Analysis of Longitudinal Earnings Data: American Scientists 1960-70

The major findings of this study are as follows: (1) Simple cross section estimates grossly underestimate cohort profiles during the period 1960-70. Furthermore the growth in earnings is not uniform across experience groups and more recent vintages tend to have steeper profiles in most fields. Consequently the rate of return or present value comparisons based on cross sections are likely to be misleading even if the standard adjustment for growth is made. (2) For purposes of estimating mean profiles and mean effects of variables estimates based on pooled independent cross sections are quite close to those based on the more expensive longitudinal data. (3) There are important persistent unmeasured individual effects on both the level and growth of earnings. Consequently, individuals with the same observed characteristics will still have a wide variance in their permanent income.