Simulation was used to examine the changes that may occur in international dairy sire evaluations when differing amounts of historical performance data are used. When the base genetic variances within each population were equal, use of all historical data in international sire evaluations gave unbiased predictions for breeding values. When some historical data from one population were discarded, the estimated genetic standard deviation for this population was reduced because of the effects of genetic selection and slope coefficients of conversion equations were biased slightly in favor of this population. When base genetic variances differed among populations, the use of all historical performance data in international sire evaluations resulted in substantial upward bias in evaluations of elite sires from the importing population. This result occurred because the estimated genetic standard deviation in the importing population was reduced by inclusion of performance data from many older bulls from the breed or strain that was being replaced. When performance data from bulls that were born prior to the beginning of importation were discarded, estimated parameters for genetic standard deviation were similar for both populations, and estimated breeding values of elite bulls were very close to the true values. This research suggests that accuracy of international dairy sire evaluations may be improved by discarding historical performance data of bulls from breeds or strains that have been replaced by imported stock.