impact on CDRGROWTH, though at varying
degree. NPA of CBI also has negative impact on CDRGROWTH with p value of
coefficient marginally higher than 0.05. R2 is low, ranging from minimum of .225 for
CBI to maximum .374 for PNB.
finally restricted to SBT. The other banks in the earlier modeling of the dependent
variable CDRGROWTH has been dropped due to presence of multi collinearity beyond
tolerance level. NPA of SBT has negative impact on CDRGROWTH with p value of
coefficient for the predictor NPA is less than 0.05, while for predictor Time (Year) is
more than 0.05. R2 in the model is moderate at .488 for SBT.
7.5.2.6. Impact of NPA on CAR

To examine the impact of NPA (GNPA/TA) over CAR, simple linear regression is tried.
7.5.2.6.1. Simple Regression between: CAR and NPA (GNPA/TA)
253
Table- 7.72.: Regression Model Assumptions Verification for CAR and NPA (GNPA/TA)
SBT 2.050 .925 .182
UCO 2.309 .871 .023
CBI 1.097 .860 .015
PNB .987 .898 .063
PSB .690 .967 .767
Dependent Variable: CAR
DW statistics are all higher than 2.04 except for the models CBI, PNB and PSB. This suggests
that the assumption of independent (no serious serial correlation) errors has been met for all the
models except for the models CBI, PNB and PSB. The p values of the SW statistic are all higher
than 0.05 for all model except, UCO and CBI. This suggests that the assumption of normality of
the error distribution has been met for these models. In other words, the model of SBT is
justified by Independence and Normality of the error distribution. So, we are keeping this model
only for further analysis.

Table- 7.73.: Regression Model Summary for CAR and NPA (GNPA/TA)
SBT .577 .333 1 7.480 .015
We see that R is 0.577 and R2 is 0.333. In other words we can say that the proportion of variance
of the CAR of SBT that is “explained” or “accounted for” by knowledge of the NPA of SBT is
approximately 33.3%. We can say an average quality of model fit for SBT is exhibited here.
Hence, multiple regressions incorporating time as a variable is examined to get better models.
The F-ratio is 7.480. The test of significance reveals that the probability of obtaining an F-stat as
the one we’ve obtained on the model degrees of freedom is very small, less than .05. Hence, we
can conclude that the goodness of fit for the model (SBT) is adequately established.

254
Table- 7.74.: Regression Model Coefficients, t stat for CAR and NPA (GNPA/TA)
B Std. error Beta
SBT (Constant) 0.120 .004 32.839 .000
GNPA/TA -0.171 .063 -.577 -2.735 .015
We notice that both the p-values (marked) are small, less than 0.05, and hence, we can conclude
by chance at 95% confidence level.

Our main objective of developing regression models is to find the impact of NPAs and other over
CAR. From the above output table we found our regression equations given below to get the
impact of predictors in the models.
Table- 7.75.: Model Equations on CAR and NPA (GNPA/TA)
Bank Equation
SBT CAR = 0.120-0.171*(GNPA/TA)
1. SBT: for one unit of change in GNPA/TA, CAR is affected by -0.171 units. Clearly the
impact on CAR is negative and nearly 17.1% of GNPA/TA.
7.5.2.6.2. Multiple Regression between CAR and NPA (GNPA/TA) & Time (Year)
7.5.2.6.2.1 Evaluating Model Fit and Verifying Assumptions
Table- 7.76.: Regression Model Assumptions Verification for CAR and GNPA/TA & Time (Year)
SBT 2.006 .931 .229
UCO 2.313 .875 .026
CBI 1.189 .860 .015
PNB .971 .896 .058
PSB .961 .943 .356
DW statistics all are higher than 1.1 and lower than 2.4 except for the model PNB and PSB. This
suggests that the assumption of independent (no serious serial correlation) errors has been met
for all the models except for PNB and PSB. The p values of the SW statistic are higher than 0.05
for SBT, PNB and PSB. This suggests that the assumption of normality of the error distribution
255
has been met for the models of the said banks only. In other words, the model SBT is justified by
independence and normality of the error distribution. So, we keep this model only for further
analysis.

Table- 7.77.: Regression Model Coefficients, t stat, Co linearity Statistics for CAR and GNPA/TA &
Time (Year)
Tolerance VIF
SBT (Constant)
Year .171 5.836
GNPA/TA .171 5.836
We interpret the tolerance for GNPA and Year for the model SBT is greater than 0.1, and VIF is
<10. Hence, there is no serious co linearity issue.

Table- 7.78.: Regression Model Summary for CAR and GNPA/TA & Time (Year)
SBT .577 0.333 2 3.497 .059
The model summary contains R is 0.577. R2 is 0.333. In other words we can say that the
proportion of variance of the CAR of SBT that is “explained” or “accounted for” by knowledge
of the NPA and Time (Year) approximately 33.3 %. We can say a below average quality of
models fit exhibit here. We also observed that after incorporating Time (Year), no dramatic
changes are found for SBT.
The F-ratio is 3.497. The test of significance reveals that the probability of obtaining an F-stat as
the one we’ve obtained on the model degrees of freedom are marginally higher than 0.05. Hence,
we can conclude the goodness of fit for the model is moderately established.

256
Table- 7.79.: Regression Model Coefficients, t stat for CAR and GNPA/TA & Time (Year)
B Std. error Beta
SBT (Constant) 0.118 .017 6.828 .000
Year 0.000 .001 .049 .092 .928
GNPA/TA -0.158 .157 -.533 -1.010 .330
We notice that all p-values are more than 0.05, and hence, we have evidence to reject the null
hypothesis that the co-efficients are zero for the related predictors and the coefficients could be
due to by chance (valid only for this sample).

Our main objective of developing regression models is to find the impact of GNPAs and other
variables over CAR. From the above output table we found our regression equations given below
to get the impact of predictors in the models.
Table- 7.80.: Model Equations on CAR and NPA (GNPA/TA) & Time (Year)
Bank Equation
SBT CAR = 0.118 - 0.158*( GNPA/TA) -0.000*(YEAR)
1. SBT: -0.158 times of GNPA/TA and 0.000 times of YEAR are responsible to predict
CAR of SBT. Clearly the quantities are the mathematical measures of the impacts of the
predictors on CAR of SBT and it is negative.
7.5.2.6.3. Summary of findings of Impact of NPA on CAR

(i) NPA of SBT has been found to have statistically significant (p value less than 0.05)
negative impact on CAR and R2 is low at .333 for SBT. As simple linear as well as
conventional alternate model are not feasible with the dataset for other banks, idea of
modeling with CAR and NPA, has been dropped.
finally restricted to SBT. NPA of SBT have negative impact on CAR with p value of
coefficient for most of the predictors are higher than 0.05, putting question mark on the
statistical validity of the such model. R2 in the SBT model is surprisingly same even with
inclusion of Time in the predictor variable.
257
1. The main objective of the research work undertaken is to obtain a quantitative assessment of
the relationship between NPA and selected strategic banking variables including profitability,
productivity and few others and to examine the impact of NPA on such dependent variables.
It is observed from above detailed study that NPA has different degree of association and
impact on selected variables among six selected PSBs.
2. Correlation Analysis given in Table 6.14 exhibits the following relationship
(i) Profitability represented by PBT/TA of all PSBs, with the exception of SBI have been found
to have statistically significant (p value less than 0.05) negative relationship with NPA
(GNPA/TA), though at varying degree. Profitability of SBI also has negative relationship
with NPA with p value marginally higher than 0.05.
(ii) Productivity represented by OPE of all PSBs, have been found to have statistically significant
(p value less than 0.05) negative relationship with NPA (GNPA/TA), though at varying
degree.
(iii)NIM/TA of SBT has been found to have statistically significant (p value less than 0.05)
negative relationship with NPA (GNPA/TA), while for PSBs similar relationship is observed
with p value marginally higher than 0.05. The relationship of other banks are not statistically
significant and hence left out from the scope of our study.
(iv) CDR of all PSBs, have been found to have statistically significant (p value less than 0.05)
negative relationship with NPA (GNPA/TA), though at varying degree.
(v) CDRGROWTH of SBT, UCO, PNB and PSB have been found to have statistically
significant (p value less than 0.05) negative relationship with NPA (GNPA/TA), while for
CBI similar relationship is observed with p value marginally higher than 0.05.
CDRGROWTH of other banks like SBI and SB though are having negative relationship with
NPA (GNPA/TA), are not statistically significant and hence left out from the scope of our
study.
(vi) CAR of all selected PSBs with the exception of SBI and SB have been found to have
statistically significant (p value less than 0.05) negative relationship with NPA (GNPA/TA),
while for SBI and SB the relationship are not statistically significant and hence left out from
the scope of our study.
3. Impact Analysis examined with the help of simple as well as multiple regression reveal
258
(i) NPA (GNPA/TA) has negative impact on profitability for all the selected banks along with
PSBs in aggregate both in respect of simple regression with NPA (GNPA/TA) as predictor and
multiple regressions where along with NPA (GNPA/TA), selected strategic banking variables
are predictors. It is also noted that impact of these selected strategic banking variables are not
same for different PSBs. The multiple regressions have also improved the precision of the
modeling as evident from the tremendous increase in the R2 value of the models in
comparison to models obtained with the help of simple regression. It has been also observed
that with time and NPA as predictors only SBI, SBT and UCO are found to have statistically
significant impact on profitability. However, with the inclusion of other statistically
significant strategic variables in predictor only SBI and SB are found to have statistically
significant impact on profitability.
(ii) In case of productivity (OPE) we observe that though it has strong negative correlation with
NPA for all the banks but simple linear model cannot be established. However, with the
inclusion of time in the predictor, models with very high R2 (between 0.917 to 0.978) have
been established for all the banks excepting CBI and PSBs in aggregate.
(iii) Since NIM/TA has high negative correlation with SBT and moderate correlation with PSBs
in aggregate, statistically significant models have been established in these cases. However,
with the inclusion of time in the predictor, model for SBT can only be developed.
(iv) In case of CDR we observe that though it has strong negative correlation with NPA for all
the banks but simple linear model cannot be established. However, with the inclusion of
time in the predictor in case of SB, model with very high R2 (.944) has been established.
(v) Since CDRGROWTH has negative correlation for all the banks except SBI and SB,
statistically significant models have been established in these cases though R2 value is not
good. With the inclusion of time in the predictor, model for SBT can only be developed with
moderate R2 value.
(vi) CAR has been found to have negative correlation for all the banks except SBI and SB, but
statistically significant models could be established in case of SBT only though R2 value is
not good. With the inclusion of time in the predictor, model for SBT can only be developed
with low R2 value
259
7.6. Conclusion
An in-depth analytical study to examine the impact of distress asset over strategic banking
variables exposes risk associated with the degree of distress assets in the banking sector. An
important observation from the study is that NPA has significant impact on various strategic
banking variables across the PSBs in general, though at varying degree for different PSBs. The
study also reveal that most of the selected PSBs are found to be badly hit by NPA in respect of
variable like profitability, while a few such PSBs are also exposed significantly to other selected
strategic banking variables. All PSBs are subjected to strict surveillance of RBI and are operating
under same economic conditions. Such wide differences demonstrate that some PSBs are more
efficient in handling the critical aspect of NPA in banks than their competitors in India.
260
ChapterVIII
Summary, Suggested Action Plan and Conclusion
8.1. Summary of the Thesis
8.2. Suggested Action Plan based on the findings emanated from chapters
4, 5, 6 and 7.
8.3. Uniqueness of the Research work
8.4. Social Benefits of the Research
8.5. Future Application of the Thesis
8.6. Scope of Future Research on the Subject
8.7. Conclusion
261
8.1. Summary of the Thesis
The first chapter begins with a brief study on status of NPA in Indian banking which reveals
thatGNPA as a percentage of GA have shown marked improvement with the exception of last
few years, though GNPA in absolute figure is still very high and is of paramount concern for
all stake holders and therefore has been identified as the subject matter of study.
A detailed survey and review of relevant chapters of different books on banking by famous
authors and research papers published in national and international journals with respect to
subject matter of the thesis have been made with reasonable precision. In view of the
seriousness of the problem, numerous research studies have been conducted on different
issues concerning credit risks in banks including NPAs.However, empirical works on NPA
problems in Public Sector Banks (PSBs) are inadequate. The exhaustive review of literature
on NPA demonstrates that majority of the research work has been undertaken on aggregate
PSBs data with primary focus on impact analysis and modeling with regression technique.
However empirical work on individual bank-wise study of movement of NPA over time and
examination of impact of NPA on banking variables has seldom been attempted.
Accordingly, examination of trend of NPA and forecasting the same for medium term with
the help univariate time series data by employing parametric (Curve Estimation Technique of
Regression Analysis), non parametric (LOESS Curve Fit Technique) and semi parametric
(Penalised Spline Curve Fit Technique) has been attempted in our study. Similarly, for
examining the impact of NPA on strategic banking variables along with time variable in
selected PSBs, simple as well as multiple regression analysis have been undertaken.
Objective, Research Methodology and Scope of the study has been documented with
reasonable precision, objectivity and simplicity. The chapter ends with scheme of the thesis,
giving a fair idea of distribution of the total research work.
The second chapter ‘Existing Prudential Norms on Asset Quality’begins with the
background of the prudential norms, as prescribed by Narasimham Committee and
implemented by RBI, followed by review of a few relevant literatures on the chapter.
Circularized instructions on prudential norms in general with NPA in particular issued by
RBI from time to time since 1992 have been examined to appreciate various norms and
concepts relating to NPA and desired procedures as are followed by individual banks.
Prudential norms on NPA essentially revolve around three vital aspects, ie, income
recognition, asset classification and provisioning. Income recognition norms tell us that
income of a bank (interest and other charges) can be booked only when the same is
262
realized. As per asset classification norms, an asset may be either a performing asset
which services interest and bank charges without any difficulty or a NPA, if the interest
charged during any quarter is not serviced fully within 90 days from the end of the
quarter. Classification of NPAs into three categories, namely sub standard asset, doubtful
asset and loss asset is done taking into account the degree of well defined credit
weaknesses, risks and extent of dependence on collateral security for realization of dues.
Provisioning norms are applicable for both performing assets as well as NPAs. In case of
Performing Assets a general provision is created depending on the nature of advance. In
case of NPAs it is based on asset classification, reckoning the period for which the asset
has remained non-performing and the availability of security with its realizable value.
Treatment of income recognition, asset classification and provisioning in respect of
various special types of advances followed by brief discussions on few operational
concepts on NPA like Out of Order’ Status, Overdue Status, Take-Out Finance, Early
Alert System, Special Mention Account, Floating Provision, GNPA and NNPA have
been captured in brief.
In chapter three, delineation on the mechanism for effective management of NPAs as initiated
by PSBs in India has been narrated based on researcher’s personal interaction with recovery
cell officials of such banks and RBI circularized instructions. All PSBs individually have
framed a Recovery Policy, duly approved by the Board with the primary objective of
adopting aggressive recovery measures and implementing case specific strategies, which may
include persuasion, restructuring, settlements, and legal methods, to ensure maximum and
timely recovery of dues in distress asset. A number of recovery mechanism including
(i)Review and Monitoring of Existing NPA Accounts, (ii) Turnaround through Restructuring
& Corporate Debt Restructuring (CDR), (iii) One Time Settlement (OTS)/ Negotiated
Settlement (NS), (iv) Compromise Settlements in Suit Filed Accounts, BIFR Accounts and
LokAdalats Cases, (v) Acquiring Real Estate Assets in satisfaction of Bank dues in NPA
accounts, (vi) Sale of Assets under SARFAESI Act, 2002, (vii) Sale of NPAs to Asset
Reconstruction Companies (ARCs)/ Banks/ FIs/ NBFCs, (viii) Change of Management or
Induction of Strategic Investors and (ix) Legal Action used by PSBs have been captured in
brief for our study. Filing a suit against the borrower and the guarantors are considered as the
last option, at an appropriate stage of the account becoming NPA, after other feasible options
of recovery of dues are exhausted.
263
The fourth chapter, entitled, ‘Dynamicity of NPA over Recent Years in Selected PSBs in
India’ is devoted to an in-depth analytical study to reveal the movement of the financial
parameters, GNPA and GA, each of which, in effect, exposes the degree of distress assets in
the banking sector, over time in respect of six selected PSBs along with PSBs-aggregated.
Durbin Watson Test has been employed to diagnose the existence of independence in the
data-sets in respect of the parameters GNPA and GA. To examine the normality or non-
normality of the distributions followed by the different data sets on the above parameters,
normal Q-Q plots are constructed apart from finding out the critical values (from table) of the
Shapiro and Wilk Test Statistic. To determine the features of the above data sets, values of
the simple statistical measures, like, mean, range, standard deviation, coefficient of variation
along with measures of the properties, namely, skewness and kurtosis have been computed.
Parametric models (using time series regression analysis) have been employed to analyze the
trends existing in the above data-sets relating to the parameters, GNPA and GA respectively
over time. Finally, the best fitted model to represent the trend has been obtained in case of
each data-set, and the forecasted values based on the said best fitted model for the respective
data-set have been generated.
The chapter five, entitled, ‘Movement of NPA over Time in Selected PSBs in India with Non
Parametric Approach’ is devoted to an in-depth study to reveal the movement of the financial
parameters, GNPA and GA over time. Diagnostic tool like Quantile Comparison Plots of the
Residuals, Residuals vs Predictor Plot, Absolute Residual vs Predictor Plot,have been
employed to detect departures from normality in the residual distribution, to check the
presence of outliers in the data and to examine appropriateness of LOESS model for our data
respectively. It is observed that residuals follow approximately normal distribution, existence
of very few outliers in few selected banks and LOESS Curve (Quadratic) Model fits well
with the dataset. LOESS model establishes non linearity in the given data. Goodness of fit
statistics represented by R2LOESS, Sig of LOESS F statistics establishes high precision of the
model and excellent fit for dataset in respect of both the parameters. Finally, the LOESS
model is extended to get the forecasted values for the respective data-set for medium term,
which clearly shows future upward trend in respect of both the parameters for all the selected
PSBs along with PSB in aggregate.
The sixth chapter, entitled, ‘Modeling NPA Time Series Data in Selected PSBs in India with
Semi Parametric Approach’ is devoted to an analytical study to reveal the movement of the
financial parameters, GNPA and GAover time. Diagnostic tool like Quantile Comparison
264
Plots of the Residuals and Residuals vs Predictor Plot and QQ Plot have been employed and
to check the presence of outliers in the data, to examine appropriateness of Penalized Spline
(Semi Parametric Curve Fit) model for the given dataset and to detect departures from
normality in the residual distribution for our data respectively. It is observed that residuals
follow approximately normal distribution, existence of very few outliers in few selected
banks and Penalized Spline (Semi Parametric Curve Fit) model fits well with the dataset.
Penalized Spline (Semi Parametric Curve Fit) Model establishes curve linearity in the given
data. Goodness of fit statistics represented by R2, Sig of F statistics establishes high precision
of the model and excellent fit for dataset in respect of both the parameters.Finally, the
Penalized Spline (Semi Parametric Curve Fit) model is extended to get the forecasted values
for the respective data-set, which clearly demonstrates future upward trend in respect of both
the parameters for all the selected PSBs along with PSB in aggregate also.
The chapter seven, entitled, ‘Examination of Impact of NPAs on Strategic Banking Variables
in Selected PSBs in India’ is devoted to examine impact of NPAs on selected banking
variables. Based on earlier studies in the field of NPA in banks which have used different
banking variables as well as information gathered through personal interview with bank
officials, a few variables like NPAratio, Profitability ratio, Productivity ratio, Net Interest
Margin ratio, Credit Deposit ratio, Capital Adequacy ratio and ratio of Credit Growth etc
have been initially identified. For the purpose of examining degree of association and
relationship between variables identified, simple pair wise correlation co-efficient has been
computed and appropriate tests have been employed to judge the significance status of these
co-efficient. This has enabled us to select the statistically significant variables to be
considered for further impact study. For the purpose of examining impact of NPA on the
profitability of banks simple linear regression have been attempted. Durbin Watson Test of
independence of the errors and Shapiro-Wilk test of normality of the error distribution has
been employed to satisfy requirement of linear regression. For few banks, which could not
qualify linearity requirement as stated above, we have examined alternate models. To
examine the impact of NPA along with time variable, multiple regression modelsare invoked.
Same regression process has alsobeen followed for other statistically significant banking
variables as stated above.
However, apart from NPA, as there are number of variables affect profitability of a bank, it is
not appropriate to judge profitability with NPA alone, few other strategic banking variables
along with NPAratio have been considered through multiple regression model which consist
265
of determining a regression equation explaining relationship between dependent variables
(profitability) with the independent ones. To diagnose the problem of multi co-linearity in
multiple regressions, the value of tolerance factor (TOL) along with variance inflating factor
(VIF) have also been computed and compared with standard.
8.2.SuggestedAction Plan based on the findings emanated from Chapters 4, 5, 6 and 7.
8.2.1. Introduction
The main purpose of the thesis is to examine movements in GNPA and GA sample dataset of
seventeen years over time and to build models (based on different techniques) to forecast
future values in respect of parameters GNPA and GA and also to undertake regression
analysis to examine impact of NPA on strategic banking variables in selected PSBs in India.
The study reveals a non linear upward trend in classified debt over the sample period and
forecast steep rise in the next three years, which is an alarming situation for the growing
economy in general with banking industry in particular. Moreover, it has been found that
NPA has statistically significant negative impact on profitability with all the selected PSBs
along-with PSBs in aggregate. In respect of other selected banking variables, NPA has
statistically significant impact only with respect to few of the selected PSBs. In short, it may
be concluded that arresting the steep growth in NPA is the prime challenge with the policy
makers in PSBs.
8.2.2. Factors Responsible for Growth in NPA
Before discussing the strategic action points to combat NPA and its growth, we may briefly
have a look at the factors responsible for such an unwarranted situation. Based on
interaction with the bankers and survey of research papers on the subject, we have identified
following major factors, which contribute heavily towards accumulation of NPA in PSBs.
i) External factors such as global recession, downturn in industry, natural calamities, etc
ii) Cumbersome legal system.
iii) Delays in settlement proposals and implementation of rehabilitation package in respect

of BIFR cases.
iv) Non co-operative and dishonest attitude of the borrower.
v) Lack of adequate follow-up in recovery of dues.
vi) Absence/ Inadequate market or resale value of assets charged.
266
vii) Advances not backed by enforceable security.
viii) Genuine business failure.
8.2.3. Strategic Action Plan to Contain NPA to a Tolerance Level
Though arresting the growth of NPA is really a challenging task but it is certainly possible to contain
it’s growth. In this direction, a few comprehensive action plans have been suggested to contain the
growth of distress asset as under.
8.2.3.1. General Action Plan
Top priority has to be accorded towards management of NPA with a view to improve overall
quality of asset. A two pronged approach comprising (i) preventive measures and (ii) curative
measures are advocated to contain the rising NPAs. It is observed that even a good appraised
account tends to become nonperforming for lack of proper monitoring. Ensuring compliance
of sanction terms, periodical verification of primary are collateral securities, proper valuation
of securities charged in favour of the bank, identification of early warning signals for
initiating timely preventive measures, monitoring and follow-up are treated as nucleus of
effective Credit Management System as it arrests degradation in the quality of assets. Signals
can be picked up at the incipient stage of sickness. The bank should identify these stress
symptoms and take corrective steps to avoid slippage into NPA category. It is pertinent to
emphasize that every loan should be under regular surveillance when it is under performing
category. Therefore, proactive steps should be initiated at the right time to prevent fresh
addition to NPAs. The branches/ offices should put more impetus on credit appraisal/
sanctioning/ documentation/ monitoring/ review/ renewal etc to restore quality of the asset. In
cases where the accounts show symptoms of irregularity/ sickness for reasons beyond the
control of the borrower, the same should be restructured/ rescheduled/ rehabilitated in early
stage itself when it is detected in order to prevent its slippage to NPA.
An illustrative list of such signals is as under:
i) Poor credit turnover in the account.
ii) Credit turnover is not commensurate with the sales estimate.
iii) Routing of transactions through other bank /Non memberbanks.
iv) Frequent request for allowing over drawing in the account.
v) Delay in submission of stock statement/ other control statements/ financial statements.
267
vi) Return of cheques issued by borrowers.
vii) Frequent devolvement of Letter of Credit and non-payment within a reasonable period.
viii) Frequent invocation of Bank Guarantee and non-payment within a reasonable period.
ix) Return of bills/cheques discounted.
x) Poor financial performance in terms of declining sales and profits, cash losses, net
losses, erosion of net worth etc.
xi) Incomplete documentation in terms of creation/registration of charge/ mortgage etc.
xii) Non-compliance of terms and conditions of sanction.
8.2.3.2. Autonomy in credit decisions:
PSBs in general are subjected to strict compliance of government directives in all matters
including credit policy and credit decision. Under the present liberalized scenario where an
individual PSB is also responsible for its bottom lines, it should be given full autonomy in the
matter of credit decisions. Referrals by the ministry to sanction large corporate loans should
be dispensed with.
8.2.3.3.Risk based pricing of loan products:
In an increasingly competitive environment, banks need to indulge in risk based modelling

and pricing instead of following purely conservative cost plus methods. For instance, not only
historical parameters and repayment track records should be considered in decisions but also
quantified behavioural parameters should be considered. Profile calls can also be taken while
on-boarding new loan relationships with clients. Pricing policy should be dynamic and while
quoting interest rate to a particular client, net interest margin and borrower’s credit history
should also be taken into consideration. This will encouragebetter credit culture and improve
credit worthiness of the customers.
8.2.3.4. Streamlining Credit and Systematic Risk Processes
Credit and Systematic Risk processes have to be streamlined. Exhaustive Credit Risk
Assessment System for Commercial and Institutional customers segment have to be
launched. Necessary methodology has to be evolved for determining and managing target
portfolio mix, classified by industry, borrower type, credit rating, product, maturity and large
exposure aggregates. Industry risk concentration exercises have to be undertaken as a part of
ongoing review of loan portfolio.
268
8.2.3.5. Greater due diligence for Priority Sector Lending (PSL)
With RBI increasingly keeping pressure on sanction and disbursement of PSL like
agriculture, SSIetc, the quality checks in place needs to be geared up. Debt burden ratio, and
collateralisation in PSL is necessary to ensure that the customers do not default in repayment
of loan instalment and accrued interest.
8.2.3.6. Management of Time Barred Documents
The Law of Limitation prescribes the period within which the existing rights can be enforced
in a court of law. The Limitation Act lays down that every suit instituted, appeal preferred,
and application made after the prescribed period shall be dismissed although limitation has
not been set up as a defense. As a remedial mechanism, ’Debt Acknowledgement’ in writing
and duly signed and stamped by the borrower, has to obtained before the expiry of the
limitation period.
8.2.3.7. Special Care for Compromise/ OTS/ legal/ BIFR cases
Compromise and one time settlement as measure of managing NPA relating to borrowers
who have failed to meet their obligation to the banks on account of genuine reasons are to be
encouraged. Loan compromise proposal have to be handled with care and on fast truck.
Besides, to speed up legal process, suit filed cases of Rs 10 lacs and above still pending in
various courts, within the jurisdictional area of the DRT functioning at a number of centers to
be transferred to the later. To expedite disposal of cases transferred to DRT, special legal cell
may be created at District Level. Mechanism of follow up of suit filed cases, especially those
filed with DRT, has to be streamlined. For close monitory DRT cases specially designated
nodal officer with appropriate knowledge and skill be appointed in legal/DRT Cell. As
regards BIFR referred cases bank has to work in close association with BIFR for
rehabilitating the sick units with basic viability.
8.2.3.8. Motivational package and training for employees
Employees have to be motivated towards participation in recovery drive and management of

NPA by formulating scheme for award of cash prizes/ certificate of merit and trophies to
branches showing outstanding performance in this respect.In order to further improve the
quality and management of bank’s credit portfolio, conscious efforts have to be made for up
gradation of credit skills at various levels through appropriate training interventions.
269
8.2.3.9. Introduction of Credit Audit Cell
A separate Credit Audit in addition for existing Inspection & Audit Dept. has to be set up to
improve efficiency and effectiveness of the appraisal, assessment, sanctions and disbursement
process and thereby overall quality of the bank’s credit portfolio. An effective credit audit
mechanism will generate warningsignals for taking appropriate action to prevent performing
asset turning into NPA.
8.2.3.10. Special Attention on High Value NPA
Special type of branches like Rehabilitation and Recovery branch may be set up where big
existing NPA be transferred for systematic follow up and appropriate remedial measures.
Task Force, at the corporate level, comprising top executives to monitor all big NPA accounts
(say Rs.5 crores and above), local controlling offices like Regional Office / Zonal Office for
monitoringall medium size NPA account ( say Rs.1 crore and above to less than Rs. 5 crores )
should be set up.
8.2.3.11. Restructuring and Up-gradation of Potential as well as existing NPA Accounts
Restructuring of impaired standard accounts as well as of viable NPA has to be an ongoing

process and to be religiously persuaded towards up-gradation of such assets in time.
8.2.3.14. Transparency in communication between bankers and FIs.
Regular and transparent communication and prompt feed- back among bankers and FIs in
locality to be encouraged for mutual benefit of all the lenders.
8.3. Uniqueness of the Research work
Despite the importance of monitoring non-performing loans, forecasting on parameters

GNPA and GA have only received moderate attention in literature. This study contributes to
the existing literature by modeling the economic parameters GNPA and GA of the PSBs in
India, not only on an aggregate level but also on an individual bank level. The inclusion of
the individual bank models provides a greater basis for comparison. Forecasting with
different approaches using different techniques for individual PSB along with PSBs in
aggregate as stated above is the uniqueness of the research work.
Of the three types of forecasting approaches, namely, Parametric, Non parametric and Semi
parametric, in respect of parameter GNPA, precision level of the Non parametric and Semi
parametric models, have outperform the Parametric ones convincingly, while in respect of
270
parameter GA, precision level of all the models namely Parametric and Non parametric and
Semi parametric are very high. Given the forecasting models, the thesis is capable of
providing insights into the stability of the financial system and is also of practical importance
for bankers/policy makers/regulators/financial engineers/researchers in terms of developing
plans to regulate the occurrence of NPLs in the future.
8.4. Social Benefits of the Research
Unimpressive asset quality is a matter of great concern to not only the lenders but also all
concerned including the people at large (society). High level of NPA implies scarce resources
of the bank get blocked, restricting the recycling of funds in the productive sectors of the
economy. Once the credit to various sectors of the economy slows down, the economy is
badly hit, resulting in increase in unemployment and poverty. Moreover mounting menace of
NPA raises the cost of credit, makes banks more averse to credit risk and as a result, genuine
small and medium entrepreneurs are denied of credit, which unfortunately, throttle their
enterprising spirits as well.
Again, when a NPA burdened PSB becomes sick with negative net worth, government comes
forward to rescue by way of infusion of capital to the sick bank. Such financial assistance
comes out of government’s budgetary resource which is contributed by the people in general,
either in the form of tax revenues or from the hard earned savings of the investing public. In
any case, the society is bearing the cost of these NPAs.
By forecasting NPA for medium term and examining the impact of NPA on banking
variables in a PSB, as attempted in the study,a bank will be in a position to initiate corrective
actions as recommended in the thesis towards improving level of distress asset in the industry
and a great relief for the society.
8.5.Future Application of the Thesis
The main purpose of constructing a time series model is to forecast. Such forecasts enable the
policy makers to judge whether it is necessary to take any measure to influence the relevant
economic variables. Given its association with bank failure and financial crisis, the evaluation
of non-performing loans (NPLs) is of great importance and should therefore be of interest in
a developing country like India. Areas of future application of such modeling in the industry
are enumerated below:
i) Future values of GNPA and GA, along with other strategic banking variables, can be
forecasted by other banks from the past performance using all the three methods
271
(parametric, non parametric and semi parametric) which has been illustrated in the
forecasting section of the relevant chapters of the thesis.
ii) With the help of these forecasted values the banks can formulate policy initiatives
through planning and budgeting to bridge the gap between the desired level of NPA and
forecasted level of NPA.
iii) Such models can be used to mitigate risk associated with credit portfolio as an early
warning system.
iv) Such forecasts enable the policy makers to judge whether it is necessary to take any
measure to influence the relevant economic variables.
v) Motivational schemes like promotion/ fringe benefits can be employed through

performance benchmarking based on forecasted value of GNPA and GA.
vi) The models can be used as a performance measurement means for the selected PSBs.
Some of the major areas are given as under :
a. reviewing the profitability of the bank to reflect changes in policies and their
underlying economics along with harmonization of these policies with risk
management methods,
b. determining the nature and impact of the critical factor that have greatest influence
on the bank’s performance,
c. redefining the profitability of the banks by effective management of NPA and

harmonizing new policy initiatives with NPA management, managing the
criticalities to align the components of performance management processes with
forecasting.
8.6. Scope of Future Research on the Subject
NPA is one of the most important banking variables for determining soundness and
efficiency of the monitory system in the economy. NPA has impact on number of strategic
banking variables which the thesis has already established for selected PSBs and PSBs in
aggregate. The study finds that NPA in PSB in India can be adequately captured by
parametric regression (curve estimation technique), non parametric regression (lowest fit) and
semi-parametric regression (penalized spline). In future the researchers can use this study as a
reference to examine the trend of NPA for other PSBs and PrSBs and develop appropriate
forecasting models and to examine the impact of NPA on strategic banking variables and
272
further compare these results among different group of banks for examining the relative
strengths and weaknesses of different banks. The researchers may also explore multivariate
modeling taking clue from this study which primarily focuses on univariate modeling.
8.7. Conclusion
Tightening of Prudential Norms in Indian banking industry to reach international standard

clearly demonstrates intention of RBI in containing growth in NPA. With continued global
recession, volatility in market driven factors and changes in socio-economic environment of
the country, restoration of asset quality has gained paramount importance. NPA is one of the
most important banking variables, which is having direct influence over the health and
prospect of the monetary system in an economy. NPAs have inevitable burden on the banking
industry. Hence, success of a bank depends upon its ability to manage NPAs and keeping
them within tolerance level. The spiraling and devastating effect of NPAs on the economy has
made NPA problem an issue of public debate and of national priority as well.
The study undertakes regression analysis to examine impact of NPA on strategic banking
variables in six selected PSBs, along-with PSBs in aggregate in India and finds statistical
evidence in support of general perception, though at varying proportion for selected PSBs and
PSBs in aggregate. Movement of GNPA and GA over time in respect of selected PSBs and
PSBs in aggregate has been adequately captured by parametric regression (curve estimation
technique), non parametric regression (lowest fit technique) and semi-parametric regression
(penalized spline technique). Surprisingly all the technique exhibited non linear feature and
upward trend in our dataset in respect of parameter GNPA and GA for all the PSBs and PSBs
in aggregate. Models of all the banks captured, are backed by precision level of high
magnitude. Based on our findings, we suggest that highly accurate forecasts may be obtained
through the use of uni-variate time series modeling.
Though arresting the growth of NPA is a very challenging task but it is certainly possible to
contain the growth. Bankers should evolve a two prone strategy and religiously implement
the same to put a break on the tremendous growth in NPA. The strategies are : (I) Blocking
further slippage of advances to NPA category and (II) Recovery of existing NPA. Of late,
several institutional mechanisms have been developed and are being religiously persuaded by
PSBs in India, to deal with NPAs (briefly narrated in Chapter 7)and there has also been
tightening of legal provision. However, it is felteffective management of NPAs requires an
appropriate internal checks and balance systems in a bank, which primarily depends on the
273
wisdom and mindset of the top management (Board). Forecasted values of GNPA for three
years (2013, 2014 and 2015) of all the selected PSBs clearly demonstrates future upward
trend, which is undoubtedly an alarming situation and deserves immediate and serious
attention on the part of the regulators to relook into the practices of credit appraisal and
monitoring in PSBs in India.
274
Bibliography
Agarwal, S., & Mittal, P. (2012). Non-Performing Asset: Comparative Position of Public and
Private Sector Banks in India. International Journal of Business and Management
Tomorrow, 2(1), 1-7. Retrieved on 10th February, 2013 from:
http://www.scribd.com/doc/86544915/Npa-Analysis-of-Data
Agnani, J. (2010). NPAs in Bank: A Syndrome Probing Remedy. International Journal of

Research in Commerce & Management, 1(5), 62-73. Retrieved on 10th February, 2013
from: http://www.ijrcm.org.in
Ahmed, J.U. (2010). The Quality of Asset Portfolio and Loan Recovery of Commercial Banks:
an Implication in Indian Context. AIUB Journal of Business and Economics, 9(2), 31-52.
Retrieved on 15th February, 2013 from:
http://dspace.nehu.ac.in/bitstream/1/1277/1/nj003.pdf
Ahluwalia, M.S. (2001, January 15-17). Second Generation Reforms in Banks: Major Issues. In
K.S. Shetty, J. Jain, K.C. Chakrabarty, N.V. Rao, R.K. Takkar, B.R. Prasad, & K. Ravi
(Eds.). Banking in the New Millennium. Paper presented at the Proceedings of Bank
Economists’ Conference 2000, New Delhi (pp. 114-117). New Delhi, India: Oriental
Bank of Commerce.
Basu, P. (2005). How Distressed is India’s Banking System. In P. Basu (Ed.), India’s Financial
Sector, Recent Reforms, Future Challenges (pp. 89 - 110). New Delhi, India: Macmillan
India Ltd.
Banerjee, A.V., Cole, S., & Duflo, E. (2004). Banking Reforms in India. India Policy Forum,
1(1), 277-332. Retrieved on 15th February, 2013 from:
http://www.google.co.in/url?sa=t&rct=j&q=&esrc=s&frm=1&source=web&cd=1&cad=r
ja&ved=0CCgQFjAA&url=http%3A%2F%2Fwww.brookings.edu%2F~%2Fmedia%2Fr
esearch%2Ffiles%2Farticles%2F2004%2F3%2Findia%2520policy%2520forum%2F200
4_banerjee_cole_duflo.pdf&ei=45x0Uu1eiIquB5GtgfgP&usg=AFQjCNFvwumK9A_8W
5IhkDfCvc-Xte3-lw
275
Berger, A.N., & Young, R.D. (1997). Problem loans and cost efficiency in commercial banks.
Journal of Banking and Finance, 21(6), 849-870. Retrieved on 25th February, 2013 from:
http://www.federalreserve.gov/pubs/feds/1997/199708/199708pap.pdf
Bernstein, D. (1996). Asset Quality and Scale Economies in Banking. Journal of Economics and
Business, 48(2), 157-166. Retrieved on 25th February, 2013 from:
http://ideas.repec.org/a/eee/jebusi/v48y1996i2p157-166.html
Bhatia, S., & Verma, S. (1998-99). Factors Determining Profitability of Public Sector Banks in
India: An Application of Multiple regression Model. Prajnan, XXVII(4), 433-445.
Bhide, M.G., Prasad, A., & Ghosh, S. (2002). Banking Sector Reforms A Critical Overview.
Economic and Political Weekly, 37(5), 399-407.
Bodla, B.S., & Verma, R. (2006). Determinants of Profitability of Banks in India: A Multivariate
Analysis. Journal of Services Research, 6(2), 75-89. Retrieved on 16th February, 2013
from: http://connection.ebscohost.com/c/articles/23654497/determinants-profitability-
banks-india-multivariate-analysis.
Casas, I. & Gao, J. (2007). Nonparametric Methods in Continuous Time Model Specification.
Econometric Reviews, 26(1), 91–106. Retrieved on 25th April, 2013 from:
http://www.jitigao.com/uploads/3/1/8/5/3185830/paper07c.pdf
Chaitanya, K.V. (2004). Causes for Non Performing Assets in Public Sector Banks. Ekonomska
istraživanja : znanstveno-stručni časopis. – Pula, 17, 19-30. Retrieved on 25th February,
2013 from: http://www.econbiz.de/en/search/detailed-view/doc/all/causes-for-non-
performing-assets-in-public-sector-banks-chaitanya-krishna/10002514085/?no_cache=1
Chansarkar, A.R. (2001, January 15-17). Prospects for PSBs. In K.S. Shetty, J. Jain, K.C.
Chakrabarty, N.V. Rao, R.K. Takkar, B.R. Prasad, & K. Ravi (Eds.). Banking in the New
Millennium. Paper presented at the Proceedings of Bank Economists’ Conference 2000,
New Delhi (pp. 154 - 157). New Delhi, India: Oriental Bank of Commerce.
Chaudhary, K., & Sharma, M. (2011). Performance of Indian Public Sector Banks and Private
Sector Banks: A Comparative Study. International Journal of Innovation, Management
and Technology, 2(3), 249-256. Retrieved on 25th March, 2013 from:
http://www.ijimt.org/show-31-397-1.html
276
Chipalkatti, N., & Rishi, M. (2007). Do Indian Banks Understate Their Bad Loans. The Journal
of Developing Areas, XL(2), 75-91. Retrieved on 25th March, 2013 from:
http://www.jstor.org/discover/10.2307/4193031?uid=3738256&uid=2129&uid=2&uid=7
0&uid=4&sid=21102803484327
Cleveland, W.S. (1993). Visualizing Data. Hobert Press, Summit, NJ, USA.
Dash, M.K., & Kabra, G. (2010). The Determinants of Non –Performing Assets in Indian
Commercial Banks: An Econometric Study. Middle Eastern Finance and Economics, 7,
94-106. Retrieved on 25th March, 2013 from:
http://www.wbaonline.co.uk/abstract.asp?id=54833&rev=
Dayal, D. (2001, January 15-17). Issues in Indian Banking. In K.S. Shetty, J. Jain, K.C.
Millennium. Paper presented at Proceedings of Bank Economists’ Conference 2000, New
Delhi (pp. 54 - 59). New Delhi, India: Oriental Bank of Commerce.
Dias, R. (n.d.). Nonparametric Regression: Lowess/Loess. Retrieved on 25th April, 2013 from:
http://www.ime.unicamp.br/~dias/loess.pdf
Eilers, P.H.C., & Marx, B.D. (2004). Splines, Knots, and Penalties. Retrieved on 25th March,
2013 from:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.128.8312&rep=rep1&type=pdf
Field, A.P. (2009). Discovering Statistics Using SPSS: and sex and drugs and rock ‘n’ roll.
London, UK: Sage.
Ghosh, D., & Ghosh, S. (2011, June 13-14). Management of Non-Performing Assets in Public
Sector Banks: Evidence From India. Paper presented at International Conference on
Management (ICM 2011).
Gopalakrishnan, T.V. (2006). Management of Non Performing Asset: A Study with Reference to
Public Sector Banks. New Delhi, India: Northern Book Centre.
Govil, A. (2001, January 15-17). Progress and Prospects in Indian Banking. In K.S. Shetty, J.
Jain, K.C. Chakrabarty, N.V. Rao, R.K. Takkar, B.R. Prasad, & K. Ravi (Eds.). Banking
in the New Millennium. Paper presented at the Proceedings of Bank Economists’
277
Conference 2000, New Delhi (pp. 127 - 130). New Delhi, India: Oriental Bank of
Commerce.
Goyal, K. (2010). Empirical Study of Non Performing Assets Management of Indian Public
Sector Banks. APJRBM , 1(1), 114-131.
Graham, D.R., & Humphrey, D.B. (1978). Bank Examination Data as Predictor of Bank Net
Loan Losses. Journal of Money, Credit and Banking, 10(4), 491-504. Retrieved on 25th
April, 2013 from:
http://www.jstor.org/discover/10.2307/1991578?uid=3738256&uid=2129&uid=2&uid=7
0&uid=4&sid=21102814014587
Gupta, P.K., & Jain, A. (2010). Factor Modeling for Indian Private Sector Banks’ Problem
Loans. Paper presented at International Conference on Applied Economics – ICOAE
2010. Retrieved on 20th February, 2013 from:
http://kastoria.teikoz.gr/icoae2/wordpress/wp-content/uploads/2011/10/0271.pdf
Hosmani, A.P., & Hudagi, J. (2011). Unearthing the Epidemic of Non Performing Assets –A
Study with Reference to Public Sector Banks in India. International Journal of
Multidisciplinary Research, 1(8), 447-459. Retrieved on 20th February, 2013 from:
http://zenithresearch.org.in/images/stories/pdf/2011/Dec/zijmr/37_VOL%201_ISSUE8_Z
EN.pdf
Hou, Y. (2007, August). The Non-Performing Loans Some Bank Level Evidences. Paper
presented at the Research Conference on Safety and Efficiency of The Financial System,
France. Retrieved on 25th April, 2013 from:
http://cba.upd.edu.ph/asialink/Hou_Dickinson.pdf
ICRA. (2004). Report 1: Global Benchmarking. Mumbai, India: IBA Bulletin Special Issue.
Jacoby, B. (n.d.). Lecture 14: Non-normal Error Distribution and Nonlinear Functional Forms –
Regression III: Advanced Methods [pdf]. Michigan, USA: Michigan State University.
Retrieved on 22nd April, 2013 from:
http://www.waikato.ac.nz/__data/assets/pdf_file/0017/51632/APA_Referencing_6th_ed.p
df
278
Jacoby, W.G. (2000). Loess: A Nonparametric, Graphical Tool for Depicting Relationships
between Variables. Electoral Studies, 19, 577–613. Retrieved on 22nd April, 2013 from:
http://www.med.upenn.edu/beat/docs/Jacoby2000.pdf
Jayasree, M., & Radhika, R. (2011). Non Performing Assets: A Study of Scheduled Commercial
Banks in India. International Journal of Research in Commerce, Economics and
Management, 1(1), 60-63. Retrieved on 15th February, 2013 from:
http://www.google.co.in/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&ved
=0CDEQFjAB&url=http%3A%2F%2Fijrcm.org.in%2Fdownload.php%3Fname%3Dijrc
m-3-Evol-1_issue-1_art-10.pdf%26path%3Duploaddata%2Fijrcm-3-Evol-1_issue-1_art-
10.pdf&ei=NHJ8UqvyAcOHrQf4q4DQBw&usg=AFQjCNEz-
C3Lrv3ljqRo3WyZobR4QrOs-w&bvm=bv.56146854,d.bmk
Joshi, V., & Little, I.M.D. (1996). India's Economic Reforms: 1991-2001. New York, USA:
Oxford University Press.
Kageyama, S., Pal, S., Pal, S., Medda, C., & Basu, T.K. (2004). Power of Parametric and Semi-
parametric Models to represent real life Data Situations. Bulletin of the Graduate School
of Education, Hiroshima University. Part. II, Arts and science education. 53, 49-56.
Retrieved on 25th March, 2013 from: http://ir.lib.hiroshima-u.ac.jp/00018251
Kannan, R., & Swamy, B.N.A. (2000). Profitability of Indian Banks. In K.S. Shetty, J. Jain, K.C.
Millennium. Paper presented at the Proceedings of Bank Economists’ Conference 2000,
New Delhi (pp. 154 - 157). New Delhi, India: Oriental Bank of Commerce.
Khasnobis, S. (2006). NPAs: Emerging Challenges. The Indian Banker, 1(2), 17-20.
Kwan, S., & Elsenbeis, R.A. (1997). Bank Risk, Capitalization, and Operating Efficiency.
Journal of Financial Services research, 12(2/3), 117-131. Retrieved on 15th November,
2011 from:
http://link.springer.com/article/10.1023/A:1007970618648#page-2
Machiraju, H.R. (2003). Modern Commercial Banking. New Delhi, India: Vikas Publishing
House Pvt. Ltd.
279
Malyadri, P., & Sirisha, S. (2011). A Comparative Study of Non Performing Assets in Indian
Banking Industry. International Journal of Economic Practices and Theories, 1(2), 77 -
87. Retrieved on 25th February, 2013 from:
http://www.ijept.org/index.php/ijept/article/view/15/15
McLeod, A.I. (n.d.). Statistics 259b -Robust Loess: S lowess [pdf document]. Ontario, Canada:
University of Western Ontario. Retrieved on 20th April, 2013 from:
http://www.stats.uwo.ca/faculty/aim/2004/04-259/notes/RL.pdf
Meeker, L.G., &. Gray, L. (1987). A Note on Non-Performing Loans as an Indicator of Asset
Quality. Journal of Banking and Finance, 11, 161-168. Retrieved on 27th February, 2013
from: http://www.sciencedirect.com/science/article/pii/0378426687900288
Misra, B.M., & Dhal, S. (2001). Pro Cyclical Management of Banks’ Non – Performing Loans
by the Indian Public Sector Banks. RBI Access Paper, 24(3), 81-82. Retrieved on 27th
February, 2013 from https://www.bis.org/repofficepubl/arpresearch201003.08.pdf
Mohan, A., & Kapoor, P. (1996). A Practical Guide to Non Performing Bank Advances.
Luknow, India: Vinod Law Publication.
Naik, J.G. (2006). NPA Management, Challenges before Banking Sector. The Management
Accountant, 41(5), 355-360.
Nandy, D. (2011). A Multivariate Analysis Approach of Selecting Profitability Indicators –An

Empirical Study of Commercial Banks in India. International Journal of
Multidisciplinary Research, 1( 6), 1-18. Retrieved on 28th February, 2013 from:
http://zenithresearch.org.in/images/stories/pdf/2011/Oct/ZIJMR/1_vol-1_issue-
6%20_%20Debaprosanna%20Nandy%20Research%20Article%20by%20Dr%20Debapro
sanna%20Nandy.pdf
Nettleton, D. (n.d.) Smoothing Scatterplots Using Penalized Splines. [pdf]. Iowa, USA: Iowa
State University. Retrieved on 10th July, 2013 from:
http://www.public.iastate.edu/~dnett/S511/34SmoothingPenSpline.pdf
280
Pal, P., Pal, S., Pal, S., & Kageyama, S. (2012). Non-parametric Modelling and Its Application
on Real-life Data. Bulletin of Hiroshima Institute of Technology Research. 46, 173-179.
Retrieved on 25th March, 2013 from:
www.it-hiroshima.ac.jp/institution/library/pdf/research46_173-179.pdf
Pal, V., & Malik, N.S. (2007). A Multivariate Analysis of the Financial Characteristics of
Commercial Banks in India. The Icfai Journal of Bank Management, VI(3), 29-42.
Poongavanam, S. (2011). Non Performing Assets: Issues, Causes and Remedial Solutions. Asian
Journal of Management Research, 2(1), 123-132. Retrieved on 1st March, 2013 from:
http://www.ipublishing.co.in/ajmrvol1no1//voltwo/EIJMRS2012.pdf
Pradhan, T.K. (2012). A Study of Trends of Non Performing assets in Public Sector Banks
during Post Reform Period. Research Journal of Economics & Business Studies, 1(11),
30-37.
Rajamohan, S. (2012). Analysis of NPAs of the Public Sector Banks in India. International
Journal of Business and Management Tomorrow, 2(9), 1-6.
Rajeev, M., & Mahesh, H.P. (2010). Banking Sector Reforms and NPA: A Study of Indian
Commercial Banks (252). The Institute for Social and Economic Change, Bangalore,
Working Paper. Retrieved on 12th October, 2012 from:
http://www.isec.ac.in/WP%20252%20-
%20Meenakshi%20Rajeev%20and%20H%20P%20Mahesh.pdf
Rajput, N., Gupta, M., & Chauhan, A.K. (2012). Profitability And Credit Culture of NPAS: An
Empirical Analysis of PSBs. International Journal of Marketing, Financial Services and
Management Research, 1(9), 91-109.
Rangarajan, C. (1997). The broad objective of the reform has thus been to create a banking
system that is both viable and efficient. IBA Bulletin, October, pp. 14 – 19.
Ranjan, R., & Dhal, S.C. (2003). Non Performing Loans and Terms of Credit of Public Sector
Banks in India: An Empirical Assessment. Reserve Bank of India Occasional Papers,
24(3), 81-120. Retrieved on 5th September, 2011 from:
http://rbidocs.rbi.org.in/rdocs/Publications/PDFs/60613.pdf
281
Rao, D.S. (2007). Reforms in Indian Banking Sector: Evaluation Study of the Performance of
Commercial Banks. Finance India, XXI(2), 591-597.
Rawlin, R., Sharan, M.S., & Lakhshmipathy, B.P. (2012). Modeling the NPA of a Midsized
Indian Nationalised Bank as a Function of Advances. European Journal of Business and
Management, 4(5), 78-88. Retrieved on 5th August, 2013 from:
=0CCgQFjAA&url=http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F235602
120_Modeling_the_NPA_of_a_Midsized_Indian_Nationalized_Bank_as_a_Function_of
_Advances%2Ffile%2Fd912f511db5423e160.pdf&ei=knp8UtPOFovOrQfO1oGIDQ&us
g=AFQjCNGX4ZxtVX5fujysFa_9ssWYZXum2g&bvm=bv.56146854,d.bmk
Rawlin, R., & Sharan, M.S. (2011). Modeling the NPA of a Large Indian Public Sector Bank as
a Function of Total Assets. Norderstedt, Germany: GRIN.
Mohan, A., & Kapoor, P. (1996). A Practical Guide to Non Performing Bank Advances.
Luknow, India: Vinod Law Publication.
Reddy, B.K., Babu, P.P., Mallikarjuna, V., & Viswanath, P. (2006). Non Performing Assets in
Public Sector Banks: An Investigation. The IUP Journal of Financial Economics, IV(I),
69-79. Retrieved on 4th November, 2011 from:
http://ideas.repec.org/a/icf/icfjfe/v04y2006i1p69-79.html
Report on Trends and Progress of Banking in India 2011-12. (2012). Trends and Progress of
Banking in India, RBI, p. 54. Retrieved on 9th July, 2013 from
http://rbidocs.rbi.org.in/rdocs/Publications/PDFs/0TPB021112FLS.pdf
Report of the Committee on the Financial System. (1991). Mumbai, India: Reserve Bank of
India.
Report of the Committee on Banking Sector Reforms. (1998). New Delhi, India: Government of
India.
Ruppert, D. (2002). Selecting the Number of Knots for Penalized Splines. Journal of
Computational and Graphical Statistics, 11(4), 735–757. Retrieved on 6th July, 2013
from: http://www.sta.cuhk.edu.hk/tlee/sta5050/forassign1.pdf
282
Ruppert, D., Wand, M.P., & Carroll, R.J. (2003). Semiparametric Regression. Cambridge, UK:
Cambridge University Press
Salkowski, N.J. (2008). Using the SemiPar Package. Manuscript. Retrieved on 6th July, 2013
from:
http://www.biostat.umn.edu/~hodges/RPLMBook/Manuscripts/SalkowskiClassProject.pd
f
Saluja, R., & Lal, R. (2010). Comparative Analysis on Non Performing Assets of Public Sector,
Private Sector and Foreign Banks in India. International Journal of Research in
Commerce and Management, 1(7), 80-88. Retrieved on 19th March, 2012 from:
=0CDAQFjAB&url=http%3A%2F%2Fijrcm.org.in%2Fdownload.php%3Fname%3Dvol-
1_issue-7_art-11.pdf%26path%3Duploaddata%2Fvol-1_issue-7_art-
11.pdf&ei=qnt8Uq7yCIWKrQeF14HwBg&usg=AFQjCNE2RqJxSLDlK72ezccQcaZwX
AOqeQ&bvm=bv.56146854,d.bmk
Selvarajan, B., & Vadivalagan, G. (2012). Cost of Non Performing Assets in Indian Banks.
European Journal of Economics, Finance and Administrative Sciences, 51, 28-44.
Retrieved on 6th July, 2013 from:
http://connection.ebscohost.com/c/articles/79274562/cost-non-performing-assets-indian-
bank
Selvarajan, B., & Vadivalagan, G. (2013). A Study on Management of Non-Performing Assets in

Priority Sector reference to Indian Bank and Public Sector Banks (PSBs). International
Journal of Finance & Banking Studies, 2(1), 31-42. Retrieved on 19th August, 2013 from:
http://www.ssbfnet.com/ojs/index.php/ijfbs/article/view/52/39
Sethi, J., & Bhatia, N. (2007). Elements of Banking and Insurance. New Delhi, India: PHI
Learning Pvt. Ltd.
Shajahan, K.M. (1998). Non-Performing Assets of Banks-Have They Really Declined And on
Whose Account. Economic and Political Weekly, XXXIII(12), 671-674.
Sharma, J.K. (1985). Bank Credit & Economic Development in India. New Delhi, India:
Classical Publisher Co.
283
Sharma, M. (2005). Problem of NPAs and its Impact on Strategic Banking Variables. Finance
India, X1X(3), 953-967.
Shekhar, K.C., & Shekhar, L. (2005). Banking Theory and Practice. New Delhi, India: Vikas
Publishing House Pvt. Ltd.
Shukla, S.C. (1995). Capital Adequacy and Prudential Accounting Norms. Central Bank of India
Economic Bulletin, 15(1), 1-7.
Siraj, K.K., & Pillai, P.S. (2012). A Study on the Performance of Non Performing Assets
(NPAs) of Indian Banking During Post Millenium Period. International Journal of
Business and Management Tomorrow, 2(3), 1-12. Retrieved on 15th February, 2013 from:
http://rjmarketings.com/themes/default/images/Effectiveness%20of%20SARFAESI%20
Act.pdf
Singla, H.K. (2008). Financial Performance of Banks in India. The ICFAI Journal of Bank
Management, VII(1), 50-62.
Srivastava, V., Bansal, D., & Sharma, A. (2013). A Study on The Performance of NPAs
in Commercials Banks in India. VSRD International Journal of Business and
Management Research, 3(3). Retrieved on 19th August, 2013 from:
http://vsrdjournals.com/MBA/Issue/2013_03_March/Vivek_Srivastava_1470_Research_Article_
VSRDIJBMR_March_2013.docx.
Suneja, H.R. (1992). Management of Bank Credit. Mumbai, India: Himalaya Publishing House.
Tarapore, S.S. (1998). Need for Second Generation Banking Sector Reforms. The Journal of
Indian Institute of Bankers, 69(2), 51-58.
Thiagarajan, S., Ayyappan, S., Ramachandran, A., & Sakthivadivel, M. (2011). An Analysis of
Determinants of Profitability in Public and Private Sector Banks in India. The
International Journal’s Research Journal of Social Science and Management, 1(6), 140-
152. Retrieved on 19th November, 2012 from:
http://www.theinternationaljournal.org/ojs/index.php?journal=tij&page=article&op=view
&path[]=262
284
Thiagarajan, S., Ayyappan, S., & Ramachandran, A. (2011). Credit Risk Determinants of Public
and Private Sector Banks in India. European Journal of Economics, Finance and
Administrative Sciences, 34, 147-154. Retrieved on 12th September, 2012 from:
http://connection.ebscohost.com/c/articles/67737764/credit-risk-determinants-public-
private-sector-banks-india
Toor, N.S. (1994). Non Performing Advances in Banks Concept, Practice And Management.
New Delhi, India: Skylark Publication.
Vallabh, G., Bhatia, A., & Mishra, S. (2007). Non Performing Assets of Indian Public, Private
and Foreign Sector Banks – An empirical Assessment. The ICFAI Journal of Bank
Management, VI(3), 7-28.
Vashishtha, G., & Rajaraman, I. (2002). Non Performing Loans of PSU Banks - Some Panel
Results. Economic and Political Weekly, XXXVII(05), 429-433. Retrieved on 9th August,
2011 from:
http://www.epw.in/system/files/pdf/2002_37/05/NonPerforming_Loans_of_PSU_Banks.
Veerakumar, K. (2012). Non Performing Assets in Priority Sector: A Threat to Indian Scheduled
Commercials Banks. International Research Journal of Finance and Economies, 93, 6-
23. Retrieved on 5th August, 2013 from:
http://connection.ebscohost.com/c/articles/78308435/non-performing-assets-priority-
sector-threat-indian-scheduled-commercial-banks
Vyas, R.K., Singh, M., & Kashif, Md. (2007). Capital Regulation and Credit Growth in Indian
Banks. Amity Business Review, 8(2), 21-33.
Yadav, M.S. (2011). Impact of Non Performing Assets on Profitability and Productivity of
Public Sector Banks In India, AFBE Journal, 4(1), 232-241.Retrieved on 15th September,
2012 from: http://www.afbe.biz/main/wp-content/uploads/afbejournal-
issue6vol4no11.pdf
Yao, F. & Lee, T.C.M. (2008). On Knot Placement for Penalized Spline Regression. Journal of
the Korean Statistical Society, 37, 259-267. Retrieved on 15th March, 2013 from:
http://www.utstat.utoronto.ca/fyao/2008-jkss.pdf
285
Weisberg, S. (1996). The R-Code: A Graphical Paradigm for Regression Analysis. In R. Stine &
J. Fox (Eds.), Statistical Computing Environments for Social Research. Thousand Oaks,
CA, USA: Sage Publications.
Wu, L. (2010). Mixed Effects Models for Complex Data. New York, USA: CRC Press.
286
Annexure 1A.:Status of GA, GNPAs, and GNPA/GA Ratio of SCBs & PSBs from 1993 – 2012 (Rs in crores)
Year (end- Schedule Commercial Banks ( SCBs) Public Sector Banks ( PSBs)
March) Gross Advances Gross NPA Gross NPA as Gross Advances Gross NPA Gross NPA as
% of GA % of GA
1992-93 ** ** ** 169340 39253 23.2
1993-94 ** ** ** 165621 41041 24.8
1994-95 ** ** ** 197352 38385 19.5
1995-96 ** ** ** 231321 41661 18.0
1996-97 301698 47300 15.7 244213 43576 17.8
1997-98 352696 50815 14.4 284971 45653 16.0
1998-99 399436 58722 14.7 325328 51710 15.9
1999-00 475113 60408 12.7 380077 53294 14.0
2000-01 558766 63741 11.4 442134 54774 12.4
2001-02 680868 70953 10.4 509369 56507 11.1
2002-03 776507 70313 9.1 577813 54089 9.4
2003-04 902026 64897 7.2 661976 51541 7.8
2004-05 1171210 57546 4.9 870850 46597 5.4
2005-06 1473528 51243 3.5 1070872 41379 3.9
2006-07 1893513 50116 2.6 1373777 38425 2.8
2007-08 2331678 55699 2.4 1696333 39606 2.3
2008-09 2789307 68213 2.4 2099938 44032 2.1
2009-10 3264372 81805 2.5 2512358 57293 2.3
2010-11 3991501 94117 2.4 3059870 71080 2.3
2011-12 4665544 137096 2.9 3550389 112489 3.2
** implies ‘data not available’
Source: Statistical Tables Relating to Banks in India, RBI various issues
Annexure 1B.: List of Scheduled Commercial Banks as on 31st March 2012 (excluding RRBs)
1. Public Sector Banks: (26) 2. Private Sector Banks: (20) 3. Foreign Banks in India: (36)
1.1 State Bank Group: (6) 2.1 Old Private Sector Banks: (13)
1. State Bank of Bikaner and 1. Catholic Syrian Bank Ltd. 1. AB Bank Ltd.
Jaipur 2. City Union Bank Ltd. 2. Abu Dhabi Commercial Bank Ltd.
2. State Bank of Hyderabad 3. Dhanalakshmi Bank Ltd. 3. American Express Bank Ltd.
3. State Bank of Mysore 4. Federal Bank Ltd. 4. Antwerp Diamond Bank Nv
4. State Bank of India 5. ING Vysya Bank Ltd. 5. ANZ Banking Group
5. State Bank of Patiala 6. Jammu and Kashmir Bank Ltd. 6. Bank of America NA
6. State Bank of Travancore 7. Karnataka Bank Ltd. 7. Bank of Bahrain and Kuwait B.S.C.
8. Karur Vysya Bank Ltd. 8. Bank of Ceylon
9. Lakshmi Vilas Bank Ltd. 9. Bank of Nova Scotia
10.Nainital Bank Ltd. 10.Bank of Tokyo-Mitsubishi UFJ Ltd.
11.Ratnakar Bank Ltd. 11.Barclays Bank PLC
12.South Indian Bank Ltd. 12.BNP Paribas
13.Tamilnadu Mercantile Bank Ltd. 13.Chinatrust Commercial Bank
14.Citibank N.A.
1.2 Nationalized Bank Group(19) 2.2 New Private Sector Banks:(7) 15.Commonwealth Bank of Australia
1. Allahabad Bank 1.Axis Bank 16.Credit Agricole Corporate and Investment
2. Andhra Bank 2.Development Credit Bank Ltd. Bank
3. Bank of Baroda 3.HDFC Bank Ltd. 17.Credit Suisse Ag
287
4. Bank of India 4.ICICI Bank Ltd. 18.DBS Bank Ltd.
5. Bank of Maharashtra 5.IndusInd Bank Ltd. 19.Deutsche Bank AG
6. Canara Bank 6.Kotak Mahindra Bank Ltd. 20.Firstrand Bank Ltd
7. Central Bank of India 7.Yes Bank Ltd. 21.HSBC Ltd
8. Corporation Bank 22.Industrial and Commercial Bank of China
9. Dena Bank 23.JP Morgan Chase Bank
10. Indian Bank 24.JSC VTB Bank
11. Indian Overseas Bank 25.Krung Thai Bank Public Co. Ltd.
12. Oriental Bank of Commerce 26.Mashreqbank psc
13. Punjab and Sind Bank 27.Mizuho Corporate Bank Ltd.
14. Punjab National Bank 28.Oman International Bank S.A.O.G.
15. Syndicate Bank 29.Rabo Bank International
16. UCO Bank 30.Royal Bank of Scotland
17. Union Bank of India 31.Shinhan Bank
18. United Bank of India 32.Societe Generale
19. Vijaya Bank 33.Sonali Bank
34.Standard Chartered Bank
1.3 Other Public Sector Bank(1) 35.State Bank of Mauritius Ltd.
1. IDBI Bank Ltd. 36.UBS AG
Source: www.rbi.org.in
Annexure 1C.: Bank wise GNPA/GA Ratio as on March 31st, 2008 of all PSBs. (in
percentage)
Sr. No. Name of the Bank 2007-08
Nationalised Banks
1 Allahabad Bank 2.01
2 Andhra Bank 1.08
3 Bank of Baroda 1.84
4 Bank of India 1.68
5 Bank of Maharashtra 2.57
6 Canara Bank 1.31
7 Central Bank of India 3.16
8 Corporation Bank 1.47
9 Dena Bank 2.45
10 Indian Bank 1.21
11 Indian Overseas Bank 1.63
12 Oriental Bank of Commerce 2.31
13 Punjab and Sind Bank 0.74
14 Punjab National Bank 2.74
15 Syndicate Bank 2.71
16 UCO Bank 2.97
17 Union Bank of India 2.18
18 United Bank of India 2.70
19 Vijaya Bank 1.60
State Bank Group
20 State Bank of India 3.04
21 State Bank of Bikaner and Jaipur 1.73
22 State Bank of Hyderabad 0.87
23 State Bank of Indore 1.45
24 State Bank of Mysore 1.69
25 State Bank of Patiala 1.42
26 State Bank of Saurashtra 1.42
27 State Bank of Travancore 2.01
Other Public Sector Bank
28 IDBI Bank Ltd. 1.87
Source: Appendix Table III.28: Non-Performing Assets as percentage of Advances - Scheduled Commercial Banks (http://www.rbi.org.in)
288
Annexure 7A.: Average Total Asset (TA) Dataset (Rs. in crores)
1996 133232.96 3322.90 13873.98 13096.75 20719.40 30039.18 445115.09
1997 150471.51 7217.99 15511.09 14389.10 23578.65 33264.34 508188.56
1998 168072.92 8461.64 17521.30 15777.15 27354.27 37424.51 593820.02
1999 201090.85 9989.11 20213.63 17899.18 31975.45 43064.66 701556.69
2000 242006.99 11639.85 24528.67 20354.76 37546.29 50226.04 823046.14
2001 288574.57 13458.62 27703.01 23658.96 43610.36 58823.91 952916.58
2002 332092.67 15509.90 29999.71 27593.60 49596.15 68249.69 1085743.61
2003 362346.64 17800.59 33095.81 32272.15 55019.95 79630.62 1215409.65
2004 392112.29 21541.26 40829.31 39355.95 60553.57 94327.11 1374469.28
2005 433977.65 26457.46 49675.73 49193.65 66279.48 114320.92 1578175.94
2006 477021.72 30394.69 56609.50 58214.44 71973.44 135809.25 1893356.71
2007 530483.37 34973.43 75184.04 68351.66 84206.24 153939.46 2228931.75
2008 644465.62 41084.04 98204.82 82329.43 108903.61 180788.95 2732992.18
2009 843584.03 46732.58 118693.98 100895.43 136240.22 222994.22 3385607.03
2010 1009499.79 54403.64 134653.31 124902.09 165558.82 271798.80 4071830.46
2011 1139325.21 65220.77 147794.87 150848.94 196615.70 337506.04 4853163.46
2012 1281051.63 78534.68 169503.43 172478.91 220094.98 418288.13 5670398.81
Sources: Prowess Database
Annexure 7B.: Profit before Tax (PBT) Dataset (Rs. in crores)

1996 831.60 26.21 20.17 -236.66 -73.53 -95.92 -104.22
1997 2478.85 74.25 66.96 -176.23 156.83 360.56 4957.20
1998 2989.05 121.40 90.66 -96.22 218.02 777.45 7558.78
1999 1550.55 48.87 155.99 -67.77 158.12 530.99 4597.31
2000 3191.53 95.54 241.65 52.00 199.03 532.14 7708.43
2001 2575.61 122.59 257.14 35.70 51.69 575.07 6638.34
2002 3699.59 179.90 273.05 169.60 179.44 761.43 11831.17
2003 4569.78 272.33 456.63 231.96 445.23 1183.59 17531.44
2004 4970.78 386.80 704.67 468.51 874.19 1769.48 23844.48
2005 6521.60 377.79 433.40 373.37 589.62 2047.08 20023.93
2006 6906.15 338.99 582.21 210.44 375.93 2034.83 22447.70
2007 7625.08 468.13 785.97 358.51 695.31 2169.13 28658.90
2008 10438.89 527.74 1005.54 470.02 873.34 3295.91 37699.35
2009 14180.65 891.39 1036.34 564.03 924.98 4681.65 49730.49
2010 13926.08 943.02 1174.03 971.56 1549.54 5798.94 55425.03
2011 14954.23 882.03 1285.42 914.05 1659.17 6608.42 65564.89
2012 18504.59 797.58 1427.32 1071.33 758.72 7044.92 68194.26
Annexure 7C.: Aggregate Credit Dataset (Rs. in crores)
1995 48530.21 3126.37 4011.47 4876.15 7809.60 11616.95 169595.38
1996 59825.64 3158.26 4908.31 4109.42 8902.58 12679.88 200123.79
1997 55576.79 3417.30 5476.13 3984.12 8790.30 12688.48 207471.55
1998 66303.86 3643.11 6638.38 5610.53 10713.54 14678.83 243224.64
1999 82359.84 3824.03 8916.74 6222.21 12799.84 17557.83 286348.15
2000 98101.97 5131.21 12206.31 7630.27 14925.86 20668.73 349375.02
2001 113590.27 6397.50 13116.15 10085.44 18833.38 28043.16 413173.52
2002 120806.46 7435.51 14884.66 12805.37 21287.51 34369.41 476715.35
2003 137758.46 9170.66 16305.35 15923.11 21208.65 40228.12 543777.64
2004 157933.54 11132.43 20646.92 20626.44 21960.29 47224.72 628418.01
2005 202374.46 14848.34 26729.20 27655.71 27277.32 60412.75 806517.33
2006 261800.93 18866.40 36466.24 37377.58 37483.48 74627.36 1106287.72
2007 337336.49 24629.89 51670.44 46988.92 51795.47 96596.52 1440546.50
2008 416768.20 28136.62 64051.01 55081.90 72997.42 119501.57 1797400.80
2009 542503.21 32601.42 81532.27 68803.86 85483.19 154702.99 2259099.68
2010 631914.15 38461.27 90406.36 82504.54 105383.49 186601.21 2677261.85
2011 756719.44 46044.23 106781.92 99070.82 129725.41 242106.66 3304427.68
2012 867578.89 55345.96 123620.18 115540.02 147512.85 293774.76 3878312.11
289
Annexure 7D.: Aggregate Deposit Dataset (Rs. in crores)
1996 96395.45 5424.14 12718.25 11443.16 19751.60 27122.89 381465.88
1997 110701.17 6463.68 14946.30 12614.10 23050.98 30806.42 449339.79
1998 131091.32 7468.06 16816.15 14462.46 26373.49 35173.55 531723.21
1999 169041.93 8658.07 19914.33 16251.21 30649.31 40777.13 636867.91
2000 196821.07 10182.61 23655.42 18359.95 35871.71 47483.23 737280.48
2001 242828.37 11572.80 25094.84 21535.66 41517.92 56131.13 859461.98
2002 270560.14 13459.68 28548.33 26848.77 47137.38 64123.48 968623.62
2003 296123.28 15926.28 30660.54 31343.39 51165.12 75813.51 1079167.38
2004 318618.67 19721.37 42584.82 39244.26 55908.60 87916.40 1229273.41
2005 367047.53 24133.01 46294.57 49470.24 60751.68 103166.89 1421437.89
2006 380046.05 25996.52 53624.40 54543.73 66482.65 119684.92 1622481.13
2007 435521.09 30984.01 78633.57 64860.01 82776.28 139859.68 1994199.59
2008 537403.95 35353.89 95170.80 79908.95 110319.66 166457.22 2453867.69
2009 742073.12 42041.07 115885.14 100221.56 131271.86 209760.51 3112747.13
2010 804116.23 50883.39 117025.80 122415.54 162107.47 249329.80 3661394.95
2011 933932.81 58157.92 135596.08 145277.60 179356.02 312898.73 4372448.66
2012 1043647.36 71469.82 157941.06 154003.48 196173.32 379588.48 5002013.43
Annexure 7E.: Priority Sector Advance (PSADV) Dataset (Rs. in crores)
1996 14945.26 1128.56 1347.22 1330.36 2781.35 4091.67 59915.09
1997 16945.76 1209.43 1507.24 1591.56 3074.57 4379.91 68889.32
1998 19522.82 1302.44 1800.51 1871.25 3655.14 5275.35 80560.54
1999 23090.38 1408.55 2810.68 1819.67 4152.68 6595.05 93709.31
2000 25877.55 1705.26 3131.87 2193.14 5814.43 8323.23 108584.90
2001 30153.35 2015.50 3810.38 2748.30 7170.07 10856.74 126451.08
2002 31591.48 2260.42 4101.35 3975.46 8203.28 13441.28 146864.87
2003 35111.80 3081.33 5041.51 4844.11 8763.58 16033.63 175737.54
2004 42705.87 4143.40 6725.39 6549.53 9925.32 20734.52 217398.19
2005 57864.82 5695.32 9693.92 10013.91 12215.55 28268.45 289915.52
2006 80018.23 7754.56 13488.39 13077.68 16299.57 34093.43 390831.33
2007 102015.85 9738.47 16953.55 16272.79 20262.76 36527.61 490090.92
2008 119230.51 11741.78 20580.36 18191.10 23998.51 46216.88 582087.64
2009 143637.56 13014.62 26393.25 21525.88 26830.21 49212.35 681330.55
2010 170568.21 14091.45 31085.63 24359.47 33873.13 66615.47 824465.31
2011 231597.87 17254.10 32175.89 24089.67 40509.51 78637.01 1009650.94
2012 250176.96 20042.36 36421.11 28924.55 38522.18 92032.95 1117574.50
Annexure 7F.: Number of Employee
1996 233000 12855 37809 34205 51156 71263 834673
1997 236204 12990 36905 33574 50372 67616 883003
1998 241714 13049 36266 32903 49702 67616 809834
1999 239023 13234 35749 32086 48933 65705 820903
2000 235804 12953 34581 31233 48933 64733 723180
2001 214845 12172 28977 29166 47474 58309 759956
2002 209462 12137 25570 25567 39631 57859 734900
2003 208998 12005 25452 25060 39330 58981 733553
2004 207039 12007 25165 25116 38933 58839 697157
2005 205515 11828 24834 24969 38303 58329 724958
2006 198774 11642 24624 24622 37241 58047 702618
2007 185388 11607 24360 24773 39055 57316 683451
2008 179205 11365 26637 23828 37488 56025 690789
2009 205896 11365 27121 23736 35543 58205 719125
2010 200299 12192 27713 23370 34826 56928 716213
2011 222933 12072 28509 23148 34015 57020 696622
2012 215481 12597 26904 23259 35901 62127 669296
290
Annexure 7G (1) .: Computation of Net Interest Margin (NIM) and Burden (SBI) (Rs. in crores)
Year end Total Interest Total Interest Non Interest Non Burden
Net
March Income Income Expense Expense Income Interest (NII - NIE)
Interest
(TI) (II) (TE) (IE) NII = Expense
Margin
(TI - II) NIE =
[II-IE]
(TE - IE)
1996 15748.27 12958.57 14916.67 8225.93 2789.7 6690.74 4732.64 3901.04
1997 17593.72 9429.14 16244.47 9591.43 8164.58 6653.04 -162.29 -1511.54
1998 19663.16 9487.24 17801.96 10473.2 10175.92 7328.76 -985.96 -2847.16
1999 22406.5 19107.54 21378.7 13044.44 3298.96 8334.26 6063.10 5035.3
2000 26308.63 22200.92 24257.08 15272.58 4107.71 8984.5 6928.34 4876.79
2001 30166.76 26138.59 28562.51 17756.02 4028.17 10806.49 8382.57 6778.32
2002 34087.84 29810.08 31656.22 20728.84 4277.76 10927.38 9081.24 6649.62
2003 37582.8 31087.02 34477.8 21109.46 6495.78 13368.34 9977.56 6872.56
2004 38186.56 30460.49 34505.56 19274.18 7726.07 15231.38 11186.31 7505.31
2005 39582.04 31682.73 35277.52 18483.37 7899.31 16794.15 13199.36 8894.84
2006 43507.73 35979.57 39101.06 20390.45 7528.16 18710.61 15589.12 11182.45
2007 44688.84 37242.33 40147.53 21919.38 7446.51 18228.15 15322.95 10781.64
2008 58440.49 48950.3 51711.37 31929.08 9490.19 19782.29 17021.22 10292.1
2009 76540.55 63788.44 67419.32 42915.3 12752.11 24504.02 20873.14 11751.91
2010 86968.51 70993.92 77802.46 47322.48 15974.59 30479.98 23671.44 14505.39
2011 98250.54 81394.37 89986.02 48867.96 16856.17 41118.06 32526.41 24261.89
2012 122016.89 106521.45 110309.6 63230.37 15495.44 47079.23 43291.08 31583.79
Annexure 7G (2) .: Computation of Net Interest Margin (NIM) and Burden (SBT) (Rs. in crores)
Year end Total Interest Total Interest Non Interest Non Burden
Net
March Income Income Expense Expense Income Interest (NII - NIE)
Interest
(TI) (II) (TE) (IE) NII = Expense
Margin
(TI - II) NIE =
[II-IE]
(TE - IE)
1996 888.24 541.85 862.03 546 346.39 316.03 -4.15 -30.36
1997 1032.91 639.41 992.66 661.32 393.5 331.34 -21.91 -62.16
1998 1199.24 610.21 1135.94 713.57 589.03 422.37 -103.36 -166.66
1999 1202.42 549.1 1159.15 781.08 653.32 378.07 -231.98 -275.25
2000 1373.19 1159.52 1306.75 877.59 213.67 429.16 281.93 215.49
2001 1509.62 1315.36 1412.13 920.24 194.26 491.89 395.12 297.63
2002 1699.37 1453.89 1578.44 1029.46 245.48 548.98 424.43 303.5
2003 1887.53 1584.44 1716.49 1061.59 303.09 654.9 522.85 351.81
2004 2211.38 1739.85 1966.78 1056.52 471.53 910.26 683.33 438.73
2005 2476.84 2008.32 2229.71 1112.16 468.52 1117.55 896.16 649.03
2006 2687.57 2298.62 2428.89 1343.49 388.95 1085.4 955.13 696.45
2007 3054.27 2690.62 2727.99 1698.44 363.65 1029.55 992.18 665.9
2008 3880.49 3434.11 3494.38 2476.83 446.38 1017.55 957.28 571.17
2009 4811.72 4123.15 4203.88 2840.6 688.57 1363.28 1282.55 674.71
2010 5001.1 4378.11 4316.83 2977.95 622.99 1338.88 1400.16 715.89
2011 6080.76 5228.76 5353.03 3532.72 852 1820.31 1696.04 968.31
2012 7649.31 6828.76 7138.85 4998.39 820.55 2140.46 1830.37 1319.91
Annexure 7G (3) .: Computation of Net Interest Margin (NIM) and Burden (SB) (Rs. in crores)
Non Interest Non Burden
Net
Total Interest Total Interest Income Interest (NII - NIE)
Year end Interest
Income Income Expense Expense NII = Expense
March Margin
(TI) (II) (TE) (IE) (TI - II) NIE =
[II-IE]
(TE - IE)
1996 1511.82 1367.92 1491.65 870.59 143.9 621.06 497.33 477.16
1997 1755.89 795.93 1688.93 1042.9 959.96 646.03 -246.97 -313.93
1998 1934.94 808.46 1852.28 1137.22 1126.48 715.06 -328.76 -411.42
1999 2396.43 1179.74 2253.84 1425.16 1216.69 828.68 -245.42 -388.01
2000 2760.63 2437.63 2544.98 1612.33 323 932.65 825.30 609.65
291
2001 3111.16 2792.17 2876.22 1698.53 318.99 1177.69 1093.64 858.7
2002 3200.62 2882.42 2950.07 1774.87 318.2 1175.2 1107.55 857
2003 3393.85 2875.17 3049.72 1665.45 518.68 1384.27 1209.72 865.59
2004 3908.68 3084.85 3474.55 1655.64 823.83 1818.91 1429.21 995.08
2005 4348.31 3757.63 3945.41 2063.8 590.68 1881.61 1693.83 1290.93
2006 4673.22 4050.42 4136.73 2169.55 622.8 1967.18 1880.87 1344.38
2007 6750.11 6040.07 6034.05 3890.02 710.04 2144.03 2150.05 1433.99
2008 8846.74 7906.31 7998.67 5833.56 940.43 2165.11 2072.75 1224.68
2009 10555.14 9573.17 9642.32 6977.6 981.97 2664.72 2595.57 1682.75
2010 11340.66 10128.05 10527.34 7307.37 1212.61 3219.97 2820.68 2007.36
2011 12427.23 11511.63 11379.28 7068.1 915.6 4311.18 4443.53 3395.58
2012 16419.89 15268.35 15106.5 10183.32 1151.54 4923.18 5085.03 3771.64
Annexure 7G (4) .: Computation of Net Interest Margin (NIM) and Burden (UCO) (Rs. in crores)
Net
Year end Interest
March Margin
[II-IE]
(TE - IE)
1996 1352.29 1203.44 1588.95 867.73 148.85 721.22 335.71 572.37
1997 1481.27 705.28 1657.5 986.82 775.99 670.68 -281.54 -105.31
1998 1656.59 1445.19 1752.81 1093.75 211.4 659.06 351.44 447.66
1999 1913.06 1692.76 1980.83 1246.29 220.3 734.54 446.47 514.24
2000 2252.18 1977.5 2215.54 1424.94 274.68 790.6 552.56 515.92
2001 2584.78 2274.7 2551.78 1612.65 310.08 939.13 662.05 629.05
2002 3147.07 2541.78 2982.55 1812.03 605.29 1170.52 729.75 565.23
2003 3419.75 2792.7 3212.26 1910.68 627.05 1301.58 882.02 674.53
2004 3738.91 3096.29 3303.49 1901.69 642.62 1401.8 1194.60 759.18
2005 4075.54 3548.73 3729.89 2140.49 526.81 1589.4 1408.24 1062.59
2006 4821.76 4354.59 4625.11 2788.83 467.17 1836.28 1565.76 1369.11
2007 5922.77 5317.85 5606.67 3623.17 604.92 1983.5 1694.68 1378.58
2008 7426.26 6553.34 7014.1 5020.81 872.92 1993.29 1532.53 1120.37
2009 9280.93 8243.8 8723.21 6476.68 1037.13 2246.53 1767.12 1209.4
2010 10500.04 9526.32 9487.85 7202.2 973.72 2285.65 2324.12 1311.93
2011 12300.47 11370.8 11393.93 7525.89 929.67 3868.04 3844.91 2938.37
2012 15625.77 14632.38 14517.1 10730.27 993.39 3786.83 3902.11 2793.44
Annexure 7G (5) .: Computation of Net Interest Margin (NIM) and Burden (CBI) (Rs. in crores)
Net
Year end Interest
March Margin
[II-IE]
(TE - IE)
1996 2423.75 2152.65 2497.28 1422.74 271.1 1074.54 729.91 803.44
1997 2914.71 2530.93 2763.88 1695.15 383.78 1068.73 835.78 684.95
1998 3185.43 2842.26 3010.54 1892.1 343.17 1118.44 950.16 775.27
1999 3726.22 3282.04 3600.23 2233.26 444.18 1366.97 1048.78 922.79
2000 4203.76 1889.75 4053.07 2521.6 2314.01 1531.47 -631.85 -782.54
2001 4742.37 4265.26 4695.91 2816.07 477.11 1879.84 1449.19 1402.73
2002 5304.74 4657.5 5141.44 3122.52 647.24 2018.92 1534.98 1371.68
2003 5673.78 2533.89 5368.26 3175.57 3139.89 2192.69 -641.68 -947.2
2004 6054.24 2406.23 5436.13 2941.53 3648.01 2494.6 -535.30 -1153.41
2005 6144.07 5204.88 5786.66 2829.93 939.19 2956.73 2374.95 2017.54
2006 6215.37 5385.59 5957.95 3005.51 829.78 2952.44 2380.08 2122.66
2007 6959.66 6234.21 6461.65 3759.79 725.45 2701.86 2474.42 1976.41
2008 8839.59 7884.27 8289.43 5772.47 955.32 2516.96 2111.80 1561.64
2009 11525.51 10455.19 10954.27 8226.72 1070.32 2727.55 2228.47 1657.23
2010 13822.06 12064.31 12763.83 9519.01 1757.75 3244.82 2545.30 1487.07
2011 16485.72 15220.57 15233.31 9895.23 1265.15 5338.08 5325.34 4072.93
2012 20657.61 19149.51 20124.57 13980.86 1508.1 6143.71 5168.65 4635.61
292
Annexure 7G (6) .: Computation of Net Interest Margin (NIM) and Burden (PNB) (Rs. in crores)
Non Non Burden
Year Interest Total Interest Interest Interest Net Interest (NII - NIE)
Total Income
end Income Expense Expense Income Expense Margin
(TI)
March (II) (TE) (IE) NII = NIE = [II-IE]
(TI - II) (TE - IE)
1996 3497.33 3167.73 3593.25 2089 329.6 1504.25 1078.73 1174.65
1997 4128.75 2128.59 3891.04 2438.7 2000.16 1452.34 -310.11 -547.82
1998 5000.71 2171.67 4499.46 2699.07 2829.04 1800.39 -527.40 -1028.65
1999 5057.45 4367.23 4681.09 2795.36 690.22 1885.73 1571.87 1195.51
2000 5931.21 5137.69 5523.07 3538.2 793.52 1984.87 1599.49 1191.35
2001 6692.19 5863.48 6228.55 3825.05 828.71 2403.5 2038.43 1574.79
2002 7724.62 6647.87 7162.23 4352.58 1076.75 2809.65 2295.29 1732.9
2003 8914.44 7485.01 8072.24 4361.29 1429.43 3710.95 3123.72 2281.52
2004 9836.11 7778.95 8727.42 4154.99 2057.16 4572.43 3623.96 2515.27
2005 10597.8 8459.85 9187.68 4453.12 2137.95 4734.56 4006.73 2596.61
2006 11471.76 9584.16 10032.45 4917.39 1887.6 5115.06 4666.77 3227.46
2007 13168.85 11236.14 11628.77 6022.91 1932.71 5605.86 5213.23 3673.15
2008 16291.49 14265.02 14242.73 8730.86 2026.47 5511.87 5534.16 3485.4
2009 22556.13 19127.22 19465.25 12295.31 3428.91 7169.94 6831.91 3741.03
2010 25104.19 21422.1 21198.83 12944.02 3682.09 8254.81 8478.08 4572.72
2011 30676.83 26986.48 26243.33 15179.14 3690.35 11064.19 11807.34 7373.84
2012 40713.41 36428.03 35821.33 23013.59 4285.38 12807.74 13414.44 8522.36
Annexure 7G (7) .: Computation of Net Interest Margin (NIM) and Burden (PSB) (Rs. in crores)
Year Interest Total Interest Income Interest Net Interest (NII - NIE)
Total Income
end Income Expense Expense NII = Expense Margin
(TI)
March (II) (TE) (IE) (TI - II) NIE = [II-IE]
(TE - IE)
1996 52430.8 43678.92 52619.98 30233.71 8751.88 22386.27 13445.21 13634.39
1997 61749.61 42213.7 58641.94 36338.59 19535.91 22303.35 5875.11 2767.44
1998 70228.06 44470.23 65144.88 40173.53 25757.83 24971.35 4296.70 -786.48
1999 79338.42 60509.1 76304.89 47839.77 18829.32 28465.12 12669.33 9635.8
2000 91915.8 74906.62 86750.09 55374.51 17009.18 31375.58 19532.11 14366.4
2001 103865.25 89418.41 99553.1 61693.18 14446.84 37859.92 27725.23 23413.08
2002 117849.01 97807.95 109717.78 69153.74 20041.06 40564.04 28654.21 20522.98
2003 129805.36 99885.53 117536.13 69852.6 29919.83 47683.53 30032.93 17763.7
2004 138940.95 101890.68 122421.88 65764.56 37050.27 56657.32 36126.12 19607.05
2005 142982.62 115695.26 127848.44 66146.76 27287.36 61701.68 49548.50 34414.32
2006 165905.93 137841.07 149369.36 80260.56 28064.86 69108.8 57580.51 41043.94
2007 192066.8 164695.97 171978.44 101942.6 27370.83 70035.84 62753.37 42665.01
2008 249037.01 213487.14 222446.78 149096.77 35549.87 73350.01 64390.37 37800.14
2009 318146.05 273601.77 283774.98 193595.92 44544.28 90179.06 80005.85 45634.78
2010 356646.2 303594.58 317698.52 210015.21 53051.62 107683.31 93579.37 54631.69
2011 419250.03 366370.16 374355.88 231153.02 52879.87 143202.86 135217.14 90322.99
2012 566059.97 484888.27 516549.52 328539.06 81171.7 188010.46 156349.21 106838.76
Annexure 7H(1) .: Computation of Total Overhead Expense (SBI) (Rs. in crores)
Selling
Outsou
Repair Insura & Miscellan
Year Rent & rced Direct Commun Printing & Total
s& nce distribu eous Depreci
end lease profess ors' ications stationery Overhead
mainte premiu tion expenditu ation
March rent ional fees expenses expenses Expense
nance m paid expense re
jobs
s
1993 198.18 18.19 121.85 11.29 0.33 5.49 39.35 49.64 208.14 55.10 707.56
1994 221.15 21.15 199.40 27.37 0.30 12.67 47.46 59.49 329.80 62.16 980.95
1995 249.50 23.36 238.22 19.05 0.47 9.33 46.42 57.95 259.91 66.47 970.68
1996 271.42 29.33 255.75 23.80 0.56 13.87 52.18 65.02 308.98 86.98 1107.89
1997 303.08 37.38 265.58 16.25 0.00 11.72 0.00 0.00 550.39 97.33 1281.73
1998 344.82 41.27 60.15 11.68 0.00 15.03 0.00 0.00 526.10 164.07 1163.12
1999 379.90 44.21 74.32 35.60 0.72 22.20 68.90 92.58 414.36 310.57 1443.36
293
2000 425.55 50.54 98.30 65.12 0.66 17.56 60.43 96.96 636.54 365.65 1817.31
2001 471.31 54.77 107.94 66.19 0.69 30.39 56.76 103.32 993.89 401.92 2287.18
2002 524.44 58.75 142.50 75.67 0.72 24.32 62.07 119.90 624.80 424.95 2058.12
2003 564.74 68.92 151.44 84.55 0.64 34.44 56.82 122.17 676.30 493.69 2253.71
2004 641.77 97.67 168.87 95.18 0.74 67.36 81.55 146.13 800.01 698.34 2797.62
2005 723.12 133.95 239.96 104.02 0.99 67.08 74.43 162.51 908.56 752.21 3166.83
2006 796.35 170.27 340.77 113.05 1.23 109.44 102.25 175.64 1063.93 729.13 3602.06
2007 896.50 189.15 355.29 119.64 1.08 88.43 118.17 173.87 1346.40 619.86 3908.39
2008 993.42 235.83 415.84 157.80 1.23 173.23 216.58 188.88 1759.95 679.98 4822.74
2009 1295.14 160.59 529.02 178.31 1.00 251.23 279.73 232.82 2210.42 763.14 5901.40
2010 1589.57 327.91 683.83 208.22 0.61 224.04 321.58 242.32 3033.28 932.67 7564.03
2011 1794.49 374.25 800.91 242.83 0.74 257.88 363.36 255.40 2723.46 990.50 7803.82
2012 2065.41 373.30 963.46 245.79 0.48 206.63 433.26 276.48 3522.96 1007.17 9094.94
Annexure 7H(2) .: Computation of Total Overhead Expense (SBT) (Rs. in crores)

Selling
Outsou
jobs
s
1996 11.84 1.06 16.21 0.08 0.00 1.05 0.00 0.00 25.13 4.89 60.26
1997 13.76 1.57 4.22 0.18 0.00 0.51 0.00 0.00 26.54 7.16 53.94
1998 15.93 2.31 3.47 0.10 0.00 0.60 0.00 0.00 35.10 7.97 65.48
1999 16.83 2.24 4.66 0.10 0.00 0.75 0.00 0.00 31.84 10.74 67.16
2000 19.50 2.55 5.56 2.22 0.11 0.87 2.92 4.04 24.03 11.11 72.91
2001 21.03 2.83 6.74 2.59 0.09 1.26 2.62 4.49 26.89 13.63 82.17
2002 22.79 3.32 7.46 3.37 0.12 1.28 2.17 4.52 26.72 16.59 88.34
2003 27.29 3.91 6.40 3.42 0.08 1.72 0.98 5.08 29.85 19.88 98.61
2004 31.18 4.36 10.11 6.91 0.18 1.20 1.02 5.92 41.13 33.68 135.69
2005 37.92 5.24 18.69 8.00 0.18 1.36 1.54 6.27 48.48 37.41 165.09
2006 44.46 6.16 23.76 8.73 0.23 3.60 4.49 6.28 115.00 52.58 265.29
2007 49.53 6.70 26.24 11.06 0.17 3.35 1.99 6.97 87.30 41.90 235.21
2008 55.35 9.37 32.63 13.58 0.17 2.48 3.53 7.35 98.61 48.25 271.32
2009 67.95 11.37 37.59 17.40 0.32 3.99 5.93 8.02 117.62 41.34 311.53
2010 74.58 11.75 45.67 15.43 0.39 8.60 5.15 7.68 128.66 42.82 340.73
2011 86.14 13.69 51.67 18.92 0.43 7.88 7.11 8.84 159.30 46.94 400.92
2012 96.21 14.54 61.91 20.25 0.43 7.21 5.95 9.94 175.40 50.20 442.04
Annexure 7H (3) .: Computation of Total Overhead Expense (SB) (Rs. in crores)

Selling
Outsou
jobs
s
1996 31.22 2.98 19.67 0.23 0.00 0.48 0.00 0.00 57.67 8.50 120.75
1997 33.84 3.79 19.33 0.83 0.00 0.49 0.00 0.00 41.45 8.67 108.40
1998 39.03 4.39 7.08 0.54 0.00 0.83 0.00 0.00 54.07 9.14 115.08
1999 45.60 5.62 8.50 0.65 0.00 0.87 0.00 0.00 58.77 13.33 133.34
2000 48.74 6.94 10.14 4.65 0.16 27.87 10.50 6.16 48.57 18.27 182.00
2001 53.13 7.87 11.12 5.36 0.15 31.35 11.48 7.11 48.68 19.59 195.84
2002 58.22 9.51 12.38 8.60 0.24 26.38 12.26 8.02 52.38 24.58 212.57
2003 64.31 12.04 13.64 7.54 0.22 23.34 12.93 9.03 59.46 33.74 236.25
2004 63.50 39.65 15.72 13.68 0.38 24.70 14.45 11.38 69.76 26.52 279.74
2005 64.85 20.93 33.28 14.04 0.36 27.78 15.13 10.31 78.62 37.58 302.88
2006 75.00 41.28 44.18 16.04 0.42 13.09 15.50 10.46 136.20 45.43 397.60
2007 87.38 45.34 57.84 17.59 0.46 17.73 18.90 13.04 149.82 83.40 491.50
2008 101.58 44.57 76.38 25.58 0.84 12.82 19.53 14.25 167.03 103.05 565.63
294
2009 120.69 57.94 91.21 27.99 0.93 62.45 19.37 14.27 162.93 113.02 670.80
2010 133.60 48.57 105.12 28.91 0.74 63.19 18.92 15.86 192.72 88.17 695.80
2011 147.86 62.83 104.84 28.07 0.44 70.41 18.30 14.12 256.96 70.96 774.79
2012 167.47 94.57 125.14 30.34 0.85 14.62 28.69 15.56 379.34 66.04 922.62
Annexure 7H (4) .: Computation of Total Overhead Expense (UCO) (Rs. in crores)

Selling
Outsou
Repairs Insura & Miscellan
& nce distribu eous Depreci
mainten premiu tion expenditu ation
ance m paid expense re
jobs
s
1996 25.86 1.27 4.86 2.36 0.11 0.69 2.54 5.49 32.18 6.46 81.82
1997 26.22 1.34 5.77 0.89 0.00 0.70 0.00 0.00 52.03 6.62 93.57
1998 28.96 1.84 6.79 3.80 0.22 0.77 2.64 6.96 35.78 6.73 94.49
1999 31.39 2.82 7.95 4.62 0.18 0.89 2.50 6.92 43.11 11.02 111.40
2000 32.42 2.60 8.34 4.54 0.14 1.23 2.56 8.20 44.91 13.46 118.40
2001 33.36 3.07 10.19 4.39 0.04 0.98 2.92 7.59 47.18 18.31 128.03
2002 36.55 4.04 11.54 5.99 0.23 1.70 5.82 8.04 53.89 22.24 150.04
2003 39.73 4.75 13.66 6.24 0.00 1.82 6.92 10.27 57.39 36.80 177.58
2004 43.68 6.57 17.45 10.16 0.33 5.53 6.23 11.48 77.88 24.69 204.00
2005 46.22 5.95 31.05 18.11 0.18 6.00 6.83 12.70 63.30 35.56 225.90
2006 53.31 4.84 45.57 16.01 0.48 8.43 8.08 15.89 101.80 43.03 297.44
2007 62.89 5.49 50.25 17.31 0.52 12.66 12.51 16.10 121.16 60.56 359.45
2008 79.28 5.22 57.98 19.62 0.77 11.16 24.69 18.56 124.23 69.87 411.38
2009 85.22 5.46 69.83 21.55 0.68 19.30 18.91 19.70 141.17 83.61 465.43
2010 94.66 5.63 92.77 26.70 0.70 18.87 30.01 19.78 163.49 74.19 526.80
2011 105.72 6.46 100.92 29.78 0.67 21.68 34.83 21.36 193.26 80.55 595.23
2012 122.78 6.62 115.10 37.96 0.78 24.01 37.25 26.69 223.87 78.12 673.18
Annexure 7H (5) .: Computation of Total Overhead Expense (CBI) (Rs. in crores)

Selling
Outsou
jobs
s
1996 39.48 5.52 8.99 3.98 0.06 1.13 5.10 9.86 80.66 11.57 166.35
1997 41.48 6.45 10.43 9.66 0.11 1.37 3.53 13.29 91.26 15.92 193.50
1998 49.54 7.53 13.68 9.22 0.13 2.33 5.30 11.72 99.01 19.69 218.15
1999 52.80 9.28 14.64 10.43 0.15 2.65 6.77 11.64 134.17 25.16 267.69
2000 55.08 10.90 16.57 9.97 0.12 3.09 4.89 13.30 105.77 31.72 251.41
2001 62.59 12.93 19.43 13.04 0.07 3.36 3.87 13.39 121.86 45.40 295.94
2002 67.09 16.73 21.60 15.37 0.06 4.56 4.50 13.46 123.59 45.49 312.45
2003 74.64 17.56 23.74 16.09 0.34 5.47 4.55 14.36 134.29 49.42 340.46
2004 84.00 20.56 24.24 22.75 0.60 6.93 6.09 17.51 135.08 56.33 374.09
2005 94.31 23.62 41.06 24.09 0.54 7.04 5.93 15.77 142.82 53.17 408.35
2006 100.11 25.16 57.29 23.40 0.53 7.79 5.88 16.25 153.18 50.90 440.49
2007 109.00 33.86 64.38 22.99 0.45 6.01 10.77 16.61 179.87 65.00 508.94
2008 124.80 20.66 75.72 29.27 1.08 6.87 11.32 17.21 170.96 73.61 531.50
2009 142.13 32.89 91.41 31.18 1.45 15.33 9.72 17.88 163.53 83.59 589.11
2010 162.69 24.27 121.20 32.01 1.54 17.22 4.71 18.93 216.52 78.75 677.84
2011 191.50 64.86 155.75 37.48 1.30 96.94 35.33 21.57 311.35 118.69 1034.77
2012 256.78 83.31 164.60 32.08 0.94 35.22 39.45 26.48 460.36 143.54 1242.76
295
Annexure 7H (6) .: Computation of Total Overhead Expense (PNB) (Rs. in crores)
Outsou Selling
Repairs Insuran Miscellan
Year Rent & rced & Communi Printing & Total
& ce Directo eous Deprecia
end lease profess distributi cations stationery Overhead
maintena premiu rs' fees expenditu tion
March rent ional on expenses expenses Expense
nce m paid re
jobs expenses
1996 41.45 7.44 45.14 8.96 0.13 1.66 14.24 15.29 83.15 24.78 242.24
1997 47.34 8.81 41.99 4.80 0.00 1.59 0.00 0.00 120.08 28.81 253.42
1998 55.29 10.34 18.14 6.05 0.00 3.04 0.00 0.00 129.60 34.75 257.21
1999 64.70 13.57 23.61 11.28 0.22 3.92 22.86 20.31 106.65 39.97 307.09
2000 71.98 13.68 30.55 13.03 0.12 4.34 23.66 22.34 117.48 42.99 340.17
2001 82.41 16.97 33.11 14.88 0.17 4.71 25.67 24.45 135.56 74.62 412.55
2002 94.31 18.98 38.88 20.92 0.31 8.42 28.50 28.56 158.72 85.29 482.89
2003 105.57 19.93 46.10 24.15 0.29 7.84 31.31 52.51 164.00 128.97 580.67
2004 119.96 24.42 54.66 31.13 0.35 10.84 45.11 35.09 213.66 181.45 716.67
2005 133.96 29.28 84.08 33.10 0.69 19.16 47.52 36.65 286.25 183.28 853.97
2006 154.32 39.07 116.66 29.90 0.73 20.15 55.71 38.70 266.31 186.64 908.19
2007 173.38 40.86 127.81 33.09 0.82 18.03 51.51 39.51 293.97 194.79 973.77
2008 193.94 55.39 140.82 49.28 0.94 23.32 55.57 44.35 330.09 170.24 1063.94
2009 226.27 66.87 169.65 46.35 1.35 31.24 76.59 45.43 427.00 191.07 1281.82
2010 260.39 77.19 220.91 55.88 1.42 40.11 101.05 47.56 613.46 222.82 1640.79
2011 307.57 84.66 261.65 60.49 1.46 39.68 103.27 50.96 737.52 255.85 1903.11
2012 363.05 100.85 315.94 74.03 1.25 37.16 105.63 59.46 929.64 292.26 2279.27
Annexure 7H(7) .: Computation of Total Overhead Expense (PSB) (Rs. in crores)

Selling
Outsou
Year Rent & rced Direc Commun Printing & Total
end lease professi tors' ications stationery Overhead
March rent onal fees expenses expenses Expense
jobs
s
1996 947.43 136.17 610.50 109.26 2.31 51.63 158.12 197.06 1306.88 358.45 3877.81
1997 1081.08 164.36 559.85 115.71 1.84 49.06 75.13 123.04 1744.74 452.62 4367.43
1998 1235.38 191.11 318.58 110.23 1.90 65.29 75.79 108.30 1896.49 592.97 4596.04
1999 1348.57 217.78 354.88 144.31 3.16 72.83 168.40 218.09 1927.78 829.92 5285.72
2000 1515.53 245.03 375.15 225.91 4.89 96.03 232.79 305.95 2086.11 1009.18 6096.57
2001 1672.70 267.60 421.94 236.04 3.43 114.44 243.94 328.24 2550.81 1166.56 7005.70
2002 1864.02 290.29 497.44 279.62 5.43 117.65 259.32 363.45 2316.91 1328.02 7322.15
2003 2046.62 340.91 556.49 300.72 7.30 152.07 257.00 414.89 2745.33 1594.97 8416.30
2004 2253.93 436.69 636.47 432.67 9.84 251.55 334.39 453.15 3187.73 1931.20 9927.62
2005 2542.76 500.41 1028.69 432.91 9.91 290.90 355.20 490.65 3734.47 2183.36 11569.26
2006 2889.42 653.40 1443.15 471.32 11.20 385.35 465.05 539.95 4677.67 2483.19 14019.70
2007 3292.04 728.14 1674.24 525.99 12.00 427.65 504.54 552.89 5251.94 2518.33 15487.76
2008 3705.35 869.66 2002.56 771.12 14.96 548.04 682.83 612.58 5988.41 2713.39 17908.90
2009 4443.63 948.05 2443.89 888.88 18.36 814.46 825.86 687.94 6905.56 2794.07 20770.70
2010 5148.70 1200.62 3026.28 1008.92 18.39 741.15 1002.44 717.16 8449.41 3139.63 24452.70
2011 5900.88 1362.13 3564.87 1145.58 17.43 987.48 1096.87 783.78 9519.98 3364.04 27743.04
2012 6880.53 1630.36 4200.48 1237.51 19.56 784.27 1269.56 872.15 11996.38 3502.97 32393.77
Annexure 7I (1) .: Computation of Operating Profit (SBI) (Rs. in crores)

Total Total Total
Extra- Non Staff
Year end Net Interest Operating Overhead Operating Operating
ordinary Interest Expenses
March Margin (1) Income Expenses Expenses Profit (4-7)
Income (2) Income (3) (5)
[4=(1+2+3)] (6) [7=(5+6)]
1996 4732.64 0.49 2789.21 7522.34 3351.64 1107.89 4459.53 3062.81
1997 -162.29 0.34 8164.24 8002.29 3322.93 1281.73 4604.66 3397.63
1998 -985.96 965.20 9210.72 9189.96 3557.75 1163.12 4720.87 4469.09
1999 6063.10 0.00 3298.96 9362.06 4462.71 1443.36 5906.07 3455.99
2000 6928.34 0.82 4106.89 11036.05 4598.87 1817.31 6416.18 4619.87
2001 8382.57 0.03 4028.14 12410.74 6011.65 2287.18 8298.83 4111.91
296
2002 9081.24 72.75 4205.01 13359.00 5152.78 2058.12 7210.90 6148.10
2003 9977.56 708.76 5787.02 16473.34 5688.72 2253.71 7942.43 8530.91
2004 11186.31 45.61 7680.46 18912.38 6447.69 2797.62 9245.31 9667.07
2005 13199.36 746.28 7153.03 21098.67 6907.35 3166.83 10074.18 11024.49
2006 15589.12 60.35 7467.81 23117.28 8123.04 3602.06 11725.10 11392.18
2007 15322.95 50.42 7396.09 22769.46 7932.59 3908.39 11840.98 10928.48
2008 17021.22 14.11 9476.08 26511.41 7788.94 4822.74 12611.68 13899.73
2009 20873.14 57.80 12694.31 33625.25 9805.11 5901.40 15706.51 17918.74
2010 23671.44 7.99 15966.60 39646.03 12762.64 7564.03 20326.67 19319.36
2011 32526.41 1008.39 15847.78 49382.58 16185.91 7803.82 23989.73 25392.85
2012 43291.08 160.53 15334.91 58786.52 17015.15 9094.94 26110.09 32676.43
Sources: Prowess Database, NIM given in Annexure 7G (1) and Total Overhead Expenses given in Annexure 7H (1)
Annexure 7I (2) .: Computation of Operating Profit (SBT) (Rs. in crores)

Non Total Total
Net Extra- Staff Total
Year end Interest Operating Operating Operating
Interest ordinary Expenses Overhead
March Income Income Expenses Profit (4-7)
Margin (1) Income (2) (5) Expenses (6)
(3) [4=(1+2+3)] [7=(5+6)]
1996 -4.15 0.05 346.34 342.24 137.35 60.26 197.61 144.63
1997 -21.91 0.05 393.45 371.59 167.07 53.94 221.01 150.58
1998 -103.36 68.75 520.28 485.67 151.54 65.48 217.02 268.65
1999 -231.98 30.45 622.87 421.34 173.31 67.16 240.47 180.87
2000 281.93 15.06 198.61 495.60 221.98 72.91 294.89 200.71
2001 395.12 0.31 193.95 589.38 277.47 82.17 359.64 229.74
2002 424.43 12.54 232.94 669.91 245.50 88.34 333.84 336.07
2003 522.85 1.10 301.99 825.94 270.40 98.61 369.01 456.93
2004 683.33 0.54 470.99 1154.86 318.26 135.69 453.95 700.91
2005 896.16 60.02 408.50 1364.68 389.04 165.09 554.13 810.55
2006 955.13 37.85 351.10 1344.08 406.36 265.29 671.65 672.43
2007 992.18 0.36 363.29 1355.83 411.53 235.21 646.74 709.09
2008 957.28 7.85 438.53 1403.66 431.62 271.32 702.94 700.72
2009 1282.55 115.26 573.31 1971.12 531.45 311.53 842.98 1128.14
2010 1400.16 60.40 562.59 2023.15 615.14 340.73 955.87 1067.28
2011 1696.04 10.99 841.01 2548.04 566.01 400.92 966.93 1581.11
2012 1830.37 141.99 678.56 2650.92 743.12 442.04 1185.16 1465.76
Annexure 7I(3) .: Computation of Operating Profit (SB) (Rs. in crores)

Net Non Total Total
Extra- Staff Total
Year end Interest Interest Operating Operating Operating
ordinary Expenses Overhead
March Margin Income Income Expenses Profit (4-7)
Income (2) (5) Expenses (6)
(1) (3) [4=(1+2+3)] [7=(5+6)]
1996 497.33 0.27 143.63 641.23 419.66 120.75 540.41 100.82
1997 -246.97 0.23 959.73 712.99 483.12 108.40 591.52 121.47
1998 -328.76 19.41 1107.07 797.72 503.46 115.08 618.54 179.18
1999 -245.42 38.37 1178.32 971.27 589.31 133.34 722.65 248.62
2000 825.30 16.61 306.39 1148.30 669.21 182.00 851.21 297.09
2001 1093.64 24.07 294.92 1412.63 881.43 195.84 1077.27 335.36
2002 1107.55 19.07 299.13 1425.75 815.78 212.57 1028.35 397.40
2003 1209.72 5.21 513.47 1728.40 849.76 236.25 1086.01 642.39
2004 1429.21 31.23 792.60 2253.04 906.62 279.74 1186.36 1066.68
2005 1693.83 1.55 589.13 2284.51 961.33 302.88 1264.21 1020.30
2006 1880.87 30.86 591.94 2503.67 1037.22 397.60 1434.82 1068.85
2007 2150.05 58.42 651.62 2860.09 894.47 491.50 1385.97 1474.12
2008 2072.75 0.43 940.00 3013.18 928.91 565.63 1494.54 1518.64
2009 2595.57 67.42 914.55 3577.54 1120.16 670.80 1790.96 1786.58
2010 2820.68 0.24 1212.37 4033.29 1337.76 695.80 2033.56 1999.73
2011 4443.53 0.18 915.42 5359.13 1773.31 774.79 2548.10 2811.03
2012 5085.03 0.53 1151.01 6236.57 1891.50 922.62 2814.12 3422.45
297
Annexure 7I (4) .: Computation of Operating Profit (UCO) (Rs. in crores)
Net Non Total Total
Extra- Staff Total
(1) (3) [4=(1+2+3)] [7=(5+6)]
1996 335.71 0.00 148.85 484.56 424.67 81.82 506.49 -21.93
1997 -281.54 34.53 741.46 494.45 422.58 93.57 516.15 -21.70
1998 351.44 0.20 211.20 562.84 443.22 94.49 537.71 25.13
1999 446.47 4.81 215.49 666.77 483.96 111.40 595.36 71.41
2000 552.56 5.85 268.83 827.24 506.23 118.40 624.63 202.61
2001 662.05 0.02 310.06 972.13 617.95 128.03 745.98 226.15
2002 729.75 0.10 605.19 1335.04 686.73 150.04 836.77 498.27
2003 882.02 0.17 626.88 1509.07 689.69 177.58 867.27 641.80
2004 1194.60 12.63 629.99 1837.22 817.10 204.00 1021.10 816.12
2005 1408.24 0.13 526.68 1935.05 859.41 225.90 1085.31 849.74
2006 1565.76 2.59 464.58 2032.93 879.94 297.44 1177.38 855.55
2007 1694.68 54.70 550.22 2299.60 833.13 359.45 1192.58 1107.02
2008 1532.53 95.17 777.75 2405.45 896.95 411.38 1308.33 1097.12
2009 1767.12 1.12 1036.01 2804.25 1147.54 465.43 1612.97 1191.28
2010 2324.12 1.11 972.61 3297.84 1191.67 526.80 1718.47 1579.37
2011 3844.91 1.54 928.13 4774.58 1333.05 595.23 1928.28 2846.30
2012 3902.11 1.20 992.19 4895.50 1222.99 673.18 1896.17 2999.33
Annexure 7I (5) .: Computation of Operating Profit (CBI) (Rs. in crores)

Net Non Total Total
Extra- Staff Total
(1) (3) [4=(1+2+3)] [7=(5+6)]
1996 729.91 0.34 270.76 1001.01 622.72 166.35 789.07 211.94
1997 835.78 78.17 305.61 1219.56 647.55 193.50 841.05 378.51
1998 950.16 0.16 343.01 1293.33 713.42 218.15 931.57 361.76
1999 1048.78 95.52 348.66 1492.96 831.19 267.69 1098.88 394.08
2000 -631.85 0.10 2313.91 1682.16 990.84 251.41 1242.25 439.91
2001 1449.19 0.06 477.05 1926.30 1152.22 295.94 1448.16 478.14
2002 1534.98 0.03 647.21 2182.22 1118.80 312.45 1431.25 750.97
2003 -641.68 0.03 3139.86 2498.21 1186.65 340.46 1527.11 971.10
2004 -535.30 0.38 3647.63 3112.71 1183.47 374.09 1557.56 1555.15
2005 2374.95 0.00 939.19 3314.14 1277.56 408.35 1685.91 1628.23
2006 2380.08 278.42 551.36 3209.86 1275.74 440.49 1716.23 1493.63
2007 2474.42 163.33 562.12 3199.87 1175.42 508.94 1684.36 1515.51
2008 2111.80 52.52 902.80 3067.12 1214.34 531.50 1745.84 1321.28
2009 2228.47 0.00 1070.32 3298.79 1272.59 589.11 1861.70 1437.09
2010 2545.30 53.69 1704.06 4303.05 1566.37 677.84 2244.21 2058.84
2011 5325.34 0.00 1265.15 6590.49 2668.84 1034.77 3703.61 2886.88
2012 5168.65 112.38 1395.72 6676.75 2210.86 1242.76 3453.62 3223.13
Sources: Prowess Database, NIM given in Annexure 7G (5) and Total Overhead Expenses given in Annexure 7H(5)
Annexure 7I (6) .: Computation of Operating Profit (PNB) (Rs. in crores)

Non Total Total
Net Extra- Staff Total
Year end Interest Operating Operating Operating
Interest ordinary Expenses Overhead
March Income Income Expenses Profit (4-7)
Margin (1) Income (2) (5) Expenses (6)
(3) [4=(1+2+3)] [7=(5+6)]
1996 1078.73 0.64 328.96 1408.33 757.84 242.24 1000.08 408.25
1997 -310.11 80.94 1919.22 1690.05 811.02 253.42 1064.44 625.61
1998 -527.40 386.21 2442.83 2301.64 871.44 257.21 1128.65 1172.99
1999 1571.87 81.50 608.72 2262.09 1068.93 307.09 1376.02 886.07
2000 1599.49 17.69 775.83 2393.01 1183.67 340.17 1523.84 869.17
2001 2038.43 0.54 828.17 2867.14 1459.08 412.55 1871.63 995.51
2002 2295.29 0.67 1076.08 3372.04 1316.32 482.89 1799.21 1572.83
2003 3123.72 102.91 1326.52 4553.15 1476.08 580.67 2056.75 2496.40
2004 3623.96 0.80 2056.36 5681.12 1654.06 716.67 2370.73 3310.39
298
2005 4006.73 1.05 2136.90 6144.68 2423.98 853.97 3277.95 2866.73
2006 4666.77 339.47 1548.13 6554.37 2114.97 908.19 3023.16 3531.21
2007 5213.23 1.02 1931.69 7145.94 2352.45 973.77 3326.22 3819.72
2008 5534.16 1.47 2025.00 7560.63 2461.54 1063.94 3525.48 4035.15
2009 6831.91 63.54 3365.37 10260.82 2924.38 1281.82 4206.20 6054.62
2010 8478.08 23.83 3658.26 12160.17 3136.69 1640.79 4777.48 7382.69
2011 11807.34 48.11 3642.24 15497.69 3752.57 1903.11 5655.68 9842.01
2012 13414.44 12.39 4272.99 17699.82 4015.15 2279.27 6294.42 11405.40
Annexure 7I (7) .: Computation of Operating Profit (PSB) (Rs. in crores)

Total Total
Year Extra- Non Total
Net Interest Operating Staff Operating Operating
end ordinary Interest Overhead
Margin (1) Income Expenses (5) Expenses Profit (4-7)
March Income (2) Income (3) Expenses (6)
[4=(1+2+3)] [7=(5+6)]
1996 13445.21 7.83 8744.05 22197.09 10890.45 3877.81 14768.26 7428.83
1997 5875.11 317.10 19218.81 25411.02 11631.52 4367.43 15998.95 9412.07
1998 4296.70 2101.20 23656.63 30054.53 12623.51 4596.04 17219.55 12834.98
1999 12669.33 585.76 18243.56 31498.65 15155.08 5285.72 20440.80 11057.85
2000 19532.11 171.96 16837.22 36541.29 16547.18 6096.57 22643.75 13897.54
2001 27725.23 81.75 14365.09 42172.07 20992.63 7005.7 27998.33 14173.74
2002 28654.21 207.81 19833.25 48695.27 19099.39 7322.15 26421.54 22273.73
2003 30032.93 1071.38 28848.45 59952.76 20483.92 8416.3 28900.22 31052.54
2004 36126.12 597.66 36452.61 73176.39 22716.19 9927.62 32643.81 40532.58
2005 49548.50 1805.75 25481.61 76835.86 25683.75 11569.26 37253.01 39582.85
2006 57580.51 4000.71 24064.15 85645.37 27401.15 14019.7 41420.85 44224.52
2007 62753.37 1353.09 26017.74 90124.20 27822.26 15487.76 43310.02 46814.18
2008 64390.37 1639.18 33910.69 99940.24 28673.61 17908.9 46582.51 53357.73
2009 80005.85 1561.88 42982.40 124550.13 35038.89 20770.7 55809.59 68740.54
2010 93579.37 1100.09 51951.53 146630.99 41495.23 24452.7 65947.93 80683.06
2011 135217.14 2272.43 50607.44 188097.01 53488.44 27743.04 81231.48 106865.53
2012 156349.21 1596.26 79575.44 237520.91 54534.76 32393.77 86928.53 150592.38
299

THESIS Gour Bandyopadhyay 2016 PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

THESIS Gour Bandyopadhyay 2016 PDF

Hochgeladen von

Copyright:

Verfügbare Formate

An Analysis of the Non-Performing

Assets of Some Selected Public

Maulana Abul Kalam Azad University of Technology,

Prof. (Dr.) Malayendu Saha

IV Dynamicity of NPA over Recent Years in Selected PSBs in 86-138

Bibliography 275– 286

First of all I would like to express my sincere gratitude to my Research Supervisor

1.3 Review of Literatures

1.4 Identification of the Research Gap

1.5 Objective of the Study

1.6 Research Methodology

1.7 Scope of the Study

1.8 Scheme of Research Work

Chart-1.1.: GNPA of SCB and PSB

Chart-1.3.: GNPA Ratio of SCB and PSB

Macroeconomic/ industry impact of NPA :

Indirectly the burden of NPA is to be borne by the society as a whole. When

Subsequently, a different committee [Committee on Banking Sector Reforms (CBSR)] was

Singla (2008) examines empirically the financial performance in terms of profitability of

Thiagarajan, Ayyappan, Ramachandran and Sakthivadivel (2011) in their paper empirically

Veerakumar (2012) in an attempt to study NPA in priority sector advances (PSA) in a

1.4. Identification of Research Gap

1.5. Objective of the Study

i) to review the existing norms and concepts related to NPA,

1.6. Research Methodology

1.7. Scope of the Study

Commercial Banks in India

Schedule Commercial Non Schedule Commercial

Nationalised Banks (19) Old Private Banks (13)

State Bank of India and its New Private Banks (7)

Other Public Sector Bank

Figures in the bracket represent number of banks in the group.

Chapter 3 delineates on the mechanism for effective management of NPAs as initiated by

Existing Prudential Norms on Asset Quality

2.2 Review of literature

2.3 Norms Relating to Income Recognition.

2.4 Norms Relating to Asset Classification

2.5 Norms Relating to Provisioning

2.6 Treatment of Prudential norms on various types of Advances

2.7 Operational Concepts Related to NPAs

A high-power Committee on the Financial System, with Shri M. Narasimham as the

2.3. Norms Relating to Income Recognition

Table- 2.1.: Summary of Progressive Changes in Income Recognition Norms

2.3.1. Reversal of Income

2.3.2. Appropriation of Recoveries

2.3.3. Interest Application

2.4. Norms Relating to Asset Classification

• Asset classification under Health Code System.

• Asset classification under Prudential Norms.

2.4.1. Asset classification under Health Code (HC) System

This category covers all borrowers where

This category of borrowers covers those accounts where

Code 4: Sick: Non-Viable / Sticky

Code 5: Advances Recalled

Code 6: Suit-filed Accounts

Code 8: Debts classified by the bank as Bad and Doubtful

2.4.2. Asset classification under Prudential Norms

Table- 2.2.: Summary of Changes in NPA Norms

iii) Bills Purchased/ Discounted : In case of Bills Purchased/ Discounted ( Inland or

2.4.4. Norms Relating to Classification of NPAs

2.4.4.1. Sub Standard Asset

2.4.4.2. Doubtful Asset

2.4.4.3. Loss Asset