SUBHASH SHARMA & VIJAY MAHAJAN

Only recently have academicians and practitioners realizedthe importance of a systematic analysis of business failures.The authors develop a failure process model that can helpmanagers to understand and predict failure. Causes andsymptoms of failure are reviewed briefly as methods ofanalyzing failures. A parsimonious financial corporate failuremodel based on symptoms of failure (performance indicators)is developed and used to predict the business failure ofretail establishments over the five-year period prior to theactual failure.

EARLY WARNINGINDICATORS OF BUSINESS

FAILURE

THE high failure rate of tiew products is well known(Crawford 1977). Similarly, every year several

hundred thousand firms are started and an almost equalnumber are discontinued. Even more firms transfer own-ership or control. For example, in 1977 about 8,000concerns failed. Their aggregate liabilities totaled about$3 billion (The Business Failure Record 1978).

Failure is not restricted to products and firms.Military failures, failure of economics (Hazlitt 1959;Schoeffler 1955), and failure of technology (Juenger1949) also occur. Yet academicians and practitionersalike have all but neglected the failure phenomenon. Injournals and books very few pages have been devoted tothe study of failures. Only recently has any seriousattempt been made to study business failures and theircauses (e.g., Altman 1968; Altman, Haldeman, andNarayanan 1977; Argenti 1975, 1976a, b, c; Cooper1975, 1979; Hartley 1976). The reasons for lack of

Subhash Sharma is Assistant Professor of Marketing,The University of South Carolina. Vijay Mahajan is Asso-ciate Professor of Marketing, The Wharton School, Uni-versity of Pennsylvania.

interest in studying business failures may be many, andthey include:

* The very negative connotation of the term "fail-ure."

* The notion that the failure process for each prod-uct or firm is atypical and hence does not lend to ascientific study.

* Lack of a systematic body of knowledge related tothe failure process.

* The nature of the reward criteria used by man-agement.

* The belief that failure is a sudden rather than agradual process. Two examples substantiate thispoint: first, directors of Rolls Royce were unawareof its failure until the last minute; second, a well-known British merchant bank bought a Japanese firmwhich failed two weeks later (Argenti 1976b).

Why Study Failures?The importance of systematically analyzing failures haslong been recognized in fields such as engineering.After World War II, Aeronautical Radio, Inc. and Cornell

80 / Journal of Marketing, Fall 1980Journal of MarketingVol. 44 (FaU 1980^ R

University analyzed 45.000 and 100,000 vacuum tubes,respectively (Shooman 1966). The objectives of theanalysis were to determine causes of failure and developmathematical models for predicting failures. The resultsand the subsequent interest in failure analysis led toreliability engineering.

As mentioned earlier, very few studies related tosystematic analysis of business failures have been under-taken. The results of the few major attenipts to analyzebusiness failure (e.g., Altman, Haldeman, and Narayanan1977; Argenti 1976; Cooper 1979) have been encourag-ing. Thus, rather than accepting failures as they come,we need to develop a systematic study of failures thatwill enhance the body of knowledge and perhaps reducethe number of failures. Such study will eventually leadto;

• Identification of causes of failures.• Identification ofthe indicators of failures.• Development of mathematical models for predict-

ing failures.Therefore, the major purposes of our article are (1)

to develop a failure process model, (2) to review brieflythe major research studies involving analysis of failures,and (3) to illustrate a parsimonious financial corporatefailure model to predict business failure.

What is a Failure?One of the most difficult tasks of researchers in analyz-ing failures is to define the term "failure." The word

has many meanings. The McGraw-Hill Dictionary ofModern Economics (1973) defines business failures as,"The cessation of operations by a business concernbecause of involvement in court procedures or voluntaryactions which will result in loss to its creditors." Ac-cording to The Business Failure Record (1978), failuresinclude concerns involved in court proceedings or vol-untary actions involving loss to creditors. An entrepre-neur may discontinue operations for a variety of reasons,such as loss of capital, inadequate profits, ill health, andretirement, but if the creditors are paid in full, thebusiness is not tallied as a failure. In contrast, Webster'sThird New International Dictionary (1961) defines fail-ure as "the fact of a certain action or process not havingoccurred . . . the fact of nonoccurrence."' In the first twodefinitions of failure a business is viewed essentially asa reservoir of cash (Walters 1957). The firm is consid-ered to be bankrupt (failed) when the reservoir becomesempty. In other words, an enterprise may be regardedas a "failure" when it cannot meet its liabilities (VanHome 1977). However, if one takes a broader view offailure, as given in Webster's dictionary, a firm will beconsidered a failure if it does not meet the objectives setforth by management. On the basis of this defmitionone may classify Sears as a failure (Business Week1975; Time 1980) because it has failed to achieve theoverall goal set by management (e.g., image, positioning).In most of the studies pertaining to corporate failures,bankruptcy has been used as the definition of failure,whereas in studies on products the inability of the prod-

FIGURE 1Failure Process

Ineffective orBad Management Leads to

Mistakes inStrategic Planand/or its Imple-mentation

CauseDeteriorationm PerformanceIndicators

In absence of no

corrective action,or an ineffectivecorrective action

a /

Leads toFailure

Unanticipatedor

UnforseeableEvents

Early Warning Indicators of Business Failure / 81

TABLE 1Sample Description

Firm

GiantBeckSimonAy resFederalsNational BellasInterstateAncorpFishmanGrantHartfield-Zody'sArlansMiller-WohlKentonBig-BearBohackHarvestPenn-FruitMangelUnishopsCoitBotanyHorn and Hardart

^Information not available

YearChapter 11

Filed

19731970197019721973197319741973197419751974197319731974197619751974197519751973197519711971

Failed Firms

State & YearIncorporated

OH (1950)DE (1932)CA (1935)IN (1896)Ml (1932)DE (1932)DE (1928)DE (1934)DE (1927)DE (1937)DE (1945)NY (1957)DE (1932)DE (1968)OH (1933)NY (1913)NY (1961)PA (1952)DE (1929)NY (1947)TX (1967)DE (1966)NJ (1898)

PrimaryBusiness

Discount storesShoes and ApparelVariety and hardv^rareDepartment storesDepartment storesDepartment storesDepartment storesFast food restaurantDepartment storeVariety storesWomen's apparelDepartment storeWomen's apparelVariety and jewelrySupermarket-groceryGrocery storesSupermarket-grocerySupermarket-groceryWomen's apparelMen's and boys' apparelFabrics and sewing notionsMen's and boys' wearFast food restaurants

No. ofStores

15257

a

15284

120400

521086

621112122343

41470

150326

69101

a

uct to perform as expected has been used as the defini-tion.

Failure ProcessA systematic study of failures requires a model of theunderlying failure process. To understand the failureprocess one needs to know the planning process of agiven firm. The success of any business firm is a resultof the interaction of two major sets of factors. First, theperformance of an enterprise is influenced by a varietyof factors emanating from outside the business itself andthus beyond the control of business managers (uncon-trollable variables). Such environmental conditions asthe rate of growth of the economy, shifting preferences,attitudes, and behavior of consumers, and changingstructure and operating characteristics of the market-place clearly influence the profitability and market strengthof individual businesses.

The other major factors influencing the performanceof a business enterprise emanate from inside the firm.They determine the firm's ability to use its resources toadapt to and take advantage of the constantly changingenvironment. Through a continuous process of formu-lating strategic market plans and executing, monitoring.

and evaluating those plans, management attempts to keepperformance of the enterprise consistent with its envi-ronment and its resources.'

An enterprise has multiple responsibilities in termsof producing performance results. First, the firm mustachieve certain market performance results such as salesvolume, sales growth, competitive market share, andstrength of market position. Second, the owners (i.e.,stockholders and creditors) expect the firm to producecertain financial performance results in terms of profit-ability, growth, and liquidity. Finally, a variety of otherstakeholders in the business, such as employees, suppli-ers, and the community, expect certain performanceresults in terms of employment stability and advance-ment, creditworthiness, and good "corporate citizen-ship" (Argenti 1976a). A firm may be regarded as afailure if it cannot meet one or more of its responsibili-ties. The degree to which a firm meets the responsibili-ties can be measured by performance indicators. Thus,the firm meets its responsibilities by formulating and

'We recognize that all firms do not have a strategic market plan. Infact, some firms may not have a formal "plan." The absence of astrategic market plan or a "plan" may itself be a cause of poor perform-ance (see, for example, Karger and Malik 1975).

82 / Journal of Marketing, Fall 1980

TABLE 1 (continued)

Firm

JameswayGenescoA-Dry GoodsCaldorKingsOutletZayreGinosHecksWoolworthCornwallFed-MartWinklemanRichtonKrogerLuckyPenn-TrafficSupermarketLane-BryantMaysHouse-of-FabricsNew ProcessHost

Nonfailed

State & YearIncorporated

NY (1966)TN (1968)VA (1916)DE {19611DE (1961)Rl (1925)DE (1962)MD (1960)VA (1959)NY (1911)DE (1935)CA (1954)Ml (1928)DE (1969)OH (1902)CA (1931)PA (1903)DE (1966)DE (1920)NY (1927)CA (1946)DE (1924)DE (1914)

Firms

PrimaryBusiness

Discount storesFootwear, men's wearDepartment storesVariety storesVariety storesDepartment storesDiscount storesFast food restaurantDiscount storesVariety storesVariety storesLow-margin retail storesWomen's apparelVariety and jewelrySupermarket-grocerySupermarket-groceryDepartment storesSupermarket-groceryWomen's apparelDepartment storesFabrics and sewing notionsMail orderFast food restaurants

No. ofStores

501500

3

5018721

251506160

37961935887a

12021317

44109188

8560—188

executing strategic market plans, the outcome of whichcan be evaluated by the performance indicators.

The planning process described can be used to de-velop a failure process model. The model shown inFigure 1 is adapted from Argenti (1976a). Ineffective orpoor management usually leads to mistakes in formulat-ing a strategic market plan and/or its implementation. Apoor strategic plan will be ineffective no matter howwell it is executed. In contrast, an excellent strategicplan can be ruined by improper execution. Such mis-takes affect the performance indicators. Thus, basicallymanagement mistakes are the major causes of failuresand the performance indicators are the symptoms ofpossible failures. In fact, according to most studiesrelating to failure of firms and products, about 90% ofali failures can be traced to lack of adequate manage-ment (e.g., Argenti 1975; Business Week 1971; Cooper1975; Houston 1972; Kaye and Garter 1979; Lazo 1965;The Business Failure Record 1978; The CPA Journal1975; Worthing 1966; Wyant 1972).

Methods for Predicting FailuresThe failure process suggests that failures can be predictedeither by analyzing the strategic plan and/or its imple-

mentation (i.e., causes of failures) or by observing theperformance indicators (i.e., symptoms of failures).

Causes of FailuresIn almost all of the studies on the causes of failure a listof causes contributing to the failure has been developed.The list is obtained by analyzing case histories of thefailed firms or products (see, for example, Ross andKami 1973).

Although these studies have provided useful insightsinto the causes of failures, they have certain limitations.First, the lists of causes may be, and are, differentacross studies. For instance, Smith (1966) lists six causesof failure, none of which are the same as those of Rossand Kami (1973). Second, one does not know whichcauses discriminate best between success and failure.Third, there is no indication as to how many causesmust occur before a firm or product will eventually fail.Fourth, one does not know how to use the list of causesto predict or avoid failures.

Some of the limitations have been overcome instudies by Miller (1977), Miller and Friesen (1977), andCooper (1979), who essentially used factor analysis toreduce a number of causes of failure to a few underly-ing causes. Cooper (1979) extended the objective not

Early Warning Indicators of Business Failure / 83

only to measure the underlying causes of failures butalso to determine which causes discriminate best be-tween failures and successes. He first asked managersto choose two successful and two unsuccessful newproducts. The managers were then given a list of 77potential causes of success or failure and were asked toindicate the degree (on a 0-10 scale) to which eachcause contributed to success or failure of each of thechosen products. The resulting data were factor ana-lyzed to obtain the underlying dimensions (causes) ofsuccess and failure. Discriminant analysis showed 11dimensions to be significant determinants of successand failure. The power of discrimination was strong,with about 84% of the products correctly classified.

Cooper's model has important implications for theproduct manager. First, the estimated discriminantfunction can be used to predict success or failure. Sec-ond, because the causes for failure are known, man-agement can take the necessary corrective action. How-ever, the usefulness of the model will depend on theaccuracy of management's rating of the product on the77 items. This rating process may be the major limita-tion, because it is basically judgmental. Another limita-tion is that the factor structure defining the underlyingdimensions may change. In spite of these limitations.Cooper's study shows how the causes of failures andsuccesses can be analyzed systematically to develop afailure prediction model.

Symptoms of Failures

The second method which can be used to predict fail-ures is analysis of the performance indicators (i.e.,symptoms of failures). The usefulness and significanceof this approach can be effectively illustrated by aparsimonious financial corporate failure model. Unlikeearlier financial corporate failure models (e.g., Altman,Haldeman, and Narayanan 1977), this model is specifi-cally developed for retailing firms because the determi-nants of success and failure could be different acrossindustries (Business Week 1980a, b, c). Furthermore, ofthe businesses failing in 1977, 43% were retailing firms;they constituted the largest segment of all commercialand industrial business failures. Sixty percent of thefailed retailing concerns were within their first fiveyears of operation, 25% were within six to 10 years ofoperation, and 15% had operated more than 10 years(The Business Failure Record 1978). The model isdeveloped to predict the business failure over the five-year period prior to failure. For the purpose of modeldevelopment, failed firms are defined as ones that filedunder Chapter XI of the bankruptcy act (Van Home1977). Note that the n:iajor objective of the businessfailure model is not to explain the success or failure ofbusiness performance but rather to predict success or

TABLE 2Description of Financial PerformanceIndicatorsProfitability

1. Return on Assets:Earnings Before Interest and Taxesn'otal Assets

Leverage Ratios2. Debt Service:

Earnings Before Interest and Taxes/InterestCoverage

3. Cash Flow:Cash Flow/Total Debt

4. Capitalization:Market Value of Equity/Total Capital

Liquidity Ratios5. Current Ratio:

Current Assets/Current Liabilities6. Cash Turnover:

Net Sales/Cash7. Receivables Turnover:

Net Sales/Receivables8. Inventory Turnover:

Net Sales/Inventories9. Sales Per Dollar Working Capital:

Net Sales/(Current Assets - Current Liabilities)Miscellaneous10. Retained Earnings/Total Assets11. Total Assets (in thousands of dollars)

failure from performance indicators.^The underlying rationale of the model development

is that an enterprise is a reservoir of cash and theobjective of the firm is to manage its cash flowseffectively (Day 1977; Walters 1957). The firm is con-sidered to be bankrupt when this reservoir becomesempty. The "amount"' of cash in the reservoir, how-ever, can be measured by certain financial performanceindicators reflecting the firm's profitability, growth, andliquidity (see, for example. Table 2). Therefore, anexamination of these indicators should provide signalsof possible failure (Beaver 1966; Fitzpatrick 1931).

Such a framework has been utilized in several stud-ies to predict business failure. In an extensive researchstudy. Beaver (1966) used financial performance ratiosto predict business failure. The study encompassed a

In recent years some attempts have been made to study empirically therelationships among various strategic variables and to identify the deter-minants of market and financial performance. The scope of these sUidieshas ranged from the identification of determinants of profitability for aspecific industry (e.g.. Schendel and Patton 1978) to the formulation of"propositions" of corporate strategy (e.g.. PIMS). The most substantialattempt yet in the field of business policy and corporate strategy is anongoing study being conducted by the Strategic Planning Institute (re-fened to as the PIMS program). See Wind and Mahajan (forthcoming)for details of this program.

84 / Journal of Marketing, Fall 1980

FIGURE 2Mean Current Ratio andfor Failed and Nonfailed

3 -

.Q 2 -

Cur

rent

0 -

0.2-

asse

tsp

Ret

urn

on

o 1

-0.1 -

-0-2-

Return on AssetsFirms

Figure 2Mean current ratio and retum on assets tor tailed and non-tailed finns

— -

4 3Year to failure

- V5 4 3 ^

1 1

Year to failure

— - • — •

Non-failed

2 - 1

\ 2 1

\

\

sample of 79 relatively large firms that failed during the1954—1964 period. For each of these companies, an-other firm was selected that did not fail but was in thesame industry and was of approximately the same sizeas the firm that failed (referred to as the paired design).The data collected for the nonfailed companies were forthe same years as those for the failed firms. Thesesamples were used to test the predictive ability of 30fmancial ratios by univariate analysis. Beaver's workhas been extended by Altman (1968) and others (e.g.,Blum 1974, Taffler and Tisshaw 1977). Whereas Bea-ver used univariate analysis to determine the predictiveability of individual financial ratios, researchers whoextended his work used discriminant analysis (Joy andTollefson 1975). The discriminant scores were used todistinguish between failed and nonfailed firms. Althoughindividual financial performance indicators measurecertain important aspects of the firm's performance,discriminant analysis is a means of capturing the infor-mation provided by individual indicators into one com-posite score.

In our proposed model for retailing firms, we em-ploy a series of discriminant analyses to predict busi-ness failure. The runs are conducted separately foreach year over a five-year period prior to failure. Weuse a paired design approach to select the sample ofnonfailed firms. The validity of the model is establishedby the Lachenbruch holdout method.

The ModelTo develop the business failure model for retailing firms,we selected a sample of 46 firms. The sample was

FIGURE 3Scores of Failed and Nonfailed FirmsOne Year Prior to Failure

Arlans " ^Mange) •.^ ~--,Hardart •.- ^ ~~- -Kent • . - ^ " ~ - -Grant •- -, J ^ -Federa ls» -_^^Ancorp •~~^__

• Bohack

Failed nmis

~ "^ ^ "^ Cr "--^ •

^——_ ~

NahonaJ-BellasMjller-WohlHartfield-ZodyUnishopsBotanyBeckPenn FruitFishmanCoitBig Bear

2.0 25 3.0

-3.5 -3.0 -2.5 -2-0 -1.5

New Process *^'GenescoKings

House c1 FabricsHecksLane BryantOutletHost

Early Warning Indicators of Business Failure / 85

TABLE 3Means and Standard Deviations of

indicator^

1. Return on assets

2. Debt service

3. Cash flow

4. Capitalization

5. Current ratio

6. Cash turnover

7. Receivables turnover

8. Inventory turnover

9. Sales per dollarworking capital

10. Retained earnings/total assets

11. Total assets

^Numbers in the parentheses are stancJard

IndicatorsYear to

1

Failed

-.130(.226)

-5.092(12.587)

-1.551(5.292)

.382(2.014)

1.628(.994)

43.667(28.240)

44.436(63.201)

18.630(35.883)

17.894(48.442)

-.179(.532)

127.878(265.720)

deviations.

Failure

Nonfailed

.139(.050)

17.885(38.352)

1.342(3.419)

1.364(1.113)

2.295(.967)

69.680(106.660)

81.329(92.081)

11.734(16.304)

11.554(8.815)

.302(.129)

302.568(534.346)

Year

Failed

-.020(.240)

-.567(10.243)

.162(.603)

1.125(1.087)

1.658(.571)

52.562(50.735)

35.527(56.001)

9.198(12.388)

9.997(38.708)

.159(.193)

126.977(262.901)

to Failure2

Nonfailed

.135(.047)

39.686(79.33)

1.207(2.806)

1.909(2.806)

2.549(1.184)

49.732(50.385)

65.914(72.181)

10.138(11.580)

10.050(8.506)

.304(.132)

253.738(452.683)

limited to 46 firms because they were the only ones forwhich financial data were readily available from publishedsources such as Moody's Industrial Manual. Half of thesample consisted of failed firms. All failures occurredduring 1970-1976. Table 1 is a description of the failedand nonfaiied firms.

The firms were paired on the basis of firm type asclassified by Moody's Industrial Manual (e.g., discountstore, department store) and size (total assets). For ex-ample. Giant (discount store) was paired with Jamesway(also a discount store); Miller-Wohl (women's apparel)was paired with Winkleman (also a women's apparelstore).

Eleven fmancial performance indicators were selectedto discriminate the failed from nonfailed firms. Table 2is a description of the indicators. The selection of theseindicators was governed by the following considerations.

1. Results of past research which established theimportance of various financial indicators as de-terminants of success or failure (e.g., Altman1968; Beaver 1966; Blum 1974).

2. The development of a comprehensive set of indi-cators reflecting the fmancial performance in termsof profitability, leverage, and liquidity. Theseindicators have been shown to measure the fi-nancial health of firms (Van Home 1977).

3. Data availability that permitted the calculation ofcertain indicators across firms and across years.

Data on 11 performance indicators were taken fromMoody's Industrial Manual for five years prior to fail-ure for failed firms and for a corresponding five-yearperiod for each nonfailed firm. Table 3 gives the meanand standard deviation for each indicator across the fiveyears for failed and nonfailed firms, respectively. Meanvalues of two indicators—return on assets and currentratio—are depicted in Figure 2. Visual examination ofmean values suggests the differences between failed andnonfailed firms on these indicators.

A two-group discriminant analysis was conducted tofind the "best'' linear discriminant functions for thefive years prior to failure. Table 4 is a summary of theresults. Wilk's lambda values, canonical correlations,

86 / Journal of Marketing, Fall 1980

TABLE 3 (continued)

Year to Failure3

Failed

.058(.055)

5.400(5.955)

.426(.868)

.946(.690)

1.751(.537)

78.848(98.628)

46.462(49.368)

10.606(12.201)

16.220(21.779)

.203(.149)

123.734(246.50)

Nonfailed

.146(.058)

16.529(17.774)

1.159(2.355)

2.575(2.195)

2.508(1.213)

42.105(32.667)

89.948(136.478)

10.496(10.562)

10.923(9.939)

.311(.126)

223.794(400.900)

Year

Failed

.073(.054)

6.445(9.202)

.352(.587)

1.203(1.0061

1.915(.781)

44.44(58.190)

44.710(51.559)

8.961(8.949)

13.000(16.074)

.232(.142)

114.848(216.691)

to Failure4

Nonfailed

.153(.060)

15.083(12.887)

.710(.677)

3.531(3.241)

2.273(.838)

35.884(19.962)

68.757(77.579)

10.555(12.121)

10.653(7.878)

.319(.129)

206.051(371.346)

Year

Failed

.0922(.050)

7.559(9.441)

.536(.679)

1.338(.897)

1.923(.613)

47.939(32.942)

84.072(101.195)

11.359(10.794)

16.972(14.419)

.289(.122)

137.281(246.707)

to Failure5

Nonfailed

.155(.051)

13.455(10.829)

.742(.639)

3.856(4.119)

2.401(1.236)

53.563(73.590)

57.610(92.199)

9.522(5.494)

12.193(12.665)

.337(.127)

266.937(426.135)

and the percentage correct classification of the sampleitself indicate a high degree of separation between thetwo groups across the five years. The discriminantfunctions for each of the five years include only twoindicators—return on assets and current ratio. Thediscriminant scores for one year pnor to failure areplotted in Figure 3.

The appearance of current ratio and return on assetsas two important indicators of failure is not surprisingas both ratios indirectly or directly affect the cash flowof a firm. Argenti (1976a), after analyzing a number ofbusiness failures, concluded that high leverage (meas-ured by current ratio) and the inability to make a profit(measured by return on assets) were among the majorsymptoms of failure.

The fact that ratios such as inventory and receiv-ables turnover do not enter the discriminant function issurprising because the model was developed specificallyfor retailing firms. A reason may be that these ratios affectthe profitability of a firm and return on assets essen-tially captures most of the information they provide.However, the reason why some ratios did or did notenter is not important because the objective of analyzingperformance indicators is not to explain why they are

different but to provide an early warning signal of apossible failure.

Before applying the results to predict business fail-ure, one must validate the model. Validation can bedone by means of a holdout sample. As our smallsample size precluded such an analysis, we validatedthe model by using the Lachenbruch holdout method.This procedure involves holding out one group memberat a time and predicting its membership by using thediscriminant function derived from the remainder of thesample. The validated correct classifications, as shownin Table 4, for years one through five are 92%, 78%,74%, 73%, and 77%>, respectively. These results indi-cate the strong separation power of the model and sug-gest that the model is at least partially validated.

The model and its concepts have several managerialimplications. First, the discriminant weights of Table 4can be used along with the appropriate financial ratiosto calculate the discriminant score. If the discriminantscore is less than the cutoff score (which happens to bezero), the firm is a potential candidate for bankruptcy.A negative score does not predict that the firm willbecome bankrupt. Rather, it is an early warning signalof a potential problem calling for a detailed marketing

Early Warning Indicators of Business Failure / 87

TABLE 4Discriminant Analysis Results

Return on AssetsCurrent RatioConstant

Canonical CorrelationWilk's Lambda

Chi squareSignificance

Number of Companies AvailableClassification Rate (%)Classification Rate (%)

(Lachenbruch Holdout Method)

1

3.6790.380

-0.862

0.6510.576

18.230.000

3691.67

91.67

2

3.0310.630

-1.534

0.5310.718

14.259.001

4678.26

78.26

Year Before Failure

3

Weights

7.9050.443

-1.875

0.5390.710

14.745.001

4673.91

73.91

4

10.8310.201

-1.779

0.4910.759

11.286.004

4475.00

12.12

5

10.2980.417

-2.277

0.5640.6828.790

.0122680.77

76.92

audit (Kotler, Gregor, and Rogers 1977).Second, the model can be incorporated in the man-

agement information system and used as a control modelfor monitoring a firm's performance over time. Thismonitoring can be accomplished by an analysis of thetrend of discriminant scores. A positive trend (i.e., thescores increase over time) is indicative of a healthyfirm. A negative trend is an early warning signal for anailing firm that may be heading for bankruptcy andneeding some sort of detailed performance evaluation bymanagement.

Third, the model is not limited to the prediction ofpotential bankruptcy of the firm as a whole. It can beused for decision making at a micro level, such asevaluating products, salespeople, and good and bad creditrisks. For example. Cheek (1979) analyzed various per-formance indicators of successful and failed products todevelop a lexicographic rule for predicting failures. Heclaims a classification rate of about 88%. The details ofhis study are not available to discuss or evaluate.

Obviously, an empirical model such as ours hascertain limitations. First, the choice of performanceindicators is not based on any "theory" but on thefindings of previous research studies. Second, it maynot be possible to obtain a perfect pairing of failuresand successes. The effect of a paired design is notknown and needs to be investigated. However, theselimitations may not be severe given the objective of themodel—to predict business failures by using perform-ance indicators rather than to explain it.

ConclusionWe stress the importance of analyzing failures and pro-pose a failure process model. The model suggests thatfailures can be predicted by analysis of either the causes

of failure or the performance indicators. The majorlimitation of the former is that the inputs to the modelrely on managers' judgments and therefore are subjectto bias and error. Failure prediction models which useperformance indicators do not have this limitation. Theirinputs are completely objective. These models, howev-er, do not tell the causes of failure. They only predictthe possibility of failure. Because each model has itsown advantages and disadvantages and their objectivesare different, management could use both as comple-mentary models. Argenti (1977) describes a subjectivemethod by which the two models might be combined.We believe that combining Cooper's method for analyz-ing causes of failures with the method we suggest foranalyzing symptoms of failures should be an interestingtopic for further research.

Finally, in spite of the limitations of analyzing fail-ures, the identification of determinants of market andfinancial performance of business firms is of interest toboth practitioners and academicians. The recent empiri-cal investigations, however, seem to have been limitedto going concerns (e.g., PIMS program). An importantobjective of our article is to suggest that the scope ofthese performance investigations and the value of theirresults can be enhanced by including failures. Onlythrough the investigation of failures and successes candeterminants of business performance be identified andpredicted.

REFERENCESAltman. E. I. (1968), "Financial Ratios. Discriminant Analysis

and the Prediction of Corporate Bankruptcy," Journal ofFinance, 23 (September), 58^-609.

88 / Journal of Marketing, Fall 1980

, R. Haldeman, and P. Narayanan (1977), "ZETAAnalysis: A New Model to Identify Bankruptcy Risk of COT-porations.^' Journal of Banking and Finance, 1 (June), 29-54.

Argenti, J. (1975). "Company Failure; The Tell Tale Signs at theTop," The Director, 28 (September), 287-9.

(1976a), Corporate Collapse: The Causes and Symp-toms. New York; John Wiley & Sons, Inc.

(1976b), "Corporate Planning and Corporate Col-lapse," Long Range Planning, 9 (December), 12-17.

(1976c), "How Can You Tell If They Are GoingBust?" Accountancy, 87 (October). 42-4.

(1977). "Company Failure—Long Range PredictionNol Enough," Accountancy, 88 (August), 46-8. 50-2.

Beaver, W. H. (1966), "Financial Ratios as Predictors of Fail-ure," in Empirical Research in Accounting: Selected Studies.Supplement to Journal of Accounting Research. 4. 71-1II.

Blum. M. (1974), "Failing Company Discriminant Analysis,"Journal of Accounting Research. 12 (Spring), 1-25.

The Business Failure Record (1978), New York; Dun andBradstreet, Inc.

Business Week (1971), "Hard Look at Why Brokers Fail."(August 7). 54-5.

(1975). "Sears Identity Crisis." (December 8), 52-7.(1980a) "The Economic Case Against Govemment

Bailouts . . . and Who May Need Them." (March 24).(1980b), "In Excellent Shape," (April 7). 4-5.(1980c), "Dealing in Cash Commodities," (April 14),

7.Cheek. M.L. (1979). "Predicting Your Product's Profitability,"

Managerial Planning. 27 (May-June), 1-8,40.Cooper, R. G. (1975), "Why New Industrial Products Fail,"

Industrial Marketing Management. 4 (January), 315-26.— (1979), "The Dimensions of Industrial New Product

Success and Failure." Journal of Marketing, 43 (Summer),93-103.

The CPA Journal (\915), "Some Common Reasons for BusinessFailure," 45 (April). 74.

Crawford. M. C. (1977). "Marketing Research and the NewProduct Failure Rate," Journal of Marketing, 41 (April),51-61.

Day, G. S. (1977). "Diagnosing the Product Portfolio," Journalof Marketing. 41 (April), 29-38.

Fitzpatrick, P. J. (1931), Symptoms of Industrial Failures asRevealed hy an Analysis of the Financial Statements of theFailed Companies, Washington, D. C ; Catholic Universityof America.

Hartley, R. F. (1976). Marketing Mistakes. Columbus, Ohio;Grid Publishing Inc.

Hazlitt. H. (1959), "The Failure of the 'New Economies': AnAnalysis of the Keynesian Fallacies." Princeton, New Jersey:D. Van Nostrand Company, Inc.

Houston. B. (1972). "Let's Put More Espirit in De Corpora-tion," Harvard Business Review, 50 (November-December),55-61.

Joy, O. M. and J. O. Tollefson (1975), "On the Financial Ap-plications of Discriminant Analysis," Journal of Financialand Quantitive Analysis. 10 (December). 723-39.

Juenger, E. G. (1949). The Failure of Technology: PerfectionWithout Purpose. Hinsdale, Illinois; Henry Regnery Compa-ny.

Karger, D. W. and Z. A. Malik (1975), "Long-Range Planningand Organizational Performance." Long Range Planning, 8(December), 6(}-4,

Kaye, J. L. and L. N. Garter (1979), "Early Warning Signs ofBusiness Failure," Credit and Financial Management, 81(June), 36-7.

Kotler. P.. W. Gregor and W. Rogers (1977), "The MarketingAudit Comes of Age," Sloan Management Review. 18 (Win-ter), 25^2 .

Lazo. H. (1965). "Finding a Key to Success in New ProductFailures," Industrial Marketing. 50 (November), 74-7.

The McGraw-Hill Dictionary of Modern Economics: A Handbookof Terms and Organizations (1973). New York: McGraw-HillBook Company.

Miller. D. (1977). "Common Syndromes of Business Failure,"Business Horizons, 20 (December). 43—53.

and P. H. Friesen (1977), "Strategy-Making in Context;Ten Empirical Archetypes.'' The Journal of Management Stud-ies, 14 (October), 253-80.

Ross, J. E. and M. J. Kami (1973), Corporate Management inCrisis: Why The Mighty Fail, Princeton, New Jersey: Prentice-Hall, Inc.'

Schendel, D. and G. R. Patton (1978). "A Simultaneous Equa-tion Model of Corporate Strategy," Management Science, 24(November). 1611-21.

Schoeffler. S. (1955). The Failures of Economics: A DiagnosticStudy. Cambridge: Harvard University Press.

Shooman, M. L. (1968), Probabilistic Reliability: An Engineer-ing Approach. New York; McGraw-Hill Book Company.

Smith. R. A. (1966), Corporations in Crisis, New York: Double-day and Co, Inc.

Taffler, R. and H. Tisshaw (1977), "Going. Going. Gone—FourFactors Which Predict," Accountancy. 88 (March), 50-4.

Time (1980). "Sears Searches for Success," 115 (May 5), 88.Van Home, J. C. (1977), Financial Management and Policy, 4th

edition. New York; Prentiee-Hall, Inc.Walters. J. E. (1957), "The Determination of Technical Solven-

cy." Journai of Business, 30 (January), 30-45.Webster's Third New International Dictionary (1961). Chicago;

R. R. Donnelly and Sons Inc.Wind, Y. and V. Mahajan. Portfolio Analysis and Strate-

gy. Reading, Massachusetts: Addison-Wesley, forthcoming.Worthing, J. C. (1966). "Who Fails and Why." y\<ivance(/Man-

agement Journal, 31 (January), 21-6,Wyant, R. (1972), "Business Failures," Dun's, 99 (June), 131-2.

Early Warning Indicators of Business Failure / 89