Document Sample

FINANCIAL RATIOS AS THE PREDICTOR OF CORPORATE DISTRESS IN MALAYSIA NAZRINN FARISS IDRIS FACULTY OF BUSINESS AND ACCOUNTANCY UNIVERSITY OF MALAYA JUNE 2008 FINANCIAL RATIOS AS THE PREDICTOR OF CORPORATE DISTRESS IN MALAYSIA NAZRINN FARISS IDRIS FACULTY OF BUSINESS AND ACCOUNTANCY UNIVERSITY OF MALAYA JUNE 2008 2 FINANCIAL RATIOS AS THE PREDICTOR OF CORPORATE DISTRESS IN MALAYSIA NAZRINN FARISS IDRIS Bachelor of Business Administration (Finance) Wichita State University, USA 1995 / 1998 Submitted to the Graduate School of Business Faculty of Business and Accountancy University of Malaya, in partial fulfillment Of the requirement for the Masters of Business Administration JUNE 2008 3 ABSTRACT The 1997 Asian Financial Crisis severely impacted Malaysia’s domestic economy. In the space of just 3 short weeks, almost RM300 billion in market capitalization was wiped off the local bourse. While some argue the situation had been predicted, clearly the depth and suddenness of the event had caught many by surprise. To many Malaysians, the occurrence of the financial crisis highlighted the need for firm action and activities to be constantly monitored and regulated. As the value of listed firms are often supported by the use of funds supplied by the general public, the potential loss in equity value as a direct or indirect result of firm distress, bankruptcy, and reorganization could have wide ranging effects. Clearly, a pre-emptive tool to identify potential problems had to be developed. One tool often cited was through the use of financial ratios. First developed around the turn of the 20th century, research argued that it could predict firm health with a 78% accuracy rate up to five years before distress or failure. In order to evaluate the usefulness of financial ratios to predict firm status in Malaysia, data for firms listed on the Industrial sector of Bursa Malaysia’s Main Board were collected. By compiling the data in the form of financial ratios, we hoped that differences between those from healthy firms and those from distressed firms could be evaluated. Through the use of a ratio analysis, tests of significance, and logit analysis, we found that financial ratios did indeed discriminate the two groups. For almost all ratios tested, the difference between the two groups was found to be statistically significant. Furthermore, by comparing the two groups based on their liquidity and profitability, results indicate that firm distress or failure could be predicted based on these characteristics. 4 ACKNOWLEDGEMENT First and foremost, I would like to express my deepest appreciation to my project paper supervisor, Dr. Rubi Ahmad from the Faculty of Business and Accountancy, University of Malaya, for her invaluable time, patience, and ideas in guiding me to complete this project. Dr. Rubi, given your already heavy workload, I truly appreciate all the help you had given me. In addition, I would also like to take this opportunity to thank all MBA City Campus staff, the staff of University of Malaya’s library, and those attached with KLSE’s Resource Centre. You have all aided me in ways that I can hardly express, and for thank I thank you. Without your kind help, advice, and encouragement, I truly believe that this paper would not have been possible. 5 LIST OF TABLES TABLE 1 GROUP STATISTIC APPENDIX TABLE 3: BURSA MALAYSIA PN4 RECLASSIFICATION CRITERIA Page 43 TABLE 4.1.1 SUMMARY OF PROFITABILITY RATIO STATISTICS Page 51 TABLE 4.1.2 SUMMARY OF CASH FLOW RATIO STATISTICS Page 53 TABLE 4.1.3 SUMMARY OF LIQUIDITY RATIO STATISTICS Page 54 TABLE 4.1.4 SUMMARY OF LONG-TERM RATIO STATISTICS Page 57 TABLE 4.2 MULTICOLLINEARITY DIAGNOSTIC Page 60 TABLE 4.3 COEFFICIENT CORRELATION MATRIX Page 61 TABLE 4.4 LOGIT ANALYSIS MODEL 1 Page 62 TABLE 4.5 LOGIT ANALYSIS MODEL 1 ODDS RATIO Page 64 TABLE 4.6 MODEL 1 PREDICTIVE ACCURACY Page 67 TABLE 4.7 LOGIT ANALYSIS MODEL 2 Page 68 TABLE 4.8 MODEL 2 PREDICTIVE ACCURACY Page 69 6 LIST OF APPENDICES Appendix 1: TABLE 1 – GROUP STATISTICS Appendix 2: SUMMARY STATISTICS Appendix 3: Independent Samples Test 1998 Appendix 4: Independent Samples Test 1999 Appendix 5: Independent Samples Test 2000 Appendix 6: Independent Samples Test 2001 Appendix 7: PN4 Criteria Appendix 8: t-TEST Results Appendix 9: LOGIT ANALYSIS – ALL VARIABLES Appendix 10: LOGIT ANALYSIS – MODEL 1 Appendix 11: LOGIT ANALYSIS – STEPWISE REGRESSION Appendix 12: SAMPLE POPULATION 7 TABLE OF CONTENTS ABSTRACT 4 ACKNOWLEDGEMENT 5 LIST OF TABLES 6 LIST OF APPENDICES 7 CHAPTER 1: INTRODUCTION 1.1 BACKGROUND 10 1.2 SIGNIFICANCE OF THE STUDY 12 1.3 PURPOSE OF THE STUDY 15 1.4 LIMITATIONS OF THE STUDY 17 1.5 ORGANIZATION OF THE STUDY 18 CHAPTER 2: LITERATURE REVIEW 2.1 INTRODUCTION 19 2.2 PREDICTING BANKRUPTCY: UNIVARIATE ANALYSIS 20 2.3 PREDICTING BANKRUPTCY: DISCRIMINANT ANALYSIS 26 2.4 PREDICTING BANKRUPTCY: MULTIPLE DISCRIMINANT 28 ANALYSIS 2.5 PREDICTING BANKRUPTCY: LOGIT, PROBIT & LOGISTIC 34 ANALYSIS CHAPTER 3: RESEARCH METHODOLOGY 3.1 HYPOTHESIS 40 3.2 RESEARCH DESIGN 41 3.3 VARIABLES UNDER STUDY: DEPENDENT VARIABLE 43 3.4 INDEPENDENT VARIABLES 44 3.4.1 LIQUIDITY RATIOS 44 3.4.2 PROFITABILITY & PRODUCTIVITY RATIO 46 3.4.3 CASH FLOW RATIO 47 3.4.4 LONG-TERM SOLVENCY RATIO 48 8 3.5 DATA COLLECTION 49 3.6 DATA ANALYSIS 49 CHAPTER 4: RESULTS & ANALYSIS 4.1.1 RATIO ANALYSIS & T-TEST: Profitability ratios 51 4.1.2 RATIO ANALYSIS & T-TEST: Cash Flow ratios 53 4.1.3 RATIO ANALYSIS & T-TEST: Liquidity ratios 54 4.1.4 RATIO ANALYSIS & T-TEST: Solvency ratios 57 4.2 LOGIT ANALYSIS 58 4.2.1 LOGIT ANALYSIS: STREAMLINING THE MODEL 59 4.2.2 LOGIT ANALYSIS: MODEL 1 62 4.2.3 LOGIT ANALYSIS: MODEL 2 68 CHAPTER 5: CONCLUSION 5.1 SUMMARY 71 REFERENCES APPENDIX 9 CHAPTER 1: INTRODUCTION 1.1 BACKGROUND In mid to late 1960‘s, William H. Beaver (1966) and Edward I. Altman (1968) published what financial experts today would refer to as the pioneering work in the study and prediction of corporate bankruptcy. Although the works of these authors were developed, written and published independently of one another, as these works came out at about the same time, the different but complimentary manner in which each study addressed the concerns at hand resonated throughout the financial community of 1960‘s America. Today, almost forty-years after these articles were first published, the works of Beaver and Altman is continually cited by contemporary researchers as the basis of their own studies. And with the spurt of interest created and generated, academicians and contemporary researchers use it as a foundation to develop, create, and identify newer and more accurate approaches to understanding and predicting corporate bankruptcy. Today, while the spurt of interests surrounding the understanding and prediction of corporate bankruptcy is to be expected given the volatility and uncertainties of the global economic climate, when Beaver and Altman first wrote their paper, these reasons were not their motivation. If we were to read the first page of either author’s work, ironically, their study in the area of corporate bankruptcy was driven more chance than by design. For both authors, the initial purpose of their works was driven more by their need to understand and evaluate the effectiveness of financial ratios. In the 1960’s, 10 although the use of ratio analysis was widespread, its significance was doubted by many. In the years leading up to these works, academicians questioned the purpose and value of ratio analysis as an analytical tool through which business performance could be evaluated. With theorists downgrading the arbitrary rules of thumbs such as comparisons of one company’s ratio with that of another’s, these attacks constituted more than an attack on the usefulness of ratio analysis as an analytical tool. Seen from a different angle, the attacks on the validity and usefulness of the information presented through ratio analysis could also be constituted as an attack on the validity and usefulness of corporate financial statements as a source of financial information. As financial statements are reports that state a company’s financial condition and performance sometime during the past, many academicians were increasingly disillusioned with using historical information to predict the future. In a sense, the prediction of corporate bankruptcy was seen as an illustrative example through which an assessment of the value and function of ratio analysis could established and determined. By proving the usefulness of ratio analysis, doing so would give credence to the use of past information to predict future performances. 11 1.2 SIGNIFICANCE OF THE STUDY In looking back at the reasons cited by Beaver and Altman as to what constitutes the motivation and purpose for their research, it is amusing that this was driven more by chance and convenience than by market demand. Given the volatility and uncertainties prevalent in the global economy, we would have expected these issues to drive any study on corporate distress and failure. Perhaps the economic climate between the 1960’s and that of the new millennium are quite different. But, given the high costs of financial failure and restructuring, the understanding of financial distress and bankruptcy is as vital today as it would have been then. Through the understanding of financial distress and bankruptcy, businesses could be better managed and evaluated as stakeholders would have at their disposal the knowledge and tools required to monitor and evaluate firm action. According to Aharony (1980) corporate failure is an indication of resource misallocation deemed undesirable from a social standpoint. Given the finite resources available, it is therefore imperative that firm action maximizes the returns to stakeholders. From the point of view of business managers, by understanding of the topic better, the insights provided exposes them to the challenges that lie ahead. Through proper planning and resource allocation, courses of action can be put into place. The incentive here is that “good companies” can differentiate themselves from the rest, allowing them to capitalize on opportunities to 12 maximize their profits whilst minimizing their costs. Owners or shareholders stand to benefit directly through share value maximization. Investors on the other hand can also use financial ratios as a tool through which investment choices can be identified, evaluated, and subsequently monitored. Doing so ensures that the potential returns from investments reflect the risks borne by investors. From a lender’s point of view, by having a better grasp of the factors affecting corporate distress and bankruptcy, firm specific risks can be determined. By more accurately identifying the factors that can drive a company to distress and bankruptcy, lenders can evaluate firm financial positions more confidently. Chartkou (2005) 1states that while lenders are concerned with the burden of bad loans and the premium value needed to undertake those risks, borrowers want to borrow at lowest possible rates. As a result, this benefits the lender as they are able to “price” their investment to reflect the risks bared. Given the high financial costs associated with financial default and corporate failure and the volatility of today’s economic environment, the price of failure is sometimes too great. While the reasons mentioned above are the positive reasons why the study is beneficial to us all, perhaps the more important reasons why this study should be undertaken is due to the potential adverse effects that comes as a result not being able to identify the characteristics of financial distress. According to Beaver (1968), evidence suggests that a large portion of firm value is lost 1 Working paper 13 during a corporate reorganization and liquidation process. This point is echoed by Russell (1999) who found that by comparing the pre-bankruptcy equity value with the value after bankruptcy declaration, firm equity value can on average fall by approximately 70%. For many of us in Malaysia, the 1997 Asian Financial Crisis (AFC) illustrates the point. In 1997, due to the financial crisis that had afflicted most economies within the region, companies listed on the Bursa Saham Kuala Lumpur (BSKL) lost hundreds of billions of dollars in market value almost overnight. According to Isa (2005), as a result of the AFC, approximately 80% or RM300 billion of the market value of the local bourse was lost in the span of just 3 weeks. While some may argue that this crisis was predicted in many ways, it was clear that due to the lack of research and understanding of the factors that trigger distress, the event had caught almost everyone by surprise. While the effects of the AFC were unprecedented, the reality is that it clearly adversely affected public listed firms in Malaysia. Though some were more affected than others, clearly more had to be done to distinguish and regulate high risk firm activities. If we look back at past records, for some, even though more than 3 years had past from the time the crisis initially began, its impact and effects could still be seen. With mounting debts, huge accumulated losses, and poor cash flows, by 2001, 91 companies from various sectors would need to be reclassified as financially distressed under BSKL’s newly introduced Practice Note 4/2001 provision2. Under a PN4 status, troubled companies would be given the time 2 Practice note 4/2001 - Please refer to Appendix. 14 and opportunity to regularize their activities in-line with the requirements and provisions of the law. Failing which, these companies would be delisted and removed from the official list of companies on the local bourse. 1.3 PURPOSE OF THE STUDY In light of the evidence previously stated, it should by now be clear that in order for investors, companies, lenders, and government regulators to safeguard the value of the public’s investment, understanding the characteristics of financially distressed companies is the vital first step to developing a relevant and accurate distress prediction model. While most past and current literature have focused their attention on the identification and prediction of corporate bankruptcy models, this paper will instead will try to look at it from a financial distress point of view. According to Isa (2005), the lack of work on financial distress stems from the difficulties in objectively defining distress. In contrast, bankruptcy is legally defined. Through Bursa Malaysia’s efforts however, we here in Malaysia are blessed as conditions of distress can be legally defined. By being able to define the conditions legally, data can be selected objectively. As just mentioned, while the majority of predictive models reviewed focus their investigation on the ability to predict bankruptcy, there are a few reasons why the lessons and implications are also applicable to corporate distress. Firstly, Ward (1997) suggests that financial distress or economic bankruptcy is often described by one of the following circumstances: (1) a condition of negative net 15 worth, (2) an inability to pay debts as they are due, and (3) as a legal definition under which a firm continues to operate or liquidate under court protection. Given these conditions, it is clear that the occurrence of distress supercedes bankruptcy. Secondly, while firms that experience distress are often more likely to fail and go bankrupt, not all distressed firms actually do. Based on his study, Ward (1997) found that the amount of time a firm can remain distressed before filing for bankruptcy can stretch up to 7 years. This finding is verified by Shirata (1999) who studied the predictors of distress in a Japanese setting. The reason cited for this is that while financial distress is a critical event, it is seldom a fatal event given the wide range of preventive and prescriptive options available to the firm. Gilbert et al (1990) puts it best when she states that “Bankruptcy filing can be viewed as a strategic and voluntary response by management to financial problems”. So given the huge window of opportunity available to rectify the situation and the high costs of bankruptcy, the high cost of failure should provide reason enough for us to want to identify the situation early. Based on empirical evidence, it has been suggested that financial ratios can signal potential bankruptcy up to 5 years before its occurrence. Thirdly, Taffler (1984) argues that the results of bankruptcy prediction studies should be interpreted as a description of distress rather than of bankruptcy. The reason is that prior research has succeeded in identifying differences between healthy and distressed companies. Gilbert et al (1990) found it difficult to differentiate between distressed and bankrupt firms. To date, as very little 16 research on the topic has been written within the context of Malaysia, we hope doing so will spur further interest on the topic. With the exception of Mohd Isa (2005) who chose to apply Altman’s bankruptcy predictive model to Malaysian based companies, little else is written on the topic. 1.4 LIMITATIONS OF THE STUDY In order to discover the variables at play and their significance, we hope to do this by analyzing and comparing the characteristics of healthy and distressed firms as given by their financial ratios. Given the availability of a relatively large sample of distressed or PN4 companies, we hope to use this to identify specific characteristics that act as pre-cursors to financial distress. According to Needles (1995) the use of ratios is not without its limitations. Firstly, the use of financial ratios by themselves is meaningless as the values themselves do not mean much. In order for it to be useful, it should be compared with past values or against other companies as a point of reference. As our study focuses on firms listed on the Industrial Sector of Bursa Malaysia’s Main Board, the second limitation is that inferences are limited to other similar type firms. Thirdly, Altman (1968) and Beaver (1966) argue that the use of ratio comparisons should be limited to companies of comparative sizes. Thus, using the results to form judgments on smaller non-listed entities may be misleading and inappropriate as companies of different size and industry operate and behave differently. Lastly, according to Needles (1995), factors like accounting 17 methodologies employed and the date when the analysis is conducted can also dictate the results from ratio analysis. As Balance Sheet items are often snapshots of values at a particular point time, variations stemming from different dates can significantly affect the conclusions observed. As data collected is grouped by the fiscal year, certain biases are unavoidable and must therefore be interpreted carefully. 1.5 ORGANIZATION OF THE STUDY The rest of the paper will be organized as follows. Chapter 2 discusses the literature review related to financial distress and bankruptcy prediction models. Chapter 3 elaborates on the manner in which research methodology is designed and conducted. Chapter 4 presents and discusses the results of the research. And finally, Chapter 5 concludes the study. 18 CHAPTER 2: LITERATURE REVIEW 2.1 INTRODUCTION In this section we will be reviewing past literature to identify the current body of knowledge on the subject matter. As we are interested in identifying and understanding the characteristics of distressed firms, let us first begin by reviewing the manner in which previous works were conducted. By understanding the differing methodologies used in the past to explore the relationship between distressed and bankrupt firms, this should allow us to understand the context and limitations of the studies themselves. This is extremely important as researchers often complain that the most difficult issue afflicting any investigation on topic is its lack of theoretical underpinning. As Ohlson (1980) states, in the absence of any theory on bankruptcy, the selection of appropriate functions to study is a huge problem for researchers. As a practical solution, all a researcher can do is to choose based on computational and interpretive simplicity. This view is also shared by Zmijewski (1984) and Gilbert et al (1990). For most researchers, while determining the variables of interest and their implied relationship to financial distress or bankruptcy is very important, the methodologies used to asses the relationships are often viewed more critically than the variables themselves. As shall be seen, given the variables under study, a large degree of the differences between differing works is driven more by differing methodologies than by differences in the variables of interest. For 19 this reason, the review of literature will be partitioned according to the approaches used. Hopefully, by understanding the context and limitations of the studies, perhaps this will give us a clue on why certain variables were chosen, what are their significance, and the relationship these variables have with financial distress3. 2.2 PREDICTING BANKRUPTCY: UNIVARIATE ANALYSIS Beginning with the pioneering works of Beaver (1966) who sought to determine the usefulness and effectiveness of financial ratios as an analytical tool through which corporate bankruptcy could be predicted, the primary concern Beaver had with the use of ratios was not with so much related to use of ratios as a method or form through which financial-statement data could be presented, but rather with the underlying predictive ability of the financial statements themselves. In effect, Beaver wanted to obtain empirical verification of the usefulness of accounting data. To do this, he took the predictive ability of ratios to signal business failure as a proxy on the predictive powers of the ratios themselves. In order to do this, he began by defining “failure” as the inability of the firm to meet its maturing financial obligations. Operationally, a firm is said to have failed when any of the following events occur; bankruptcy, bond default, an overdrawn bank account, or the non-payment of a preferred stock dividend. 3 While a large percentage of the previous works reviewed center their discussions on the topic of bankruptcy, as most firms will undergo a period of distress before filing for bankruptcy, understanding the knowledge and factors 20 This definition of failure is shared by Altman (1968), Charitou (2000), and a host of other researchers. Having defined the dependent variable of interest, he then went about selecting the independent variables. In looking at the independent variables under his study, what made Beaver’s study a pioneering look into corporate bankruptcy was his use of multiple ratios as the independent variable. By looking at 30 financial ratios from several distinct groups, his study differed from past research as he used the variables to identify ratios that could best explain and predict bankruptcy. Before then, studies focused their efforts on testing one explanatory variable against the dependent variable. By pitting a singular explanatory variable against the dependent variable, comparisons of predictive power and relative significance often proved to be a futile exercise. By combining the use of multiple variables, their relative effect and significance could be ranked and identified. Out of all possible combinations, the criteria used to identify the variables were driven by several factors. The first criterion was popularity as given by its frequent appearance in financial literature. The second criterion was dictated by the past performances of the ratios themselves. The third and final selection criterion was dictated by its ability to be defined as a “cash-flow” concept. Any ratio which met these criteria was included in the study. Although the manner used seems void of any clear and objective rules, the lack of theoretical underpinning necessitates the approach. Similar approaches were also used by researchers like Deakin (1972), Ohlson (1980), Altman (1968), and Dichev afflicting bankruptcy should be a good proxy to advancing our knowledge of distressed companies. 21 (1998). Having identified the variables of interest, they were then tested for their ability to predict bankruptcy. As a result, the number of independent variables chosen was narrowed down to seven which exhibited the best performance. Amongst them were six accounting ratios and one accounting measure. The reason he did this was that for his multi-ratio analysis to convey as much additional information possible, common elements had to be eliminated4. The selected ratios were now confined to; Cash Flow to Total Debt, Net Income to Total Asset, Total Debt to Total Asset, Working Capital to Total Asset, Current Ratio, No Credit Interval5, and Total Assets. To analyze the ratios, their predictive power was tested through their ability to correctly classify firms. Based on the accuracy of ratios to predict group classification6, ratios were differentiated and ranked. The higher the accuracy (or the lower the misclassification rate), the better it was as a predictor of failure. The ability to predict failure was best given by the ratio of Cash Flow to Total Debt. Used by itself, it could correctly predict and classify firm status 87% of the time in the year preceding failure, and 78% of the time five years prior to failure. The next best indicator is given by the ratio of Net Income to Total Assets. This ratio correctly classified firms 87% of the time the year prior to failure and 72% of the time five years prior to failure. As both these ratios were based on a cash flow concept and were highly correlated to one another, he deemed the results to be as expected. 4 ratios with common numerators and or denominators were dropped 5 No Credit Interval = (Defensive Assets - Current Liabilities)/ Operational Expenditures 22 In analyzing the comparisons of means between failed and non-failed companies, Beaver found that given the mean values for failed firms, a trend could be seen. With the mean ratios for failed firms being significantly lower up to five years before failure, the decline in value grew increasingly worse as bankruptcy loomed. As the mean values for non-failed firms exhibited a more stable value throughout the period under study, no such trend or pattern could be seen. As this pattern is evident in all ratios under the study, he suggests that the use of ratios to predict failure is not without its merits. In a follow-up to his initial paper released just several years after the first, Beaver (1968) advanced his look into the prediction of bankruptcy to include the use of market returns. By doing so, he sought to investigate the extent to which changes in the market value of stocks could be used to predict failure. By using the same sample population to conduct his study, the annual rates of returns for both failed and non-failed firms were computed for up to five years prior to failure. Through a cross sectional analysis which sought to investigate the variables’ value at a particular point in time, Beaver had instinctively expected failed firms to carry a higher expected rate of return compared to their healthier counterparts. The reason for this was simple. As failing firms would be expected to carry a higher probability of failure, the relatively higher inherent risks should compensate risk averse investors with a higher return on their 6 Firms are classified into two groups; fail or not failed. 23 investment. Based on the evidence presented however, it was clear that at no point in time did ex post returns differ between failed and non-failed firms. This to him was extremely surprising. As he had expected investors to behave in a logical and rational manner, it was assumed that at the start each period, the solvency and inherent risk of the firm will be evaluated to ensure that expected returns would commensurate the risks of the firm. If at any time the solvency of the firm deteriorated, stock prices would fall as investors adjusted their positions, giving a rate of return lower than would have been initially expected. A longitudinal analysis however found that the mean distribution of firm returns did indeed vary with time. Although non-failed firms exhibited no significant upward or downward trend, the median returns from failed firms were lower in the years approaching failure. As failure approached, these differences grew increasingly larger. This indicated that on average, an unexpected decline in solvency position does indeed impact the ex post returns of failed firms. Furthermore, as data also showed a larger return dispersion, failed firms carried a greater risk of variability7. These findings were the same whether the returns were adjusted or unadjusted for market wide economic effects. In comparing the predictive powers of his ratio model to that of the market model, Beaver used a dichotomous classification test to measure the rate of misclassification errors8. By measuring the percentage misclassification error rates, the lower misclassification error rate given by ratios than market returns 7 Both the Avg. and Marginal rates of returns were lower for failed firms. The dispersion of returns were also much greater 8 Misclassification rates were given by how accurately each model would correctly predict group membership. Barniv (2000) gives the following example: if a cutoff probability value of 0.5 is used, firms with a failure 24 clearly suggests that it is a superior predictor. Through the use financial ratios, 78% of his sample could accurately classified five years before failure. Continuing on with the research conducted by Beaver, Deakin (1972) sought to advance the knowledge on bankruptcy prediction by first verifying Beaver’s results. This was done by replicating Beaver’s original study. Although the general procedure used by Deakin to evaluate and estimate the probability of bankruptcy is somewhat similar to that employed by Beaver, the differences amongst these approaches mainly centered on how the authors chose to define failure or bankruptcy. Unlike Beaver who had chosen to define “failed” firms to include those that missed-out on fulfilling their preferred dividend payments and those which defaulted on their loan obligations, Deakin was more stringent on his definition. Only firms that were bankrupt, insolvent, or liquated by creditors would be deemed as failed in his study. The independent variables used by Deakin were taken directly from those highlighted by Beaver. With a sample consisting of 32 failed and non-failed firms, like Beaver before him, the firms were matched on the basis of industry classification, asset size, and of course, year of financial data. To test his findings, percentage error rates were calculated. Although some slight differences could be seen between the two, Deakin admits that his small sample size prohibits him from determining if these differences were indeed significant. probability > 0.5 would be classified as a bankrupt firm. 25 To work around this, he uses a Spearman rank-order correlation coefficient to allow a comparison of the predictive power of ratios in the two studies. As the rank-order correlation coefficient was very high in four of the five years under study, this confirmed Beaver’s findings. Interestingly, the low correlation coefficient in year 3 before failure was attributed to the rapid expansion failed firms undertook three to four years before failure. This was financed primarily through the use of increased debt and preferred stock instead of common equity or retained earnings. As these firms would later be unable to generate the income required to support higher debts, these “assets” would be quickly lost. Soon after, asset and debt values would fall back to expected levels. Based on the error classification rates obtained through this methodology, although the 20% 9 misclassification rate the year preceding failure would indicate the appearance of a fairly accurate model, Deakin felt that through the use of an alternative model to predict bankruptcy, a lower misclassification rate could be obtained. The methodology he chose to accomplish this was through the use of a Discriminant Analysis (DA). 2.3 PREDICTING BANKRUPTCY: DISCRIMINANT ANALYSIS The purpose of Discriminant Analysis (DA) is to find a linear combination of ratios which would best discriminate between groups under study. A major assumption to DA is that samples must be drawn randomly. To avoid bias, 9 As given by the ratio of Cash Flow to Total Debt. 26 Deakin chose a second sample consisting of 32 non-failed firms chosen randomly. By inputting the ratios used by Beaver into the DA model, a scaled vector would indicate the relative contribution of each variable. This would then be used produce a score that maximizes the distinction between the two groups. The significance of each of the discriminant functions is then measured by Wilks lambda. This is used to test the hypothesis that the means of the ratio vector from both groups are similar. By converting it into an F value, the ratio value indicates the probability of a significant separation between the scores of a failed and non-failed firm. Based on the results generated, the variables that carry a high value would indicate a significant contribution to the predictive power of the function. Those with a low value indicate the opposite. From Deakin’s results, he finds that of his original variables, the ratios deemed as significant predictors of failure varied depending on the year of study. In year five before failure for example, the three most significant predictors of failure were given by Working Capital to Total Asset, Total Debt to Total Asset, and Quick Asset to Sales. In the year prior to failure, they were; Working Capital to Total Asset10, the Current Ratio, and the ratio of Current Asset to Total Asset11. Interestingly, when Deakin tried to omit variables with little predictive power to create a more concise model, misclassification errors increased significantly. In light of the evidence presented by Deakin, the question that is often asked 10 Scaled vector score for years 2-4 prior to failure indicate low predictive abilities. 27 now is how DA compares to Univariate Analysis. Like before, model accuracy is measured by comparing their misclassification error rates. With misclassification error rates of 14%12 one year prior to failure and 10% two years prior to failure, the use of the DA model appears to produce better results. 2.4 PREDICTING BANKRUPTCY: MULTIPLE DISCRIMINANT ANALYSIS Like Beaver before him, Altman’s (1968) study was initially motivated by the need to asses the quality of ratio analysis as an analytical tool. By using corporate bankruptcy as an illustrative example, he hoped to determine the value of financial ratios through the use of a Multiple Discriminant Analysis (MDA) technique. MDA is a statistical technique used to classify observations into several a priori (formal) groupings dependent upon the observations of individual characteristics. MDA is primarily used when the dependent variable appears in qualitative form, which in this case it would appear as either bankrupt or non-bankrupt. For Altman, the univariate approach originally employed by Beaver suffered from too many deficiencies. Firstly, as the use of a univariate approach centers on the identification of a singular signal of impending problem, information presented in this form is confusing and easily misinterpreted13. Secondly, as 11 While the ratio of Working Capital to Total Asset is significant one year and five years before bankruptcy, this is an anomaly. Again, the predictors deemed significant varied every year studied. 12 Recall earlier that Deakin’s replication of Beaver’s original study produce 20% misclassification error rate the year prior to failure. 13 Example: Imagine a company with poor profitability. Taken solely, we would conclude that it is a higher risk than another firm with superior profitability levels. But, given that it also has better liquidity, does it still remain the 28 there are many ratios available, the selection and combination of differing ratios would each result its own unique implications. Due to this, ranking and selecting the appropriate combination to use is both tricky and nearly impossible. As such, Altman argued that for financial ratios to be relevant it has to be an extension of past findings, building upon and combining the factors deemed significant in predicting failure. Now, the focus shifts from identifying a ratio that can explain or predict failure to identifying the combination of ratios and their weights (as a measure of factoring their overall impact) that would best explain the situation. Looking at the variables under study, initially, 22 ratios were chosen based upon their popularity and potential relevancy. They were classified into 5 standard ratio categories; liquidity, profitability, leverage, solvency, and activity ratios. Of the original 22, 5 were chosen as doing the best overall job. This was done through; (i) observation of their statistical significance, (ii) evaluation of inter-correlations amongst independent variables, (iii) observation of the predictive accuracy of the various profiles, and, (iv) the judgment of the analyst. The final set included; (i) Working Capital to Total Asset, ii) Retained Earnings to Total Asset, iii) EBIT to Total Asset, iv) Market Value of Equity to Book Value of Debt, and (v) Sales to Total Asset. In looking at the final ratios chosen, the variable profile did not necessarily contain the ones with the most significant predictive power as measured riskier of the two or is it the opposite. This is the flaw to the use of a univariate approach. 29 independently14. In order to improve upon Beaver’s univariate approach, the variables had to be chosen based on their overall impact. This was done by determining the relative contribution each variable made to the total discriminating power of the model. In fact, although “Sales to Total Asset” is the least significant of all ratios measured individually, due to its relationship with other variables in the model, it has the second highest discriminating ability. A probable reason for this could be due to the high negative correlation observed between EBIT to Total Asset and Sales to Total Asset. There are several advantages to using MDA. Firstly, when the coefficients are applied to the ratio under study, firms can be classified into mutually exclusive groups. By doing so, this allows us to consider common characteristics in the entire profile and observe the manner in which they interact. Secondly, because a two-state MDA model is analyzed in only one dimension15, the classification process is easily interpreted. The resulting single discriminant score or Z value can then be used to classify firms. The final discriminant function developed by Altman is as follows; Z = .012 (WC/TA) + .014(RE/TA) + .033(EBIT/TA) + .006(Market Value Equity/ Book Value Debt) + .999(Sales/TA) By inputting data of a particular firm into the model and calculating its Z-score, a Z-score value below 1.2 would imply a high probability of bankruptcy and a Z-score above 2.9 would imply a low probability of bankruptcy. For those with 14 To test the discriminating abilities of these 5 variables, an F-test was employed. For variables (i) to (iv), the high F ratio at the .001 level indicates that there is significant differences between bankrupt and non-bankrupt group. The f ratio for variable (v) however does not show significant difference between the two groups. 15 MDA reduces the analyst’s space dimension by G-1. In a two group model, analysis is reduced to a single dimension as given by; Z=v1x1+v2x2+…+vNxN -where; vN represents discriminant 30 scores between 1.2 and 2.9, results were deemed inconclusive. Much like before, an evaluation of the model’s predictive ability is measured through the use of determining its classification error rates. Through the use of the Z-score, Altman found that this predictor could correctly classify firms 95% of the time one year prior to failure. Beyond that however, the model’s accuracy deteriorates significantly16 suggesting that its predictive ability is unreliable. Other significant findings from the study also confirmed the deterioration of financial ratios as bankruptcy approached, and that for the majority of firms, the largest changes occurred two to three years prior to bankruptcy. Given the simplicity of Altman’s Z model, Isa (2005) argues that its popularity is to be expected. As it provided a benchmark through which similar firms could be compared, its indicators of financial strength can and has often been used to predict distress. However, as many detractors will argue, the model is not without its flaws. By using Altman’s original model to study predictors of financial distress in Japan, Shirata (1999) argued that the manner in which Altman conducted his research was flawed given certain limitations. Firstly, as Altman never specified the manner in which variables under his study were identified, it is difficult to state for certain if the variables he chose contained the best set possible. By using powerful data mining tools like Classification And Regression Tree model (CART) and Stepwise, these approaches allowed her uncover explanatory variables more objectively. Unlike Stepwise procedure which assumes that variables are normally distributed, CART selects variables coefficient xN represents independent variables 31 without needing such assumptions. Beginning with 61 financial ratios as the basis of her study, the use of the Stepwise and CART helped identify four variables deemed significant predictors of distress. This was given by the ratios of Retained Earnings to Total Assets, Accounts plus Notes Payable to Sales, Current Gross Payable to Previous Gross Payable, and the ratio of Interest Expense to Borrowing. Correlation analysis ensured that these variables were independent of each other. Amazingly, none of the ratios selected included any profitability or liquidity ratios. Although they were widely believed to be predictors of distress, her study proved that in the case of Japanese firms, they were not significant predictors. Secondly, as Altman and many researchers after him would choose to conduct their studies through a paired sampling method, Shirata argued that the biases introduced can lead to misleading conclusions. In Altman’s defense, the paired sampling approach used was done in order to minimize the effects of firm size. For Altman, although he acknowledges that size would be a significant factor in determining distress, in order to easily evaluate the impact of ratios, size would have to be a controlled variable. The reason he does this is due to his inability to clearly define its relationship to bankruptcy. This view is also shared by Beaver. Beaver realized that while a paired design sample selection methodology mitigates the disruptive influence of asset size and industry, its use would also virtually eliminate any predictive power these factors might have had. Another method to control the effect of size on the independent variables as suggested by Peat (2003) is by simply dividing the variables of interest with total assets. In 16 Accuracy for years 2 to 5 are 72%, 48%, 29%, and 36% respectively. 32 doing so, total asset will ceases to be an explanatory variable. Shirata’s study however proves that given the four variables chosen through CART and Stepwise, the impact of industry and size does not influence the discriminant power of the variables. Given the significantly larger sample population used by Shirata, this finding is unique. But with a 13.84% misclassification error rate, her model does appear to be universal in its application. As Altman’s model was based on research conducted with financial data from the 1940s through to 1960s, Begley et al (1996) feared that using this model well after it was first developed could result in high misclassification errors. The reason for this is that in the time since it was first developed, economic conditions have changed significantly. Through changes in bankruptcy laws and changes in the acceptance of higher leverage, he feared the model was now obsolete. By applying the financial data from the 1980s to Altman’s original model, Begley found that Type I and Type II misclassification rates17 were much higher than those from the initial study. Interestingly, unlike Shirata, Begley’s study found size to be a significant predictor of distress18. Furthermore, with the increasing acceptance of higher leverage, debt level is not as significant a predictor as before. In line with financial expectations, however, given the higher debt level 17 Type I error is 18.5% and Type II error is 25.1% giving an overall misclassification rate of 21.8% one year prior to failure. 18 The effect identified by Begley is shared by Dichev (1998) who also found that size effect to be weak to nonexistent for firms in the 1980s and 1990s 33 carried by firms in the 1980s, the significance of liquidity is greater now than before as the size of a firm’s interim obligations were now much larger. In order to update Altman’s original model, Begley re-estimated the values of Altman’s Z model. As a result, Working Capital to Total Assets is now the most powerful predictor of probable distress. In Altman’s original model, this variable was interestingly the variable with the least predictive power. This result however is in line with the increasing importance of liquidity given higher use of leverage. Furthermore, while the ratio of Sales to Total Assets was the second most important predictor in the original model, Begley’s re-estimation found this to be the least important predictor. As a result of the re-estimation process, the misclassification rates were now lower, but not by much. 2.5 PREDICTING BANKRUPTCY: LOGIT, PROBIT & LOGISTIC ANALYSIS Continuing with our review of past research on bankruptcy and bankruptcy prediction, the next breakthrough for research in this field came through the efforts of Ohlson (1980). Given the simplicity and accuracy of the methodology prescribed by Altman’s (1968) MDA Approach, its popularity and usage was widespread. However, to Ohlson, MDA was not without its drawbacks. Firstly, the use of MDA imposes certain statistical requirements on the distributional properties of the predictors. As Altman’s model clearly violated this condition, Ohlson sees this as a limitation of its predictive abilities. Secondly, as the output 34 from the use of MDA is a score, its output allows for little else than an intuitive interpretation. And thirdly, as most users of MDA have used a paired sampling design to choose their sample population, the bias introduced could lead to misleading results. In order to avoid the problems mentioned above, Ohlson instead chose to study the topic through the use of a conditional logit analysis. Through the use of logit, the fundamental estimation problem can be reduced to finding out the probability of a firm failing within a specified frame of time given the population it belongs to. No assumptions regarding prior probability of failure is needed nor do we need to determine the distribution of predictors. According to Barniv (2000) logit or probit models can be seen as a limited dependent variable where only two possible outcomes can occur. It normally takes the following form: P = P (y=1)/X = F(X’B) Where; P is the probability of bankruptcy, Y = 1 if bankruptcy has occurred, Y = 0 if otherwise X’ = (X1, …, Xk) is the vector of independent variables B’ = (B1,…, Bk) is the coefficient corresponding to values of X And, F(X’B) is the cumulative distribution function given by logit as: P = F(X’B) = (eX’B)/ 1+eX’B = 1/(1+e-X’B) Furthermore, unlike most past research that use paired sampling methodologies to select firm sample population, Ohlson’s study relies on observations from 105 bankrupt firms and 2058 non-bankrupt firms with financial data from 1970 to 1976. 35 Unlike earlier studies which derive their data through Moody’s Manual, Ohlson’s data is extracted from the company’s 10-K financial statements. This is an important difference because through a 10-K filing, the public can check whether the bankruptcy occurs before or after the filing date. For Ohlson, this issue of timing matters because if the purpose of the investigation is to forecast relationships then such details are important. Using predictors with statements released after the bankruptcy date overstates the predictive power of the models19. Dichev (1998) points out that as some firms are delisted long after entering bankruptcy the predictive powers of models which do not address the issue of bankruptcy timing must be interpreted with caution. In looking at the manner in which Ohlson chose his ratios or independent variables, again, simplicity was his most important criteria. These variables included; Size, Total Liabilities to Total Asset, Working Capital to Total Asset, Current Ratio, Net Income to Total Asset, comparisons of Total Liabilities to Total Assets 20 , funds from operations/ Total Liabilities, a measure of Net Income represented by INTWO21, and a measure of change in net income represented by CHIN. By comparing the means of failed firms one and two years prior to failure with those from non-bankrupt firms, ratios exhibited deterioration as it moves closer to bankruptcy. Although the data used is not comparable to Beaver (1966), in 19 Under normal circumstances, it is possible for a company to file for bankruptcy at some point in time after the fiscal year date but before releasing the financial statements. Neglecting this possibility may lead to back-casting for many of the failed firms. 20 a value of 1 if TL>TA, O otherwise 36 general, the results are similar. With the exception of Size, the standard deviations of failed firms were larger one year prior to failure than that of their non-bankrupt counterparts. Based on the coefficient estimates and t-statistics calculated for the independent variables under study, Size appears to be an important predictor. With a t-statistic value exceeding 3.7 in all models used22, it is a statistically significant predictor of bankruptcy. Other significant predictors include financial structure, performance and liquidity given by the ratios of; Total Liabilities to Total Assets, Net Income to Total Assets, and Current Liabilities to Current Assets. And, given the fact that financial state variables and performance variables show little correlation to one another, this would indicate that the contributions of these variables are significant and independent of one another. Looking at the results of the logit model used, as it was able to predict bankruptcy 96.12% of the time within one year and 95.55% within two years of its application, initial results would indicate a more accurate model than those previously developed (example given, Altman’s Z model). These findings are shared by Dichev (1998) and Begley et al (1996). Through the efforts of Begley (1996), the result from Ohlson’s original model was retested with the use of more current data. By applying financial data from the 1980s to Ohlson’s model and cutoffs, surprisingly, Type I error rates 21 1 if net income positive in last 2 years, and 0 otherwise 22 Ohlson uses 3 models. Model 1 predicts failure within 1 year. Model 2 predicts failure within 2 years. Model 3 predicts failure within 1- 2 years. 37 remained the same. But as Type II error rate has increased, overall misclassification rates 23 were now higher than before. This indicated that changes in the operating environment affected the model. To update Ohlson’s model for changes in the environment, Begley began a re-estimation process by replicating Ohlson’s original logit analysis with data from the 1980s. As expected, the increasing acceptance of leverage reduced the significance of debt level as given by the ratio of Total Debt to Total Assets. Given the higher debt levels carried by firms, this situation is expected to put more strain on their liquidity positions. The increase in the coefficient of Working Capital to Total Assets implies that working capital is now more important than before. More significant however was the finding that two variables had switched signs indicating an opposite relationship with distress than was previously expected. The ratio of Current Liabilities to Current Assets was now inversely related to distress, and the ratio of Net Income to Total Assets was now positively related to distress. Begley admits that these results were clearly against economic convention. While the use of the re-estimated model has resulted in a lower Type II error rate, the Type I and overall error rate is now higher than before.24 This suggests that the new model does not improve the user’s ability to classify bankrupt and 23 In Ohlson’s original study, misclassification rate was 14.9%. By using it with data from the 1980s, misclassification rates were now 18.7%. 24 Misclassification error rate using Ohlson’s original model is 18.7%. The overall re-estimated model misclassification error rate is now 22.1%. 38 non-bankrupt firms. Its use however can still be merited. If users are more inclined to minimize Type II error classifications, the re-estimated model is by far superior. But, as the costs of Type I errors are often more costly, then Begley admits that the use of Ohlson’s original model is empirically justified. 39 CHAPTER 3: RESEARCH METHODOLOGY 3.1 HYPOTHESIS Based on the evidence deduced from past literature, numerous works have identified and confirmed the existence of financial differences between failed and non-failed firms. While most previous works reviewed and cited focus on the characteristics of failed and non-failed firms, we differ slightly as we will use the knowledge to identify and investigate firms in financial distress. Gilbert (1990) demonstrated that when bankrupt and distressed firms were investigated, the model used was unable to distinguish between the mean from one group to that from another. According to Ward (1997), previous works were prevented from such an investigation due to the unavailability of a clear and objective definition for distress. But as Isa (2005) pointed out, with the conditions set forth in Bursa Malaysia’s Listing Requirement, distressed listed firms in Malaysia can be legally defined, identified, and classified. From the pioneering works of Beaver (1966) and Altman (1968) all the way to Peat (2003), research has shown that through the use of various analytical tools, these differences can be identified. Through the use of financial ratios as an investigative tool, researchers have been able to use these unique ratios to identify specifically what the differences are. However, it must be noted that although differences were found to exist, the characteristics or the ratios deemed predictive were often as unique as the research themselves. 40 In this study, we hope to first identify the characteristics of distressed firms and to investigate if these characteristics as given by their financial ratios are indeed different than that of healthy firms. Given past results, we expect it to differ. Having identified the characteristics of firms in distress, we ultimately hope to use these explanatory variables to predict financial distress. As such, the following null hypothesis will be tested: H01: There is no significant difference between the ratios of healthy and distressed firms. 3.2 RESEARCH DESIGN The purpose of this study is to (a) determine the characteristics of distressed firms through the use of financial ratios and (b) to investigate the predictive powers of the ratios (independent variables) to correctly classify firms as being healthy or distressed. The sample period under study will be from 1998 to 2001. This sample period was chosen for several reasons. Firstly, the occurrence of the Asian Financial Crisis in 1997 severely impacted most firms in Malaysia and in the region. Given the unusual occurrence, we begin with data from 1998 to ensure that the broad ranging effects of the regional financial crisis are minimized. Secondly, given the small number of firms entering conditions of financial distress post 2002, it is expected that any comparisons between the financial data of healthy 41 and distressed firms beyond this period could lead to misleading and inaccurate conclusions as the small sample size will introduce new bias. In determining the population of firms under study, much like the earlier works of Beaver (1966) and Altman (1968), we focus our investigative look into Malaysian public listed firms classified on the Industrial Products sector of Bursa Malaysia’s Main Board. While all firms listed on the Industrial Sector of the Main Board will be eligible, inclusion will only occur if it meets the following selection criterion; a) The firm must be classified under Bursa Malaysia’s Industrial Products sector as at 31st December 2001. b) Healthy firms must have complete financial data for the periods between 1998 and 2002 accessible through Bloomberg Professional’s online financial database. c) As PN4 companies are often excluded from Bloomberg Professional’s coverage, distressed firms must have been classified as a PN4/2001 company by Bursa Malaysia and have their financial records stored within Bursa Malaysia’s Resource Centre. Through the filtering process mentioned above, the sample size for healthy firms listed on the Industrial Products sector has been reduced 53 firms. The sample size for distressed firms given the aforementioned criterion is 13. Therefore the total number of firms in our sample is 66 firms. 42 3.3 VARIABLES UNDER STUDY: DEPENDENT VARIABLE For the purpose of this study, the dependent variable is the financial classification of public listed companies in Malaysia. Firms are either classified as normal healthy companies or distressed companies. Unless specified, companies are assumed to be normal and healthy. To identify distressed firms, we use the conditions set forth in Bursa Malaysia’s Listing Requirement. Through the conditions stipulated by Practice Note 4/200125 issued in relation to paragraph 8.14 of the listing requirements, this provision aimed at ensuring the financial condition of listed companies warrants trading is used to classify firms as distressed. From here on, these firms will sometimes be referred to as PN4 companies. Table 3: Bursa Malaysia PN4 reclassification criteria summary – The shareholders’ equity is equal to or less than 25% of the issued and paid-up capital. – Receivers and/or managers have been appointed – Winding up of subsidiary or associated company which accounts 50% of the total assets. – Auditors have expressed an adverse or disclaimer opinion. – Auditors have expressed a modified opinion with emphasis on the listed issuer’s going concern. – A default in payment. – The listed issuer has suspended or ceased:- (i) all of its business or its major business; or (ii) its entire or major operations, 43 3.4 VARIABLES UNDER STUDY: INDEPENDENT VARIABLES Independent variables under study within this research will comprise of 13 financial ratios from four specific groups. These 13 ratios comprise of those representing liquidity ratios, productivity and profitability ratios, cash flow ratios, and long-term solvency ratios. As the problem of a lack of theoretical underpinning as a guide to variable selection has been covered earlier, the use of the independent variables under our study is based on the popularity of the ratios from past research and their past performance in reviewed literature. 3.4.1 LIQUIDITY RATIOS Liquidity ratios measure a firm’s ability to meet its short-term obligations. Through the use of cash and other easily sellable liquid assets, the size and composition of these assets can be used to cover payables, short-term debt, and other liabilities. As these ratios basically measure the coverage and cushion provided by the firm’s more liquid assets, it is expected that higher liquidity values should provide a better buffer to distress and insolvency. This suggests an inverse relationship with distress. For this study, the ratios used will be as follows: a) Working Capital to Sales: (Current Asset-Current Liabilities)/ Sales – Working Capital measures the “current” position of the company. By definition, current liabilities are paid out of current assets. A positive 25 Refer to Appendix for more on Practice Note 4/2001. 44 working capital indicates assets can be used to continue operations. By dividing against asset size, this allows comparison of working capital against firms of differing sizes. b) Cash to Total Asset: Cash/ Total Asset - measures the portion of a company’s assets held in cash or marketable securities. A high ratio acts as a buffer to safety. c) Cash to Current Liabilities: Cash/ Current Liabilities – measures a company’s ability to meet short-term obligations immediately. A higher ratio indicates greater ability. d) Current Asset to Current Liabilities: Current Assets/ Current Liabilities – also known as Current Ratio, measures the firm’s ability to pay near-term obligations. Oftentimes, “Current” refers to periods of 1 year or less. The higher the ratio, the larger the firm’s safety cushion against default. e) Current Asset to Total Asset: Current Asset/ Total Asset – measures the proportion of assets that can be easily sold or converted to cash. A higher value provides a larger cushion should an unexpected obligation surface. f) Working Capital to Total Asset: (Current assets-Current Liabilities)/Total Assets – measures liquidity. Given the extensive research coverage liquidity ratios have received in the past, evidence suggest that liquidity can be a significant predictor of distress. While its’ overall importance was not particularly significant in Altman’s (1968) original study, the increasing acceptance of leverage in the 1980s and 1990s has 45 meant an increase in significance given the larger debt maintenance costs. This view is also shared by Begley (1996) and Deakin (1972). Shirata (1998) however disagrees with this. To her, liquidity is not significant. Given the results presented by Begley (1990) and Deakin (1972) the ratio of Working Capital to Total Asset could be an important explanatory variable. 3.4.2 PROFITABILITY & PRODUCTIVITY RATIOS According to Fazilah (2000), the firm’s earning capacity and its continued ability to generate and improve profits is usually its most important objective. Through profitability ratios, we can investigate how well the firm performs given the resources available to it. It is normally presented in percentages and the higher the percentage, the better its performance. Therefore, profitability is expected to be inversely related with distress. For this study, the ratios used will be as follows: a) Earnings to Total Assets: EBIT/ Total Assets – measures the productivity of the company’s assets. The higher the utilization percentage, the better its productivity. b) Return On Assets: Net Income/ Total Asset – measures the effectiveness of the company’s use of assets. A higher value indicates better utilization of resources and could indicate effective management. 46 c) Sales to Total Asset: Sales/ Total Assets - measures revenue generation. By dividing sales value against asset size, this allows comparison of revenue generation against firms of differing sizes. Based on past research surrounding profitability and productivity ratios, evidence suggests explanatory significance. The ratio of Net Income to Total Assets studied by Beaver (1966) for example could be used to correctly classify firms 87% of the time one year prior to bankruptcy and 72% 5 years before. While this was also verified by Ohlson (1980), Begley’s (1996) re-estimation of Ohlson’s study revealed profitability to be positively related to distress. Though he admitted this to be puzzling, he could offer no explanation for the relationship. 3.4.3 CASH FLOW RATIOS According To Needles, because cash flows are needed to pay debts when they are due, cash flow ratios can indicate solvency and liquidity. Through the efforts of Beaver (1966), cash flow ratios could accurately classify firms up 78% of the time 5 years before bankruptcy. This suggests significant predictive potential. Because a higher ratio will indicate more cash flows will be available to meet financial obligations, we expect higher ratio values to translate into lower distress risk. The ratios to be included from this grouping are: a) Cash Flows to Sales: Cash Flows/Sales – measures the ability of the firm’s sales to generate operating cash flows. 47 b) Cash Flows to Total Debt: Cash Flows/ Total Debt – measures the length of time a company will need to pay off its debts with just its cash flows. Although not realistic, a high ratio signals a less risky company better able to payoff its debt. c) Cash Flows to Current Liabilities: Cash Flows/ Current Liabilities – measures whether or not the company is earning enough to cover its current liabilities. If the ratio is below 1.0, then the company must either find an alternative way to finance its commitments or slow down the rate it is spending its cash. 3.4.4 SOLVENCY RATIOS One of the best ways for us to measure a firm’s riskiness is to examine its solvency ratios. According to Fazilah (2000), unlike liquidity ratios which is concerned with the company’s ability to meet near-term obligations, solvency ratios tend to take a long-run point of view. By measuring long-term debt for degree of financial leverage, this can identify future problems. The ratio to be used here is: a) Total Debt to Total Assets: Total Debt/ Total Assets – measures the company’s financial risk by determining the amount of assets financed by debt. A higher ratio signals greater financial risks. 48 Looking at empirical evidence to support its inclusion, Deakin (1972) and Ohlson (1980) prove the ratio to be significant. However, these studies were conducted with pre-1980s data. With the acceptance of higher debt use in the 1980s and 1990s, its power as a descriptive variable has declined. This view is supported by Begley (1990). Regardless, as the use of debt increases the amount needed to service it, positive increases in debt should increase the risk of distress. 3.5 DATA COLLECTION This study makes use of secondary data available in Bloomberg Professional online database for the collection of financial data for healthy firms. With a huge database available, Bloomberg Professional service provides comprehensive financial data for most public firms listed any of the major indices in the world. Available data for public listed firms includes; balance sheet data, income statement data, statement of cash flows, commonly used financial ratios, share price, and dividend records. As was earlier mentioned, the financial data used for PN4 companies are either those included in Bloomberg Professional service or those with hardcopy filings maintained by Bursa Malaysia’s Resource Center. The financial data found in hardcopy filings maintained by Bursa Malaysia are taken from their Annual Reports. 3.6 DATA ANALYSIS In line with the findings of Beaver (1966), Altman (1968), and Ohlson (1980), 49 comparisons of the financial ratios used in this study are expected to highlight differences in the average mean values for healthy firms and those of distressed firms. To investigate if these earlier findings are merited, the first step of our analysis will be to conduct a ratio analysis. Having done so, we will then test these differences for significance. Next, through the use of a logit regression model to further analyze the relationship between the dependent and independent firms, this should aid us on discovering the probability of a healthy firm becoming distressed. Data will be analyzed through the use of Microsoft’s Excel (Excel) spreadsheet, Statistical Package for Social Sciences (SPSS) Version 14 for Windows, and the use of Intercooled Stata Version 8.0. The use of Excel will be primarily to calculate the values of financial ratios for all firms under study for the period between 1998 and 2001. Thereafter, SPSS and Stata will be used to perform the various statistical analyses on the dependent and independent variables. Again, based on the findings presented by previous researchers, differences between the two groups could be possible. 50 CHAPTER 4: RESULTS AND ANALYSIS 4.1 RATIO ANALYSIS & T-TEST 4.1.1 RATIO ANALYSIS & T-TEST: Profitability ratios Table 4.1.1 : Profitability Ratio - Two sample t-test with equal variances Variable Group Observations Mean Std Dev t df Status nita 0 219 .0276142 .2623657 5.7604 268 1 51 -.2618334 .50961.38 ebitta 0 219 .0623366 .2177557 5.2390 268 1 51 -.1763142 .5033522 salesta 0 215 9.694038 100.3764 0.4987 261 1 48 2.444032 13.27714 Note: Non-failed group distinguished by status 0 and failed group distinguished by status 1 Looking at the mean value for the ratios under study, data from Table 4.1.1 suggest that differences between the two groups do appear to exist. If we compare the average cumulative mean values for profitability ratios given by NITA and EBITTA throughout the period under study, the results presented clearly shows that healthy non-distressed firms are more profitable than their distressed PN4 counterparts. With NITA and EBITTA for healthy firms averaging 2.77% and 6.23%, the positive value of these ratios indicates that profits are generated throughout the period. For PN4 companies on the other hand, as its NITA and EBITTA ratios are -26.2% and -17.6% respectively, the negative value of these ratios suggests that these firms had suffered significant losses during the period. This 51 corresponds to 29% (NITA) and 24% (EBITTA) difference between the two groups. If we observe the mean values each year within our period of study as found in Table 1 of the Appendix, several observations can be made. Firstly, for each year under study, the average profitability of the healthy group is significantly higher than those of the PN4 group. Secondly, while the profitability of healthy firms is positive every year, as a whole, the PN4 groups suffer losses each and every year. Thirdly, the mean values for healthy firms do appear to be stable with little variation over time. For the PN4 group, although the mean values are consistently negative, its values do appear to be more erratic. All these factors do suggest that differences between the two groups do exist. To see if these differences are indeed significant or just a result of random errors, a t-test will be used to asses if the differences between the two groups are statistically significant. Based on the results of the t-test presented in the appendix, the t-values for NITA and EBITTA are 5.76 and 5.24.By comparing these values with those from the Critical Values of the t Distribution Table whose tabled t-value obtained is 1.645 given an alpha of 0.05 and 268 degrees of freedom, the larger value clearly suggests that these differences are significant. Given statistical convention, the large disparity between the two t-values indicates that the probability of the means being similar is small and unlikely due to the impact of random errors or fluctuations. This result is in-line with the 52 findings presented by Beaver (1966) and Ohlson (1980) both of whom found profitability as given by NITA to be a significant predictor of distress. 4.1.2 RATIO ANALYSIS & T-TEST: Cash Flow Ratios Table 4.1.2 : Cash Flow Ratio - Two sample t-test with equal variances Variable Group Observations Mean Std Dev T df Status cfsales 0 214 .0698125 .3363745 1.5917 260 1 48 -.0170025 .3639241 cftd 0 219 .1438238 .3597432 2.9745 268 1 51 -.0086326 .1360677 cfcl 0 219 .2269034 .5013146 3.8070 270 1 51 -.0513442 .2977745 Note: Non-failed group distinguished by status 0 and failed group distinguished by status 1 For Cash Flow ratios, again we find that differences in cumulative mean values between the two groups do appear to be significant. If we look at the average cumulative mean values for the Cash Flow ratios throughout the period under study, the largest differences in means is given by the CFCL and CFTD ratios. For CFCL, the cumulative difference in means between the two groups is a staggering 27.8%. For CFTD, the average difference is 15.3%. Looking at the means values in greater detail, the positive Cash Flow ratios of healthy firms does indicate that cash is generated from its operations. As cash is often how firms meet expected and unexpected obligations, a positive cash flow is an asset to the firms. As the mean values for CFCL and CFTD for distressed PN4 firms are -5.1% and -1%, the negative cumulative means suggests that should the need for cash arise, other sources of income (sometimes external sources) will be needed. 53 If we look at the movement of the mean values over time as given by Table 1 in the Appendix, the following observations can be made. Firstly, the mean Cash Flow ratios for PN4 group are lower than those from healthy firms each year under study. Secondly, the CFCL mean values for PN4 firms are consistently negative throughout the study period (almost every year for CFTD). As we can see, for the healthy group, the ratios are positive throughout with the mean value being stable with no noticeable upward or downward trend. Lastly, for PN4 group, the means are slightly more erratic with a tendency to get worse as distressed reclassification date looms. Given these factors, a statistical difference between groups does appear to exist. To confirm this, we again turn to the t-test. By comparing the t-value calculated for CFCL and CFTD with the tabled t-value found in the Critical Values of the t Distribution Table, the comparisons again verifies the differences to be significant. With calculated CFCL and CFTD t-values 3.80 and 2.97, the large disparity with the tabled t-value (t= 1.645) does suggest the differences are not due a random error. Therefore, statistically we can conclude that differences do exist, and these ratios can be a predictor of distress. While this will only be verified later through the use of a logit analysis, the findings are in accordance with those of Beaver (1966). 4.1.3 RATIO ANALYSIS & T-TEST: Liquidity ratios Table 4.1.3 : Liquidity Ratio - Two sample t-test with equal variances Variable Group Observations Mean Std Dev t df Status 54 wcsales 0 214 .0242214 1.078106 5.3526 260 1 48 -2.008213 5.099299 wcta 0 219 -.0067682 .6024754 5.3926 268 1 51 -.7115569 1.484854 cashta 0 219 .2222634 .509383 4.5434 270 1 51 -.1165082 .3185023 cata 0 219 .4343766 .231741 -.15960 268 1 51 .4941126 .2764944 cacl 0 219 1.978614 2.739189 2.0643 270 1 51 1.079808 3.052731 Note: Non-failed group distinguished by status 0 and failed group distinguished by status 1 Looking at the differences in cumulative liquidity ratios for both healthy and distressed firms, differences between the two groups does appear large and significant. If we observe the difference in cumulative means for WCTA, WCSALES, and CASHTA, the approximate difference in mean values are 70.5%, 203%, and 33.9%. Given the values of Working Capital either as a percentage of Total Assets or as a percentage of Sales, the differences between the two groups suggests that unlike most healthy firms, distressed firms suffer significantly from a lack of liquidity. With mean values of -.6% (WCTA), 2.4% (WCSALES), and 22.2% (CASHTA) for healthy firms, the amount of liquidity carried by these healthy firms are significantly greater than those of the PN4 group. Liquidity is important as it is the ability to meet unexpected needs for cash and to pay bills when they are due. Given the poor profitability and cash flow ratios of distressed firms, distressed firms are expected to have poor liquidity values. 55 As the value of these ratios for distressed firms is negative, distressed firms do not carry the enough liquid assets (as given by Current Assets) to cover short-term obligations (as given by Current Liabilities). Should an unexpected obligation or situation suddenly arise, the need for cash will require external sources of capital as the level of cash and near cash assets held will not be enough. If we look at the mean values over time as given by Table 1 in the Appendix, the following observations can be made. Firstly, while Working Capital (either as a percentage of Sales or of Total Assets) for both distressed and healthy firms were poor in 1998, the mean for the healthy group did improve with time. For distressed firms, Working Capital continually decreased over time. Secondly, looking at the percentage of cash held, those from the healthy group had positive cash holdings. As the mean value for thePN4 group is negative, this again suggests poor liquidity and a need for external credit in order for the firm to operate on a day to day basis. Finally, while the level of cash held by healthy firms as a proportion of total assets grew each year, the decline in cash position for PN4 companies does suggest a higher probability of default or distress. Given that the cash flow generated from operations is negative, this situation was to be expected. Given all these factors, we would therefore expect that the differences between the two groups are statistically significant. Based on the calculated t-value 56 presented in Table 4.1.3, comparison of the calculated value with those from the Critical Values of the t Distribution Table appears to confirm our suspicions. As the calculated t-values for WCTA, WCSALES, and CASHTA are 5.4, 5.36, and 4.54, comparing this with the tabled value of 1.645 proves that the probability of the means from the two groups being similar is small. Statistically speaking, the differences do appear significant and could be a good predictor of possible distress. Let us confirm this later through the use of a logit regression analysis. 4.1.4 RATIO ANALYSIS & T-TEST: Solvency Ratio Table 4.1.4 : Long-term Solvency Ratio - Two sample t-test with equal variances Variable Group Observations Mean Std Dev t df Status tdta 0 219 .7246381 1.157173 -3.7863 268 1 51 1.518509 1.977196 Note: Non-failed group distinguished by status 0 and failed group distinguished by status 1 While the liquidity ratios highlight the problems faced by distressed companies in the near future, its long term riskiness is often measured by examining its solvency ratios. Looking at the cumulative TDTA ratio throughout the period under study, we can see that PN4 firms are highly leveraged. With a cumulative TDTA ratio of 1.52 times, the significant use of long term debt does appear to be a strain for PN4 companies. As the use of Total Debt to Total Assets indicates the percentage of assets funded by debt, a ratio greater than 1 suggests that distressed firms have 57 overextended itself through its use of debt. As the larger use of long term debt is positively correlated to increasing interest and principal obligations, we do expect the firms to carry significantly higher financial risks. While this increases the possibility of interest or principal default, as our earlier analysis had highlighted negative cash flows and poor profitability records, perhaps the only way these firms can continue to operate is through increasing their borrowing activities. If we look at the behavior of the mean value over time as given by Table 1, while the ratio of Total Debt to Total Assets for the PN4 group in 1998 is 1.19, by 2001, this ratio had grown to 1.9 times. Unlike the mean for the PN4 group which grew increasingly worse as distress looms, the mean for healthy firms was relatively stable over the period. While the results do suggest that the differences between the two groups are significant, verifying the statement will require the use of a t-test. With a calculated t-value of -3.79 as seen in TABLE 4.1.4, the big difference with the tabled t-value of 1.645 indicates that the differences are statistically significant. 4.2 LOGIT ANALYSIS Based on the analysis done thus far, the results presented clearly suggest that given the different measures available, differences between the mean values of healthy and distressed firms can be significant. Much like the conclusions presented by Altman (1968) and Beaver (1966), this point proves that financial ratios can in fact be used to distinguish between healthy and potentially 58 unhealthy firms. While this can clearly be done through the use of a simple ratio and t-test analysis, perhaps we have gotten a little ahead of ourselves as we have yet to determine if these ratios do infact influence financial distress. The analytical tools presented earlier measures the significance of the differences amongst the healthy and distressed means. It however fails to evaluate the relationship these factors or ratios have with the dependent variable; group status. In order for us to evaluate the factors that can influence group status and understand the nature of their connection, the perfect tool available is through the use of a logit regression analysis. A logit regression analysis is a form of regression used when the dependent is a dichotomy and therefore can take one of two possible outcomes; distressed or non-distressed. Its goal therefore is to estimate or predict the outcome based on the value of variables of interest. Before we can conduct a logit analysis however, we must first begin the process by evaluating the variables of interest. By determining which factors should be included and which factors should be omitted from our logit models, doing so ensures that our model is robust. 4.2.1 LOGIT ANALYSIS: STREAMLINING THE MODEL In determining the variables of interest that should be included in our model, we will discriminate the variables on the basis of their correlation and collinearity. 59 Based on the data presented in Table 4.2, the Variance Inflation Factor (VIF) generated by Stata shows us how much the variance of the coefficient estimate is inflated by collinearity. Collinearity is the situation that arises when there is close to a near perfect relationship between some or all of the variables of interest. In practical terms, high collinearity indicates that variable redundancy or overlap does exist. Although this flaw is not fatal to our model, it can cause a loss in power as the analysis will suffer from the difficulties associated with disentangling the effects of the variables. This in turn makes model interpretation more difficult. Table 4.2: Multicollinearity Diagnostic MULTICOLLINEARITY DIAGNOSTICS SQRT R- Variable VIF VIF Tolerance Squared ------------------------------------------------------------ wcta 9.05 3.01 0.1105 0.8895 wcsales 4.58 2.14 0.2181 0.7819 cashta 3.61 1.90 0.2768 0.7232 nita 25.51 5.05 0.0392 0.9608 ebitta 24.38 4.94 0.0410 0.9590 cfcl 9.32 3.05 0.1073 0.8927 cftd 7.27 2.70 0.1376 0.8624 tdta 11.11 3.33 0.0900 0.9100 cacl 1.60 1.26 0.6258 0.3742 salesta 1.02 1.01 0.9852 0.0148 cata 2.23 1.49 0.4481 0.5519 cashcl 1.81 1.35 0.5515 0.4485 cfsales 1.93 1.39 0.5175 0.4825 ------------------------------------------------------------ Mean VIF 7.96 In assessing the collinearity amongst variables, this can be measured by the square root of VIF. This in essence tells us how much larger the standard error is compared to what it would be if the variables were uncorrelated to one another. For statisticians, though VIFs of 10 or more is a definite reason for 60 concern, values greater than 2.5 should also be noted and reviewed. Based on the data from Table 4.2, the following variables required further investigation: NITA and EBITTA, CFCL and CFTD, and CASHCL and CACL. Table 4.3: Coefficient Correlation Matrix COEFFICIENT CORRELATION MATRIX | wcta wcsales cashta nita ebitta cfcl cftd -------------+--------------------------------------------------------------------------------------------- wcta | 1.0000 wcsales | -0.2324 1.0000 cashta | 0.4768 0.1003 1.0000 nita | 0.1306 0.0650 0.4504 1.0000 ebitta | -0.0534 -0.2794 -0.4604 -0.8483 1.0000 cfcl | -0.0203 -0.0185 0.0212 0.2182 -0.1451 1.0000 cftd | 0.0281 0.0571 -0.0316 -0.2732 0.1032 -0.8367 1.0000 tdta | 0.3079 -0.1388 -0.5926 -0.3817 0.2658 -0.0578 0.1320 cacl | -0.6023 -0.1377 -0.6284 -0.1570 0.0879 0.1235 -0.1240 salesta | -0.0041 -0.0218 -0.1148 -0.1491 0.1204 -0.0797 0.1970 cata | -0.2417 -0.1199 0.0805 -0.0137 0.0885 0.0271 -0.0008 cashcl | -0.6127 -0.0958 -0.7578 -0.2528 0.1900 0.0858 -0.0836 cfsales | 0.0359 -0.0219 -0.0187 -0.0687 0.0767 -0.6479 0.2221 _cons | 0.4745 0.2526 0.8232 0.4894 -0.3500 0.0069 -0.0919 | tdta cacl salesta cata cashcl cfsales _cons -------------+--------------------------------------------------------------------------------------------- tdta | 1.0000 cacl | 0.2540 1.0000 salesta | 0.1181 -0.0259 1.0000 cata | -0.4779 -0.3178 0.0312 1.0000 cashcl | 0.3491 0.9568 0.0060 -0.2679 1.0000 cfsales | 0.0334 -0.0047 -0.0936 -0.1452 0.0041 1.0000 _cons | -0.5445 -0.7568 -0.1121 -0.0233 -0.8141 0.0482 1.0000 To investigate the collinearity of the variables, let us now look at the correlation matrix of the estimated coefficient (not the variables) as presented in Table 4.3. As high correlation coefficients between pairs indicate potential collinearity problems, those with high correlation amongst their coefficients will be taken out of the model. On the basis of this, the high correlation between EBITTA and NITA, and CFCL with CFTD results in us removing EBITTA and CFTD from our 61 logit Model. The low correlation between CASHCL and CACL on the other hand for now justifies their inclusion. 4.2.2 LOGIT ANALYSIS: MODEL 1 While the complete results of the logit analysis can be found in the appendix, the following table highlights the results of the logit analysis. Table 4.4: Logit Analysis of Model 1 MODEL 1 – ALL VARIABLES INCLUDED Logit estimates Number of obs = 262 LR chi2(11) = 149.91 Prob > chi2 = 0.0000 Log likelihood = -49.813862 Pseudo R2 = 0.6008 ------------------------------------------------------------------------------------------------------------------ var1 | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+--------------------------------------------------------------------------------------------------- wcta -3.757966 2.167542 -1.73 0.083 -8.00627 .4903389 wcsales .1517615 .229463 0.66 0.508 -.2979777 .6015007 cashta -13.14618 5.826237 -2.26 0.024 -24.56539 -1.726966 nita -4.493003 2.075364 -2.16 0.030 -8.560642 -.4253632 cfcl -2.501012 1.944906 -1.29 0.198 -6.312958 1.310934 tdta 1.580787 1.472999 1.07 0.283 -1.306237 4.467812 cacl -.945844 1.266378 -0.75 0.455 -3.427899 1.536211 salesta -.0006405 .003489 -0.18 0.854 -.0074788 .0061979 cata 4.181737 2.2948 1.82 0.068 -.3159889 8.679463 cashcl -.2390359 .9935026 -0.24 0.810 -2.186265 1.708193 cfsales 1.846088 1.506245 1.23 0.220 -1.106097 4.798274 _cons -3.59651 1.516353 -2.37 0.018 -6.568507 -.6245127 As we can see from the data presented in the table above, the result of the logit analysis on Model 1 takes into account all 11 variables initially identified from popular literature. Based on the result of 262 complete observations, the goodness-of-fit of the model as measured by Pseudo R2 of 0.6008 does indicate a well fitting model. 4 financial ratios were deemed to be significant predictors of corporate distress as given by their z-statistic. With a z- statistic 62 greater than 1.64 (signifying .05 acceptance level) the variables deemed statistically significant were; WCTA, CASHTA, NITA, CATA. So based on the variables or ratios deemed significant, we can therefore argue that liquidity (as given by WCTA, CASHTA and CATA), and profitability (as given by NITA) are important characteristics that can be used to differentiate and identify problematic firms. In analyzing the results found in the Table 4.4, we can either interpret the logit through coefficient value or through the odds ratio. For the logit model, the coefficients are calculated through the use of Maximum Likelihood Estimation (MLE) method as opposed the Original least Squares (OLS) method employed by linear regression. While OLS seeks to minimize the sum of squared distances between the data points and the regression line, MME seeks to maximize the log likelihood. This reflects how likely the observed values of the dependent variable can be predicted from the observable values of the independent variables. Furthermore, in OLS, the value of the coefficient refers to the change in the dependent variable as the independent variable changes. In MLE, the slope of the coefficient is the change in “log odds” of the dependent as the independent changes. So, from the coefficients presented above, we can see that for one unit decrease in WCTA, CASHTA and NITA, the log odds of the firm being reclassified from healthy to distressed or PN4 increases by 3.76, 13.15, and 4.5. For a unit decrease in CATA, the results state that the log odds of a firm being reclassified decreases by 4.18. 63 Table 4.5: Logit Analysis Model 1 - Odds ------------------------------------------------------------------------------------------------ var1 | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+--------------------------------------------------------------------------------- wcta .0233312 .0505713 -1.73 0.083 .0003334 1.63287 wcsales 1.163883 .267068 0.66 0.508 .7423179 1.824855 cashta 1.95e-06 .0000114 -2.26 0.024 2.14e-11 .1778232 nita .011187 .0232171 -2.16 0.030 .0001915 .6535324 cfcl .082002 .1594861 -1.29 0.198 .0018127 3.709638 tdta 4.85878 7.156976 1.07 0.283 .2708373 87.16577 cacl .3883517 .4917999 -0.75 0.455 .0324551 4.646948 salesta .9993597 .0034868 -0.18 0.854 .9925491 1.006217 cata 65.47949 150.2624 1.82 0.068 .7290675 5880.887 cashcl .7873866 .7822706 -0.24 0.810 .1123355 5.518982 cfsales 6.334991 9.542047 1.23 0.220 .3308477 121.3009 Since the measure and interpretation of coefficients can sometimes be confusing, a more natural interpretation of the logit results can be achieved through the use of Stata’s Odds Ratio. This can be found in Table 4.5 above. Odds ratio is the probability of the event occurring divided by the probability of the non-event. In this case, the odds for WCTA, CASHTA, and NITA are 0.023, 1.95e-06, 0.011 respectively. What this means is that for every unit decrease in the variable, the odds of distress occurring increases by the value for the odds ratio. For CATA, the odds ratio is approximately 65.48. So, for every unit increase in CATA, our logit analysis argues that the odds of distress occurring is 64 times more likely. In looking at comparisons between the variables this study found to be significant with those previously conducted, the predictive abilities of liquidity and profitability ratios is inline with our expectations. Deakin (1972) found that 64 liquidity as proxied by WCTA was the best predictor of potential distress reclassification both in the near-term (1 year prior to event) and in the long-term (5 years prior to event). Although Altman (1968) found WCTA to be the least predictive of the variables under his study, Deakin’s conclusion is verified by Begley (1996). In his study, Begley reestimated Altman’s original Z model through the use of more current data. For profitability ratios, NITA was deemed the second best predictor by Beaver (1966). Able to accurately predict potential distress 87% of the time 1 year prior to the event and 72% of the time 5 years before, the ability to generate profits appears to be significant in ensuring a firms’ continued survival. Both these results appear to refute Shirata’s (1998) findings that liquidity and profitability ratios have no significant importance in determining firm health. As we look at the analysis just presented, the negative coefficient values observed for WCTA, CASHTA, and NITA indicates an inverse relationship with the dependent variable. The value of CATA coefficient indicates a positive relationship with distress. While the negative relationship is to be expected given the greater likelihood of distress as the value of these ratios deteriorates, the positive relationship between CATA and distress is troubling. The reason is that a positive coefficient indicates that as the proportion of current assets as a percentage of total assets increase so does the likelihood of a firm being reclassified as a PN4 company. Although this finding does indicate an error in either theory or in application, no reason for this finding will be offered. 65 Another interesting finding that we can observe based on the results of the logit analysis is the exclusion of the ratio of total debt to total asset (TDTA). As the study by Deakin (1972) found that long-term solvency ratios as proxied by TDTA could significantly predict potential distress up to five years prior to bankruptcy filing, debt ratio was expected to be a significant predictor. In Deakin’s study, debt ratio could be used to accurately classify PN4 firms 67% of the time 5 years before the event26. This was verified by Ohlson (1980). But, as we can see based on our results, long-term solvency is not a good predictor of distress for Industrial firms in Malaysia. Perhaps Begley (1996) was right. As investors and firms in the 1980s and 1990s had grown more comfortable with the increasing use of leverage as given by debt levels, perhaps this explains the decreasing significance of the debt ratio. While the exclusion of long-term solvency measures as significant influencers of distress were already stumbled upon by earlier researchers, the insignificance of any cash flow measure is really surprising. While earlier studies by Altman (1968) argued for the inclusion of these ratios, based on the results obtained, perhaps the role of cash is not as important as we initially expected for Industrial firms in Malaysia. In his original study, Altman determined that CFTD would accurately predict PN4 reclassification with 87% one year before the event and with 78% accuracy up to 5 years before the event. 26 Altman (1968) – TDTA can predict PN4 reclassification 72% of the time 5 years prior to event. 66 This is significant. As most financial obligation are paid or met through the use of cash, its exclusion implies that the unavailability of internal cash to meet financial obligations will require firms to raise cash from external sources. Given the high degree of short-term and long-term credit carried by firms within our study, perhaps this explains why financial leverage and cash levels are not significant factors in influencing distress. Table 4.6: Model 1 Predictive Accuracy Logistic model for var1 number of observations = 262 area under ROC curve = 0.9561 Classification Rate -------- True -------- Classified | D ~D | Total -----------+--------------------------+----------- + | 32 7 | 39 - | 16 207 | 223 -----------+--------------------------+----------- Total | 48 214 | 262 Classified + if predicted Pr(D) >= .5 True D defined as var1 != 0 -------------------------------------------------- Sensitivity Pr( +| D) 66.67% Specificity Pr( -|~D) 96.73% Positive predictive value Pr( D| +) 82.05% Negative predictive value Pr(~D| -) 92.83% -------------------------------------------------- False + rate for true ~D Pr( +|~D) 3.27% False - rate for true D Pr( -| D) 33.33% False + rate for classified + Pr(~D| +) 17.95% False - rate for classified - Pr( D| -) 7.17% -------------------------------------------------- Correctly classified 91.22% -------------------------------------------------- For Model 1, the area under the under Receiver Operating Characteristic (ROC) curve is .9576. Area under the curve can be used as a measure to test 67 the accuracy of the model in separating and classifying healthy and distressed firms. With area .9576, statistical convention deems this an excellent discriminator of group membership. Furthermore, as Model 1 can accurately classify firm group membership 91.22% of the time, the high classification accuracy indicates that this model is robust enough to be applied to Industrial firms listed on the Main Board of Bursa Malaysia. 4.2.3 LOGIT ANALYSIS: MODEL 2 Table 4.7: Logit Analysis Model 2 MODEL 2 – STEPWISE REGRESSION Logit estimates Number of obs = 262 LR chi2(4) = 144.25 Prob > chi2 = 0.0000 Log likelihood = -52.645412 Pseudo R2 = 0.5781 ------------------------------------------------------------------------------ var1 | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- wcta -6.178122 1.109093 -5.57 0.000 -8.351904 -4.00434 cata 5.904799 1.72847 3.42 0.001 2.517061 9.292537 cashta -12.70224 3.422159 -3.71 0.000 -19.40955 -5.994934 nita -4.537302 1.658816 -2.74 0.006 -7.78852 -1.286083 cons -4.406335 .777508 -5.67 0.000 -5.930222 -2.882447 Beginning with 11 variables of interest, Model 2 uses Stepwise regression with a p-value equal to .05 to automatically determine which variables should be added or dropped from the model. Although the procedure runs the risk of modeling the noise in the data, it is still considered useful particularly for exploratory purposes. As our study into the factors influencing financial distress lack a theoretical underpinning to guide research, stepwise regression allows us to explore possible relationships. 68 The results presented indicate that based on the stepwise procedure, factors deemed significant predictors of distress centers mainly around liquidity (as given by WCTA, CATA, and CASHTA) and profitability measures (as given by NITA). As similar findings were found in our analysis of Model 1, reanalyzing the results appear redundant. Interestingly, positive coefficient value for CATA is consistent with our earlier finding. As further investigation of this is beyond the scope of this paper, let us just note that the positive correlation between current asset level and distress does appear to defy both logic and convention. The goodness-of-fit of this model as measured by Pseudo R2 of 0.5781 indicates a respectable fitting model. Table 4.8: Model 2 Predictive Accuracy Logistic model for var1 number of observations = 262 area under ROC curve = 0.9479 Classification Rate Logistic model for var1 -------- True -------- Classified | D ~D | Total -----------+--------------------------+----------- + | 32 6 | 38 - | 16 208 | 224 -----------+--------------------------+----------- Total | 48 214 | 262 Classified + if predicted Pr(D) >= .5 True D defined as var1 != 0 -------------------------------------------------- Sensitivity Pr( +| D) 66.67% Specificity Pr( -|~D) 97.20% Positive predictive value Pr( D| +) 84.21% Negative predictive value Pr(~D| -) 92.86% -------------------------------------------------- False + rate for true ~D Pr( +|~D) 2.80% False - rate for true D Pr( -| D) 33.33% 69 False + rate for classified + Pr(~D| +) 15.79% False - rate for classified - Pr( D| -) 7.14% -------------------------------------------------- Correctly classified 91.60% For Model 2, the area under the under Receiver Operating Characteristic (ROC) curve is .9479. Area under the curve can be used as a measure to test the accuracy of the model in separating and classifying healthy and distressed firms. With area .9479, statistical convention deems this an excellent discriminator of group membership. Based on the data from the table above, we see that this model can correctly classify group membership accurately 91.6% of the time. 70 5.0 CONCLUSION 5.1 SUMMARY Since the pioneering works of Beaver (1966) and Altman (1968), a lot of research has been conducted to try and understand corporate bankruptcy. Despite the large volume of research on the topic however, findings from the studies have proven to be inconsistent and inconclusive. Furthermore, as most research had focused their studies on understanding and predicting corporate bankruptcy, little effort has been paid to understanding corporate financial distress. For researchers, a common reason often cited to explain this is due to the unavailability of any consistent measure of what defines corporate distress. For those of us in Malaysia, due to the availability of Bursa Malaysia’s Practice Note 4/2001 pursuant to paragraph 8.14C(2) of the Listing Requirements, it addresses the issue of a lack of legal definition for financial distress. By having a legal definition of distress, our study benefits as we can consistently define what it is that constitutes financial distress. As past research have found that financial distress can precede failure by up to 7 years, identifying the problems early through the help of financial ratios can help firms to avoid costly bankruptcy and restructuring efforts. As we look back to the results of the research, financial ratios can help us identify potential problems. By comparing the mean values of healthy firms with distressed firms, differences between the two do appear to exist. Through the 71 use of a t-test comparison of group mean values, these differences can be statistically significant and not just down to random errors. This finding is valuable because if we can correctly identify the factors or characteristics that differentiate between healthy and distressed, financial ratios can be used to predict troubled firms. Therefore we fail to reject the null hypothesis that differences between the means of healthy and distressed firms are similar. As we review back the results of the logit analysis, the factors that differentiate healthy and distressed firms mainly centers on their levels of liquidity and profitability. Proxied by the ratios of WCTA, CATA, and CASHTA, low liquidity levels was found to be detrimental to the continued operations of a firm as the level of liquid assets available to the firm might be insufficient should an unexpected need arise. As liquidity ratios measure a firms’ ability to meet its short-term obligations given its level of sellable assets, high liquidity levels provide the firms with a greater cushion to absorb unexpected events. Another important characteristic found to differentiate between healthy and potentially distressed firms is given by their levels of profitability. Firms with better profitability are often seen as being better managed. By productively utilizing a firm’s assets and resources, the profits generated by these efforts are often seen as an indication of firm performance. Based on our results, it is therefore proven that performance and firm health are positively correlated. In looking at the factors excluded from our model, it is interesting that cash flow generation and debt levels are insignificant influencers of firm status. As Beaver 72 (1966) had found cash flow ratios to be a good predictor of potential problems, we had expected cash flows to be an important characteristic. Perhaps, this also explains why debt structure is also deemed insignificant. While the studies conducted in the 1960s and 1970s determined debt level to be positively related to distress, perhaps the greater acceptance of leverage in the 1980s and 1990s has allowed firms to reduce their dependence on internal cash. 73 APPENDIX 7 – PN4 CRITERIA Pursuant to paragraph 8.14C(2) of the Listing Requirements, the Exchange prescribes the following criteria (hereinafter referred to as the “Prescribed Criteria”), the fulfillment of one or more of which will require a listed issuer to comply with the provisions of paragraph 8.14C and this Practice Note :- (a) the shareholders’ equity of the listed issuer on a consolidated basis is equal to or less than 25% of the issued and paid-up capital of the listed issuer and such shareholders’ equity is less than the minimum issued and paid-up capital as required under paragraph 8.16A(1) of the Listing Requirements; (b) receivers and/or managers have been appointed over the asset of the listed issuer, its subsidiary or associated company which asset accounts for at least 50% of the total assets employed of the listed issuer on a consolidated basis; (c) a winding up of a listed issuer’s subsidiary or associated company which accounts for at least 50% of the total assets employed of the listed issuer on a consolidated basis; (d) the auditors have expressed an adverse or disclaimer opinion in the listed issuer’s latest audited accounts; (e) the auditors have expressed a modified opinion with emphasis on the listed issuer’s going concern in the listed issuer’s latest audited accounts and the shareholders’ equity of the listed issuer on a consolidated basis is equal to or less than 50% of the issued and paid-up capital of the listed issuer; (f) a default in payment by a listed issuer, its major subsidiary or major associated company, as the case may be, as announced by a listed issuer pursuant to Practice Note No 1/2001 and the listed issuer is unable to provide a solvency declaration to the Exchange. (g) the listed issuer has suspended or ceased:- (i) all of its business or its major business; or (ii) its entire or major operations, for any reasons whatsoever including, amongst others, due to or as a result of:- (aa) the cancellation, loss or non-renewal of a licence, concession or such other rights necessary to conduct its business activities; (bb) the disposal of the listed issuer's business or major business; or (cc) a court order or judgment obtained against the listed issuer prohibiting the listed issuer from conducting its major operations on grounds of infringement of copyright of products etc; or (h) the listed issuer has an insignificant business or operations. 74 Appendix 3: Independent Samples Test 1998 Levene's Test for Equality of Variances t-test for Equality of Means Std. Sig. Mean Error 95% Confidence (2-taile Differe Differe Interval of the F Sig. t df d) nce nce Difference Lower Upper Lower Upper Lower Upper Lower Upper Lower Equal variances assumed NITA 7.288 .009 3.085 66 .003 .46086 .14937 .16263 .75910 Equal variances not assumed 1.785 12.676 .098 .46086 .25821 -.09842 1.02014 Equal variances assumed -.4482 TDTA .004 .953 -1.093 66 .278 .41000 -1.26687 .37032 7 Equal variances not -.4482 assumed -1.277 22.458 .215 .35096 -1.17526 .27871 7 Equal variances assumed WCTA .939 .336 1.258 66 .213 .33940 .26982 -.19931 .87811 Equal variances not assumed 1.047 15.275 .311 .33940 .32424 -.35062 1.02942 Equal variances assumed -.5607 CACL 5.424 .023 -.569 66 .571 .98582 -2.52900 1.40751 5 Equal variances not -.5607 1.6819 assumed -.333 12.732 .744 -4.20222 3.08073 5 8 Equal variances assumed EBITTA 8.331 .005 3.061 66 .003 .44673 .14593 .15537 .73810 Equal variances not assumed 1.704 12.519 .113 .44673 .26211 -.12174 1.01520 Equal variances assumed SALESTA 2.255 .138 1.140 65 .258 .32079 .28131 -.24103 .88261 Equal variances not assumed 1.813 47.358 .076 .32079 .17693 -.03508 .67666 Equal variances assumed -.1037 CATA .589 .446 -1.472 66 .146 .07050 -.24455 .03697 9 Equal variances not -.1037 assumed -1.352 16.596 .194 .07674 -.26600 .05842 9 Equal variances assumed CASHTA .048 .828 2.356 66 .021 .28447 .12076 .04336 .52557 Equal variances not assumed 2.511 19.602 .021 .28447 .11329 .04785 .52109 Equal variances assumed 3.0817 1.2850 CASHCL 11.073 .001 2.398 66 .019 .51610 5.64740 5 3 Equal variances not 3.0817 2.4316 assumed 1.267 12.329 .228 -2.20070 8.36419 5 3 Equal variances assumed WCSALES 8.499 .005 .176 65 .861 .07601 .43291 -.78857 .94059 Equal variances not assumed .094 12.352 .927 .07601 .80919 -1.68151 1.83352 Equal variances assumed CFTD 1.749 .191 1.946 66 .056 .15024 .07720 -.00389 .30436 Equal variances not assumed 2.161 20.751 .043 .15024 .06952 .00556 .29492 Equal variances assumed CFCL .516 .475 2.237 66 .029 .26863 .12006 .02893 .50834 Equal variances not assumed 2.282 18.539 .035 .26863 .11772 .02182 .51545 Equal variances assumed -.0360 CFSALES .742 .392 -.238 65 .813 .15124 -.33807 .26603 2 Equal variances not -.0360 assumed -.224 17.089 .825 .16053 -.37457 .30253 2 75 Appendix 4: Independent Samples Test 1999 Levene's Test for Equality of Variances t-test for Equality of Means Std. Sig. Mean Error 95% Confidence (2-taile Differe Differe Interval of the F Sig. t df d) nce nce Difference Lower Upper Lower Upper Lower Upper Lower Upper Lower Equal variances assumed NITA .660 .420 2.976 66 .004 .23695 .07962 .07798 .39591 Equal variances not assumed 2.657 16.173 .017 .23695 .08917 .04808 .42581 Equal variances assumed -.3993 -1.0042 TDTA .699 .406 -1.318 66 .192 .30297 .20554 6 7 Equal variances not -.3993 assumed -2.111 47.424 .040 .18922 -.77994 -.01879 6 Equal variances assumed WCTA .015 .902 2.171 66 .034 .35364 .16287 .02846 .67883 Equal variances not assumed 2.559 22.792 .018 .35364 .13819 .06763 .63966 Equal variances assumed 1.2451 CACL 3.865 .054 1.799 66 .077 .69203 -.13650 2.62688 9 Equal variances not 1.2451 assumed 3.537 63.338 .001 .35209 .54166 1.94872 9 Equal variances assumed EBITTA 1.242 .269 3.216 66 .002 .23799 .07399 .09026 .38572 Equal variances not assumed 2.676 15.272 .017 .23799 .08894 .04871 .42727 Equal variances assumed 10.115 20.946 -31.729 SALESTA .882 .351 .483 64 .631 51.95952 05 01 42 Equal variances not 10.115 9.8156 -9.5723 assumed 1.031 53.023 .307 29.80248 05 2 8 Equal variances assumed -.0541 CATA .002 .965 -.731 66 .467 .07402 -.20191 .09367 2 Equal variances not -.0541 assumed -.721 17.840 .480 .07505 -.21190 .10365 2 Equal variances assumed CASHTA .240 .626 2.337 66 .022 .32617 .13956 .04754 .60480 Equal variances not assumed 3.145 29.639 .004 .32617 .10373 .11423 .53812 Equal variances assumed 1.2255 CASHCL .063 .802 1.849 66 .069 .66274 -.09761 2.54878 8 Equal variances not 1.2255 assumed 2.516 30.408 .017 .48719 .23116 2.22001 8 Equal variances assumed 1.0740 WCSALES 2.860 .096 3.797 64 .000 .28289 .50893 1.63921 7 Equal variances not 1.0740 assumed 3.095 13.682 .008 .34701 .32818 1.81996 7 Equal variances assumed CFTD 1.939 .168 1.415 66 .162 .19248 .13601 -.07908 .46404 Equal variances not assumed 2.705 65.673 .009 .19248 .07116 .05038 .33458 Equal variances assumed CFCL .776 .382 1.771 66 .081 .30383 .17160 -.03878 .64644 Equal variances not assumed 2.463 32.102 .019 .30383 .12335 .05260 .55506 Equal variances assumed CFSALES 2.000 .162 2.159 64 .035 .19658 .09104 .01470 .37846 Equal variances not assumed 1.538 12.660 .149 .19658 .12784 -.08035 .47350 76 Appendix 5: Independent Samples Test 2000 Levene's Test for Equality of Variances t-test for Equality of Means Std. Sig. Mean Error 95% Confidence (2-taile Differe Differe Interval of the F Sig. t df d) nce nce Difference Lower Upper Lower Upper Lower Upper Lower Upper Lower Equal variances assumed NITA .198 .658 3.347 65 .001 .18492 .05525 .07458 .29525 Equal variances not assumed 3.195 17.321 .005 .18492 .05788 .06298 .30685 Equal variances assumed -1.182 -2.214 TDTA 6.476 .013 -2.291 65 .025 .51640 -.15152 84 16 Equal variances not -1.182 -3.104 assumed -1.333 12.689 .206 .88717 .73856 84 24 Equal variances assumed WCTA 9.687 .003 3.045 65 .003 .83479 .27417 .28725 1.38234 Equal variances not -.2119 assumed 1.729 12.582 .108 .83479 .48288 1.88152 3 Equal variances assumed 1.2548 -.2255 CACL 2.470 .121 1.693 65 .095 .74122 2.73515 2 1 Equal variances not 1.2548 assumed 3.247 63.827 .002 .38650 .48265 2.02699 2 Equal variances assumed EBITTA 1.131 .292 2.038 65 .046 .10964 .05380 .00218 .21709 Equal variances not assumed 3.343 51.837 .002 .10964 .03279 .04383 .17545 Equal variances assumed 18.575 54.990 -91.31 128.4646 SALESTA .487 .488 .338 63 .737 73 09 321 7 Equal variances not 18.575 27.058 -35.56 assumed .686 59.322 .495 72.71385 73 67 239 Equal variances assumed -.0738 -.2263 CATA .848 .360 -.966 65 .338 .07640 .07877 1 9 Equal variances not -.0738 -.2546 assumed -.864 16.264 .400 .08540 .10700 1 2 Equal variances assumed CASHTA .603 .440 2.112 65 .039 .35711 .16912 .01937 .69486 Equal variances not assumed 3.210 41.568 .003 .35711 .11124 .13256 .58167 Equal variances assumed 1.1477 -.3085 CASHCL .977 .327 1.574 65 .120 .72918 2.60398 1 6 Equal variances not 1.1477 assumed 2.785 62.006 .007 .41215 .32383 1.97159 1 Equal variances assumed 2.8094 1.0368 WCSALES 16.043 .000 3.167 63 .002 .88701 4.58199 4 9 Equal variances not 2.8094 1.7805 -1.101 assumed 1.578 11.178 .142 6.72072 4 1 84 Equal variances assumed -.0808 CFTD 4.405 .040 1.129 65 .263 .10508 .09309 .29100 3 Equal variances not -.0019 assumed 1.964 60.067 .054 .10508 .05351 .21211 4 Equal variances assumed -.0394 CFCL 4.894 .030 1.695 65 .095 .22133 .13056 .48207 2 Equal variances not assumed 2.650 45.080 .011 .22133 .08351 .05313 .38952 Equal variances assumed -.0393 CFSALES .001 .982 1.417 63 .161 .09588 .06766 .23110 3 Equal variances not -.0524 assumed 1.372 15.821 .189 .09588 .06990 .24420 4 77 Appendix 6: Independent Samples Test 2001 Levene's Test for Equality of Variances t-test for Equality of Means Std. Sig. Mean Error 95% Confidence (2-taile Differe Differe Interval of the F Sig. t df d) nce nce Difference Lower Upper Lower Upper Lower Upper Lower Upper Lower Equal variances assumed NITA 2.132 .149 3.121 65 .003 .27940 .08952 .10062 .45818 Equal variances not assumed 3.314 17.285 .004 .27940 .08432 .10172 .45707 Equal variances assumed -1.174 -2.031 TDTA 4.506 .038 -2.734 65 .008 .42936 -.31658 07 56 Equal variances not -1.174 -2.583 assumed -1.811 12.243 .095 .64834 .23543 07 57 Equal variances assumed 1.3381 WCTA 18.801 .000 4.313 65 .000 .31021 .71856 1.95764 0 Equal variances not 1.3381 -.0278 assumed 2.151 11.214 .054 .62205 2.70405 0 5 Equal variances assumed 1.6970 1.0352 -.3705 CACL 2.376 .128 1.639 65 .106 3.76460 4 6 3 Equal variances not 1.6970 assumed 3.357 62.876 .001 .50553 .68678 2.70729 4 Equal variances assumed EBITTA .775 .382 3.157 65 .002 .15694 .04971 .05767 .25621 Equal variances not assumed 3.274 16.816 .005 .15694 .04794 .05571 .25817 Equal variances assumed -.1955 SALESTA 2.515 .118 1.223 63 .226 .30867 .25234 .81292 9 Equal variances not -.0847 assumed 1.630 21.378 .118 .30867 .18940 .70212 8 Equal variances assumed -.0015 -.1629 CATA 4.634 .035 -.019 65 .985 .08082 .15986 6 7 Equal variances not -.0015 -.2226 assumed -.015 13.357 .988 .10260 .21949 6 1 Equal variances assumed CASHTA .053 .819 2.302 65 .025 .38976 .16933 .05158 .72793 Equal variances not assumed 2.944 22.915 .007 .38976 .13238 .11586 .66366 Equal variances assumed 1.9174 1.0330 -.1456 CASHCL .149 .701 1.856 65 .068 3.98060 7 4 6 Equal variances not 1.9174 assumed 2.518 25.818 .018 .76136 .35193 3.48301 7 Equal variances assumed 4.5447 1.0872 2.3713 WCSALES 22.519 .000 4.180 62 .000 6.71801 0 1 9 Equal variances not 4.5447 2.2519 -.4618 assumed 2.018 10.165 .071 9.55126 0 3 6 Equal variances assumed -.0315 CFTD 5.225 .026 1.673 65 .099 .16285 .09732 .35721 1 Equal variances not assumed 3.159 62.626 .002 .16285 .05154 .05984 .26586 Equal variances assumed -.0048 CFCL 3.944 .051 1.968 65 .053 .32101 .16315 .64685 3 Equal variances not assumed 2.960 33.239 .006 .32101 .10844 .10046 .54156 Equal variances assumed -.1145 CFSALES .560 .457 .923 62 .360 .09818 .10641 .31088 3 Equal variances not -.0272 assumed 1.580 41.966 .122 .09818 .06213 .22356 0 78 Appendix 2: SUMMARY STATISTICS Table 4.1.1: Profitability Ratio - Two sample t-test with equal variances Variable Group Observations Mean Std Dev t df Status nita 0 219 .0276142 .2623657 5.7604 268 1 51 -.2618334 .50961.38 ebitta 0 219 .0623366 .2177557 5.2390 268 1 51 -.1763142 .5033522 salesta 0 215 9.694038 100.3764 0.4987 261 1 48 2.444032 13.27714 Note: Non-failed group distinguished by status 0 and failed group distinguished by status 1 Table 4.1.2: Cash Flow Ratio - Two sample t-test with equal variances Variable Group Observations Mean Std Dev T df Status cfsales 0 214 .0698125 .3363745 1.5917 260 1 48 -.0170025 .3639241 cftd 0 219 .1438238 .3597432 2.9745 268 1 51 -.0086326 .1360677 cfcl 0 219 .2269034 .5013146 3.8070 270 1 51 -.0513442 .2977745 Note: Non-failed group distinguished by status 0 and failed group distinguished by status 1 79 Table 4.1.3: Liquidity Ratio - Two sample t-test with equal variances Variable Group Observations Mean Std Dev t df Status wcsales 0 214 .0242214 1.078106 5.3526 260 1 48 -2.008213 5.099299 wcta 0 219 -.0067682 .6024754 5.3926 268 1 51 -.7115569 1.484854 cashta 0 219 .2222634 .509383 4.5434 270 1 51 -.1165082 .3185023 cata 0 219 .4343766 .231741 -.15960 268 1 51 .4941126 .2764944 cacl 0 219 1.978614 2.739189 2.0643 270 1 51 1.079808 3.052731 Note: Non-failed group distinguished by status 0 and failed group distinguished by status 1 Table 4.1.4: Long-term Solvency Ratio - Two sample t-test with equal variances Variable Group Observations Mean Std Dev t df Status tdta 0 219 .7246381 1.157173 -3.7863 268 1 51 1.518509 1.977196 Note: Non-failed group distinguished by status 0 and failed group distinguished by status 1 80 Appendix 12: SAMPLE POPULATION – INDUSTRIAL PRODUCTS – MAIN BOARD BURSA MALAYSIA27 COMPANY - TICKER Status ASB - 1481 0 ALCOM - 2674 0 AISB - 2682 0 ANCOM - 4758 0 ANJOO - 6556 0 CIHLDG - 2828 0 CAMERLN - 3751 0 CIMA - 2844 0 CCM - 2879 0 DELLOYD - 6505 0 DIJAENT - 5401 0 DRB - 1619 0 ESSO - 3042 0 FACBIND - 2984 0 FCW - 2755 0 GBH - 3611 0 GOPENG - 2135 0 GUH - 3247 0 HEXZA - 3298 0 HUMEIND - 3328 0 ICP - 6829 0 JTIASA - 4383 0 KSENG - 3476 0 KIALIM - 6211 0 KRAMAT - 2151 0 KYM - 8362 0 LEADER - 4529 0 LIONCORP - 3581 1 MUIIND - 3891 0 MAICA - 3743 0 MSC - 5916 0 MPI - 3867 0 MENTIGA - 5223 1 METROD - 6149 0 MIECO - 5001 0 MINHO - 5576 0 MUDA - 3883 0 NYLEX - 4944 0 PMI - 4103 0 PETGAS - 6033 0 PNEPCB - 6637 0 SAPURA - 7811 0 SCIENTX - 4731 0 SEAL - 4286 0 SHELL - 4324 0 27 STATUS DENOTES HEALTHY (0) OR DISTRESSES/ PN4 (1) 81 SINORA - 6262 0 SITATT - 4359 0 SAB - 5134 0 SUBUR - 6904 0 SSTEEL - 5665 0 TASEK - 4448 0 TENGARA - 8257 0 UAC - 4537 0 UNISEM - 5005 0 VS - 6963 0 WEMBLEY - 5428 1 WTK - 4243 0 YTLCMT - 8737 0 JAVA - 2747 1 DOLMITE - 5835 1 AMSTEEL 1 CHG 1 CASH 1 TONGKAH 1 WING TEIK HLDG 1 RNC CORPORATION 1 FORESWOOD GROUP 1 TRU-TECH HOLDINGS 1 82 Appendix 8: t-TEST Results Current Assets to Current Liabilities Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 1.978614 .1850973 2.739189 1.613805 2.343423 1| 51 1.079808 .4274679 3.052731 .2212134 1.938402 ---------+-------------------------------------------------------------------- combined | 270 1.80884 .171454 2.817276 1.471277 2.146402 ---------+-------------------------------------------------------------------- diff | .8988063 .4353988 .0415692 1.756043 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 2.0643 t = 2.0643 t = 2.0643 P < t = 0.9800 P > |t| = 0.0399 P > t = 0.0200 Net Income to Total Assets Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 .0276142 .017729 .2623657 -.007328 .0625565 1| 51 -.2618334 .0713602 .5096138 -.4051646 -.1185022 ---------+-------------------------------------------------------------------- combined | 270 -.0270592 .0208114 .3419668 -.0680332 .0139148 ---------+-------------------------------------------------------------------- diff | .2894476 .0502482 .1905162 .3883791 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 5.7604 t = 5.7604 t = 5.7604 P < t = 1.0000 P > |t| = 0.0000 P > t = 0.0000 83 Appendix 8: t-TEST Results Total Debt to total Assets Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 .7246381 .0781945 1.157173 .5705241 .8787521 1| 51 1.518509 .2768628 1.977196 .962414 2.074604 ---------+-------------------------------------------------------------------- combined | 270 .8745916 .0840796 1.381568 .7090539 1.040129 ---------+-------------------------------------------------------------------- diff | -.7938711 .2096719 -1.206685 -.3810574 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = -3.7863 t = -3.7863 t = -3.7863 P < t = 0.0001 P > |t| = 0.0002 P > t = 0.9999 Working Capital to Total Assets Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 -.0067682 .0407115 .6024754 -.0870068 .0734703 1| 51 -.7115569 .2079211 1.484854 -1.129179 -.2939351 ---------+-------------------------------------------------------------------- combined | 270 -.139895 .0537607 .8833789 -.2457403 -.0340497 ---------+-------------------------------------------------------------------- diff | .7047887 .1306956 .4474679 .9621095 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 5.3926 t = 5.3926 t = 5.3926 P < t = 1.0000 P > |t| = 0.0000 P > t = 0.0000 84 Appendix 8: t-TEST Results Current assets to Current Liabilities Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 1.978614 .1850973 2.739189 1.613805 2.343423 1| 51 1.079808 .4274679 3.052731 .2212134 1.938402 ---------+-------------------------------------------------------------------- combined | 270 1.80884 .171454 2.817276 1.471277 2.146402 ---------+-------------------------------------------------------------------- diff | .8988063 .4353988 .0415692 1.756043 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 2.0643 t = 2.0643 t = 2.0643 P < t = 0.9800 P > |t| = 0.0399 P > t = 0.0200 Earning Before Interest and Taxes (EBIT) to Total Assets Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 .0623366 .0147146 .2177557 .0333356 .0913377 1| 51 -.1763142 .070482 .5033422 -.3178814 -.0347469 ---------+-------------------------------------------------------------------- combined | 270 .0172581 .0186863 .3070467 -.0195318 .0540481 ---------+-------------------------------------------------------------------- diff | .2386508 .0455528 .1489639 .3283378 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 5.2390 t = 5.2390 t = 5.2390 P < t = 1.0000 P > |t| = 0.0000 P > t = 0.0000 85 Appendix 8: t-TEST Results Sales to Total Assets Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 215 9.694038 6.845613 100.3764 -3.799426 23.1875 1| 48 2.444032 1.91639 13.27714 -1.411248 6.299311 ---------+-------------------------------------------------------------------- combined | 263 8.370843 5.607245 90.93424 -2.670158 19.41184 ---------+-------------------------------------------------------------------- diff | 7.250006 14.53748 -21.37566 35.87567 ------------------------------------------------------------------------------ Degrees of freedom: 261 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 0.4987 t = 0.4987 t = 0.4987 P < t = 0.6908 P > |t| = 0.6184 P > t = 0.3092 Current Assets to Total assets Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 .4343766 .0156596 .231741 .403513 .4652402 1| 51 .4941126 .038717 .2764944 .4163473 .5718779 ---------+-------------------------------------------------------------------- combined | 270 .4456601 .014692 .241414 .4167342 .474586 ---------+-------------------------------------------------------------------- diff | -.059736 .0374276 -.1334254 .0139535 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = -1.5960 t = -1.5960 t = -1.5960 P < t = 0.0558 P > |t| = 0.1117 P > t = 0.9442 86 Appendix 8: t-TEST Results Cash to Total Assets Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 .2222634 .0344209 .509383 .154423 .2901038 1| 51 -.1165082 .0445992 .3185023 -.2060884 -.026928 ---------+-------------------------------------------------------------------- combined | 270 .1582732 .0302326 .4967716 .0987507 .2177957 ---------+-------------------------------------------------------------------- diff | .3387716 .0745637 .1919664 .4855768 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 4.5434 t = 4.5434 t = 4.5434 P < t = 1.0000 P > |t| = 0.0000 P > t = 0.0000 Cash to Current Liabilities Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 .8907428 .178303 2.638644 .5393243 1.242161 1| 51 -.9574581 .6339817 4.527535 -2.230848 .3159316 ---------+-------------------------------------------------------------------- combined | 270 .5416382 .1922369 3.158774 .163158 .9201184 ---------+-------------------------------------------------------------------- diff | 1.848201 .4789149 .9052867 2.791115 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 3.8591 t = 3.8591 t = 3.8591 P < t = 0.9999 P > |t| = 0.0001 P > t = 0.0001 87 Appendix 8: t-TEST Results Working Capital to Sales Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 214 .0242214 .0736978 1.078106 -.121049 .1694919 1| 48 -2.008213 .7360204 5.099299 -3.488895 -.5275312 ---------+-------------------------------------------------------------------- combined | 262 -.348133 .1544699 2.500313 -.6522988 -.0439672 ---------+-------------------------------------------------------------------- diff | 2.032435 .3797098 1.284737 2.780133 ------------------------------------------------------------------------------ Degrees of freedom: 260 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 5.3526 t = 5.3526 t = 5.3526 P < t = 1.0000 P > |t| = 0.0000 P > t = 0.0000 Cash Flow to Total Debt Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 .1438238 .0243092 .3597432 .0959127 .191735 1| 51 -.0086326 .0189133 .1350677 -.046621 .0293558 ---------+-------------------------------------------------------------------- combined | 270 .1150265 .0203529 .3344315 .0749554 .1550976 ---------+-------------------------------------------------------------------- diff | .1524565 .0512552 .0515424 .2533705 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 2.9745 t = 2.9745 t = 2.9745 P < t = 0.9984 P > |t| = 0.0032 P > t = 0.0016 88 Appendix 8: t-TEST Results Cash Flow to Current Liabilities Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 219 .2269034 .0338757 .5013146 .1601376 .2936692 1| 51 -.0513442 .0416968 .2977745 -.1350946 .0324062 ---------+-------------------------------------------------------------------- combined | 270 .1743456 .0293167 .4817221 .1166262 .2320649 ---------+-------------------------------------------------------------------- diff | .2782476 .0730875 .1343489 .4221463 ------------------------------------------------------------------------------ Degrees of freedom: 268 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 3.8070 t = 3.8070 t = 3.8070 P < t = 0.9999 P > |t| = 0.0002 P > t = 0.0001 Cash Flow to Sales Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 214 .0698125 .0229941 .3363745 .0244873 .1151376 1| 48 -.0170025 .0525279 .3639241 -.1226751 .08867 ---------+-------------------------------------------------------------------- combined | 262 .0539074 .021161 .342521 .0122394 .0955755 ---------+-------------------------------------------------------------------- diff | .086815 .0545429 -.0205871 .194217 ------------------------------------------------------------------------------ Degrees of freedom: 260 Ho: mean(0) - mean(1) = diff = 0 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 t = 1.5917 t = 1.5917 t = 1.5917 P < t = 0.9437 P > |t| = 0.1127 P > t = 0.0563 89 90 Appendix 1: TABLE 1 - GROUP STATISTICS (CONTINUED) 1998 1999 2000 2001 2002 Std. Std. Std. Std. Std. Std. Std. Std. Std. Std. Deviati Error Deviatio Error Deviati Error Deviati Error Deviatio Error STATUS N Mean on Mean Mean n Mean Mean on Mean Mean on Mean Mean n Mean 0 0.0379 0.0212 EBITTA 55 0.0789 0.28176 9 0.0653 0.22392 0.03019 0.0781 0.18973 0.02582 0.0273 0.15742 3 0.0345 0.1492 0.02012 1 0.2593 -0.129 0.0429 -0.079 13 -0.3679 0.93507 4 -0.1727 0.30164 0.08366 -0.0315 0.0729 0.02022 6 0.1489 8 7 . . 0 0.1345 0.1090 SALESTA 54 0.8571 0.98897 8 10.6019 72.1218 9.81454 26.8251 188.939 25.9528 0.8093 0.80118 3 0.7775 0.81477 0.11088 1 0.1148 0.1548 13 0.5363 0.41413 6 0.4869 0.50379 0.14543 8.2494 26.5237 7.65672 0.5006 0.51364 7 0.4497 . . 0 0.1860 0.1484 TDTA 55 0.746 1.38005 9 0.7032 1.06577 0.14371 0.7266 1.08349 0.14744 0.7227 1.10102 6 0.739 1.22083 0.16462 1 0.2975 0.6311 13 1.1943 1.07288 6 1.1026 0.44384 0.1231 1.9095 3.15426 0.87483 1.8968 2.18624 1 0.8482 . . 0 0.0346 0.0448 CFTD 55 0.1487 0.25708 7 0.1854 0.48526 0.06543 0.1059 0.32994 0.0449 0.1346 0.33276 7 0.1411 0.23456 0.03192 1 0.0602 -0.028 0.0253 13 -0.0015 0.21727 6 -0.0071 0.10089 0.02798 0.0008 0.10492 0.0291 3 0.08788 7 0.0428 . . 0 0.0527 0.0738 CFCL 55 0.2328 0.39148 9 0.2476 0.59416 0.08012 0.1892 0.4582 0.06235 0.2373 0.54793 8 0.2599 0.50252 0.06838 1 0.1052 -0.083 0.0793 13 -0.0358 0.3794 3 -0.0562 0.33817 0.09379 -0.0321 0.2003 0.05555 7 0.27494 7 0.0999 . . 0 0.0653 CFSALES 54 0.0254 0.48025 5 0.1038 0.24568 0.03343 0.0765 0.20971 0.02881 0.0737 0.34581 0.0475 0.0953 0.19722 0.02735 1 0.1466 -0.024 0.0400 13 0.0615 0.52866 2 -0.0927 0.42742 0.12339 -0.0194 0.22063 0.06369 5 0.13281 4 0.0808 . . *Healthy firms denoted by STATUS 0, Distress Firms Denoted by STATUS 1. Appendix 1: TABLE 1 - GROUP STATISTICS 1998 1999 2000 2001 2002 Std. Std. Std. Std. Std. Std. Error Std. Error Std. Error Std. Error Std. Error STATUS N Mean Deviation Mean Mean Deviation Mean Mean Deviation Mean Mean Deviation Mean Mean Deviation Mean 0 55 -0.0412 0.81664 0.11012 -0.0161 0.54917 0.07405 0.0086 0.54378 0.074 0.022 0.45195 0.06094 0.0033 0.60721 0.08188 WCTA 1 13 -0.3806 1.09958 0.30497 -0.3698 0.42069 0.11668 -0.8262 1.72048 0.47717 -1.3161 2.14449 0.61906 0.4024 . . 0 55 0.4307 0.2224 0.02999 0.4311 0.23905 0.03223 0.4528 0.23795 0.03238 0.4232 0.23267 0.03137 0.4122 0.23039 0.03107 CATA 1 13 0.5345 0.25469 0.07064 0.4852 0.24437 0.06778 0.5266 0.28493 0.07903 0.4248 0.33839 0.09768 0.766 . . 0 55 0.1856 0.39833 0.05371 0.2093 0.481 0.06486 0.2471 0.59176 0.08053 0.2476 0.55764 0.07519 0.2637 0.63103 0.08509 CASHTA 1 13 -0.0989 0.35963 0.09974 -0.1169 0.29186 0.08095 -0.11 0.27668 0.07674 -0.1422 0.37741 0.10895 -1.1245 . . 0 55 0.6856 2.09033 0.28186 0.7659 2.28774 0.30848 0.8882 2.58793 0.35217 1.2232 3.42923 0.4624 1.4107 5.09994 0.69401 CASHCL 1 13 -2.3962 8.70827 2.41524 -0.4597 1.35963 0.37709 -0.2595 0.77198 0.21411 -0.6942 2.0953 0.60486 -3.092 . . 0 54 0.0779 0.71318 0.09705 -0.0125 0.82479 0.11224 -0.018 1.15925 0.15923 0.0491 1.48275 0.20367 -0.0961 1.83614 0.25463 WCSALES 1 13 0.0019 2.89652 0.80335 -1.0866 1.13746 0.32836 -2.8274 6.14315 1.77337 -4.4956 7.43819 2.2427 0.8947 . . 0 55 1.8062 2.13446 0.28781 1.9454 2.47342 0.33352 1.9305 2.64611 0.36009 2.2315 3.55608 0.4795 2.4089 5.08496 0.69198 CACL 1 13 2.3669 5.97501 1.65717 0.7002 0.40693 0.11286 0.6756 0.50633 0.14043 0.5345 0.55463 0.16011 2.1064 . . 0 55 0.0444 0.31521 0.0425 0.0491 0.24848 0.03351 0.0506 0.17622 0.02398 -0.0333 0.28504 0.03844 -0.009 0.18825 0.02538 NITA 1 13 -0.4165 0.91829 0.25469 -0.1878 0.29794 0.08263 -0.1343 0.18992 0.05268 -0.3127 0.25998 0.07505 -0.1761 . . *Healthy firms denoted by STATUS 0, Distress Firms Denoted by STATUS 1.

DOCUMENT INFO

Shared By:

Categories:

Tags:
financial distress, financial ratios, logit model, corporate distress, corporate governance, cash flow, distressed firms, prediction models, explanatory variables, listed companies, the firm, bankruptcy prediction, logit models, prediction model, independent variables

Stats:

views: | 413 |

posted: | 5/26/2010 |

language: | English |

pages: | 92 |

OTHER DOCS BY xkv17320

How are you planning on using Docstoc?
BUSINESS
PERSONAL

By registering with docstoc.com you agree to our
privacy policy and
terms of service, and to receive content and offer notifications.

Docstoc is the premier online destination to start and grow small businesses. It hosts the best quality and widest selection of professional documents (over 20 million) and resources including expert videos, articles and productivity tools to make every small business better.

Search or Browse for any specific document or resource you need for your business. Or explore our curated resources for Starting a Business, Growing a Business or for Professional Development.

Feel free to Contact Us with any questions you might have.