Complete longevity data are available too late for most sire selection. Earlier selection is possible using correlated traits, nonlinear evaluation of censored data, or predicted longevities for live cows in addition to completed longevity data. Completed longevity was defined as total months in milk by 84 mo of age. Predicted longevity was computed by multiple regression from cows alive at six different ages. Variables included age at first calving, standardized first lactation milk yield (optional), lactation status (dry or milking), current months in milk, current months dry, and cumulative months in milk. Completed longevity data for dead cows were then merged with predicted longevity data for live cows. A total of 1,984,038 Holstein cows born from 1979 to 1983 were included and represented 1911 sires, each with at least 70 daughters. Heritability of longevity increased gradually from .03 at 36 mo to .08 at 84 mo. Phenotypic correlations of early with completed longevity ranged from .59 to .99; genetic correlations ranged from .92 to 1.00. Inclusion of yield for projection increased heritability at 36 mo substantially but decreased genetic correlation with complete longevity information. Expansion and weighting factors will allow predicted records to be used in longevity evaluations, which is similar to the procedures that allow predicted 305-d yields to be included in yield evaluations.