Employer and Occupational Instability in Two Cohorts of the National Longitudinal Surveys

Abstract

Previous research on trends in employer and occupational stability found evidence of declines in stability among men but contradictory results for women. I provide new insights into these patterns by simultaneously analyzing employer and occupation changes, and by examining a more detailed set of transition types. I show that the patterns for women are quite similar to those of men but are masked by declines in transitions from employment to out of the labor force. Finally, I find that while some of the changes may bring increased opportunities for wage increases, they bring even greater risks of wage losses.

NOTES

Notes

¹ Bernhardt et al. (2001) found that individuals who left the 1966 cohort sample had higher employer separation rates before they left the survey than nonattritors during this same period. Hollister (2011c) found that members of the 1979 cohort who would have left the sample under the original cohort conditions had higher employer separation rates.

² Another reason to examine the results under multiple sample restrictions is that the men in the 1966 cohort came of age during the Vietnam War, and a number of respondents in the sample spent one or more years in the military, especially at younger ages (13.6 percent of the 1966 cohort at age 22 compared with 5 percent in the 1979 cohort). I include the cases involving military service in the full sample of all two-year observations. There are two additional reasons one might be concerned about the impact of the Vietnam War. First, the war may have disrupted the careers of those who served, leading them to take longer to settle into employment. If this is the case, though, then the 1966 cohort should have particularly high levels of job changes, and therefore the higher rate of changes in the 1979 cohort is even more surprising. A second concern is that efforts to avoid the draft led to a particularly high level of college education among both men and women in the original cohorts (Angrist and Krueger 1992). If individuals with higher levels of education also have more stable careers, this could have reduced job changes among this cohort. In order to assess this issue, I ran the analysis separately by education group (not shown). I found the same patterns of job changes within each education group as in the overall sample. Differences in educational attainment, therefore, cannot explain the results.

³ The median two-year wage change in each of the NLS cohorts was about a 5 percent increase, and so the threshold for a wage increase is greater than the threshold for a decline.

⁴ Fixed effects are not possible with these samples because the key independent variable, cohort, does not vary within individuals.

⁵ Since the surveys involved oversampling of individuals, and I am not controlling for the factors related to this oversampling, I use the initial sampling weights and jackknifed standard errors for these estimates.

⁶ Evidence from other data sets, such as the PSID and CPS, also support the finding that the change is not cohort-specific.

⁷ To test for significance, I conducted estimates with a pooled sample of both cohorts. The model was similar to equation 2 but included interactions between the transition types and a cohort dummy variable. The interactions tested for differences between the two cohorts in the probability that a transition would lead to a wage increase (or decrease). None of the interactions were statistically significant.