Beruflich Dokumente
Kultur Dokumente
To cite this article: R. J. Taffler (1983) The Assessment of Company Solvency and Performance Using a
Statistical Model, Accounting and Business Research, 13:52, 295-308, DOI: 10.1080/00014788.1983.9729767
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”)
contained in the publications on our platform. However, Taylor & Francis, our agents, and our
licensors make no representations or warranties whatsoever as to the accuracy, completeness, or
suitability for any purpose of the Content. Any opinions and views expressed in this publication
are the opinions and views of the authors, and are not the views of or endorsed by Taylor &
Francis. The accuracy of the Content should not be relied upon and should be independently
verified with primary sources of information. Taylor and Francis shall not be liable for any
losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities
whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial
or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or
distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use
can be found at http://www.tandfonline.com/page/terms-and-conditions
AUTUMN 1983 295
can be traced to the incorrect manner in which they risk index is also described. The next section
are conventionally used. Concentration on individ-
discusses why the derived model appears to work
ual accounting numbers or ratios obscures the true
so well within its frame of reference. This is
value of the published accounts: that they measure
followed in section V by a description of the
a number of distinct and fundamental aspects of a
PAS-score transform. The paper is concluded with
company simultaneously, i.e. they are multivariate
a summary and some thoughts on the value of such
documents. Thus, for example, to evaluate an
models to the financial statement user.
enterprise’s performance in terms of its
profitability alone, ignoring other characteristics
such as its financial risk, working capital and
degree of liquidity, can result only in an incomplete
I The methodological approach
and potentially misleading view. Once a number of The application of multivariate statistical tech-
different facets of the firm are considered together, niques to problems in accounting and business
and particularly when these are viewed relative to finance constitutes a relatively new but very
other firms rather than on an absolute basis, then a significant advance in available methodology. In
company’s published statements become extremely 1968, E. I. Altman published the seminal paper:
valuable documents for many basic decision pur- ‘Financial Ratios, Discriminant Analysis and the
poses and different types of user. Prediction of Corporate Bankruptcy’. In this he
This paper illustrates the benefits that accrue described a discriminant analysis approach to the
when published accounting information is used problem of corporate bankruptcy prediction and
correctly in an holistic and relative manner. It clearly demonstrated the value of considering
describes a UK-based linear discriminant model, financial ratio data within a multivariate decision
which has been in operational use for some years, model framework. Since Altman’s original work, a
requiring a number of different accounting inputs voluminous literature has evolved using a related
derived from conventional financial statements for approach to identify potential insolvent concerns
the determination of a fundamental decision, uiz- in such diverse industries as railroads (Altman,
the evaluation of an enterprise’s financial viability. 1973), banks (Sinkey, 1979), insurance companies
The track record since the model’s development is (Trieschmann and Pinches, 1973), over-the-
discussed, and the degree of true predictive ability counter broker dealers (Altman and Loris, 1976)
is assessed. Two related developments to the and savings and home loan associations (Altman,
approach are also described and evaluated here. 1977) etc.] Such techniques have also been widely
The first, the risk index, appears to provide an used for many years in the USA by banks, govern-
accurate measure of the actual degree of risk of ment regulatory agencies, stockbrokers, auditors
financial distress faced by a company. The second, and the like as the sine qua non of an efficient
a transformation of the z-score, provides a single monitoring process. However, interest in the ap-
readily interpretable measure of relative company proach has been slow to develop outside the USA,
performance, permitting valid cross-sectional and
inter-temporal performance comparison. ‘A fuller list of references is provided in Taffler (1982), which
also provides a critique of the extant US-based studies and
In the first section of this paper, the approach h c f i b e s the statistical methodology and points arising in
adopted in developing the original solvency model detail.
296 ACCOUNTING A N D BUSINESS RESEARCH
although by now there have been a number of resembles more that of previous failures, which is
related studies in other environments. For example a necessary but not sufficient condition for financial
Altman et al. (1974) developed a model using distress, or that of typical financially healthy com-
French data, and related work has. also been panies. This latter is effectively both a necessary
undertaken in The Netherlands by van and sufficient condition for continuity, at least until
Frederikslust (1978), in Australia by Castagna and publication of a later set of accounts. An analysis
Matolcsy (1981) and in Brazil by Altman et al. of the results from application of the model since
(1 979), who also provide other references to work it was developed is provided in section 111.
outside the USA.
Here in the UK an early model completed in the Discriminant Analysis
summer of 1974 and used for many years sub- The discriminant technique is well described in
sequently by a firm of London stockbrokers to the literature and need not detain us here.4 Essen-
provide a service to their clients is described in tially it aims at distinguishing between two or
Tamer (1982). Two more recent UK studies are more distinct populations on the basis of certain
those of Mason and Harris (1979) for the analysis of the characteristics of their members and the
Downloaded by [University of Glasgow] at 09:02 18 December 2014
the dataSTREAM model is only effectively measuring the 4See, for example, Eisenbeis and Avery (1972). Lachenbruch
profitability of an enterprise, with the balance sheet largely (1975) or Green (1978). Detailed reviews of the problems likely
ignored. In fact, an investigation of a large sample of to be experienced in the application of multiple discriminant
companies indicates that the model’s profitability ratio (profit analysis in business applications are provided by Eisenbeis
+
before tax depreciation/current liabilities) is correlated at (1977) and Pinches (1980).
0.986 with its calculated z-score and, as such, can be used as 5These are, specifically, that the two samples are each
a direct proxy. Thus a decision rule that classified a company representative of their underlying populations, are separately
with a ratio value below 0.20 as ‘failing’ and vice versa would distributed multivariate normal and have common dispersion
apparently provide the same classification as the data- matrices. If, however, dispersion matrices are unequal, then
STREAM model 95% of the time. theory suggests a more complex quadratic formulation. How-
‘Because of the very different financial structures of indus- ever, as discussed in Taffler (1982), departures from multi-
trial and distribution companies in the UK. separate models variate normality and small sample sizes, which are typical in
are needed. A related model for the analysis of retail and financial applications, render application of a quadratic model
wholesale concerns has also been developed by this author. inappropriate, despite the presence of unequal dispersion
However, since the methodology used and results obtained are matrices. This, again. is almost inevitable in such applications.
similar to those described here, discussion of this model will As a result the quadratic model will not be considered further
not be pursued. Experiments applying each model to the data here.
set of the other suggest the need for distinct models. They also ‘Although, strictly speaking, the term bankruptcy relates
suggest that it is likely that the pooling of disparate enterprises only to actions against individuals and partnerships under the
in the construction of such a solvency model will reduce its 1914 Bankruptcy Act, it is used here in accordance with
effectiveness in subsequent application. American parlance to refer to corporate failure.
AUTUMN 1983 297
met various criteria to ensure reliable source data.' ,ere made for such things as revaluation and
Half the companies failed in 1975 or 1976. The company growth through acquisition, to improve
average time between publication of last accounts compatibility and consistency.
and date of failure was just under 8 months.
Bankruptcy status for our purposes was indicated The Variables
by the appointment of a receiver, entry into cred- The lack of any real theory relating to the use of
itors' voluntary liquidation, winding up by Order financial ratio analysis constitutes a serious gap in
of the Court or clear action on the part of the the accounting literature.12 By reference to what
government.8 One company was also included be- literature there is in the area and a comparison of
cause a substantial equity injection was provided measures found useful in related or in some sense
to it by the City in 1975 to avoid threatened similar studies, 80 potentially useful ratios were
government intervention of a similar nature.' identified and computed for each of the 92 con-
Virtually all related studies appear to view non- c e r n ~ .Each
' ~ of these ratios underwent logarithmic
failed (i.e. continuing) concerns as being distinct or reciprocal transformations as appropriate to
from bankrupt enterprises. However, it was recog- improve normality and any outlying observations
nised here that a proportion of presently con- were winsorised, i.e. replaced by limiting value^.'^
Downloaded by [University of Glasgow] at 09:02 18 December 2014
tinuing companies possess financial profiles more To understand the data better, to avoid multi-
similar to those of previous bankrupts, and many collinearity, and to aid formulation and inter-
of these will in fact subsequently fail. Thus, sam- pretation of the resulting model, varimax rotated
pling at random from presently continuing con- principal component analysis (PCA) was under-
cerns is likely to result in less distinct samples and t a k e r ~ .In
' ~ this way the intercorrelations within the
consequently a less powerful model.'" To overcome 80 variable set could be handled readily and the
such problems to some degree, a two stage sam- underlying dimensionality of the published ac-
pling process was adopted to obtain solvent firms. counting information could be identified. The re-
Initially 46 companies were drawn from the Jordan sults, provided in Taffler (1981), are very similar to
Dataquest data-base, randomly matched by size those of the much more representative study of
and industry with the failed sample. A subjective Taffler and Sudarsanam (1980) who undertook a
screening process using conventional techniques of related analysis using a slightly augmented ratio set
financial analysis was then applied to the accounts and a large sample of 525 continuing UK quoted
of these companies, and 5 were considered very companies for a later period. Reification of the
unhealthy. These were replaced with randomly components identified by examining the ratios,
selected enterprises, none of which was judged to highly loaded on (correlated with) the same factor
be problematical." Where necessary, for both the and thus describing similar aspects of a company's
failed and solvent firms, adjustments to the data
financial structure, indicated the company ac-
counts examined to be measuring profitability,
financial risk, working capital position, liquidity,
'Only a handful of concerns however had to be rejected.
asset turnover, value added contribution, and cred-
Failing companies engaged in construction and related activ- itors' position. 91.6% of the variance in the data set
ities, numbering about 10, were excluded because of potential was accounted for by these seven factors. Similar
differences in the structure of their accounts compared to those dimensions are also identified by Pinches et al.
of manufacturing concerns. However. subsequent application
of the resulting solvency function to contracting and construc-
tion companies indicated such worries were unfounded. -. -~ ____ . ~ _ _ _ _ ~ _
'As in the cases of British Leyland, Alfred Herbert and ''This has. however, been partially corrected by the recent
Ferranti. articles of Lev and Sunder (1979) and Whittington (1980).
'This company, Fodens, after a further large injection of "Because of arguments in the literature attesting to the
long term funds by its bankers, ultimately failed in 1980, utility of funds Row and trend measures in such situations as
illustrating the maxim that the provision of funds alone can those considered here, both types of variable were computed
only postpone the arrival of the bailiffs. in an earlier investigation (Taffler. 1982). However, no useful
'"These issues are discussed further in Taffler (1982). contribution was made by either class and they were dropped
"Lachenbruch (1974) provides statistical justification for from further consideration at an early stage in the analysis.
this attempt to work with distinct samples. He demonstrates As a result of this experience, no such measures were employed
that, when sample observations are initially incorrectly in this study. The list of ratios used here is given in Taffler
classified to discriminant groups in a non-random fashion, there (1981).
are serious problems in interpreting the classification matrix I4Univariate normality of the constituent variables is a
obtained by applying the model to the original data. Such 'necessary but not sufficient condition for multivariate
difficulties, inter alia, may go some way towards explaining the normality. Lachenbruch et 01. (1973) suggest the linear
size of the type I1 errors-classification of a non-failed as discriminant model suffers least from non-normality when the
failed-in the original sample of Marais (1 979) and particularly constituent variables are bounded.
with his hold-out data. l5See for example Green (1978).
298 A C C O U N T I N G A N D BUSINESS RESEARCH
Table 1
What the constituent ratios are measuring
Dimension of Information Contribution
Ratio ' Represented to Model (%)*
~~
1. PBTiAVCL Profitability 53
2. CA/TL Working Capital 13
3 . CL/TA Financial Risk 18
1 4. No-Credit Interval Liquidity 16
I *Mostelk-Wallace measure. I
(1975) and Johnson (1979) in US published ac- derived, prior probability estimates of the propor-
counts.16 tions of failing and solvent concerns in the UK
quoted manufacturing population have to be in-
corporated into the function. Since, strictly speak-
Downloaded by [University of Glasgow] at 09:02 18 December 2014
Figure 1
Z-score distribution
!
Downloaded by [University of Glasgow] at 09:02 18 December 2014
centroids of the two constituent groups accounted be true misclassifications. In the case of Rolls
for by each variable according to the following Royce, it is doubtful whether it was actually in-
formula: solvent at the date of appointment of the receiver,
4th February 1971, since all creditors and de-
‘J (tf- 5 7 )
benture holders were subsequently repaid in full
pJ=
c 4ff-
4
Table 2
Industrial companies on the EXSTAT data-base with negative z-scores as at
30.11.76: summary of subsequent events over the next 4 years
Event No. of companies ”/, of companies
Financial Distress
Receivership 14 12.2
Going concern qualification’ 8 7.0
or government, bank or other
support, etc.
Acquisition or reverse 12 10.4
takeover as an apparent
alternative to bankruptcy
Major closures and disposals etc. -
6 5.2
-
40 34.8
Still ‘at risk’2 31 27.0
71 61.8
Downloaded by [University of Glasgow] at 09:02 18 December 2014
Recoveries
Still in lowest 25% of companies’ 19 16.5
by z-score
Outside lowest z-score quartile - 25 ~
21.7
(clear recoveries) 115 100.0
Figure 2
Lachenbruch holdout test results
Classified as
Actual Group
Membership’
Faded
that approximate economically to insolvency and 1980 and that were outside the ‘at risk’ region
a further 27% were still at risk.23 produced only one company that had suffered
Downloaded by [University of Glasgow] at 09:02 18 December 2014
During 1980, a very unhappy year for British financial distress according to the definition of
industry, it was possible to identify clearly 30 of the Table 2.26
‘at risk’ major manufacturing and construction On this basis, then, let us estimate that a maxi-
companies on file as suffering bankruptcy or an mum of 5% of the total population of 700 live
equivalent insolvent event.24 An average of 13% major quoted manufacturing and construction
(92) of the population of 700 live companies had companies on file experienced some form of
latest z-scores below - 1.95.2sAn examination of financial distress during 1980. In this case the
the histories of firms that were in the lowest half conditional probability of companies failing in the
of the population by z-score at some time during year, given a potential failure profile
(30/92 = 0.33), and the conditional probability of
companies not failing, given a solvent profile
(effectively 1.O but say 0.99), are both significantly
different at c( = 0.001 to the null positions (5% and
23The 10 cases of acquisition and 2 reverse takeovers of
companies with ‘at risk’ profiles in table 2 were identified from
95%).*’Similar results also obtain for the compara-
press comment and fundamental analysis as likely alternatives tive year of 1977 examined when British industry
to receivership. Three other ‘at risk’ firms taken over could was recovering from the 1974/75 recession.28As
not be so identified. The definition used for ‘major closures
and disposals etc.’ was reduction in employees and/or capacity,
measured by turnover, in each case by half or more. The
‘going concern qualification or government, bank or other
support etc.’ heading covered those concerns which could be
26This company, Sidlaw Industries, with z-score for year
identified readily by press comment or by statements from the
ending 30.9.79 of -0.90 closed down the major part of its
companies themselves, etc. as being in receipt of emergency aid
textile activities to concentrate on its highly profitable North
from the government, their bankers, financial institutions or
Sea oil service side. However, inspection indicated that
other parties as an alternative to receivership. Because of the
impending bankruptcy was not the cause.
avoidance of publicity by banks when involved in rescue
27The appropriate test statistic is
operations, it was likely the number of such situations
identified was actually underestimated with probably a number P --x
z =
of ‘still at risk’ enterprises more correctly classified under this n(1- -x)’
heading. The problem of clearly defining what constitutes
failure may be noted. An interesting comparison can be made l
r
with the results of Deakin (1077) who, for example, also where p is the sample proportion, K the probability of chance
considered default on preference dividend and ordinary divi- classification (0.05), n the sample size, and z (to be dis-
dend cut or omission as failing indicators. tinguished from z-score) a standardised normal deviate. At risk
24Therewere 7 receiverships or equivalent, 9 going concern and ‘solvent’ (z-score > - 1.95) company samples are of size
qualifications, 4 explicit statements of bank support, 3 cases 92 and 608 whence respective z’s are found to be 12.3 and -4.8.
of major rationalisation involving redundancies of over half Pr( IzI > 3.3) < 0.001 (one tail test).
the workforce, 3 instances of emergency cash injections from ”Of the 115 out of 825 (14%) companies on file at the end
external parties and 4 of continued government support. of 1976 with ‘at risk’ profiles, I5 were clearly identified as
(Where there were multiple events the most serious is listed experiencing an insolvent event during 1977. If we use this
here.) figure to suggest that K was 2.3% (estimating the conditional
25Atthe beginning of 1980, there were 69 live companies on probability of an insolvent event for solvent companies apart
file ‘at risk‘ and at the end 114. As such, let us consider the from receivership, which we know occurred in no instance, as
average number of live ‘at risk’ companies on file during the at most 0.5% (4 cases)), then the respective standardised
year to be 92. Similarly in January 1980 there were 715 live normal deviates are 7.7 and -3.4. Because of the amount of
companies on file and 690 of these published accounts during work involved in replicating the analysis for 1978, 1979 and
the year. On this basis we can consider the average number 1981, the existence of equivalent results needs to be taken on
of live companies on file during 1980 to be about 700, trust, which is probably not unreasonable.
302 A C C O U N T I N G A N D BUSINESS RESEARCH
Table 3
Average percentage of companies ‘at risk’ by year
Percentage of
Companies
Note: 80 denotes firms with accounting year ends between 1.5.80 and
30.4.81 and similarly for earlier years.
such, the derived function would appear to exhibit number of sound concerns highlighted as potential
true ex ante predictive ability where the events problem^.^' Thus in evaluating the operational
predicted are the financial distress or otherwise of utility of such a function, the percentage of the
a company within the next year. company population labelled ‘failing’ is just as
Downloaded by [University of Glasgow] at 09:02 18 December 2014
Figure 3
Risk index distributions
% of Companies
50
40
Q
"
30
4
20
Downloaded by [University of Glasgow] at 09:02 18 December 2014
10
trading at that time which had received going It is instructive to consider why the basic ap-
concern qualifications at some stage. The rating proach works as well as it does. The methodology
system, although simple, appears to be of oper- essentially does little else but combine
ational utility. four appropriately selected and weighted financial
ratios, that are computed directly from the much
criticised conventional published accounting state-
IV Why the solvency model works ments, and allow interpretation of the resulting
As we have seen, the z-model not only exhibits index on the basis of whether it is above or below
discriminant ability but also true predictive ability a specified cut-off.
in a statistical sense. This is because the probability In the first place, as has been stressed, the
of an 'at risk' firm subsequently failing within a approach is principally descriptive in nature, with
year was about 6 times as great as for a firm the derived function made up of a small number of
selected at random in both 1980 and 1977, the 2 important and distinct measures of a company's
years studied. In addition, in the more than 5 years financial state. These, when taken together, pro-
that the model has been in operation to July 1982, vide a multi-dimensional profile which can be
only one firm not classified at risk (z = -0.35) had appropriately transformed into a one-dimensional
actually entered into receivership (in February analogue, uiz: a z-score. The derived function, it
1981) out of the 35 such firms on file. Even in this will be noted, is multivariate in nature, exactly as
case, the circumstances surrounding the company a company's published financial statements are
suggest it may not have been strictly insolvent.32 multivariate documents. Thus, the model is doing
little more than reflecting and condensing the
information conveyed by the set of accounts itself.
It is essentially a device for communicating this
'2All operating assets of this small firm were acquired on
a going concern basis by the company's existing major Swiss holistic set of economic information succinctly and
shareholder once the receiver was appointed. in a readily intelligible manner. By taking into
304 ACCOUNTING A N D BUSINESS RESEARCH
account the different facets of the information covered up nowever the accounts are stated (except
simultaneously, rather than one at a time as with by serious and deliberate fraud), given that they
conventional ratio analysis, the Hegelian principle can be correctly read, in a multivariate way.
of the whole being greater than the sum of 'the A theoretical explanation originates with Walter
parts convincingly applies! (1957) who views a firm as a reservoir of liquid
The ratios selected by the step-wise discriminant assets (working capital) supplied by inflows and
approach, apart from each effectively measuring a drained by outflows. Failure can be viewed in
different aspect of the company under in- terms of an exhaustion of the liquid asset reservoir
vestigation, are separately less amenable to win- which acts as a cushion against variations in the
dow dressing and creative accounting by virtue of different flows.
their construction. Additionally, because of the Essentially, other things being equal, the proba-
model's multivariate nature and the swings and bility of failure is higher:
roundabouts of double entry, the results of such (i) the smaller the size of the reservoir,
efforts are largely defeated as their effects tend to (ii) the smaller the funds flow from operations,
be manifested elsewhere in a counter-balancing (iii) the larger the claims on the resources by
Downloaded by [University of Glasgow] at 09:02 18 December 2014
Figure 4
Dunbee-Combex-Marx and toy industry PAS trajectories
PAS -Score
70
60
\ inausrry average
\
50 \
\
\
\
40 Dunbee -Combex- Marx
\
30
Solvent Region
Downloaded by [University of Glasgow] at 09:02 18 December 2014
20
10
. . Sdvency Threshold
At Risk Region
- I I I I I I I
1973 1974 1975 1976 1977 1978 1979 Year
no ratio scale applies. This means that, for exam- overcomes these problems. A company’s PAS (per-
ple, it is not possible to view a z-score of 2 as twice formance analysis score) in a particular year is
as good as a z-score of 1, only that it is better. arrived at by ranking the z-scores of all firms for
Thus, it is not valid to average z-scores to provide that year in ascending sequence and observing the
industry means, nor to work directly with z-score percentile in which the z-score of the concern of
trajectories in an attempt to measure changes in a interest lies-its PAS. Different calculations are
company’s relative performance over time.35 An carried out for each year of data for each company
additional problem is that, as the economy so that the company’s PAS trajectory indicates its
changes, the same z-score will not indicate the relative performance over time.36
same degree of relative strength or weakness. Figure 4 provides a classic trajectory, that of
However a simple transformation that converts Dunbee-Combex-Marx (DCM), the largest UK
the z-score into a ratio measure along a scale &-1OO toy company, together with the PAS arithmetic
average for companies in the toy industry (Stock
Exchange Industry Classification 65) in EXSTAT.
ISThis is the approach adopted by dataSTREAM. The
problems that can arise may be illustrated with an example DCM went into receivership in February 1980
from the dataSTREAM system at the beginning of 1981, that following a strategy of acquisition of large ailing
of a major company in the steel and chemical plant sector: companies overseas. Its PAS declined from 67 (i.e.
Year end: 3/76 3/77 3/78 3/79 3/80 only 33% of companies performing better) on the
Company z: -0.1 -0.2 0.1 -0.5 -0.6
Sector average z: 0.1 0.2 0.3 - 0.2 - 0.6 basis of its 1975 accounts, to 7 (i.e. only 7% of
Market average z: 2.4 2.6 2.3 2.7 2.2 companies doing worse) on the basis of its last
accounts prior to failure. A reflection of the prob-
Apart from the fall in the overall average z-score in 1978,
which would not accord with a priori expectations given the
lems in the toys and games sector is provided by
then state of the economy, it will be noted how difficult it is the industry picture. The ability to interpret the
to interpret both the relative improvement/decline in the PAS directly along a ratio scale (i.e. a PAS of 60
industry performance relative to the overall picture and
particularly the company’s performance relative to its industry.
is twice as good as one of 30) and as a measure of
Did the company’s performance actually improve or decline
between 1979 and 1980 and, if so, by how much? Also it will
be noted that, as the dataSTREAM z-score cut-off is 0, the ‘6Taffler (1981) shows that the PAS explicitly meets all the
average company in steel and chemical plant would be criteria necessary to constitute an efficient and operational
classified as ‘failing’ on the basis in 1979 and 1980. This, again. measure of relative economic performance as required by
does not accord with prior expectations. Shashua and Goldschmidt (1974).
A BR 13/52 t
306 A C C O U N T I N G A N D BUSINESS RESEARCH
relative performance will be noted. The solvency formal model is able to provide efficient and
threshold is where z = - 1.95 and represents the unbiased processing of the complex information
boundary between ‘at risk’ and solvent z-scores. set presented by a set of company accounts. The
The percentage of companies ‘at risk’ varies with benefits that can derive from an explicit ‘man with
the economic climate in line with Table 3. model’ approach are legion.
VI Conclusions References
This paper illustrates the multi-dimensional nature Altman, E. I., ‘Financial Ratios, Discriminant Analysis and the
of conventional financial statements and, in a Prediction of Corporate Bankruptcy’, Journal ofFinance, Vol
particular decision situation, the benefits that de- 23, No. 4 (September 1968). pp. 589-609.
Altman, E. I., ‘Predicting Railroad Bankruptcies in America’,
rive from explicit recognition of this through the Bell Journal of Economics and Management Science, Vol 4,
adoption of appropriate multivariate tools of No. I (Spring 1973), pp. 184211.
analysis. A UK-based discriminant function, Altman, E. I., ‘Predicting Performance in the Savings and Loan
Association Industry’, Journal ofMonetary Economics, Vol3,
termed a z-model, is derived for the evaluation of No. 4 (October 1977), pp. 443-466.
Downloaded by [University of Glasgow] at 09:02 18 December 2014
company solvency on the basis of published ac- Altman, E. I., ‘Commercial Bank Lending: Process, Credit
counting information alone. This exhibits both Scoring, and Costs of Errors in Lending’, Journal ofFinancia1
and Quantitative Analysis, Vol 5, No. 4 (November 1980), pp.
true ex ante predictive ability and clear operational 813-832.
utility on the basis of its track record to date. Altman, E. I. and Loris, B., ‘A Financial Early Warning System
Although the z-model described in this paper for Over-the-counter Broker-Dealers’, Journal of Finance,
Vol 31, No. 4 (September 1976), pp. 1201-1217.
and in particular its derivative risk index can Altman, E. I., Margaine, M., Schlosser, M. and Vernimmen, P.,
predict company failure with considerable accu- ‘Financial and Statistical Analysis for Commercial Loan
racy, the model is more appropriately used in EVdlUatiOn: A French Experience’, Journal of Financial and
Quantitative Analysis, Vol 9, No. 2 (March 1974), pp.
practice to highlight potential financial distress 195-21 I .
while there is still time for recovery actions to be Altman, E. I., Halderman, R. G. and Narayanan, P., ‘Zeta
taken. Use of the PAS transform even permits Analysis. A New Model to Identify Bankruptcy Risk of
Corporations’, Journal qf Banking and Finance, Vol 1, No. I
companies to be picked up on the basis of their (June 1977). pp. 29-54.
downward trajectory, well in advance of an ‘at Altman, E. I., Baidya, T. K. N. and Ribeiro Dias, L. M.,
risk’ profile being registered, thus providing even ‘Assessing Potential Financial Problems for Firms in Brazil’,
Journal of International Business Studies, Vol 10, No. 2 (Fall
earlier warning of an impending insolvency. Ide- 1979), pp. 9-24.
ally, however, such an approach should be self- Altman, E. I. and La Fleur, J. K., ‘Managing a Return to
defeating. The evidence provided of prospective Financial Health’, Journal Sf Business Strategy, (Summer
1981), pp. 31-38.
demise should concentrate the minds both of those Bank of England, ‘Techniques for Assessing Corporate Fi-
in the company on the true state of affairs, and nancial Strength’, Bank of England Quarterly Bulletin, (June
those involved with it who are in a position to 1982), pp. 221-223.
Bazley, J. D., ‘An Examination of the Ability of Alternative
enforce action to ensure short-term survival and Accounting Measurement Models to Predict Failure’, Review
long-term viability, viz. its auditors, bankers, of Business and Economic Research, Vol 12 (Fall 1976), pp.
financial advisors and the shareholding institu- 3246.
Bildersee, J. S., ‘Discussion of the Impact of Price-Level
tion~.~’ Adjustment in the Context of Risk Assessment and the
This study shows that the conventional, much Effect of General Price-Level Adjustments on the Predictive
maligned, published accounting statements can be Ability of Financial Ratios’, Journal of Accounting Research,
Supplement to Vol 16 (1978), pp. 285-292.
valuable sources of information if (i) the holistic Castagna, A. D. and Matolcsy, 2. P., ‘The Prediction of
picture that they present of a company’s financial Corporate Failure: Testing the Australian Experience’, Aus-
state is recognised, (ii) they are read and used in the tralian Journal ofManagement, Vol 6, No. 1 (June 1981), pp.
23-50.
correct way, viz. as multivariate documents, and Chen, K. H. and Shimerda. T. A., ‘An Empirical Analysis of
with the tools appropriate to reflect this, and (iii) Useful Financial Ratios’, Financial Management, Vol 15. No.
they are considered as presenting the company’s I (Spring 1981), pp. 51-60.
dataSTREAM International PLC, The Dafastream Z-score
activities and financial state essentially relative to Service: A Technical Review, published by dataSTREAM
those of other concerns not on an absolute basis. International PLC, Monmouth House, City Road, London
It can be cogently argued that, such are the EC1 (1980).
Dawes, R. M. and Corrigan, B., ‘Linear Models in Decision
cognitive problems confronting the financial state- Making’, Psychological Bulletin, Vol 81, No. 2 (February
ment user in many decision situations, only a 1974), pp. 95-106.
Dedkin, E. B., ‘Business Failure Prediction: An Empirical
j7A good example of the constructive use of the z-score Analysis’, in Financial Crises, Institutions and Markets in a
approach as an aid to corporate financial recovery is provided Fragile Environment, edited by Altman, E. 1. and Sametz, A.
by Altman and La Fleur (1981). W., Wiley-Interscience, 1977.
AUTUMN 1983 307
Dixon, W. J. (ed.), BMD: Biomedical Computer Programs, Norton, C. L. and Smith, R. E., ‘A Comparison of General
University of California Press, 1973. Price Level and Historical Cost Financial Statements in the
Eisenbeis, R. A., ‘Pitfalls in the Application of Discriminant Prediction of Bankruptcy: A Reply’, Accounting Review, Vol
Analysis in Business, Finance and Economics’, Journal of 5 5 , No. 3 (July 1980), pp. 516-521.
Finance, Vol 32, No. 3 (June 1977), pp. 875-900. Patell, J. M., ‘Discussion of the Impact of Price-Level Adjust-
Eisenbeis, R. A. and Avery, R. B., Discriminant Analysis and ment in the Context of Risk Assessment and the Effect of
Classijication Procedures: Theory and Applications, Lexi- General Price-Level Adjustments on the Predictive Ability of
ngton, Mass: D. C. Heath & Co., 1972. Financial Ratios’, Journal of Accounting Research, Supple-
Fadel, H. and Parkinson, J. M., ‘Liquidity Evaluation by ment to Vol 16 (1978), pp. 293-300.
Means of Ratio Analysis’, Accounting and Business Research, Pinches, G. E., ‘Factors Influencing Classification Results from
Vol 8, No. 30 (Spring 1978). pp. 101-107. Multiple Discriminant Analysis’; Journal of Business Re-
Frank, R. E., Massy, W. R. and Morrison, D. G., ‘Bias in search, Vol 8, No. 4 (December 1980), pp. 429456.
Multiple Discriminant Analysis’, Journal of Marketing Re- Pinches, G. E., Eubank, A. A., Mingo, K. A. and Caruthers,
search, Vol 2, No. 3 (August 1965), pp. 250-258. J. K., ‘The Hierarchical Classification of Financial Ratios’,
Green, P. E., Analyzing Multivariate Data, The Dryden Press, Journal of Business Research, Vol 3, No. 4 (October 1975),
1978. pp. 295-310.
Johnson, W. B., ‘The Cross-Sectional Stability of Financial Scott, J., ‘The Probability of Bankruptcy’, Journal qf Banking
Ratio Patterns’, Journal of Financial and Quantitative Anal- and Finance, Vol 5 (1981), pp. 317-344.
ysis, Vol 14, No. 5 (December 1979), pp. 103S-1048. Shashua, L. and Goldschmidt, Y., ‘An Index for Evaluating
Ketz, J. E., ‘The Effect of General Price-Level Adjustments on Financial Performance’, Journal of Finance, Vol 29, No. 3
the Predictive Ability of Financial Ratios’, Journal of Ac- (June 1974), pp. 797-814.
Downloaded by [University of Glasgow] at 09:02 18 December 2014
counting Research, Supplement to Vol 16 (1978), pp. 273-284. Sinkey, J. E. Jr., Problems and Failed Institutions in the
Lachenbruch, P. A,, ‘An Almost Unbiased Method of Obtain- Commercial Banking Industry, JAI Press Inc., 1979.
ing Confidence Intervals for the Probability of Solomon, I. and Beck, P. J., ‘A Comparison of General Price
Misclassification in Discriminant Analysis’, Biometries, Vol Level and Historical Cost Financial Statements in the
23, No. 4 (December 1967), pp. 639-645. Prediction of Bankruptcy: A Comment’, Accounting Review,
Lachenbruch, P. A,, ‘Discriminant Analysis when the Initial VOI 5 5 , NO. 3 (July 1980), pp. 511-515.
Samples are Misclassified, 11: Non-Random Misclassification Tamer, R. J., ‘The Assessment of Financial Viability
Models’, Technometrics, Vol 16, No. 3 (August 1974), pp. and the Measurement of Company Performance’, City
419424. University Business School, Working Paper Series No. 27
Lachenbruch, P. A., Discriminanf Analysis, Hafner Press, New (1981).
York, 1975. Taffler, R. J., ‘Forecasting Company Failure in the UK Using
Lachenbruch, P. A., Sneeringer, C. and Revo, L. T., ‘Robust- Discriminant Analysis and Financial Ratio Data’, Journal
ness of the Linear and Quadratic Discriminant Function to of the Royal Statistical Society, Series A (General), Vol 145,
Certain Types of Non-Normality’, Communications in Statis- Part 3 (1982), pp. 342-358.
tics: l(1) (1973), pp. 39-56. Taffler, R. J. and Sudarsanam, P. S., ‘Auditing the Board: A
Lev, B. and Sunder, S., ‘Methodological Issues in the Use of New Approach to the Measurement of Company Per-
Financial Ratios’, Journal of Accounting and Economics, Vol formance’, Managerial Finance, Vol 5 , No. 2 (1980). pp.
I (1979), pp. 187-210. 127-147.
Marais, D. A. J., ‘A Method of Quantifying Companies’ Trieschmann, J. S. and Pinches, G. E., ‘A Multivariate Model
Relative Financial Strength’, Bank of England, Discussion for Predicting Financially Distressed P-L Insurers’, Journal
Paper No 4 (July 1979). of Risk and Insurance, Vol 40, No. 3 (September 1973), pp.
Mason, R. J. and Harris, F. C., ‘Predicting Company Failure 327-338.
in the Construction Industry’, Proceedings of the Institution Van Frederikslust, R. A. I., Predictability of Corporate Failure:
of Civil Engineers, Vol 66, Part I (May 1979), pp. 301-307. Models for Prediction of Corporate Failure and for Evalu-
Morrison, D. G., ‘On the Interpretation of Discriminant Anal- ation of Debt Capacity, Nijhoff, 1978.
ysis’, Journal of Marketing Research, Vol 6, No. 2 (May Walter, J. E., ‘Determination of Technical Solvency’, Journal
1969), pp. 156-163. of Business, Vol 30, No. 1 (January 1957), pp. 3C43.
Norton, C. L. and Smith, R. E., ‘A Comparison of General Whittington, G., ‘Some Basic Properties of Accounting
Price Level and Historical Cost Financial Statements in the Ratios’, Journal of Business Finance and Accounting, Vol 7,
Prediction of Bankruptcy’, Accounting Review, Vol 54, No. 1 No. 2 (Summer 1980). pp. 219-232.
(January 1979), pp. 72-87.
308 A C C O U N T I N G A N D B U S I N E S S RESEARCH
Table of Contents
August 1983
Vol. 5, No. 2
Editorial Data
Robert W. Holthausen and Richard W. Leftwich
The Economic Consequences of Accounting Choice: Implications
of Costly Contracting and Monitoring
Jerold L. Zimmerman
Taxes and Firm Size
John H. Evans and James M. Patton
An Economic Analysis of Participation in the Municipal Finance
Officers Association Certificate of Conformance Program
William R. Baber
Toward Understanding the Role of Auditing in the Public Sector
Advertisements
Annual subscription Dfl. 65.00 (US $26.00) to private subscribers or Dfl. 145.00 (US $58.00) to libraries and
institutions. Orders should be sent directly to: Publisher, North-Holland Publishing Company, Journal
Division, POB 211, 1000 AE Amsterdam, The Netherlands. Student subscriptions are also available at a
special price of US $14.50 and orders should be sent to: Professors Watts and Zimmerman, Graduate School
of Management, The University of Rochester, Rochester, New York 14627, USA. Manuscripts should be
submitted in triplicate with the appropriate submission fee (various submission fees are listed on the inside
cover of 1982 Volumes) to: Professors Watts and Zimmerman, Graduate School of Management, University
of Rochester, Rochester, New York, 14627, USA.