«Corporate Failure Prediction: Some Empirical Evidence From Listed Firms in Ghana Kingsley Opoku Appiah Kwame Nkrumah University of Science and ...»
China-USA Business Review, ISSN 1537-1514
January 2011, Vol. 10, No. 1, 32-41
Corporate Failure Prediction: Some Empirical Evidence From
Listed Firms in Ghana
Kingsley Opoku Appiah
Kwame Nkrumah University of Science and Technology, Kumasi, Ghana
This paper examines the phenomenon of bankruptcy prediction from a developing economy perspective using the
Altman Z-score model. Drawing on empirical data from a sample of 15 non-failed and failed companies listed in
Ghana Stock Exchange, the author test Altman (1968) model via a cross section of different firms with dataset within 2004 to 2005. Altman’s Z-score is applicable in predicting bankruptcy in Ghana depending on the nature and size of the company in question. Since the literature on corporate failure in African contexts is rather parsimonious, this study makes an important contribution to the global discourse on corporate failure prediction in an increasingly globalised world.
Keywords: corporate failure, bankruptcy, Z-score, listed firms, Ghana Introduction Previous researchers (Gilson, 1989; Gilson, 1990; Datta & Iskandar-Datta, 1995) argued that financial distress can be associated or attributed to corporate governance. It is also argued by Daily and Dalton (1994) that there is a relationship between corporate failure and corporate governance characteristics. The global financial crisis has left many organizations in limbo about their going concern. This is evident from some of the world’s top organizations filing for bankruptcy. Notable among these global giants that have declared bankruptcy include, General Motors (GM), Chrysler, American International Group Inc., Delta Airline Limited (Mclntyre & Ogg, 2008). This has renewed the debate among various stakeholders in the quest to identify firms with bankruptcy alerts (Gerald, 2002). Interestingly, corporate failure literature and models are all based on developed countries (Beaver, 1966; Altman, 1968; Altman, 1973; Altman, 1984; Edmister, 1972; Trieschmann & Pinches, 1973; Sinkey, 1975; Deakin, 1977; Pinches & Trieschmann, 1977; Casey & Bartczak, 1985; Aly, Barlow, & Jones, 1992; Baldwin & Glezen, 1992; Dimitras, 1996; Altman & Narayanan, 1997).
Ghana has also had “her share” of corporate failure as evidence in Table 1. The collapse of Tano Agya Rural Bank, Tana Rural Bank Ltd, Meridian BIAO Bank, Bank for Credit and Commerce International were due to the collapse of their parent banks. Most recently, the Gateway Broadcasting Services, Ghana Co-operative Bank, Bank for Housing and Construction, National Savings and Credit Bank and many other corporate failures, in Ghana to some extent, indicate the urgent need for a reliable model which accurately predicts corporate health in Ghana.
Kingsley Opoku Appiah, lecturer, Accounting and Finance Department, School of Business, Kwame Nkrumah University of Science and Technology.
From Table 1, it is obvious that many parties including employees, trade vendors, trade receivables, shareholders, bankers and the government of Ghana lose substantially whenever a company fails (Warner, 1977;
Zavgren, 1983; Jones, 1987; Boritz, 1991; Laitinen & Kankaanpaa, 1999; Charalambous, Charitou, & Kaourou, 2000). Notwithstanding the cost of corporate dismay in a developing economics as at the time of writing, it is alarming to note that, no attempts have been made by any researcher in finance to develop a corporate health model or perhaps, test the applicability of existing corporate failure models using empirical data from Ghana.
The importance of the present study is to assess the most popular bankruptcy prediction models, the Z-score model developed by Edward Altman in 1968 (Morris, 1997b). Particularly, the study attempts to add to present creative writing of predicting corporate failure by answering the question, to what extent is corporate failure predictable in Ghana using Altman (1968) Z-score model?
An Overview of the Bankruptcy Prediction Literature Review Bankruptcy prediction is becoming increasingly important in corporate governance. Global economies have become cautious of the risk involved in corporate liability, especially after the collapse of giant organizations like WorldCom and Enron. There have been several reviews of this literature on predicting corporate bankruptcy—but those are now either out-of-date (Scott, 1981; Zavgren, 1983; Altman, 1984; Jones,
1987) or too narrowly focused. All the above mentioned researchers focused exclusively on statistical models while others like Jones (1987) and Dimitras, Zanakis and Zopounidis (1996) do not give full coverage of theoretical models. Zhang, Hu, Patuwo and Indro (1999) restrict their review to empirical applications of neutral network models while Crouhy, Galai and Mark (2000) cover only the important theoretic current credit risk models.
None of the studies discussed above has provided a complete comparison of the many different approaches towards bankruptcy prediction. The above studies have also failed to provide solutions to the problem of model choice in empirical applications. There have been significant contributions on the theoretical developments topic since Morris (1998). Patrick (1932), Durand (1941) and Beaver (1966) applied accounting ratios, scoring model and univariate discriminant analysis to predict firm’s financial health respectively. The afore-mentioned methods create inconsistent signals since different variables could give conflicting forecast (Altman, 1968; Zavgren, 1983). Therefore, alternatives that guarantee consistency are imperative. This has set the scene for the studies using Multiple Discriminant Analysis, which is discussed in the next paragraph onwards.
Multiple Discriminant Analysis (hereafter refers to as MDA) derives a linear discriminant function which separates the variable space into two disjoint partitions. Altman (1968) was the first to use MDA methodology,
34 CORPORATE FAILURE PREDICTION: SOME EMPIRICAL EVIDENCEto predict failure. The initial sample composed of 66 publicly held manufacturing corporations in the USA between the periods 1946-1965. Altman classifies the companies into two mutually exclusive groups-bankrupt and non-bankrupt. Failed and non-failed companies were matched by size and industry and selected on stratified random basis. The discriminant function was developed using 33 firms in each group as estimation sample. He related 22 accounting and non-accounting ratios which experiment resulted in a single cut-off point for 5 financial ratios that were statistically momentous in predicting liquidation from zero to two years before
the actual event occurred. The original Altman model took the following form:
Z=0.012 X1 + 0.014 X2 + 0.033 X3 + 0.006 X4 + 0.999 X5
X1=Working capital/total assets;
X2=Retained earnings/ total assets;
X3=Earnings before interest and taxes/total assets;
X4=Market value of equity/book value of total liabilities;
Using the above Z-score Altman used a cut-off Z-score of 2.675 resulting in 6% and 3% type I and type II error respectively for sample firms a year prior to failure. An attempt to predict bankruptcy two years in advance, increase the type I and type II errors to 28% and 6% respectively. Finally, Altman and LaFleur (1981)
used a more suitable order of the original Z-score, given as:
Z=1.2 X1 + 1.4 X2 + 3.3 X3 + 0.6 X4 + 1.0 X5 In order to test the models rigorously for both bankrupt and non-bankrupt companies, a new sample was introduced; 86 companies went bankrupt in 1969-1975, 110 in 1976-1995 and 120 in 1997-1999 resulting in Altman reducing the cut-off score to 1.81. He identified the range between 1.8 and 2.7 as “middle ground” in which the company’s failure was uncertain. Altman’s style was related to that of Beaver (1968), but for the concurrent use of multiple financial ratios in a given year, to predict an imminent collapse.
Altman, Haldeman and Narayanan (1977) used US data and covered the period 1969-1975 with a sample of fifty-three failed and fifty-eight non-failed companies. They derived a Zeta value based on seven financial ratios, where six of them were different from Altman (1968) preceding study. Like Altman (1968), to test the models rigorously for both failed and non-failed companies, a holdout sample was introduced. The study achieved an overall mis-classification of 7% for type I error and 3% higher (i.e., 10%) type II error a year prior to failure. The predictive power of the model reduced significantly five years prior to failure to 70% and 82% for failed and non-failed companies respectively. This surveillance highlights a distress that the variables are irregular across various studies. Furthermore, these two studies were exceedingly precise in the short-run, but the precision shrinks vividly when the facts were for time periods of more than two years prior to ruin.
Joy and Tollefson (1975) among other researchers have criticized Altman’s work on the basis of lack of evidence of ex-ante predictive ability of ratios. We find some merit in Joy and Tollefson’s (1975) criticism of Altman’s work on the decisive factor used by Altman to choose variables for exclusion in the model and lack of alternative comparisons with naïve alternative models. Later, Moyer (1977) tested the efficacy of Altman’s model on 27 failed and 27 non-failed firms between 1965-1975. These firms were paired on the basis of industry and assets size ranging from $15million to $1billion. Interestingly, the result indicated that the fore-casting accuracy on a genuinely post-dated sample of firm collapse was 75% a year before failure, which conflicts with the 96%, proposed by Altman (1968). In re-estimating the Altman model parameters, Moyer used
CORPORATE FAILURE PREDICTION: SOME EMPIRICAL EVIDENCEnew data set and adopted the stepwise MDA approach. He claimed that better explanatory power could be obtained if market value of equity/book value of debt and sales/total assets variables are eliminated.
In addition to the above criticism, Altman’s (1968) model is out of date since its predictive accuracy failed with the passage of time and limited in its coverage. The model is not applicable to some industries such as the retail, banks and railroads. However, these limitations are not worth mentioning since Z-score is the most widely-used model. Various equations now exist but they all follow the concepts of the original one (devised by professor Altman in America in 1968) although they are all different (Argenti, 1983; Kip, 2002).
The most important of the previous studies, which ignored the various limitations of Professor Altman’s model and modified it, are Deakin (1972), Taffler (1977, 1983), Altman et al. (1977), Gentry, Newbold and Whitford (1987), Baldwin and Glezen (1992), Keasy and Watson (1986), Aly, Barlow and Jones (1992). More so, writers in financial management textbooks noted that Altman’s model is not just providing a basis for predicting corporate failure, but also a tool to assist in credit evaluation, internal control guideline and a guide to portfolio selection (Van Horne, 1974; Bolten, 1976; Reed, 1976). Finally, Moyer (1977, p.16) stated that “the result achieved from other dynamic approaches have not been sufficiently better than static naïve model to justify their serious attention at this time”. On the basis of the above argument for and against Altman’s work, this paper is opinionated that Altman’s work represents an important effort to find ways of predicting corporate failure. Therein lays the justification of the present paper.
There are several models that seek to predict organizational bankruptcy. Among these are the univariate model (Beaver, 1966); Multiple discriminant analysis (MDA) (Klecka, 1981; Altman, 1993); Linear probability model (LPM) (Maddala, 1983; Theodossiou, 1991; Gujarati, 1998) among others.
The univariate model by (Beaver 1966) traditionally focused on financial ratio analysis. The underlining theory or rational for this model was based on the idea that if financial ratios exhibit significant differences across failing and non-failing firms, then the financial ratios can be used as a predictive variable. The Multiple Discriminant Analysis (MDA) (Klecka, 1981; Altman, 1993) is a linear combination of a certain discriminatory variables. Bankruptcy score is used to classify firms into bankrupt and non-bankrupt groups according to their individual characteristics. The Linear Probability Model (LPM) (Maddala, 1983; Theodossiou, 1991; Gujarati,
1998) expresses the probability of failure or success of a firm as a dichotomous dependent variable that is a linear function of a vector of explanatory variables. Boundary values are developed to distinguish between failing and non-failing firms.
Given the general importance of statistical techniques in corporate bankruptcy prediction, it will be natural for purely statistical models to be used frequently. Despite the ability of the statistical models to predict corporate bankruptcy, their performance however, is questionable. MDA, Logit and Probit models all suffer from restrictive assumptions in one way or the other. The frequent empirical violation of the LPM assumptions and the lack of large time series data sets required for CUSUM and partial adjustment models make it unlikely that any of these models will be of great practical value (Adnan & Humayon, 2006).
This paper acknowledges that discriminant analysis is not the only statistical technique for the development of corporate prediction model. Other suitable techniques such as linear probability model, logit analysis, probit analysis, non-parametric and regression analysis, etc., could have served the same purpose as MDA. The predictive accuracy of statistical tools such as MDA and others mentioned above are not significantly different (Zmijewski, 1983). Argument for all statistical tools mentioned in the study and those not mentioned but reviewed prior to and during this research can be summed up in one sentence. That is, “no