Professional Documents
Culture Documents
Accounting Research Center, Booth School of Business, University of Chicago
Accounting Research Center, Booth School of Business, University of Chicago
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Wiley and Accounting Research Center, Booth School of Business, University of Chicago are collaborating
with JSTOR to digitize, preserve and extend access to Journal of Accounting Research.
http://www.jstor.org
WILLIAM H. BEAVER*
dale, "Credit Department Methods," Bankers' Magazine (1908), pp. 183-84. For an
excellent review of the historyand uses of ratios: James 0. Horrigan, "An Evaluation
of the Usage of Ratios in External Financial Statement Analysis" (unpublished dis-
sertation proposal, Graduate School of Business, The University of Chicago, 1963).
'Of the 79 failed firmsstudied, 59 were bankrupt; 16 involved nonpayment of pre-
ferredstock dividends; 3 were bond defaults; and 1 was an overdrawn bank account.
For a more complete discussion of the various definitionsof failure: Charles W. Ger-
stenberg, Financial Organization and Management of Business (3rd ed., rev.; New
York: Prentice-Hall, Inc., 1951), p. 569.
71
4A frequencydistribution
of the fiscalyearof failureappearsin Table A-1 in the
Appendix.
'The industrialcompositionof the sample appearsin Table A-2 in the Appendix.
12,000 Leading U.S. Corporations (New York: News Front Magazine). The list
the ratios of the two firmsare identical (i.e., even if the ratios have the
same numericalvalues).
If firmsare viewed as aggregatesof assets and if asset returnsare less
than perfectlycorrelatedwith one another,statistical formulaesuggest
that the variability of total returnto the firmwill increase less than
proportionately to the size of the firm.10 The rate of returnto the firm
will become more stable as asset size increases.Empirical evidence indi-
cates that the variabilityof rate of returndoes behave in this manner.1"
The implicationis that larger firmsare more solvent,even if the value
of theirratios is the same as that of smallerfirms.12 Simply stated, the
ratios of firmsfromdifferent asset-size classes cannot be directlycom-
pared.
The paired-sampledesign in conjunctionwith the paired analysis is
one way of compensatingfor the effectsof industryand asset size.
Whenever ratios are directly compared, the comparison involves two
firmsfromthe same industryand asset-size class. As previouslyindi-
cated, each failed firmhas a nonfailed "mate" in the sample. In the
paired analysis, the ratios of the nonfailed firmsare subtractedfrom
the ratios of the failed firms;the analysis is then based upon the dif-
ferencesin the ratiosratherthan upon the ratios themselves.Since every
difference in the ratios always involves a comparisonof two firms(one
failed and one nonfailed) fromthe same industryand asset-size class,
the potentiallydisruptiveeffectsof industryand asset size are mitigated.
A paired design and a paired analysis were employed in the previous
study.
Althoughthe paired analysis is a legitimateapproach to investigating
predictiveability,it has one seriousdrawback.It cannot draw inferences
regardinga single observation-only about pairs of observations.The
earlier study is restrictedto such statementsas: Based upon an analy-
sis of the ratios, firmA is more (or less) solventthan firmB. However,
this is an ambiguousstatementbecause it ignoresthe obvious question:
How solventis firmB? To be less solventthan firmB may not be so bad,
if firmB is an extremelysolvent firm.On the other hand, to be more
solventthan firmB may be very bad, if firmB is extremelyinsolvent.
In a decision-makingsituation,the user of ratios wishes to make infer-
ences regardinga single firm-for example, what is the probabilityof
failureforthis firm?Such inferencesare beyondthe scope of the paired
analysis.
In this studythe paired analysis was droppedin favorof an unpaired
TABLE 1
List of Ratios Testeda
manner.The bias would arise in that the failed firmswould have dem-
onstratedan ability to survive for at least five years; the result might
be the eliminationof the most serious failures.
Computationof Ratios
For every set of financialstatementsavailable, 30 ratios were com-
puted.15Three criteriawere used to selectthe 30 ratios fromthe set of all
possible combinationsand permutationsof financial statementitems.
The firstcriterionwas popularity-frequentappearance in the litera-
ture.16The ratios advocated in the literatureare perceivedby many to
to see the extentto
reflectthe crucial relationships.It will be interesting
which popularitywill be self-defeating-thatis, the most popular ratios
" The 30 ratiosappearin Table 1.
'" The selectionof ratioswas based upon theirfrequencyof appearancein nine-
teenfinancial-statementanalysistexts.
group over the five-yearperiod. Where two ratios predicted about equally well, other
factors entered, such as frequency of appearance in the literature and superior per-
formancein previous studies. The percentage error refersto the dichotomous classifi-
cation test which is explained on pp. 83-85.
Analysis of Evidence
The differencein the mean values is in the predicteddirectionforeach
ratio in all fiveyears beforefailure.20Failed firmsnot only have lower
cash flow than nonfailed firmsbut also have a smaller reservoirof
liquid assets. Although the failed firms have less capacity to meet
obligations,theytendto incurmoredebt than do the nonfailedfirms.
The trendline of the nonfailedfirmshas a zero slope, and the devia-
'1 A moredetaileddiscussionappearsin JamesE. Walter's,"The Determination of
TechnicalSolvency,"Journalof Business,January,1957,pp. 30-43.
1 The intervalmeasuresappear in GeorgeH. Sorter'sand GeorgeBenston's"Ap-
TABLE 2
Predictionof theMean Values ofFailed and NonfailedFirms
Ratio Predictiona
tions fromthe trendline are small. Yet the deteriorationin the means of
the failed firmsis very pronouncedover the five-yearperiod.The differ-
ence in means is evident for at least five years before failure, with
the differenceincreasingas the year of failureapproaches.
The data demonstratea substantialdegreeof consistency.The evidence
overwhelmingly in the ratios of failed
suggeststhat there is a difference
and nonfailedfirms,
ComparisonwithPreviousStudies
In 1932, Fitz Patrick published a study of nineteenpairs of failed
and nonfailedfirms.21His evidence indicatedthat therewere persistent
differencesin the ratios for at least three years prior to failure. The
Winakor and Smith study of 1935 investigatedthe mean ratios of failed
firmsforten years priorto failureand found a marked deteriorationin
the mean values with the rate of deteriorationincreasing as failure
approached.22In 1942, Merwin comparedthe mean ratios of continuing
+45- +1 .79'
+.35 .78
+.25 3 0 .652
+.15. O .58
-.05 / .44.
/d
-.15. /-.2 .37
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
YR. BEFORE FAILURE YR. BEFORE FAILURE YR. BEFORE FAILURE.
WORKING CAPITAL CURRENT NO CREDIT
TOTAL ASSETS RATIO INTERVAL
.366
.24 /
.12 /
.06 2.01(15
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
YR. BEFORE FAILURE YR. BEFORE FAILURE YR. BEFORE FAILURE
TOTAL ASSETS
(IN MILLIONS OF DOLLARS)
8F
7-
/\Q
6-
1 2 3 4 5
YR. BEFORE FAILURE
of mean values.
FIG. 1. Profileanalysis,comparisoin
Limitationsof ProfileAnalysis
Profileanalysis can demonstratethat a difference betweenfailed and
nonfailedfirmsexists,but it cannot answer the crucial question: How
large is the difference?The profileconcentratesupon a single point on
the ratio'distribution-the mean. Without the additional knowledgeof
the dispersionabout that point, no meaningfulstatementcan be made
regardingthe predictiveability of a ratio.
Consider two ratio distributions,each of which is symmetrical,and
assumethat a difference in theirmeans is observed.If the dispersionabout
themeans is tight,therecould be littleor no overlap of one distributionon
the other.Little or no overlap impliesthat the ratio would be an excel-
lent predictorof failure.However, if the dispersionabout the means is
great, there could be quite a bit of overlap, which would indicate a
lesserdegreeof predictiveabilitypresent.
If the distributionsare skewed (i.e., nonsymmetrical), the extremeob-
servations,'which constitutea small proportionof the total numberof
firms,may be responsibleformost of the difference in the means. Apart
fromtwo or three extremefirms,there could be a completeoverlap of
the distributionsof failed and nonfailedfirms.Previous studies indicate
that ratio distributions
do have pronouncedskews.24
The'precedingremarksimplythat ratios may have little or no ability
to predictfailure,in spite of the differencesin the means. The discussion
suggeststhat some sort of predictivetest is needed.25
Dichotomous ClassificationTest
The dichotomousclassificationtest predicts the failure status of a
firm,based solely upon a knowledgeof the financialratios. In contrast
to the profileanalysis,it is a predictivetest, althoughit will not provide
as much insight as likelihood ratios. Conducting this intermediate
analysis has certainadvantages.
The classificationtest is an intuitivelyappealing approach. It closely
resembles the decision-makingsituation facing many users of ratios.
3Charles L. Merwin, Financing Small Corporations in Five Manufacturing Indus-
tries,1926-36 (New York: National Bureau of Economic Research, 1942).
24 The Earning Power Ratios of Public Utility Companies (Bureau of Business Re-
TABLE 3
Percentage Dichotomous
of FirmsMisclassiJieda: Classification
Test
Year beforeFailure
Ratio
1 2 3 4 5
No-creit
inerval.23 .38 .43 .38 .37
No-creditinterval (.23) (.31) (.30) (.35) (.30)
the firsttest.
IndustryEffects
One crude test of the residual effectsof industryis the amountof in-
crease in percentageerrorfromthe firsttest to the second test,since the
industrycompositiondiffersbetweenthe two subsamples. The increase
in erroris small, especially forthe betterratios,and this findingis con-
sistentwiththe assumptionthatthe residualeffectsare not great.
Another indirecttest is the increase in percentage error from the
paired analysis to the unpaired analysis.29In most ratios the increase is
small but persistent.Althoughthe reductionin predictivepoweris by no
means overwhelming, there seems to be a small residual effectof indus-
try.To the extentthat the effectsare still present,the predictiveability
of ratios is understated.Predictiveability would be expectedto improve
if the effectof industrywere fullytaken into account.
Both tests are not directmeasures of the residual effects,because the
increasein percentageerrorscould be due to otherfactorssuch as residual
asset-size effects.The tests offerevidence of an indirectsort-that is,
nothinghas been observedto suggestthat the residualeffectsare great.
TABLE 4
Comparisonof PercentageErrorsfor Paired and UnpairedClassificationTests
Year beforeFailure
Ratio
1 2 3 4 5
TABLE 5
Correlationof Total Assets withSix Ratios, First Year Before Failure
A B A B
ContingencyTables
The percentageof incorrectpredictionssuffersfromtwo limitations.
(1) In a world wherethe costs of misclassifyinga failed firmare likely
to be much greater than the costs associated with misclassifyinga
nonfailedfirm,it is importantto know the probabilityof misclassifying
a failed firm (Type I error) versus the probabilityof misclassifyinga
nonfailedfirm(Type II). (2) The difference in percentageerrorbetween
the ratios and the random-predictionmodel can vary substantially
by alteringthe probabilityof failure. If the probabilityof failure for
the sample differsfromthat of the total population, a comparisonof
percentageof total errorsis not verymeaningful.
The Type I and Type II errorsfor a given sample will be the same
as that forthe population,even if the probabilityof failureforthe pop-
ulation is differentfromthat of the sample.80The previous statement
requires one assumption-namely, that the ratio distributionsof the
sample accurately describe the ratio distributionsof the population.
Note also that the Type I and II errorsdiscussed are those obtained
whenthe cutoffpointswere selectedon the basis of minimizingthe total
numberof incorrectpredictionswithoutregardforthe Type I and Type
II errorsimplied. Since this mightbe a disastrousdecision rule where
the misclassificationcosts are asymmetrical,the cutoffpoints can be
easily alteredto any desiredratio of Type I to Type II errors.
One way of presentingType I and Type II errorsis throughcontingency
tables, as shownin Table 6.31 The row denotesthe predictionmade; the
columndenotesthe actual status of the firm;and each cell containsthe
numberof firmsfulfillingeach condition.For example,in the firstyear
beforefailure,the total numberof failed firmswas 79, of which62 were
correctlyclassifiedas failed and 17 were not. The total numberof non-
failed firmswas also 79, of which75 were correctlyclassifiedand 4 were
not. The Type I errorwas 17/79 or 22 per cent, and the Type II error
was 4/79 or 5 per cent. The total errorwas 21/158 or 13 per cent-the
numberpreviouslyreferred to as the percentageerror.
The Type I and Type II errorsare probabilitiesof errorconditional
upon the actual status of the firm.Anotherset of conditionalprobabili-
ties of erroris the probabilityof errorconditionalupon the prediction
made.The probability failure,is 4/66 or
of errorgiventhe prediction,
6 per cent.The probability
of errorgiventhe prediction,
nonfailure,is
17/72or 18 per cent.The two sets of conditionalprobabilitiesare merely
different ways of expressingthe same underlyingrelationships.Either
can be appropriatedependingupon the inferencesdesired.
It is interestingto comparethe Type I and II errorsof a ratio with
those obtained froma random-prediction model. For example, consider
a model constrainedso that the ratio of failed to nonfailedpredictions
is the same as that of the financialratio. In the firstyear beforefailure
the Type I and II errorswould be 58 per cent and 42 per cent,respec-
tively,whilein the fifthyear the Type I and II errorswould be 71 and 29
per cent,respectively.The random-prediction model could be altered in
many otherways, but the tenorof the inferenceswould remainthe same.
TABLE 6
Cash Flow to TotalDebt: Contingency
Table
Year One Year Two Year Three
21 32 34
- = .13 3 .21 34-= .23
158 153 1 150
FLD 33 2 35 FLD 31 3 34
NFLD 29 64 93 NFLD 23 60 83
26
31=.24 =22
RELATIVE
FREQUENCY
.35
.30
.25
.20 r
.15-
3 FAILED 5NONFAILED
.10 *
.05 I i
RELATIVE
FREQUENCY
.25
.20 r-
r- W E
.15 I _
.10 6 NONFAILED
3 FAILED
< ~~~~I ~ ~lI ~
.051-L J
n I Fn
FIG. 2. Histogram.
valuesoftheratio,whiletheverticalaxisindicatestherelativefrequency
withwhichthe ratiosof failed(or nonfailed)firmsfall intoeach class
interval.For example,in the firstyear,28 per centof the ratiosof the
nonfailedfirmsfallin theinterval.1 to .2, and 21 per centof theratios
offailedfirms fallin theinterval-.1 to - .2.
Consistencyof Data
Histograms can provide insightsinto financial ratios, apart from a
For example, the inferencesdrawn from
likelihoodratio interpretation.
total assets.The data appearin Table A-7 in the Appendixand containthe same in-
formation thata histogramcontains.
RELATIVE
FREQUENCY
.30
.25
.20L
.15
.10 r -5 NON-FAILED
.05
-,n raw--
L_] m~?- r-iy
-.9 -.7 -.5 -.3 -.1 0 .1 .3 .5 .7 .9 1.1
CASH FLOW TO TOTAL DEBT(YEAR 3)
RELATIVE
FREQUENCY
.40
l l
.35 I
.30 1
.25 I l
.20 I
l l
.15 - -
.05 r- - I-
_ ~r-' II!1
-.9 .7 -.5 -.3 .1 0 .1 .3 .5 .7 .9 1
CASH FLOW TO TOTALDEBT(YEAR 4)
NONFAILED FIRMS - - - - FAILED FIRMS
FIG. 2-Continued
RELATIVE
FREQUENCY
.35
.30 1
I I
.25 l I
I l
.20
I I
.15
_1 4 NONFAILED
.10.10 X II FAILED
.05 I I |
r- - - -L-- *_
-.9 -.7 -.5 -.3 -.1 0 -.1 .3 .5 .7 .9 1.1
CASH FLOW TO TOTAL DEBT (YEAR5)
NONFAILED FIRMS - - - FAILED FIRMS
FIG. 2-Continued
Normalityof Data
The normalityof the data is best analyzed throughthe use of cumula-
tive density functions (cdf). The vertical axis indicates the fractile
associated with the correspondingvalue of the ratio, which is shown
on the horizontalaxis. The fractileis the proportionof the distribution
whose values are equal to or below a given value of the ratio. The
fractileforthe ithobservationis the "i/n + 1" fractile,where "n" is the
numberof failed (or nonfailed) firms.The ith observationis determined
FRACTILE
99
,98 x + 5 NONFAILED
x +
.60-3.
.50-
.40- + ? + FAILEDFIRMS(YEAR1
XDENOTES )1
.30 X 04
X Xxx) ++++
.10 x X++
.05- X X ++ + ++ oDENOTES NONFAILEDFIRMS-(YEAR1)
<-3 ~~~~~~+0 0 XDENOTES FAILED FIRMS (YEAR 1)
.024-3FAILED (YR. 1) + DENOTES FAILED FIRMS (YEAR 5)
.01 +
<-1 FAILED (YR. 5)
by arraying the data in ascending order. The lowest value is the first
observation, while the highest value is the nth.The .70 fractile of the failed
firms(Fig. 3) in the firstyear is zero, while the .70 fractileforthe non-
failed firmsis .40. The vertical axis is calibrated accordingto the arith-
metic probabilityscale. If the data are normal, the distributionwill
appear as a straightline. Departure fromlinearitycan be interpretedas
departurefromnormality.
The cash-flowto total-debtratio has pronouncedskews in each of the
years before failure.The other five ratios also do not conformto the
normalityassumption.4 Simple transformations(e.g., log and square
root) have been performedon the ratios in the hope that the transformed
variables would be approximatelynormal. The initial results indicate
that the transformed variables are almost as badly skewedas the original
variables. The departure fromnormalityhas serious implications for
attemptsto derive a multiratiomodel, since most of the existingmulti-
variate techniquesrely upon the normalityassumption.
Likelihood Ratios
None of the discussionthus far has been directedtoward the use of
ratios as assessmentsof likelihoodratios. The study of financialratios
as predictorsof failure is placed in its broadest context throughthe
discussion of likelihoodratios. Their use is essentially a Bayesian ap-
proach and can be explained most convenientlyby referringto the
I The data in Table A-7 can be used to formcdf'sforthe otherfiveratios.Table
A-7 appearsin theAppendix.
followingtableau:
Event PriorProbability Likelihood JointProbability PosteriorProbability
P(RF) = P(F/R)
P(R?)
P(RF) P(F/R)
P(P/)
Likelihood-odds ratio P(R/F)
CONCLUDING REMARKS
Summaryand Conclusions
Beforerestatingthe conclusions,I would like to provideone finalword
of warningregardingthe uncertaintysurroundingthis empiricalstudy.
There are many factorsthat have preventeda measurementof the "true"
predictiveability of ratios. In addition to the factors previouslymen-
tioned,two more come to mind. Both stem fromthe fact that financial
ratiosare in widespreaduse as measuresof solvency.
If ratios are used to detectthe financial"illness" of a firm,theremay
be many firmswhose illnesses were detectedbeforefailureoccurred.In
these cases, the propertreatmentwas applied, and the firmsdid not fail.
The sample of failed firmswill include those firmswhose illnesses were
not detectablethroughratios. This is a biased sample for investigating
the usefulnessof ratios. An importantpiece of informationis missing-
how many firmswere saved fromfailurebecause theirproblemswere de-
tectedin timethroughthe use of ratios? Such information would be diffi-
cult, if not impossible,to obtain. However,this argumentsuggeststhat
the studymay be understating the usefulnessof ratios.
On the other hand, ratios are used by financialinstitutionsto deter-
minethe credit-worthiness of its borrowers.In many cases, lines of credit
may be severed by the lendinginstitutionsbecause the borrowingfirm
failed to improveits financialratio to a "respectable"level. This argu-
ment suggeststhat the predictiveability of ratios may be overstated,
relativeto what it would be if ratios were not so popular. There exists a
countlessnumberof argumentson both sides, regardingthe possible bi-
ases in the data. This briefdiscussionshould suggestthat the conclusions
that followare arrivedat withmuchless than perfectcertainty.
The ratio distributionsof nonfailedfirmsare quite stable throughout
the fiveyears beforefailure.The ratio distributionsof the failed firms
exhibita markeddeteriorationas failureapproaches.The resultis a wid-
eninggap betweenthe failed and nonfailedfirms.The gap producesper-
sistentdifferences in the mean ratios of failed and nonfailedfirms,and
the difference increasesas failureapproaches.
For the cash-flowto total-debtratiothe overlapof the two distributions
is small in the firstyear beforefailure.By the fifthyear,the overlap has
increasedbut the two distributionsare still distinguishable.The degree
of overlap is reflectedin the abilityto correctlypredictthe failurestatus
of firms.The cash-flowto total-debtratio has the ability to correctly
classifyboth failed and nonfailedfirmsto a much greaterextentthan
would be possible throughrandom prediction.This ability exists for at
least fiveyears beforefailure.
Althoughratio analysis may provide useful information,ratios must
be used withdiscretion:(1) Not all ratiospredictequally well. The cash-
flowto total-debt ratio has excellentdiscriminatorypower throughout
the five-yearperiod. However, the predictivepower of the liquid asset
ratios is much weaker. (2) The ratios do not predict failed and non-
1954 2
1955 6
1956 14
1957 7
1958 6
1959 4
1960 7
1961 10
1962 9
1963 10
1964 4
Total 79
103
TABLE A-2
IndustrialClassification
No. of
Industrya Failed
Firms
Manufacturing
Food products(201) 1
Grainmill products(204) 1
Bakeryproducts(205) 3
Confectionery (207) 1
Beverages (208) 3
Textile (221) 4
Women'sapparel (233) 1
Industrialchemicalsand plastics (281) 3
Petroleumrefining(291) 3
Rubber products(301) 1
Leather products(311) 2
Clay products(325) 1
Concrete,gypsum(327) 1
Iron and Steel foundries(331) 2
Nonferrousmetals (333) 1
Cutlery,etc. (342) 3
Heating and plumbing(343) 2
Structuralmetal products(344) 1
Etching,engraving(347) 2
Engines,industrialmachinery(351) 3
Service,industrialmachines(358) 1
Householdappliances,etc. (363) 5
Commercialequipment(366) 1
Electronics(367) 6
Motor vehicles (371) 5
Aircraft(372) 3
Scientificinstruments(381) 1
Watches (387) 2
Toys, sportinggoods (394) 1
Nonmanufacturing
Public warehouse(422) 1
Hardware,wholesale (507) 1
Generalmerchandise(532) 3
Grocerystores (541) 2
Apparelstores (561) 3
Furniturestores (571) 1
Drug stores (591) 2
Personal services(721) 1
Motion pictures(781) 1
a SIC number.
The numberin parenthesesindicatesthe three-digit
TABLE A-3
Number of Observations on Which Financial-Statement Data Could Be Obtained
Year beforeFailure
Item
. 2 3 4 5
Failed firms.79 76 75 62 54
Nonfailed firms.79 77 75 66 63
TABLE A-4
Percentage Error for 30 Ratios on Dichotomous Classification Test"
Year beforeFailure
Ratio
12 3 4 5
Group I
Cash flow .14 .20 .29 .37 .44
Sales (.11) (.19) (.24) (.33) (.31)
Cash flow .10 .20 .24 .28 .28
Total assets (. 10) (.17) (.20) (.26) (.25)
Cash flow .13 .21 .28 .38 .37
Net worth (. 11) (.16) (.24) (.31) (.32)
Cash flow .13 .21 .23 .24 .22
Total debt (.10) (.18) (.21) (.24) (.22)
Group11
Net income .13 .22 .28 .35 .31
Sales (. 09) (.16) (.24) (.28) (.27)
Net income .13 .21 .23 .29 .28
Total assets (.12) (.15) (.22) (.28) (.25)
Net income .13 .26 .26 .34 .40
Net worth (.10) (.18) (.26) (.31) (.28)
Net income .15 .20 .22 .26 .32
Total debt (.08) (.16) (.20) (.26) (.26)
Group III
Currentliabilities .30 .41 .36 .40 .46
Total assets (.27) (.28) (.34) (.33) (.30)
Long-termliabilities .36 .45 .42 .47 .51
Total assets (.32) (.40) (.40) (.38) (.41)
Current+ long termliabilities .23 .30 .36 .39 .38
Total assets (.19) (.26) (.29) (.31) (.33)
Current+ longterm+ preferred .19 .25 .34 .27 .28
Total assets (.19) (.24) (.28) (.24) (.27)
Group IV
Cash .28 .29 .30 .36 .38
Total assets (.25) (.28) (.30) (.34) (.31)
Quick assets .38 .42 .36 .48 .40
Total assets (.34) (.36) (.36) (.37) (.34)
a Top rowofnumbersaretheresultsofthesecondtest;thebottomrowofnumbers,
enclosedin parentheses,referto the resultsof the firsttest.
Group IV-Continued
Currentassets .38 .48 .48 .47 .49
Total assets (.37) (.44) (.43) (.43) (.38)
Workingcapital .24 .34 .33 .45 .41
Total assets (.20) (.30) (.33) (.35) (.35)
GroupV
Cash .22 .28 .36 .38 .38
Currentliabilities (.22) (.24) (.28) (.34) (.29)
Quick assets .24 .32 .40 .34 .37
Currentliabilities (.24) (.30) (.28) (.34) (.29)
Currentassets .20 .32 .36 .38 .45
Currentliabilities (.20) (.27) (.31) (.32) (.31)
GroupVI
Cash .34 .24 .36 .43 .45
Sales (.30) (.24) (.34) (.39) (.41)
Receivables .46 .45 .46 .43 .42
Sales (.40) (.39) (.38) (.37) (.38)
Inventory .47 .50 .54 .48 .53
Sales (.40) (.50) (.47) (.42) (.42)
Quick assets .46 .47 .45 .52 .44
Sales (.40) (.42) (.40) (.48) (.42)
Cur. assets .44 .51 .48 .49 .51
Sales (.42) (.40) (.42) (.47) (.47)
Workingcapital .26 .33 .42 .46 .40
Sales (.23) (.27) (.35) (.40) (.37)
Net worth .32 .36 .44 .42 .40
Sales (.28) (.33) (.38) (.38) (.38)
Total assets .37 .42 .38 .42 .44
Sales ( .34) (.40) (.34) (.38) (.39)
Defensive
Interval39 .41 .46 .48 .51
DefensiveInterval (.34) (.40) (.41) (.47) (.43)
TABLE A-5
CutoffPoints Used in Classification Test for Six Ratiosa
Ratio 1 2 3 4 5
A B A B A B A B A B
Cash flow
.03 .07 .05 .07 .10 .09 .09 .09 .11 .11
Total debt
Net income
.00 .02 .01 .02 .03 .03 .02 .02 .04 .03
Total assets
Total debt
..57 .57 .51 .49 .53 .50 .58 .57 .57 .57
Total assets
Working capital .19 .27 .33 .28 .26 .26 .40 .24 .43 .29
Total assets
Current assets 1.6 1.6 2.3 1.7 2.3 1.8 2.6 1.8 2.8 2.1
Currentliab.
No credit interval -.04 -.04 .03 -.02 .01 -.01 .00 -.01 .04 -.02
Total assetsb 4.6 4.6 5.2 5.2 4.4 4.4 5.5 4.0 4.8 2.5
ActualOutcome
Fa F T F F T F P T
Year One
F 66 9 75 56 7 63 54 13 67
F 13 70 83 23 72 95 25 66 91
T 79 79 158 79 79 158 79 79 158
Year Two
F 57 13 70 53 16 69 48 24 72
F 19 64 83 23 61 84 28 53 81
T 76 77 153 76 77 153 76 77 153
Year Three
F 49 9 58 36 12 48 38 13 51
F 26 66 92 39 63 102 37 62 99
T 75 75 150 75 75 150 75 75 150
Year Four
F 34 9 43 43 14 57 30 25 55
F 28 57 85 19 52 71 32 41 73
T 62 66 128 62 66 128 62 66 128
Year Five
F 36 15 51 25 4 29 36 30 66
F 18 48 66 29 59 88 18 33 51
T 54 63 117 54 63 117 54 63 117
ActualOutcome
F P T F F T F P T
Year One
F 55 7 62 52 9 61 46 27 73
F 24 72 96 27 70 97 33 52 85
T 79 79 158 79 79 158 79 79 158
Year Two
F 44 18 62 49 31 80 46 34 80
F 32 59 91 27 46 73 30 43 73
T 76 77 153 76 77 153 76 77 153
Year Three
F 42 21 63 46 36 82 40 32 72
F 33 54 87 29 39 68 35 43 78
T 75 75 150 75 75 150 75 75 150
Year Four
F 34 21 55 35 21 56 31 32 63
F 28 45 73 27 45 72 31 34 65
T 62 66 128 62 66 128 62 66 128
Year Five
F 33 31 64 34 23 57 25 26 51
F 21 32 53 20 40 60 29 37 66
T 54 63 117 54 63 117 54 63 117
aF denotesfailed,P denotesnonfailed,and T denotestotal. Cells containnumber
conditionsregardingpredictedand actual outcomes.
offirmsfulfilling
109
Fa | F F F F
Below -.65 .051 .000 .000 .000 0 to .1 .013 .063 .018 .032
-.65 to -.60 .025 .000 .000 .000 .1 to .2 .025 .152 .093 .206
-.60 to -.55 .000 .000 .000 .000 .2 to .3 .038 .152 .112 .143
-.55 to -.50 .000 .000 .000 .000 .3 to .4 .089 .240 .112 .143
-.50 to -.45 .038 .000 .018 .000 .4 to .5 .076 .165 .093 .221
-.45 to -.40 .013 .000 .000 .000 .5 to .6 .102 .152 .183 .201
-.40 to -.35 .013 .000 .000 .000 .6 to .7 .152 .051 .183 .015
-.35 to -.30 .051 .000 .018 .000 .7 to .8 .038 .025 .074 .032
-.30 to -.25 .038 .000 .018 .000 .8 to .9 .076 .000 .074 .000
-.25 to -.20 .089 .000 .018 .000 .9 to 1.0 .102 .000 .037 .000
-.20 to -.15 .127 .000 .037 .000 1.0 to 1.1 .063 .000 .000 .000
-.15 to -.10 .076 .000 .037 .000 1.1 to 1.2 .038 .000 .000 .000
-.10 to -.05 .102 .000 .037 .016 1.2 to 1.3 .076 .000 .000 .000
-.05 to -.00 .202 .038 .093 .032 1.3 to 1.4 .025 .000 .000 .000
.00 to .05 .152 .367 .425 .318 1.4 to 1.5 .013 .000 .018 .000
.05 to .10 .025 .418 .223 .430 1.5 to 1.6 .025 .000 .000 .000
.1O to .15 .000 .140 .037 .127 1.6 to 1.7 .000 .000 .000 .000
.15 to .20 .000 .025 .000 .048 1.7 to 1.8 .000 .000 .000 .000
.20 to .25 .000 .013 .018 .032 1.8 to 1.9 .000 .000 .000 .000
.25 to .30 .000 .000 .018 .000 1.9 to 2.0 .000 .000 .000 .000
Above .30 .000 .000 .000 .000 above 2.0 .038 .000 .000 .000
Below -1.0 .025 .000 .000 .000 Oto .5 .139 .000 .018 .000
-1.0 to -.9 .013 .000 .000 .000 .5 to 1.0 .202 .000 .055 .000
-.9 to -.8 .000 .000 .000 .000 1.0 to 1.5 .254 .013 .223 .079
-.8 to -.7 .000 .000 .000 .000 1.5 to 2.0 .165 .152 .202 .143
-.7 to -.6 .013 .000 .000 .000 2.0 to 2.5 .038 .254 .186 .159
-.6 to -.5 .025 .000 .000 .000 2.5 to 3.0 .038 .139 .129 .190
-.5 to -.4 .025 .000 .000 .000 3.0 to 3.5 .038 .152 .055 .095
-.4 to -.3 .051 .000 .000 .000 3.5 to 4.0 .013 .063 .037 .048
-.3 to -.2 .038 .000 .000 .000 4.0 to 4.5 .038 .038 .000 .048
-.2 to -.1 .038 .000 .037 .000 4.5 to 5.0 .013 .038 .055 .048
-.1 to 0 .114 .000 .037 .000 5.0 to 5.5 .000 .025 .000 .037
o to .1 .139 .025 .148 .048 5.5 to 6.0 .000 .013 .000 .048
.1 to .2 .177 .089 .130 .095 6.0 to 6.5 .000 .051 .000 .037
.2 to .3 .114 .152 .148 .111 6.5 to 7.0 .000 .013 .000 .048
.3 to .4 .063 .254 .183 .238 7.0 to 7.5 .000 .025 .000 .000
.4 to .5 .063 .228 .148 .206 7.5 to 8.0 .000 .000 .000 .000
.5 to .6 .038 .127 .111 .190 8.0 to 8.5 .038 .013 .018 .000
.6 to .7 .051 .114 .018 .048 8.5 to 9.0 .000 .000 .000 .000
.7 to .8 .013 .013 .037 .064 9.0 to 9.5 .013 .000 .000 .000
.8 to .9 .000 .000 .000 .000 9.5 to 10.0 .013 .000 .000 .000
.9 to 1.0 .000 .000 .000 .000 Abovel0.0 .000 .013 .018 .018
110
TA B LE A - 7-Continued
Below -.9 .013 .000 .037 .000 0 to 1 .051 .000 .074 .016
-.9 to -.8 .013 .000 .000 .000 1 to 2 .127 .051 .185 .111
-.8 to -.7 .013 .000 .000 .000 2 to 3 .139 .114 .111 .175
-.7 to -.6 .013 .000 .000 .000 3 to 4 .114 .127 .148 .175
-.6 to -.5 .025 .000 .000 .000 4 to 5 .238 .114 .130 .048
-.5 to -.4 .051 .000 .000 .000 5 to 6 .051 .114 .074 .064
-.4 to - .3 .102 .000 .018 .000 6 to 7 .000 .051 .055 .032
-.3 to - .2 .076 .013 .055 .000 7 to 8 .089 .063 .074 .048
-.2 to -.1 .165 .000 .129 .048 8 to 9 .038 .038 .018 .095
-.1 to 0 .263 .316 .315 .286 9 to 10 .025 .051 .037 .079
0 to .1 .190 .392 .260 .334 10 to 11 .051 .089 .018 .032
.1 to .2 .063 .165 .074 .206 11 to 12 .000 .025 .000 .016
.2 to .3 .000 .063 .055 .032 12 to 13 .000 .000 .000 .000
.3 to .4 .013 .000 .018 .032 13 to 14 .013 .025 .000 .016
.4 to .5 .000 .000 .000 .032 14 to 15 .013 .013 .000 .032
.5 to .6 .000 .025 .018 .000 15 to 16 .013 .013 .018 .016
.6 to .7 .000 .000 .000 .000 16 to 17 .000 .038 .000 .016
.7 to .8 .000 .013 .000 .000 17 to 18 .013 .013 .018 .000
.8 to .9 .000 .00 .000 .000 18 to 19 .000 .013 .000 .016
.9 to 1.0 .000 .000 .000 .000 19 to 20 .000 .000 .018 .000
1.0 to 1.1 .000 .000 .018 .000 Above2O .038 .051 .018 .016
Above 1.1 .000 .013 .000 .032
Total Relative 1. 000 1. 000 1. 000 1. 000 Total Relative 1. 000 1. 000 1. 000 1. 000
Frequency Frequency
firms.
b In millions of dollars.