| INTRODUCTION 2 MEASURING FINANCIAL HEALTH 2 | | | | is the one proposed by Edward Altman. |
| FINANCIAL DISTRESS 2 FACTORS AFFECTING | | | | Altman's z-score, or zeta model, combined |
| FINANCIAL HEALTH 3 Capital Structure and | | | | various measures of profitability or risk. |
| Capital Adequacy 3 Operating Cash Flows and | | | | The resulting model was one that demonstrated |
| Cost Structure 4 Earnings Capacity 4 | | | | a company's risk of bankruptcy relative to a |
| Liquidity 4 Asset Conversions - "Growing | | | | standard. Altman's initial study proved his |
| Broke" 5 Asset Utilisation Efficiency | | | | model to be very accurate; it correctly |
| Turnover 5 Strategic Position 5 PREDICTING | | | | predicted bankruptcy in 94% of the initial |
| FINANCIAL DISTRESS 6 FAILURE PREDICTION | | | | sample (Altman 1968). Despite the positive |
| MODELS 7 Altman's Z Score 8 Logit Analysis: | | | | results of his study, Altman's model had a |
| The Model 9 Other Statistical Failure | | | | key weakness; it assumed variables in the |
| Prediction Models 10 The Gambler's Ruin | | | | sample data to be normally distributed. If |
| Models 10 Alternative Models - Artificial | | | | all variables are not normally distributed, |
| Neural Networks 12 CONCLUSION 12 REFERENCES | | | | the methods employed may result in selection |
| 13 Introduction A company trying to achieve | | | | of an inappropriate set of predictors |
| its business plan faces problems similar to | | | | (Sheppard 1994). Chistine Zavgren developed a |
| those faced by a driver embarking on a long | | | | model that corrected for this problem. Her |
| trip. The likelihood that car and driver will | | | | model used logit analysis to predict |
| reach their destination is dependent on: 1) | | | | bankruptcy. Due to its use of logit analysis, |
| how much fuel is in the car's tank upon | | | | her model is considered more robust (Lo |
| starting out, 2) the car's fuel efficiency, | | | | 1986). Further, logit analysis actually |
| 3) how many service stations will be | | | | provides a probability (in terms of a |
| available to refill the car's fuel tank along | | | | percentage) of bankruptcy. Also, the |
| the way and 4) whether the car's fuel tank is | | | | probability calculated might be considered a |
| large enough to cover unexpected accidents, | | | | measure of the effectiveness of management |
| delays, and detours along the way. Similarly, | | | | (ie. effective management will not lead a |
| whether or not a company survives in a highly | | | | company to the verge of bankruptcy). During |
| competitive business environment is dependent | | | | the 1980s and 1990s, the trend has been to |
| upon: 1) how financially healthy the | | | | use logit analysis in favour of multiple |
| corporation is at its inception, 2) the | | | | discriminant analysis (Stickney 1996). More |
| company's ability (and relative flexibility | | | | recently, logit analysis has been compared to |
| and efficiency) in creating cash from its | | | | a more advanced analytical tool, neural |
| continuing operations, 3) the company's | | | | networks. Research has found that the |
| access to capital markets, and 4) the | | | | approaches perform similarly and should be |
| company's financial capacity and staying | | | | used in combination (Altman, Marco, and |
| power when faced with unplanned cash | | | | Varetto 1994). Altman's Z Score Based on |
| shortfalls. Measuring Financial Health There | | | | multiple discriminate analysis (MDA), the |
| is no single measure of financial health. | | | | model predicts a company's financial health |
| Ideally, solvency could be measured along a | | | | based on a discriminant function of the form: |
| continuum in the same way that fuel | | | | Z=0.012X1+0.014X2+0.033X3+0.006X4+0.999X5 |
| sufficiency can be measured using a car's | | | | Where: X1=working capital/total assets |
| petrol gauge. Full health would equate with | | | | X2=retained earnings/total assets X3=earnings |
| having a full tank of fuel. Poor health would | | | | before interest and taxes/total assets |
| be equivalent to showing an empty tank. As | | | | X4=market value of equity/book value of total |
| healthiness progressively decreased, the | | | | liabilities X5=sales/total assets The Z-Score |
| solvency gauge would register movement in the | | | | model (developed in 1968) was based on a |
| direction of relative insolvency. Ultimately, | | | | sample composed of 66 manufacturing companies |
| as healthiness continues to decline, the | | | | with 33 firms in each of two matched-pair |
| solvency gauge would hopefully flash a | | | | groups. The bankruptcy group consisted of |
| warning light. Since, in the real world, no | | | | companies that filed a bankruptcy petition |
| single measure of financial health exists, | | | | under Chapter 11 of the United States |
| proxies that measure various aspects of | | | | bankruptcy act from 1946 through 1965. Based |
| solvency are often combined to estimate a | | | | on the sample, all firms having a Z-Score |
| company's healthiness at a point in time. | | | | greater than 2.99 clearly fell into the |
| Financial Distress As a financially healthy | | | | non-bankruptcy sector, while those firms |
| company becomes more and more financially | | | | having a Z-Score below 1.81 were bankrupt. |
| distressed, it ultimately enters an area of | | | | Altman subsequently developed a revised |
| great danger. Changes to the company's | | | | Z-Score model (with revised coefficients and |
| operations and capital structure (ie. | | | | Z-Score cut-offs) which dropped variables X4 |
| restructuring) must be made to remain | | | | and X5 (above) and replaced them with a new |
| healthy. Apple Computers' attempts in recent | | | | variable X4 = net worth (book value)/total |
| years to restructure its operations to | | | | liabilities. The X5 variable was dropped to |
| survive in the highly competitive computer | | | | minimise potential industry effects related |
| hardware business is a good example of a | | | | to asset turnover. Around 1977, Altman |
| company trying to dramatically restructure | | | | developed jointly with a private financial |
| itself in order to maintain solvency. | | | | firm (ZETA Services, Inc.) a revised |
| Continued decreases in financial health | | | | seven-variable ZETA model based on a combined |
| ultimately lead to insolvency and then | | | | sample of 113 manufacturers and retailers. |
| potentially, bankruptcy. Available evidence | | | | The ZETA model is allegedly far more accurate |
| suggests many companies do not adequately | | | | in bankruptcy classification in years 2 |
| attempt to resolve their financial health | | | | through 5 with the initial year's accuracy |
| problems until it is too late to avoid | | | | about equal. However, the coefficients of the |
| bankruptcy. Factors Affecting Financial | | | | model are not specified (without retaining |
| Health Capital Structure and Capital Adequacy | | | | ZETA Services). The ZETA model is based on |
| Companies finance their long-term operations | | | | the following variables: return on |
| primarily through two sources of capital - | | | | assets stability of earnings |
| debt and equity. One of the most important | | | | debt service cumulative |
| financing decisions a company makes is the | | | | profitability liquidity/current |
| proportion of debt to owner's equity in the | | | | ratio capitalisation (five year |
| company's capital structure. Summary measures | | | | average of total market value) size |
| of a company's capital structure include the | | | | (total tangible assets) Logit Analysis: The |
| company's debt to equity ratio (D/E) and debt | | | | Model Application of the logit model requires |
| to total capital ratio (D/(D+E)). Interest | | | | four steps. 1. a series of seven financial |
| and principal payments on debt must be paid | | | | ratios are calculated. 2. each ratio is |
| from operations before any payments can be | | | | multiplied by a coefficient unique to that |
| distributed to equity holders (in the form of | | | | ratio. This coefficient can be either |
| dividends or share buy-backs). Therefore, the | | | | positive or negative. 3. the resulting values |
| interest and principal, which must be paid on | | | | are summed together (y). 4. the probability |
| debt, are considered fixed-costs of | | | | of bankruptcy for a firm is calculated as the |
| operations. From an operational | | | | inverse of (1 + ey). Explanatory variables |
| point-of-view, the extent of the burden of | | | | with a negative coefficient increase the |
| these fixed obligations can be measured | | | | probability of bankruptcy because they reduce |
| relative to the company's continuing ability | | | | ey toward zero, with the result that the |
| to pay the fixed obligations. A frequently | | | | bankruptcy probability function approaches 1 |
| used measure of a company's ability to cover | | | | 1, or 100 percent. Likewise, independent |
| its interest payments is its earnings before | | | | variables with a positive coefficient |
| interest and taxes and before depreciation | | | | decrease the probability of bankruptcy |
| and amortisation (EBITDA) to its interest | | | | (Stickney 1996). Table 1 shows the financial |
| expense. A company is financially distressed | | | | ratios used in the logit model and their |
| whenever its EBITDA is less than its interest | | | | respective coefficients. TABLE 1 - Financial |
| expense. Financial leverage involves | | | | Ratios used in Logit Model FINANCIAL RATIO |
| the substitution of fixed-cost debt for | | | | COEFFICIENT + 0.23883 Average Inventories |
| owner's equity in the hope of increasing | | | | Sales - 0.108 Average Receivables/Average |
| equity returns. As demonstrated by Higgins | | | | Inventories - 1.583 (Cash + Marketable |
| and others, financial leverage improves | | | | Securities)/Total Assets - 10.78 Quick Assets |
| financial performance when things are going | | | | Current Liabilities + 3.074 Income from |
| well but worsens financial performance when | | | | Continuing Operations/(Total Assets - Current |
| things are going poorly. Therefore, | | | | Liabilities) + 0.486 Long-Term Debt/(Total |
| increasing the ratio of debt to equity in a | | | | Assets - Current Liabilities) - 4.35 Sales |
| company's capital structure implicitly makes | | | | (Net Working Capital + Fixed Assets) + 0.11 y |
| the company relatively less solvent (on the | | | | = Sum of (Coefficient * Ratio) Probability of |
| downside) and more financially risky than a | | | | Bankruptcy = 1/(1 + ey) Other Statistical |
| company without debt. Capital | | | | Failure Prediction Models Many additional |
| adequacy relates to whether a company has | | | | bankruptcy prediction models have been |
| enough capital to finance its planned future | | | | developed since the work of Beaver and |
| operations. If the company's capital is | | | | Altman. Lev (1974), Deakin (1977), Ohlson |
| inadequate, then it must either be able to: | | | | (1980), Taffler (1980), Platt & Platt (1990), |
| 1) successfully issue new equity, or 2) | | | | Gilbert, Menon, and Schwartz (1990), and Koh |
| arrange new debt. The amount of debt a | | | | and Killough (1990) amongst others have |
| company can successfully absorb and repay | | | | continued to refine the development of |
| from its continuing operations is normally | | | | multivariate statistical models. Almost all |
| referred to as the company's debt capacity. | | | | of these traditional models have been either |
| Capital adequacy is normally evaluated by | | | | matched-pair multi-discriminate models or |
| looking at the company's operational cash | | | | logit models. A 1997 study by Begley, Ming |
| flow projections and its projections of | | | | and Watts concludes: "Given that Ohlson's |
| capital needs. When companies undertake major | | | | original model is frequently used in academic |
| new projects or undergo a significant | | | | research as an indicator of financial |
| financial restructuring they often perform | | | | distress, its strong performance in this |
| financial feasibility studies to determine | | | | study supports its use as a preferred model." |
| whether the company has the financial | | | | The Gambler's Ruin Models Wilcox (1971 and |
| capacity to undertake the project and whether | | | | 1976), Santomero (1977), Vinso (1979) and |
| the company will be able to repay all future | | | | others have adapted a gambler's ruin approach |
| debt payments once the project is built. | | | | to bankruptcy prediction. Under this |
| Operating Cash Flows and Cost Structure All | | | | approach, bankruptcy is probable when a |
| other factors being equal, companies that can | | | | company's net liquidation value (NLV) becomes |
| consistently generate positive cash flows | | | | negative. Net liquidation value is defined as |
| from operations will remain relatively more | | | | total asset liquidation value less total |
| solvent than those that cannot. This requires | | | | liabilities. From one period to the next, a |
| that operating cash inflows (collections or | | | | company's NLV is increased by cash inflows |
| sales) consistently exceed operating cash | | | | and decreased by cash outflows during the |
| outflows (costs). Companies which experience | | | | period. Wilcox combined the cash inflows and |
| erratic cash outflows and inflows are | | | | outflows and defined them as adjusted cash |
| relatively more risky because they are less | | | | flow. All other things being equal, the |
| likely, in one or more time periods, to be | | | | probability of a company's failure increases, |
| able to cover fixed expenses/outflows. | | | | the smaller the company's beginning NLV, the |
| Companies which have a higher proportion of | | | | smaller the company's adjusted (net) cash |
| fixed costs to variable costs are also | | | | flow, and the larger the variation of the |
| relatively more risky and relatively less | | | | company's adjusted cash flow over time. |
| solvent than companies with a relatively | | | | Wilcox uses the gambler's ruin formula |
| lower proportion of fixed costs in their | | | | (Feller, 1968) to show that a company's risk |
| operating cost structure. Earnings Capacity | | | | of failure is dependent on; 1) the above |
| All other things being equal, companies with | | | | factors plus, 2) the size of the company's |
| higher relative earnings and higher relative | | | | adjusted cash flow at risk each period (ie. |
| returns on investment will remain more | | | | the size of the company's bet). Using a more |
| solvent than their less fortunate | | | | robust statistical technique, Vinso (1979) |
| competitors. The most commonly used financial | | | | extended Wilcox's gambler's ruin model to |
| measures of earnings capacity are earnings | | | | develop a safety index. Based on input |
| before interest and taxes (EBIT) and net | | | | concerning the variability of expected |
| income. Liquidity Adequate liquidity is a | | | | contribution margin amounts, the index can be |
| further necessary component of solvency. | | | | used to predict the point in time when a |
| Frequently used liquidity measures include: | | | | company's ruin is most likely to occur |
| a) working capital (current assets minus | | | | (called first passage time). The statistics |
| current liabilities), b) current ratio | | | | used in gambler's ruin approaches are |
| (current assets divided by current | | | | somewhat formidable (especially to the |
| liabilities), and c) quick ratio (cash, | | | | average reader). However, both Wilcox and |
| marketable securities and accounts receivable | | | | Vinso richly describe some of the factors |
| divided by current liabilities). To evaluate | | | | which most affect business failure. For |
| liquidity, each of the assets and liabilities | | | | example, Wilcox states: "The (cash) inflow |
| on a company's balance sheet should be | | | | rate ... can be increased through higher |
| evaluated for liquidity. Current assets are | | | | average return on investment. However, having |
| those which will likely be converted to cash | | | | a major impact here usually requires |
| within one year or less. Current liabilities | | | | long-term changes in strategic position. This |
| are those which must be paid within one year. | | | | is difficult to control over a short time |
| However, when a company becomes financially | | | | period except by divestitures of peripheral |
| distressed, even assets which are normally | | | | unprofitable businesses...The average outflow |
| considered current assets (accounts | | | | rate is controlled by managing the average |
| receivable and stock, for example) may become | | | | growth rate of corporate assets. Effective |
| relatively "illiquid". Long-term assets, in | | | | capital budgeting ... requires resource |
| general, are far less liquid than current | | | | allocation emphasising those business units, |
| assets. Some longer-term assets may be very | | | | which have the highest future payoff. The |
| "illiquid". Also, as stated above, often a | | | | size of the bet is the least understood |
| company's long-term liabilities can become | | | | factor in financial risk. Yet management has |
| immediately due and payable if the company | | | | substantial control over it. Variability in |
| violates contractual debt covenants or other | | | | liquidity flows governs the size of the bet. |
| obligations. Wilcox (1976) argues that net | | | | This variability can be managed through |
| liquidation value provides a solid conceptual | | | | dividend policy, through limiting earning |
| basis for evaluating a company's liquidity. | | | | variability and investment variability, and |
| Net liquidation value is defined as total | | | | through controlling the co-variation between |
| asset liquidation value less total | | | | profits and investments...True earnings |
| liabilities. Wilcox (1976) applies what he | | | | smoothing is attained by control of exposure |
| calls typical (not definitive) valuation | | | | to volatile industries, diversification, and |
| multipliers to balance sheet assets to arrive | | | | improved strategic position." Vinso supports |
| at representative asset liquidation values: | | | | Wilcox's emphasis on cash flow processes and |
| Cash Equivalents 100% Other | | | | stresses the importance of debt capacity: |
| Current Assets 70% Long Term Assets | | | | "Before deriving a mathematical model for |
| 50% Wilcox (1976) shows that a company | | | | determining the risk of ruin, it is necessary |
| becomes bankrupt when net liquidation value | | | | to describe the process. First, a firm has |
| is reduced to zero. Asset Conversions - | | | | some pool of resources at time = 0 of some |
| "Growing Broke" Asset and liability | | | | size U0, which are available to prevent ruin |
| conversions are continuously ongoing in any | | | | (similar to Wilcox's beginning NAV). Then, |
| dynamic business. Operationally, the company | | | | earnings come to the firm from |
| is selling its products thereby creating cash | | | | revenue(s)...less the costs incurred in |
| inflows. Alternatively, sales may be made on | | | | producing the revenues. There are two types |
| credit, increasing the company's accounts | | | | of costs to be considered: variable, which |
| receivable. Concurrently, inventories are | | | | change according to the stochastic nature of |
| produced and sold and production and | | | | the revenue sources, and fixed costs, which |
| operating expenses are incurred to continue | | | | do not vary with revenue but are a function |
| operations. If a company's inventories and | | | | of the period. So, revenue less variable |
| accounts receivable grow faster than the | | | | costs...can be defined as variable profit |
| corresponding growth in the company's sales | | | | (which is available to pay fixed costs). If |
| and accounts payable, liquidity will be | | | | Ut is less than zero, ruin occurs because no |
| negatively affected. Strategic asset | | | | funds are available to meet unpaid fixed |
| conversions are also ongoing, but with less | | | | costs...These definitions, however, ignore |
| regularity. Decisions to invest in 'bricks | | | | debt capacity, if available, which must be |
| and mortar' and other long-term investments | | | | included as the firm can use this source |
| are made and debt and equity are obtained to | | | | without being forced to confront |
| supply the capital needed to pay for them. | | | | shareholders, creditors or bankruptcy,...debt |
| Slowly but surely, companies can 'go broke' | | | | holders or other creditors will force |
| when assets are converted to less liquid | | | | reorganisation if a firm is unable to meet |
| forms over a sustained time period. This can | | | | contractual obligations because working |
| happen when the company's assets grow faster | | | | capital is too low and the firm cannot obtain |
| than the company's sales (often the case for | | | | more debt." Alternative Models - Artificial |
| many start-up companies). When this happens, | | | | Neural Networks Since 1990, another promising |
| the company becomes more highly leveraged and | | | | approach to bankruptcy prediction, based on |
| less solvent. Similarly, a company whose long | | | | the use of neural networks, has evolved. |
| term investment decisions do not pay off in | | | | Artificial Neural Networks (ANN) are |
| terms of planned operating returns (thus | | | | computers constructed to process information, |
| increasing fixed cost structures and | | | | in parallel, similar to the human brain. |
| decreasing operating cash flows), will become | | | | ANN's store information in the form of |
| less solvent. Asset Utilisation Efficiency | | | | patterns and are able to learn from their |
| Turnover Those companies, which survive, use | | | | processing experience. Unlike MDA and logit |
| their human and capital assets relatively | | | | analyses, ANN's impose less restrictive data |
| efficiently. That is, they have relatively | | | | requirements (the requirement for linearity, |
| higher returns on investment (ROI) and higher | | | | for example) and are especially useful in |
| returns per employee than less successful | | | | recognising and learning complex data |
| competitors. They achieve relatively higher | | | | relationships. Recent ANN bankruptcy |
| returns through superior asset management | | | | prediction studies include those of Bell, et |
| (capital and human assets) and through | | | | al. (1990), Hansen & Messier (1991), Chung & |
| superior strategic positioning. In the | | | | Tam (1992), Liang, et al. (1992), Tam & Kiang |
| absence of aggressive asset management, | | | | (1992), Salchenberger (1993), Coats & Fant |
| companies must usually resort to wholesale | | | | (1993), Fanning & Cogger (1994), Brockett, et |
| asset divestitures and/or are forced to | | | | al. (1994), Boritz, et al. (1995), and |
| restructure to fund their continuing | | | | Etheridge & Siriam (1995 and 1997). Research |
| operations. Strategic Position Schoffler | | | | has shown that ANN's offer a viable |
| (Buzzell and Gale, 1987) and others have | | | | alternative to other more traditional methods |
| documented the high correlation between | | | | of bankruptcy prediction. The ability of the |
| positive returns on investment and such | | | | model to learn allows for the constant |
| factors as: 1) higher relative market shares, | | | | re-calibration and validation of the model, |
| 2) relative product quality and 3) lower | | | | which helps increase classification rates. |
| relative capital intensity. Companies that | | | | From a theoretical perspective, ANN's are |
| have strong strategic market positions are | | | | more desirable because they make fewer |
| more likely to experience higher relative | | | | assumptions about the data normality and |
| returns on investment than their competitors. | | | | linear separability. One of the main |
| These positive returns, in turn, increase the | | | | disadvantages of ANN's is the inability to |
| solvency of the market leaders. Those | | | | assign intuition the network weights. Another |
| competitors that have lower market shares or | | | | disadvantage is that the model might simply |
| lower product quality are less likely to | | | | memorise the data as opposed to forming a |
| achieve industry average returns and are thus | | | | general set of classification rules, which |
| more likely to become less solvent in the | | | | can cause estimates on future samples to be |
| future. Predicting Financial Distress In | | | | less reliable. Conclusion Future research in |
| America, each year approximately one percent | | | | bankruptcy prediction should analyse the |
| of all firms required to file with the | | | | economic and institutional factors that can |
| Securities and Exchange Commission file for | | | | impact the reasons for bankruptcy. Jones |
| bankruptcy. The American Bankruptcy Institute | | | | (1987) indicated that the lack of homogeneity |
| reports that around 50,000 businesses filed | | | | in the motivation for a bankruptcy filing |
| for bankruptcy in 1997. Attempts to develop | | | | might complicate the modelling effort. |
| bankruptcy prediction models began seriously | | | | Although normally motivated by an effort to |
| sometime in the late 1960's and continue | | | | resolve severe financial problems, a firm may |
| through today. At least three distinct types | | | | file for bankruptcy primarily to void a union |
| of models have been used to predict | | | | contract or for other legal reasons (Jones |
| bankruptcy: a) statistical models (univariate | | | | 1987). Another area where models can be |
| analysis, multiple discriminate analyses | | | | improved is in catering for predictor |
| [MDA]), and conditional logit regression | | | | variables other than financial ratios may |
| analyses, b) gambler's ruin-mathematical | | | | prove beneficial. For example, measures of |
| statistical models, and c) artificial neural | | | | management experience, management expertise, |
| network models. Each of these models is | | | | or other behavioural aspects that impact the |
| discussed below. Most of the publicly | | | | operations of the firm could be significant |
| available information regarding prediction | | | | in a bankruptcy prediction model. |
| models is based on research published by | | | | Additionally, including variables that |
| academics. Commercial banks, public | | | | control for a changing economic environment |
| accounting firms and other institutional | | | | may provide valuable insights for predicting |
| entities (ratings agencies, for example) | | | | bankruptcy. Bibliography References Altman, |
| appear to be the primary beneficiaries of | | | | Edward I. Corporate Financial Distress. New |
| this research, since they can use the | | | | York, NY: John Wiley and Sons, 1983. Altman, |
| information to minimise their exposure to | | | | Edward I. (1968) Financial Ratios, |
| potential client failures. While continuing | | | | Discriminate Analysis and the Prediction of |
| research has been ongoing for almost thirty | | | | Corporate Bankruptcy, The Journal of Finance. |
| years, it is interesting to note that no | | | | Altman, Edward I. Homepage of Professor |
| unified well-specified theory of how and why | | | | Edward I. Altman, New York, NY: Stern School |
| corporations fail has yet been developed. The | | | | of Business. Available at Altman, Edward I, |
| available statistical models derive merely | | | | Giancarlo Marco, and Franco Varetto (1994) |
| from the statistical optimisation of a set of | | | | Corporate Distress Diagnosis: Comparisons |
| ratios. As stated by Wilcox (1973) the lack | | | | Using Linear Discriminant Analysis and Neural |
| of conceptual framework results in the | | | | Networks (the Italian Experience), The |
| limited amount of available data on bankrupt | | | | Journal of Banking and Finance. Altman, |
| firms being statistically 'used up' by the | | | | Edward I. and Thomas P. McGough (1974) |
| search before a useful generalisation | | | | Evaluation of a Company as a Going Concern, |
| emerges. How useful are these models? Almost | | | | The Journal of Accountancy. Beaver, W., 1967, |
| universally, the decision criterion used to | | | | Financial Ratios as Predictors of Failures, |
| evaluate the usefulness of the models has | | | | in Empirical Research in Accounting, Journal |
| been how well they classify a company as | | | | of Accounting Research. Begley, J., Ming, J., |
| solvent or non-solvent compared to the | | | | Watts, S., 1997,Bankruptcy Classification |
| company's actual status known after-the-fact. | | | | Errors in the 1980s: An Empirical Analysis of |
| Most of the studies consider a type I error | | | | Altman's and Ohlson's Models, Review of |
| as the classification of a failed company as | | | | Accounting Studies. Bell, T.B., G.S. Ribar |
| healthy, and consider a type II error as the | | | | and J. Verchio, 1990, Neural Nets Versus |
| classification of a healthy company as | | | | Logistic Regression: A Comparison of Each |
| failed. In general, type I errors are | | | | Model's Ability to Predict Commercial Bank |
| considered more costly to most users than | | | | Failures," Boritz, J.E., D.B. Kennedy and |
| type II errors. The usefulness of fail | | | | A.M. Albuquerque, 1995, Predicting Corporate |
| non-fail prediction models is suggested by | | | | Failure Using a Neural Network Approach, |
| Ohlson (1980) "...real world problems concern | | | | Intelligent Systems in Accounting, Finance |
| themselves with choices which have a richer | | | | and Management. Brockett, P.L., W.W. Cooper, |
| set of possible outcomes. No decision problem | | | | L.L. Golden and U. Pitaktong, 1994, A Neural |
| I can think of has a payoff space which is | | | | Network Method for Obtaining an Early Warning |
| partitioned naturally into the binary status | | | | of Insurer Insolvency. The Journal of Risk |
| bankruptcy versus non-bankruptcy...I have | | | | and Insurance. Buzzell, R.D., Gale, B.T., |
| also refrained from making inferences | | | | 1987, The PIMS Principles Linking Strategy to |
| regarding the relative usefulness of | | | | Performance, New York: The Free Press. Chung, |
| alternative models, ratios and predictive | | | | H.M. and K. Y. Tam, 1992, A Comparative |
| systems... Most of the analysis should simply | | | | Analysis of Inductive-Learning Algorithms, |
| be viewed as descriptive statistics - which | | | | Intelligent Systems in Accounting, Finance |
| may, to some extent, include estimated | | | | and Management. Coats, P.K. and L.F. Fant, |
| prediction error-rates - and no theories of | | | | 1993, Recognizing Financial Distress Patterns |
| bankruptcy or usefulness of financial ratios | | | | Using a Neural Network Tool, Financial |
| are tested." Subject to the qualifications | | | | Management. Cook, Roy A. and Jeryl L. Nelson. |
| expressed above, bankruptcy prediction models | | | | A Conspectus of Business Failure Forecasting, |
| continue to be used to predict failure. | | | | Available at Deakin, E., Business Failure |
| Failure Prediction Models The early history | | | | Prediction: An Empirical Analysis,, 1977, in |
| of researchers' attempts to classify and | | | | E. Altman and A. Sametz, eds., Financial |
| predict business failure (and bankruptcy) is | | | | Crises: Institutions and Markets in a Fragile |
| well documented in Edward Altman's 1983 book, | | | | Environment, New York: Wiley. Etheridge, H.L. |
| Corporate Financial Distress. Statistical | | | | and R.S. Sriram, 1995, A Neural Network |
| prediction models are more generally better | | | | Approach to Financial Distress Analysis, |
| known as measures of financial distress. | | | | Advances in Accounting Information Systems. |
| Three stages in the development of | | | | Fanning, K. and K.O. Cogger, 1994, A |
| statistical financial distress models exist: | | | | Comparative Analysis of Artificial Neural |
| 1. univariate analysis, 2. multivariate (or | | | | Networks Using Financial Distress Prediction, |
| multi-discriminate [MDA]) analysis, and 3. | | | | Intelligent Systems in Accounting, Finance |
| logit analysis. Univariate analysis assumes | | | | and Management. Gilbert, L.R., Menon, K., and |
| that a single variable can be used for | | | | Schwartz, K.B., 1990, Predicting Bankruptcy |
| predictive purposes (Cook and Nelson 1998). | | | | for Firms in Financial Distress, Journal of |
| The univariate model as proposed by William | | | | Business Finance and Accounting. Hansen, J.V. |
| Beaver achieved a moderate level of | | | | and W.F. Messier, 1991,Artificial Neural |
| predictive accuracy (Sheppard 1994). | | | | Networks: Foundations and Application to a |
| Univariate analysis identified factors | | | | Decision Problem, Expert Systems with |
| related to financial distress, however, it | | | | Applications. Jones, F. L. 1987. Current |
| did not provide a measure of the relevant | | | | Techniques in Bankruptcy Prediction. Journal |
| risk (Stickney 1996). In the next stage of | | | | of Accounting Literature. Koh, H.C. and |
| financial distress measurement, multivariate | | | | Killough, L.N., 1990, The Use of Multiple |
| analysis (also known as multiple discriminant | | | | Discriminant Analysis in the Assessment of |
| analysis or MDA) attempted to overcome the | | | | the Going-concern Status of an Audit Client, |
| potentially conflicting indications that may | | | | Journal of Business Finance and Accounting. |
| result from using single variables (Cook and | | | | Lev, B., 1974, Financial Statement Analysis, |
| Nelson 1998). The best-known, and most-widely | | | | A New Approach. Englewood Cliffs, N.J.: |
| used, multiple discriminant analysis method | | | | Prentice-Hall. Liang, T.P., J.S. Chandler, I. |