Abstract This paper proposes a new method for estimating the intergenerational persistence of lifetime earnings from data that contain only short sections of individual earnings histories. The approach infers lifetime earnings persistence from the persistence of short earnings averages together with information about the stochastic process governing individual earnings. I find that lifetime earnings are substantially more persistent than previous estimates based on short panel data suggest. About 54% of lifetime earnings differences between fathers persist into their sons’ generation. This persistence estimate exceeds previous estimates based on five-year earnings averages by one third. These findings are robust against alternative assumptions about the data generating process for earnings.