This paper studies the quantitative importance of precautionary wealth accumulation relative to life-cycle saving for retirement. Section 1 examines panel data on earnings from the PSID. Using a bivariate normal model of random effects, we find that second— period—of—life earnings are strongly positively correlated with initial earnings but have a higher variance. Section 2 studies the consequences for life—cycle saving. Households know their youthful earning power as they enter the labor market, but only in midlife do they learn their actual second—period earning ability. For plausible calibrations, precautionary saving only adds 5—6% to aggregative life—cycle wealth accumulation. Nevertheless, we find that, given borrowing constraints on households’ behavior, the variety of earning profiles that our bivariate normal model generates itself stimulates more than twice as much extra wealth accumulation as precautionary saving.