Welcome to your ultimate accounting resource!
 

Welcome to our Accounting Archive. Have fun browsing!

 

(Browse for more articles)

 

Models For Predicting Corporate Financial Distress

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