Hazard functions were plotted and troughs were used to define stayability traits that best coincided with the time of culling decisions. Traits evaluated were stayability to 17, 30, 43, and 55 mo of productive life and stayability to 42, 54, 66, and 78 mo of total life. Holstein data from Record of Performance and supervised Dairy Herd Improvement and Dairy Herd Analysis Service testing programs formed two data sets with 243,348 and 118,362 daughters of 10,723 and 4,583 sires, respectively, and were used to estimate stayability heritabilities and correlations. Variance components were estimated by Henderson's new method using a model that included a fixed herd-year-season of first calving effect and random sire and error effects. Heritabilities for stayability to 17, 30, 43, and 55 mo of productive life were .025, .040, .039, and .033, respectively, for the Record of Performance data and .010, .017, .023, and .033, respectively, for the Dairy Herd Improvement-Dairy Herd Analysis Service supervised data. Heritability estimates for stayability to 42, 54, 66, and 78 mo of total life were .051, .040, .045, and .059, respectively, for the Dairy Herd Improvement-Dairy Herd Analysis Service supervised data. Genetic correlations among productive life traits were usually very high and approached unity. Genetic correlations among total life traits were lower and more variable. Most phenotypic correlations were moderately high but some were low. Stayability to 17 mo of productive life was concluded to be the best trait for incorporation into a selection program if evaluation of sires for longevity is desired.