Bankruptcy

Evaluation of Bankruptcy Prediction Models

ABSTRACT Since bankruptcy prediction became a popular research topic in the mid-1960s the model used for evaluating the research has remained largely unchanged. When evaluating the models, the assumption is that random chance will correctly classify 50% of the companies as bankrupt or not, and any model that exceeds this is doing better than chance. In the real world, of the 10,000 companies that trade on exchanges, only 600 will go bankrupt - this is a 6% failure rate, so any model should do better than 94% accuracy, not 50%. Some companies might be eligible for bankruptcy, but choose not to file. They might instead threaten to file, and negotiate concessions from creditors. If this company was classified as bankrupt, would that be a correct or incorrect prediction? The classic evaluation model would classify that as a miss, but is it really? This paper addresses the different types of models used for bankruptcy prediction with special reference to some selected Indian and Global Companies and also highlights the shortcomings of bankruptcy prediction evaluation models and suggests that bankruptcy is better represented as a continuum, rather than a dichotomous situation.

Page 1 of 16


INTRODUCTION Bankruptcy prediction is the art of predicting bankruptcy and various measures of financial distress of public firms. It is a vast area of finance and accounting research. The importance of the area is due in part to the relevance for creditors and investors in evaluating the likelihood that a firm may go bankrupt. The quantity of research is also a function of the availability of data: for public firms which went bankrupt or did not, numerous accounting ratios that might indicate danger can be calculated, and numerous other potential explanatory variables are also available. Consequently, the area is well-suited for testing of increasingly sophisticated, data-intensive forecasting approaches. Using financial statement information to predict bankruptcy has been a popular academic topic since Altman’s (1968) paper. Beaver published somewhat earlier (1966, 1968) but did not focus on bankruptcy; rather financial distress and insolvency (close cousins). When bankruptcy prediction studies are subject to replication, the results are not nearly as good as originally achieved The models developed do not stand the test of “robustness.”. One reason might be that the statistical models tend to use matched pairs of bankrupt and nonbankrupt companies. The variable data is then used to determine cutoffs (thresholds) to differentiate bankrupt and non-bankrupt companies. The data used to derive the thresholds is year specific, and tends to correctly differentiate companies in the hold-out sample (which are from the same time period). These cutoffs work less well when applied in different time periods (including use in replication studies). Another reason for the deteriorating accuracy of bankruptcy prediction studies might be the variables selected to be incorporated into the model. Starting with Beaver, most research has used variables that were selected based on prevalence and popularity in the literature. Stemming mostly from being related to or thought to be indicative of liquidity, and liquidity being synonymous with solvency, these variables have been used with minor additions, deletions and adjustments. Any practical application involves taking a company and making a prediction of whether they will go bankrupt. In the 1960s (and perhaps 70s as well) bankruptcy was a last result, so the dichotomous classification test would have been more appropriate in that time frame, as companies would use a bankruptcy filing only as a last resort. Over time bankruptcy has taken on strategic importance and lost much of its stigma, so a new evaluation framework must be developed to encompass these changes, as well as to make it useful in a practical setting.

Page 2 of 16


BANKRUPTCY :DEFINITION •

“A Federal system of statutes and courts which permits persons and businesses which are insolvent (debtors) or (in some cases) face potential insolvency, to place his/her/its financial affairs under the control of the bankruptcy court. “ -West's Encyclopedia of American Law

•

“Bankruptcy is a legally declared inability or impairment of ability of an individual or organization to pay its creditors. “ -Wikipedia “Bankruptcy, or even the possibility of bankruptcy, can cause significant trauma for a firm’s managers, investors, suppliers, customers, and community. Thus it would be beneficial to be able to predict the likelihood of bankruptcy so that steps could be taken to avoid it or at least reduce its impact.” -Brigham, Eugene F., and Michael C. Ehrhardt, Financial Management ( Thomson/South-Western, 2005)

•

TERMS RELATING TO BANKRUPTCY • •

Voluntary bankruptcy: Bankruptcy is initiated by the debtor that is filed by the insolvent individual or organization. Involuntary bankruptcy: An involuntary bankruptcy petition may not be filed against an individual consumer debtor who is not engaged in business.

DEVELOPMENT OF BANKRUPTCY LAW •

• •

The concept and origin of bankruptcy law as it is now known in the United States originated in England. The first English bankruptcy law is generally agreed to have been enacted in 1542. (34 and 35, Henry VIII, c.4 (1542) England.) Voluntary bankruptcy in the United States was established as an institution by the Acts of 1841 (Act of Aug. 19, 1841, section 1, 5 Stat. 440) and 1867 (Act of Mar. 2, 1867, section 11, 14 Stat. 521). The Bankruptcy Reform Act of 1978, commonly referred to as the Bankruptcy Code, constituted a major overhaul of the bankruptcy system. Bankruptcy Reform Act of 1978 contains eight odd-numbered (1 through 15) chapters and chapter 12. 1. Chapter 7 contains procedures to be followed when liquidating a failed firm. 2. Chapter 11 outlines the procedures for reorganizing a firm

Page 3 of 16


BANKRUPTCY PREDICTION MODEL To evaluate any model, the first and most important question is “what is a correct prediction?” and the complimentary “what is an incorrect prediction?” Using the dichotomous classification test it would seem to be simple to evaluate whether a prediction was correct or not, based on the prediction and the outcome (see Figure 1). Figure 1 Prediction Matrix Prediction Outcome Bankrupt Filed Correct Did Not File Incorrect

Not Bankrupt Incorrect Correct

Despite, the convenience with which that evaluation model can be applied to research, it is not efficacious in determining which models are effective at predicting bankruptcy. The population who would use a bankruptcy prediction model for anything other academic research would encompass, largely, investors and creditors. Because bankruptcy has costs, they want to gauge whether their equity or debt investments would be at risk. At the time they are making their assessments, they have no idea of the population of bankrupt companies. All companies that they are considering are a candidate for bankruptcy. In academic research, the individual making the prediction would know a priori that half the companies are bankrupt and half are not. No such knowledge is available in a practical setting.

TYPES OF MODELS Statistical Models, primarily Multiple Discriminant Analysis or Conditional Logit Regression Analysis: Linear discriminant analysis (LDA) is a multivariate statistical technique that leads to the development of a linear discriminant function maximizing the ratio of among-group to within-group variability, assuming that the variables follow a multivariate normal distribution and that the dispersion matrices of the groups are equal. Clearly, both of the assumptions pose a significant problem for the application of LDA in real- world situations, since they are difficult to meet (Doumpos and Zopounidis, 1999). • • •

Artificial Neural Network based Models Survival methods are now applied. Option valuation approaches involving stock price variability have been developed.

For this project we have only analysed the statistical models .

Page 4 of 16


ALTMAN Z-SCORE (1968) The Altman Z-score, perhaps also known as the "Z-score formula for predicting bankruptcy", was published in 1968 by Edward I. Altman, who was, at the time, an Assistant Professor of Finance at New York University. The formula may be used to predict the probability that a firm will go into bankruptcy within two years. Z-scores are used to predict corporate defaults and an easy-to-calculate control measure for the financial distress status of companies in academic studies. The Z-score uses multiple corporate income and balance sheet values to measure the financial health of a company. Estimation of the Formula: The Z-score is a linear combination of four or five common business ratios, weighted by coefficients. The coefficients were estimated by identifying a set of firms which had declared bankruptcy and then collecting a matched sample of firms which had survived, with matching by industry and approximate size (assets). Altman applied the statistical method of discriminant analysis to a dataset of publicly held manufacturers. The estimation was originally based on data from publicly held manufacturers, but has since been re-estimated based on other datasets for private manufacturing, non-manufacturing and service companies. The original data sample consisted of 66 firms, half of which had filed for bankruptcy under Chapter 7. All businesses in the database were manufacturers, and small firms with assets of <$1 million were eliminated. The original Z-score formula, estimated in that sample, was as follows:

Z = 0.012T1 + 0.014T2 + 0.033T3 + 0.006T4 + 0.999T5. Where, T1 = Working Capital / Total Assets. Measures liquid assets in relation to the size of the company. T2 = Retained Earnings / Total Assets. Measures profitability that reflects the company's age and earning power. T3 = Earnings before Interest and Taxes / Total Assets. Measures operating efficiency apart from tax and leveraging factors. It recognizes operating earnings as being important to long-term viability. T4 = Market Value of Equity / Book Value of Total Liabilities. Adds market dimension that can show up security price fluctuation as a possible red flag. T5 = Sales/ Total Assets. Standard measure for sales turnover (varies greatly from industry to industry). Page 5 of 16


ORIGINAL Z-SCORE FOR PUBLIC FIRMS Z Score Bankruptcy Model: Z = 1.2T1 + 1.4T2 + 3.3T3 + 0.6T4 + 0.999T5 Zones of Discrimination: Z > 2.99 -“Safe” Zone 1.8 < Z < 2.99 -“Grey” Zone Z < 1.80 -“Distress” Zone

ORIGINAL Z-SCORE FOR PRIVATE FIRMS Z' Score Bankruptcy Model: Z' = 0.717T1 + 0.847T2 + 3.107T3 + 0.420T4 + 0.998T5 Zones of Discrimination: Z' > 2.9 -“Safe” Zone 1.23 < Z' < 2. 9 -“Grey” Zone Z' < 1.23 -“Distress” Zone

ORIGINAL Z-SCORE FOR NON-MANUFACTURER INDUSTRIALS & EMERGING MARKET CREDITS Z-Score Bankruptcy Model: Z = 6.56T1 + 3.26T2 + 6.72T3 + 1.05T4 Zones of Discrimination: Z > 2.6 -“Safe” Zone 1.1 < Z < 2. 6 -“Grey” Zone Z < 1.1 -“Distress” Zone

Page 6 of 16


SPRINGATE MODEL (CANADIAN - 1978) This model was developed in 1978 at S.F.U. by Gordon L.V. Springate, following procedures developed by Altman in the U.S. Springate used step-wise multiple discriminate analysis to select four out of 19 popular financial ratios that best distinguished between sound business and those that actually failed. The Springate model takes the following form -:

Z = 1.03A+3.07B + 0.66C + 0.4D Zones of Discrimination: Z < 0.862; then the firm is classified as "failed" Where, A = Working Capital/Total Assets B = Net Profit before Interest and Taxes/Total Assets C = Net Profit before Taxes/Current Liabilities D = Sales/Total Assets

Page 7 of 16


FULMER MODEL (U.S. - 1984) Fulmer (1984) used step-wise multiple discriminate analysis to evaluate 40 financial ratios applied to a sample of 60 companies -30 failed and 30 successful. The average asset size of these firms was $455,000. The fulmer model takes the following form -:

H = 5.528 (V1) + 0.212 (V2) + 0.073 (V3) + 1.270 (V4) - 0.120 (V5) + 2.335 (V6) + 0.575 (V7) + 1.083 (V8) + 0.894 (V9) - 6.075 Zones of Discrimination: H < 0; then the firm is classified as "failed" Where, V1 = Retained Earning/Total Assets V2 = Sales/Total Assets V3 = EBT/Equity V4 = Cash Flow/Total Debt V5 = Debt/Total Assets V6 = Current Liabilities/Total Assets V7 = Log Tangible Total Assets V8 = Working Capital/Total Debt

Page 8 of 16


V9 = Log EBIT/Interest

SELECTED GLOBAL COMPANIES WHO FILED FOR BANKRUPTCY RECENTLY XEROX CORPORATION CASE: (ON THE VERGE OF BANKRUPTCY IN 2002) YEAR

2002

2003

2004

ALTMAN MODEL

0.6401

0.8815

1.3104

SPRINGATE MODEL

0.218

0.3216

0.5285

FULMER MODEL

0.9542

1.2934

1.7773

UAL CORPORATION CASE: (FILED FOR BANKRUPTCY IN DECEMBER, 2002) YEAR

2006

2007

2008

ALTMAN MODEL

1.455

0.9122

-0.4111

SPRINGATE MODEL

1.3074

0.3807

-1.08

FULMER MODEL

0.8137

0.7588

NA

GLOBAL CROSSING LIMITED CASE: (FILED FOR BANKRUPTCY IN JANUARY, 2002) YEAR

2005

2006

2007

ALTMAN MODEL

1.5135

1.002

1.1265

Page 9 of 16


SPRINGATE MODEL

0.9025

0.6173

0.5944

FULMER MODEL

0.1244

0.2267

-0.9157

DELTA AIRLINES CASE: (FILED FOR BANKRUPTCY IN SEPTEMBER, 2005) YEAR

2007

2008

2009

ALTMAN MODEL

0.7986

-0.4276

0.1951

SPRINGATE MODEL

0.5254

-1.022

-0.0405

FULMER MODEL

0.7657

NA

NA

GENERAL MOTORS CASE: (FILED FOR BANKRUPTCY IN JUNE 2009) YEAR

2006

2007

2008

ALTMAN MODEL

0.8606

0.6351

-1.0133

SPRINGATE MODEL

0.2068

0.2063

-1.0032

FULMER MODEL

NA

NA

NA

Page 10 of 16


SELECTED HEALTHY INDIAN COMPANIES ONGC: YEAR

2007

2008

2009

ALTMAN MODEL

4.4821

4.5190

3.3385

SPRINGATE MODEL

2.3734

2.0975

1.7334

FULMER MODEL

5.7061

5.1063

4.4307

YEAR

2007

2008

2009

ALTMAN MODEL

9.4918

10.9617

9.2130

SPRINGATE MODEL

2.0073

1.9753

1.9404

FULMER MODEL

4.579

4.2809

4.3509

ITC:

CIPLA PHARMACEUTICALS: YEAR

2007

2008

2009

ALTMAN MODEL

11.3920

7.0094

5.905

SPRINGATE MODEL

1.8867

1.6048

1.5836

FULMER MODEL

5.000

4.2467

3.8244

Page 11 of 16


SELECTED SICK INDIAN COMPANIES BOMBAY DYEING: YEAR

2007

2008

2009

ALTMAN MODEL

1.6408

1.5782

1.1821

SPRINGATE MODEL

0.5425

0.4412

0.3745

FULMER MODEL

0.93

0.2878

-0.6651

EASTERN COALFIELDS: YEAR

2007

2008

2009

Altman Model

-0.892

-3.1962

-5.3

SPRINGATE MODEL

-0.6167

-2.6157

-4.4667

FULMER MODEL

-3.7412

NA

NA

YEAR

2007

2008

2009

ALTMAN MODEL

1.2701

0.6069

0.4373

SPRINGATE MODEL

0.4841

0.1211

0.0305

FULMER MODEL

1.1277

-0.3987

NA

JET AIRWAYS:

Page 12 of 16


AIR INDIA: YEAR

2006

2007

2008

Altman Model

1.4731

0.941

0.1573

SPRINGATE MODEL

0.6897

0.335

-0.4673

FULMER MODEL

1.4363

NA

NA

KINGFISHER AIRLINES CASE: YEAR

2007

2008

2009

ALTMAN MODEL

0.8496

0.42

0.2752

SPRINGATE MODEL

-0.787

-0.6378

-0.0925

FULMER MODEL

NA

NA

NA

YEAR

2007

2008

2009

ALTMAN MODEL

2.6022

2.841

1.2333

SPRINGATE MODEL

1.9012

1.8694

1.0865

FULMER MODEL

5.0917

4.4537

2.76

TATA STEEL CASE:

INDEX: SAFE ZONE GREY ZONE DISTRESS ZONE Page 13 of 16


FINDINGS • • •

•

•

The Altman Model has furnished the most consistent prediction for the established bankrupt US companies The Springate Model has also reflected the predictions of the Altman model in almost all of the firms studies The Fulmer Model could not be applied to some companies, and for those where it could be applied, the results have been mixed. It could be said that this model has not been able to predict bankruptcy consistently in this study For the financially sound Indian companies studied, all the 3 models have furnished similar estimates of solvency, except for a few cases, where the Altman and Springate models have reflected the effect of the Global Financial Crisis on the firms. For all the loss making Indian companies studied, both the Altman and Springate models have estimated bankruptcy in the near future. The managers of these companies need to take notice of this fact and act accordingly.

But What If the Books Are Cooked? An interesting feature of the Z-score model is its ability to withstand certain types of accounting irregularities. Consider a recent high-profile bankruptcy of WorldCom, where management improperly recorded billions of dollars as capital expenditures instead of operating expenses. Such a treatment would have a twofold impact on financial statements: (1) overstating earnings, and (2) overstating assets. Overstated earnings would increase the T3 ratio in the Z-score model, while overstated assets would actually decrease three ratios, T1, T2, and T5 (all three are calculated with total assets in the denominator). The overall impact of these accounting improprieties on the company’s Z-score, therefore, is likely to be downward. To examine the validity of this reasoning in a limited case-study setting, we computed Z-scores for WorldCom for fiscal years ending December 31, 1999, 2000, and 2001 based on its annual 10-K reports filed with the U.S. Securities and Exchanges Commission .We found that the company indeed experienced a rapid deterioration in its Z-score. Obviously, this limited test has to be taken with a grain of salt (especially given that WorldCom is not a manufacturing company), but it does show how this particular type of accounting impropriety can affect the Z-score.

Page 14 of 16


CONCLUSIONS & RECOMMENDATIONS The current dichotomous classification test, the standard used in evaluating bankruptcy prediction research, fails to correctly classify situations that would be very important to users outside of an academic environment. A company might be eligible to file for bankruptcy, but chooses not to, using the potential as leverage to obtain concessions from creditors. A prediction model might have predicted the company to be bankrupt and since there was no filing, this would be counted as an error. However, the company used the potential of filing as a way to negotiate every benefit they might have gotten from a filing, and therefore the model should be considered to have correctly classified this. Bankruptcy is more of a continuum than a dichotomous variable. A new way of evaluating predictions has to be developed and applied to models to make them useful in a practical setting. Creditors, lenders and potential investors do not sit down with two lists of companies; one list bankrupt and one not, jumble up the companies and then select the non-bankrupt. Real-world use starts with a list of companies without a priori knowledge of how many are/will be bankrupt and then the model is asked to classify each. This is how users will apply the model, and this is how the evaluation of the model should proceed. The basic, underlying, inescapable problem with the current dichotomous evaluation model is that bankruptcy can no longer be thought of as a dichotomous variable. The purpose of the Chapter 11 bankruptcy petition is to give a company time to reorganize and return to profitability. This same end could potentially be achieved by the company threatening creditors with the possibility of having to file a bankruptcy petition. If the company can get the creditors to accept concessions, then the company might attain all of the benefits they would have received from a bankruptcy petition without having to undertake any of the costs or burdens. Given this, the model has to take into account what previously would have been considered errors: • the company meets the criteria for a filing but chooses not to • the company is not insolvent but chooses to file Incorporating this into a model is not easy. The beauty of the dichotomous classification test was that it was simple and direct to apply. The failing was that it has no value in a contemporary, practical setting. Developing a useful evaluation model will involve creating a continuum of a bankrupt company, with the difficulty them becoming assessing how well the model did. This will be a topic for future research.

Page 15 of 16


REFERENCES Books •

Brigham, Eugene F., and Michael C. Ehrhardt, Financial Management ( 11th ed, Thomson/South-Western, 2005) Articles • Altman, Edward I., "Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy”, Journal of Finance, (September 1968): pp. 589-609 • Altman, Edward I., “Predicting Financial Distress of Companies: Revisiting the Z score and Zeta Models”, Journal of Finance, (July 2000): pp. 15-22 • Beaver, William H. (1966) “Financial Ratios as Predictors of Failure,” Empirical Research in Accounting: Selected Studies, 1966, pp 71 – 111. . Annual Reports of the companies taken into consideration. Websites • http://en.wikipedia.org/wiki • http://in.finance.yahoo.com • http://moneycentral.msn.com • http://www.bankruptcyaction.com • http://www.google.com/finance • http://www.moneycontrol.com

Page 16 of 16

Bankruptcy

Recommend Documents