price-earningsratiogreaterthan 500 willbeassumedtohave
a price-earnings ratioof 500. Theconsequence is that the
averages reported by two services for the same sector or
marketindexwillalmostnevermatchupbecausetheydeal
withoutliersdifferently.InNovember2005,forinstance,the
averageP/EreportedfortheS&P 500 variedwidelyacross
servicesfromalowvalueof16.5onYahoo!Financeto24.2
onMorningstar.Itisincumbentonthoseinvestorsusingthese
numberstobeclearabouthowtheyarecomputedandtobe
consistent in their comparisons.
Biases in Estimating Multiples
Witheverymultiple,therearefirmsforwhichthemultiple
cannotbecomputed.Consideragaintheprice-earningsratio.
Whentheearningspersharearenegative,theprice-earnings
ratioforafirmisnotmeaningfulandisusuallynotreported.
When lookingat theaverage price-earningsratio across a
groupoffirms,thefirmswithnegativeearningswillalldrop
outofthesamplebecausetheirprice-earningsratioscannot
be computed. Why should this matter whenthe sample is
large?Thefactthatthefirmsthataretakenoutofthesample
aremoney-losingfirmscreatesabiasintheselectionprocess.
Infact, theaverage P/E ratioforthe groupwillbe biased
upward because of the elimination of these firms.
Therearethreesolutionstothis problem.Thefirstistobe
awareofthebias andbuilditintotheanalysis.Inpractical
terms, this will mean adjusting the average P/E down to
reflecttheeliminationofthemoney-losingfirms.Thesecond
istoaggregatethemarketvalueofequityandnetincome(or
losses) for all of the firms in the group, including the
money-losing ones, and compute the price-earnings ratio