RISK SOLUTIONS

Risk Solutions:
Katarina Antens-Miller
The Relationship Between Default
Director
New York Rates And Recovery
(1) 212-438-6679
katarina_antens-miller@
As many banks are preparing to comply with Basel II requirements, the estimation of the main
standardandpoors.com
parameters for the Basel II framework, such as probability of default (PD), has become one of
the major focuses for risk management in the banking industry. PD estimation is a relatively
well-developed area within the Basel II framework. There are volumes of academic research
and commercial solutions regarding PD estimation. However, there has not been such progress
in post-default recovery estimation due to data limitation. The default data is scarce, and
collecting it is the most challenging aspect of any PD modeling. Tracking all post-default
activities and recording the recovery value when the company has emerged from default can
often take a few years to complete. As a result, there is only minimal academic research on
recovery modeling.

Many contend that, over time, there is a “negative” relationship between PD and recovery
(i.e., the higher the default rate, the lower the recovery rate, and vice versa). While there are
not many studies that illustrate this empirical relationship, “The Link between Default and
Recovery Rates: Theory, Empirical Evidence and Implications” (1) concludes that “aggregate
recovery rates are basically a function of supply and demand for the [defaulted] securities.”
The latest data collected by Standard & Poor’s Risk Solutions and Data Center of Excellence
confirms this finding (although our underlying dataset may have been calculated somewhat
Publication Date:
differently).
Jan. 24, 2007

Data
We collected U.S. speculative-grade annual default rates from 1988 to 2005, and also U.S.
investment-grade annual default rates from 1990 to 2005. Both default rates were obtained
®
from Standard & Poor’s CreditPro version 7.5. Comparable recovery data was collected
 The Relationship Between Default Rates And Recovery

®
from Standard & Poor’s LossStats Database version 1.7. We selected investment–grade and
speculative-grade recoveries by looking at the earliest information available. The time frame of the data
is consistent with the default rate data. We included all debt types available from the LossStats
Database, though we only included recovery values from bankruptcy events. The recovery value
implied is the issuer-level overall recovery. We took yearly averages of these issuer-level overall
recoveries to obtain year-by-year numbers. Lastly, all recovery values are as of the emergence year.

Correlation Between Default Rates And Recovery Rates From Time Series Data
From our dataset, we plotted a time series of default rates and recovery data for each rating grade. As
we can see from chart 1, in the case of speculative-grade post-default recoveries, a clear negative
correlation between the two series can be observed. However, using the investment-grade data series,
the correlation is not as clear.

Year-by-year Comparison Of Default Rates And Recovery Rates

U.S. Speculative Grade
Recovery Rate Default Rate
(Right Scale)

(Recovery rate) (Default rate)

70% 12%

60% 10%

50%
8%
40%
6%
30%
4%
20%

10% 2%

0% 0%
88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05
19

19

19

19

19

19

19

19

19

19

19

19

20

20

20

20

20

20

© Standard & Poor's 2007

Standard & Poor’s | COMMENTARY 2

 The Relationship Between Default Rates And Recovery

Year-by-year Comparison Of Default Rates And Recovery Rates

U.S. Investment Grade

Recovery Rate Default Rate

(Right Scale)

(Recovery rate) (Default rate)

100% 0.60%
90%
0.50%
80%
70%
0.40%
60%
50% 0.30%
40%
0.20%
30%
20%
0.10%
10%
0% 0.00%
90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05
19

19

19

19

19

19

19

19

19

19

20

20

20

20

20

20
© Standard & Poor's 2007

Simple Linear Regression Analysis Using Time Series Data

To quantify the correlation estimates from the data, we performed a simple regression analysis for each
rating grade. Table 1 again shows the strong negative correlation between U.S. speculative-grade
default rates and post-default recoveries. According to P values of the F test and T test, we can see that
the regression model and its coefficients are statistically valid. The correlation between default rates and
recovery rates is -0.66.

Table 1

Regression Analysis On Recovery Rates Against Default Rates (cont.'d)

U.S. Speculative Grade

Regression Statistics

Multiple R 0.655987211
R-Square 0.430319221
Adjusted R-Square 0.394714172
Standard Error 0.074198273
Observations 18

ANOVA

df SS MS F Significance F
Regression 1 0.066537543 0.066537543 12.08590458 0.00311448

www.standardandpoors.com 3
 The Relationship Between Default Rates And Recovery

Table 1

Regression Analysis On Recovery Rates Against Default Rates (cont.'d)

Residual 16 0.08808614 0.005505384
Total 17 0.154623683
Coefficients Standard Error T Stat P Value
Intercept 0.55122784 0.035699365 15.44083025 4.9449E-11
X Variable 1 -2.206147885 0.634592654 -3.476478762 0.00311448

It’s harder to justify the relationship between default rates and recovery rates for the investment-grade
category. The regression analysis suggests that the relationship is not statistically meaningful. The
correlation is only -0.27 (see table 2). This weak correlation may be due, in part, to the dataset, which
does not include many examples of investment-grade defaults.

Table 2

Regression Analysis On Recovery Rates Against Default Rates

U.S. Investment Grade

Regression Statistics

Multiple R 0.274484534
R-Square 0.07534176
Adjusted R-Square 0.009294742
Standard Error 0.18107178
Observations 16

ANOVA

df SS MS F Significance F
Regression 1 0.037401076 0.037401076 1.140729177 0.303569518
Residual 14 0.459017853 0.032786989
Total 15 0.496418928
Coefficients Standard Error T Stat P Value
Intercept 0.620356505 0.060055881 10.32965452 6.24165E-08
X Variable 1 -33.54612993 31.40878595 -1.068049239 0.303569518

Practical Application
Now that we have confirmed a negative correlation between default rates and recovery rates in U.S.
speculative-grade data, we may utilize this relationship to estimate recovery rates given default rates. As
a simple example, suppose an investor holds a speculative-grade rated bond and wants to estimate its
recovery rate as of early 2006. We can use the regression equation from table 1 and come up with a
quick and simple estimate. To illustrate this, we assume the probability of default of the bond as simply
the historical speculative-grade default rate over a certain period. According to Standard & Poor’s
CreditPro, the speculative-grade default rate from 1981–2005 data was 4.7%. Using this default rate,
recovery rates can be calculated as follows:
0.55 + (-2.21 * 0.047) = 44.61%

Standard & Poor’s | COMMENTARY 4

 The Relationship Between Default Rates And Recovery

(Note that 0.55 and -2.21 are coefficients from table 1.)
We emphasize that this example illustrates a simple approach and may not be useful for many cases. In
fact, linear regression may not fit the historical observations as well as by applying a nonlinear
regression. In chart 3, a second-order polynomial regression was utilized for a closer fit.

Regression Analysis On Recovery Rates Against Default Rates

U.S. Speculative Grade
(Recovery rate)
70%
64%
60%
56%
55% 52%
50%
50%
45% 43% 49% 46%
45%
42% 41% 39%
40%
43% 37% 37%
33%
30%

20% 22%

10%

0%
0% 2% 4% 6% 8% 10% 12%
(Default rate) Default rate trendline equation:
y = 30.872x^2 - 5.8911x + 0.6342
© Standard & Poor's 2007
R^2 = 0.4893

By using nonlinear regression, we could marginally improve R-square to 0.49 from 0.43. Using the
same example described above, we can estimate recovery rates as follows:
2
30.872 * (0.047) – 5.8911 * 0.047 + 0.6342 = 42.55%

One quick way to validate how well this simple nonlinear regression approach works is by using out-
of-time data. We used data from 1988 to 2002 only, and estimated the regression equation once again.
The regression equation from this data is:
2
y = 20.446x - 4.4992x + 0.5989
(R-square of this regression was 0.47.)

Then, we estimated recovery rates from 2003 to 2005 using this model and compared them to the
observed recovery rates. With only one factor (default rates), the estimation is not quite precise. But we
can see that the regression model tracks observed recovery rates’ upward trend correctly (see chart 4).

www.standardandpoors.com 5
 The Relationship Between Default Rates And Recovery

Estimated Recovery Rates And Observed Recovery Rates

Observed Estimated

(Recovery rates)

70%

65%
64%
60%

55%
50%
52%
50%

45%
41%
40%
43%
35%
37%
30%
2003 2004 2005

© Standard & Poor's 2007

One-Factor Regression Approach Is Far From Perfect

As we have seen so far, the estimation of post-default recovery rates from the one-factor regression
approach is far from perfect. In fact, it is not enough to have only one factor (default rates) to estimate
recovery rates precisely. It is desirable to incorporate multiple factors for precise estimation, such as
instrument-specific or company-specific factors. From Standard & Poor’s experience, instrument-
specific factors including debt type, seniority, and collateral type have proven to be important
explanatory factors for recovery rates. While company-specific factors (such as financial ratios) at the
time of default/bankruptcy may not be very useful since financial data tends to be inaccurate or noisy
under distressed circumstances, metrics from one year before the default/bankruptcy may indeed have
strong explanatory power.

One may want to differentiate ultimate recoveries from recoveries based on distressed trading prices.
Most practitioners use a 30-day distressed trading price as an estimate of recovery; however, there is a
rather substantial gap between this 30-day price and ultimate recovery(2). While 30-day prices may be
more suitable when one wants to trade distressed debt on the market, using ultimate recovery values
may be better for risk management purposes.

In addition to the data issue and the factor selection issue, more advanced modeling technology will be
helpful in predicting post-default recovery values as well. Based on internal validation research, the
maximum expected utility (MEU) approach (the methodology used by Standard & Poor’s) has been
proved to be superior to any conventional modeling techniques. MEU methodology also provides the
distribution of recovery, not just a point estimate, which is far more useful for in-depth analysis or

Standard & Poor’s | COMMENTARY 6

 The Relationship Between Default Rates And Recovery

simulations. Standard & Poor’s LossStats Model utilizes MEU methodology, while Standard & Poor’s
recovery ratings go beyond quantitative modeling, providing a comprehensive qualitative approach to
estimating post-default recovery prospects.

Evidences From Standard & Poor’s Recovery Ratings

While overall historical data confirms the relationship over time between default rates and post-default
recovery in general, we often can’t find such a relationship in some specific datasets (such as in U.S.
investment-grade data, as was previously mentioned). In many of its research reports, Standard &
Poor’s Ratings Services has also noted a weak relationship between default risk and recovery.
According to “Benign Leveraged Market or Credit Amnesia? Recovery Ratings Three Years On,”(3)
there was no material correlation between default risk and recovery expectations on secured loans. In
fact, this finding has been confirmed by the data used in this report (see charts 5 and 6). “In large part,
this is because debt in the secured loans are often structured to ensure greater recovery prospects as
default risk rises,” explains the article.

Regression Analysis On Recovery Rates Against Default Rates

U.S. Speculative Grade (Secured Loans/Bonds)
(Recovery rates)
120%

100% 100%
82% 91%
81%
80%
77%
65% 72%
67% 57%
63% 63%
60%
55% 56%

40%
35% 39% 33%
20% 18%

0%
0% 2% 4% 6% 8% 10% 12%
(Default rates)
y = 148.92x2 - 17.644x + 1.0088
© Standard & Poor's 2007 R2 = 0.2682

www.standardandpoors.com 7
 The Relationship Between Default Rates And Recovery

Regression Analysis On Recovery Rates Against Default Rates

U.S. Speculative Grade (Unsecured Loans/Bonds)

(Recovery rates)
45%
40% 39%
40%
36% 35% 34%
35%
34% 30% 30%
28%
30%
26%
21% 22% 22%
25%
19% 19%
20%
14%
15% 12%
16%
10%

5%

0%
0% 2% 4% 6% 8% 10% 12%
(Default rates)
y = 31.263x2 - 5.5164x + 0.4367
© Standard & Poor's 2007
R2 = 0.4086

Still, this point does not weaken our previous findings from broader data, because recovery ratings
have been focused on secured debt since its inception. As Standard & Poor’s Ratings Services expands
its global recovery rating coverage to unsecured and subordinated debt, we will soon be able to provide
more comprehensive analysis on this topic by incorporating the added data.

References
(1)”The Link between Default and Recovery Rates: Theory, Empirical Evidence and Implications,”
published March 2003 by Edward I. Altman, Brooks Brady, Andrea Resti, and Andrea Sironi,
available at http://www.defaultrisk.com/pp_recov_35.htm
(2)See published Dec. 13, 2005 by David C. Schwartz and Jane Zennario of Standard & Poor’s
Risk Solutions
(3) “Benign Leveraged Market or Credit Amnesia? Recovery Ratings Three Years On,” published
Jan. 4, 2007 by William H. Chew of Standard & Poor’s Ratings Services

Acknowledgements
The author thanks Arthur Caramichael, William H. Chew, Perry Sass, Jitendra Sharma, and the
Standard & Poor’s Risk Solutions Editorial Board for their comments and insights. Yong Chi, Emily
Ramdehaul, and Jong Park provided excellent support on statistical analysis and data collection.

Standard & Poor’s | COMMENTARY 8

 The Relationship Between Default Rates And Recovery

Published by Standard & Poor's, a Division of The McGraw-Hill Companies, Inc. Executive offices: 1221 Avenue of the Americas, New York, NY 10020.
Editorial offices: 55 Water Street, New York, NY 10041. Subscriber services: (1) 212-438-7280. Copyright 2007 by The McGraw-Hill Companies, Inc.
Reproduction in whole or in part prohibited except by permission. All rights reserved. Information has been obtained by Standard & Poor's from sources
believed to be reliable. However, because of the possibility of human or mechanical error by our sources, Standard & Poor's or others, Standard & Poor's
does not guarantee the accuracy, adequacy, or completeness of any information and is not responsible for any errors or omissions or the result obtained
from the use of such information. Ratings are statements of opinion, not statements of fact or recommendations to buy, hold, or sell any securities.

Standard & Poor's uses billing and contact data collected from subscribers for billing and order fulfillment purposes, and occasionally to inform subscribers
about products or services from Standard & Poor's, our parent, The McGraw-Hill Companies, and reputable third parties that may be of interest to them. All
subscriber billing and contact data collected is stored in a secure database in the U.S. and access is limited to authorized persons. If you would prefer not to
have your information used as outlined in this notice, if you wish to review your information for accuracy, or for more information on our privacy practices,
please call us at (1) 212-438-7280 or write us at: privacy@standardandpoors.com. For more information about The McGraw-Hill Companies Privacy Policy
please visit www.mcgraw-hill.com/privacy.html.

Analytic services provided by Standard & Poor's Ratings Services ("Ratings Services") are the result of separate activities designed to
preserve the independence and objectivity of ratings opinions. Credit ratings issued by Ratings Services are solely statements of opinion
and not statements of fact or recommendations to purchase, hold, or sell any securities or make any other investment decisions.
Accordingly, any user of credit ratings issued by Ratings Services should not rely on any such ratings or other opinion issued by Ratings
Services in making any investment decision. Ratings are based on information received by Ratings Services. Other divisions of Standard
& Poor's may have information that is not available to Ratings Services. Standard & Poor's has established policies and procedures to
maintain the confidentiality of non-public information received during the ratings process.

Ratings Services receives compensation for its ratings. Such compensation is normally paid either by the issuers of such securities or by
the underwriters participating in the distribution thereof. The fees generally vary from US$2,000 to over US$1,500,000. While Standard
& Poor's reserves the right to disseminate the rating, it receives no payment for doing so, except for subscriptions to its publications.

Permissions: To reprint, translate, or quote Standard & Poor's publications, contact: Client Services, 55 Water Street, New York, NY
10041; (1) 212-438-9823; or by e-mail to: research_request@standardandpoors.com.

www.standardandpoors.com 9