Discussion of Financial Ratios as Predictors of Failure

COURSE TITLE: SEMINOR IN FINANCE COURSE CODE: MPH 622

Presentation on

Discussion of Financial Ratios as Predictors of Failure

Written by: Prof. John Neter, University of Minnesota, USA

The main article was written by Prof. William H. Beaver

Published in: Empirical Research in Accounting: Selected Studies, 1966, Supplement to Journal of Accounting Research, pp. 71-111

30th March, 2011

Presentation Outline

Discussion produced by John Neter

Discussion presented by Preston Mears

Professor Beaver’s Reply to Professor Neter

Conclusions

Comments

Strength

Methodology : Substantial care to avoid the bias in sample

Much progression analysis : comparison of means, simple discriminant analysis, likelihood ratios and Bayesian inference.

Calibrating samples : use of half of the samples for developing the criterions and the other half is used to test the predictive power of the criterion.

Discussion of Fin..... by John Neter

Interpretation : assessing the predictive power of ratios by comparing it with the simple naïve model.

Implications and area of study : accounting data and preference of research area.

Realization of limitations : On data and analysis that states the possible existence of major pitfalls.

Discussion of Fin..... by John Neter

Weaknesses (methodology; analysis and implications)

Retrospective study : biases possible in the selection of the sample,

matching the populations , measurements

(retrospective, prospective and an experimental study)

Matching factors : asset size and the industry vs age of the business

Matching procedure : how the matching is not clear.

Discussion of Fin..... by John Neter

Problems on the dichotomous test : First, how “naïve” the naïve model should be. Secondly, the importance of the context.

Bayes’ decision rule : new decision rule as;

If , predict failure.

Where,

F = failed firm, NF = Non-failed firm,

LR = Likelihood ratio, = loss if failure predicted, and no failure occurs and

= loss if non-failure predicted, and failure occurs.

Discussion of Fin..... by John Neter

Unrealistic proportion : 1:99 vs. 50:50 i.e. 99 non-failed firms for every failed firm

to make the predictions of failure

Interpretation of likelihood ratio: Problem of determining the loss ratio & likelihood ratio in varying contexts.

Small samples : do not provide enough information about the tails of

the distributions.

Discussion of Fin..... by John Neter

Prof. Neter has raised the issues:

The dangers of the selectivity of choosing the best indicator on the basis of a given set of data

How the matching was done? Was it done according to asset size as of the first year before failure, as of some other year or how?

How to determine the loss ratio?

How effective is the use of multivariate analysis?

Discussion of Fin..... by John Neter

Comments by John Neter

Predicting failure in case of non-failure would be much less serious than predicting non-failure in case of failure.

In some contexts the financial ratios could turn out to be very useful whereas in other contexts the predictability is fairly low.

Thus, any general assessment of a ratio as an excellent (or poor) predictor may be misleading.

Finally, the study might have been improved if it had not used a sample of non-failed firms of the same size as that for failed firms.

Discussion of Fin..... by John Neter

Discussion of Financial Ratios as Predictors of Failure

Written by: Prof. Preston K. Mears

Discussion of Fin... by Preston Mears

The theme of the story of two MBA graduates and Johnny

“I buy for one dollar and I sell for two dollars. I must make 1 per cent.”

The replication of the idea with Beaver’s work is the worth of his work is to explore the predictive ability of accounting data or financial ratios. Thus, make your 1 per cent and you are all right.

(Beaver did well in his job)

Discussion of Fin... by Preston Mears

Conclusions in favor of Beaver’s work

Overwhelming support the definition of “failure” and its “operational” criteria

Underline the Beaver’s primarily focus as ‘the underlying predictive ability of the financial statements themselves’ rather ratios.

Utility of the ratios to expose the area where further information is needed and help to save from failure.

Discussion of Fin... by Preston Mears

Use the trend analysis to identify the trend effect as early as five years before is a most significant observation of the study

Conclusion against the Beaver’s study

The criterion for identifying the business failure can also be the liabilities of the firm as used in the earlier works.

Discussion of Fin... by Preston Mears

In summaryi) Prediction of failure or success is a daily business

decision.ii) Financial data are useful for them. iii) Ratios of themselves will not predict failure

absolutely. iv) The most useful way of using the financial ratios

are to finding the “poor” rati0 – then find out why it is poor and be guided by the answer to curve them.

v) Ratios analysis are the routine professional scope of work.

vi) Further study along the lines of Professor Beaver are welcome.

Discussion of Fin... by Preston Mears

Professor Beaver’s Reply to Professor Neter

Beaver’s Reply...

Broad questions

How does the context of analysis changes when the probability of failure is different from 0.50?

When the costs of misclassification are asymmetrical?

Beaver’s Reply...

Beaver accepted that Neter’s idea of 99:1 rather 50:50 has correctly pointed out.

i) The optimal cutoff point will shift when the probability of failure changes

ii) An analysis of total percentage error can be misleading and limited in its insights.

The type I and II errors will also change when the cutoff point is changed.

Beaver’s Reply...

Loss when a failed firm is misclassified is greater than the loss when a non-failed firm is misclassified.

Using the likelihood ratio, the decision criterion for minimization of expected loss is;

Predict failure if,

prior-odds ratio x likelihood ratio > 1/loss ratio

Beaver’s Reply...

Does the impact of the ratio analysis (likelihood ratio) leads to a change in the behavior of the decision makers?

Answer depends upon the loss ratio and the likelihood ratio. The loss ratio is not easily quantified but will be considered from the point of view of a bank’s lending decision.

The analysis tentatively suggest that ratios are useful predictors, in the sense that they change the decision behavior.

Beaver’s Reply...

Critical likelihood ratio of the multidimensional decision is lower than that implied by a dichotomous decision model.

Conclusions

The primary interest of the study is; how useful ratios are? Rather, are ratios useful?

The answer depends upon several sub-questions and affected by the multiple factors. Thus, the clear demarcation of the factors and their impact analysis is desirable prior to conclude the predictive ability of the accounting data.

Beaver’s Reply...

Comments

Beaver’s pioneering work has raised the different frontier of knowledge and scope of the new studies. It also creates the platform for researchers to generate the new contributions in this field and helps to learn how the pioneering has been done.

The discussion of ratios as predictors of failure and Beaver’s reply clarify the many confusions pertaining in the earlier study.

Beaver’s Reply...

Thank you.