heprovidedevidencetobackuphisclaim.Sincehisstudy,
both academicsandpractitionershavedeveloped theirown
versions of these credit scores.
Notwithstanding its usefulness in predicting bankruptcy,
lineardiscriminantanalysisdoesnotprovideaprobabilityof
bankruptcy. To arriveatsuch an estimate, weuse a close
variant—aprobit.Inaprobit,webeginwiththesamedata
that wasused in linear discriminant analysis,a sample of
firmsthatsurvivedaspecificperiodandfirmsthatdidnot.
Wedevelopanindicatorvariablethattakesonavalueof 0 or
1:
Wethenconsiderinformationthatwouldhavebeenavailable
atthebeginningoftheperiodthatmighthaveallowedusto
separatethefirmsthatwentbankruptfromthefirmsthatdid
not.Forinstance,wecouldlookatthedebt-to-capitalratios,
cashbalances,andoperatingmarginsofallofthefirmsinthe
sampleatthestartoftheperiod;wewouldexpectfirmswith
high debt-to-capitalratios,lowcash balances,and negative
marginstobemorelikelytogobankrupt.Finally,usingthe
distressdummyvariableasour dependentvariableand the
financial ratios (debt-to-capital and operating margin) as
independent variables, we look for a relationship:
Iftherelationshipisstatisticallyandeconomicallysignificant,
we have the basis for estimating probabilities of bankruptcy.
12