TABLE 2.1Standard Errors in Risk Premium Estimates
Notethattogetreasonablestandarderrors,weneedverylong
time periods ofhistorical returns.Conversely, thestandard
errors from10-year and 20-yearestimatesarelikely to be
almost aslarge asor largerthan theactualrisk premiums
estimated.Thiscostofusingshortertimeperiodsseems,in
our view, to overwhelm any advantages associated with
getting a more updated premium.
2.Choiceofrisk-freesecurity.TheIbbotsondatabasereports
returnson bothTreasurybillsand Treasurybonds,and the
risk premiumfor stocks canbe estimatedrelative to each.
Given that the yield curve in the United States has been
upward-slopingformost ofthepast eightdecades,therisk
premium is larger when estimated relative to shorter-term
governmentsecurities(suchasTreasurybills).Therisk-free
ratechosenin computingthepremiumhastobe consistent
withtherisk-freerateusedtocomputeexpectedreturns.For
themostpart,incorporatefinanceandvaluation,therisk-free
ratewillbealong-termdefault-free(government)bondrate
and not a Treasurybill rate. Thus,the risk premium used