Women's occupational attainment: The effects of work interruptions, self-selection, and unobserved characteristics

This paper addresses specification issues concerning the estimation of women's occupational attainment. Attainment models generally assume independence of error terms across equations, and whether this assumption is made or whether correlation of error terms is assumed is found to have important substantive implications, particularly with respect to education. While education affects first-job status, when first-job status is controlled, education has no significant effect on current attainment. A woman's first-job status appears, then, to be of major importance, and if she enters a job for which she is overqualified, education may not be relied upon to improve her occuppational status in the future. As in the analysis of earnings, it is important not only to include years of experience, as is usual in attainment models, but to include whether a recent work interruption has occurred. This is a significant and important variable in both the correlated and uncorrelated errors models, but the uncorrelated errors model appears to understate the magnitude of effect.