havemovedupordownsignificantlyinthoseperiods).Using
thesenontrading-periodreturnsintheregressionwillreduce
thecorrelationbetweenstockreturnsandmarketreturnsand
the beta of the stock.
27 Thebiascanalsobereducedusingstatisticaltechniques
suggested by Dimson and Scholes-Williams (E. Dimson,
“RiskMeasurementWhenSharesAreSubjecttoInfrequent
Trading,”JournalofFinancialEconomics 7 [1997]:197–226;
M. Scholes and J. T. Williams, “Estimating Betas from
Nonsynchronous Data,” Journal of Financial Economics 5
28 Thereturnsonboththestockandthemarketindexinclude
dividends. For Disney, the dividends are shown only in
ex-dividendmonths.Fortheindex,weusethetotaldividends
paid during the month on stocks in the index.
29 Manyservicesadjustregressionbetastoward 1 toreflect
thelong-termtendencyofthebetasofallcompaniestomove
toward the market average. Others adjust for the
characteristicsofthecompanies;businessmixes,debtratios,
dividend yields, and market capitalization are considered.
30 To seewhy,compare twofirms withrevenuesof $100
millionandoperatingincomeof$10million,butassumethat
thefirstfirm’s costsareallfixedwhereasonly halfofthe
second firm’s costs arefixed.If revenuesincrease atboth
firmsby$10million,thefirstfirmwillreportadoublingof
operating income (from $10 to $20 million) whereas the
secondfirmwillreportariseof 55 percentinitsoperating
income(sincecostswillriseby$4.5million, 45 percent of
the revenue increment).