One way to assess explanations regarding unobservables, such as this one, is to look at evidence on observables that should behave in the same way as the unobservable. In this particular case, it seems reasonable to suppose that education and the unobserved wage or productivity component should have similar relationships with early job stability. However, Tables 2-5 show that education is positively correlated with early job stability (as is AFQT, although these results are not reported in the tables). It is possible that the relationship is different for the unobservable fixed effect, if the unobservable matters mainly within education and ability levels. But it is probably best to be skeptical about an hypothesized unobservable related to wages or productivity that is inversely related to education and measured ability.

# Monthly Archives: May 2014

May

May

However, the IV estimate of the effect of early job stability should be biased upward more than the OLS estimate only if the positive correlation between the regressor (S) and the match quality component of the error term in equation (1) is exacerbated. This would require, for example, that individuals with high predicted S based on high early unemployment rates are in good matches to a greater extent than are those with high observed S (which may be influenced by a number of factors not limited to early labor market conditions). But if the observed variation in S also reflects self-selection of those with good matches into stable early jobs, it is difficult to see why the bias would not, on net, be lessened-although perhaps still present-in the IV estimates.

May

Relatively consistent evidence emerges from most of the estimations. For both men and women, most of the IV estimates for specifications that are not rejected by the data indicate a positive return to early job stability, in contrast to the OLS estimates that indicate essentially no return. Interestingly, these IV results are contrary to the hypotheses with which this study began. The presumption was that OLS estimates of the effects of early job stability were biased towards finding a positive effect of early job stability because of omitted job match quality that is positively associated with this stability; the IV results, as just noted, instead indicate that the OLS estimates of the effects of early job stability are biased downward. Thus, the evidence requires an alternative explanation. This section discusses a few possible alternative explanations of the results, and considers the consistency of the evidence with each of them.

May

The final analysis addresses the issue of measurement of local unemployment rates. As explained in Section III, the unemployment rates for individuals residing outside of metropolitan areas (as well as those residing in metropolitan areas for which unemployment rates are not reported in Employment and Earnings) are rates for the entire non-metropolitan area of their state of residence. Consequently, unemployment rates may measure local labor market conditions much more accurately for those residing in SMSAs for which separate unemployment rates are reported in Employment and Earnings. The estimates of the key specifications, both with and without adult tenure included, were therefore recomputed using only observations on individuals residing in this subset of SMSAs in each of the early labor market years as well as 1992 (since the unemployment rate for each of these years is required).

May

Columns (6)-(9) report the estimates first using the cohort average unemployment rates and individual-level deviations from these, defined over each of the five post-schooling years, and then using the five-year averages. In columns (7) and (9) the IV estimates of the effect of number of jobs held are smaller than in column (5), and insignificant. However, in these estimations the p-values for the overidentifying restrictions are quite low (.18 and .08), suggesting that the estimates in column (5) are preferred, presumably because the estimates in columns (6)-(9) use the individual-level unemployment rates that may be partly endogenous. The estimate of -.08 in column (5) implies that a one standard deviation (2.63) increase in the number of jobs held in the immediate post-schooling period lowers adult wages by 21 percent.

May

Tables 4 and 5 turn to results for the number of jobs held during the five-year post-schooling period. For men, the OLS results in column (1) of Table 4 indicate a small (-.014) negative, significant effect of number of jobs held. Using the individual-level unemployment rates to instrument for number of jobs held, in columns (2) and (3), results in a small change in the estimated effect of the number of jobs held, with the estimate falling to zero. However, the F-statistic for the instruments in the first-stage regression is only 1.9, and the overidentifying restrictions are rejected (at the ten-percent level) even when the minimum unemployment rate on the current job is included. Thus, these IV estimates are not reliable.

May

The only difference relative to men comes in columns (8) and (9), when the unemployment rate variables averaged over the five post-schooling years are used. In this case, the sign of the first-stage estimate of the cohort average unemployment rate is negative and insignificant, whereas it was positive and significant for men. In addition, the F-statistic for the instruments in the first-stage regression is only 2.4, compared with much larger values in the other columns. This suggests that small sample bias may be non-negligible, and that, more generally, the instruments in this specification do not provide much identifying information. This is also reflected in the much larger standard error of the estimated coefficient of longest tenure attained in column (9). Thus, the estimates in columns (8) and (9) are not very informative. Electronic Payday Loans Online

May

Table 3 repeats this analysis for the sample of women. Having described the analysis in Table 2 in detail, these results can be discussed more briefly. In the OLS estimates in column (1), the effect of longest tenure attained is small (.02) as it is for men, although in this case the estimate is statistically significant. Looking at columns (2)-(7), the IV estimates indicate substantially larger returns to early tenure, ranging from 12 to 24 percent.

May

The final two columns of the table use the averages of the unemployment rate variables over the five years, Ujt and (Uy-U0_). In this case, in the first-stage estimation the cohort average unemployment rate averaged over the five post-schooling years provides all of the explanatory power. Its estimated coefficient is positive, again indicating that high unemployment during the immediate post-schooling period results in longer job attachment during this period. In the wage equation estimation, again, the IV estimate of the effect of longest tenure attained is positive (.13) and statistically significant. The Hausman test rejects the exogeneity of longest tenure attained in the wage equation at the six-percent level. review

May

Columns (4) and (5) report results from using the average unemployment rate experienced by each entry cohort, in each year, rather than the unemployment rate experienced by the individual, in order to obtain a more exogenous instrument. The first-stage results in column (4) are quite similar, suggesting that early unemployment raises longest tenure attained in the immediate post-schooling period, and later unemployment lowers it, again by less. Note also that the estimated effects of unemployment are larger in column (4) than in column (2), consistent with a migration response to local unemployment rates that biases the estimates in column (2) towards weaker effects of variation in labor market conditions. The IV estimate of the effect of longest tenure attained on the wage (.08) is again considerably above the OLS estimate, and statistically significant. The specification test results are also similar, although in this case-perhaps reflecting the greater aggregation (and exogeneity) of the cohort average unemployment rates-the p-values for the overidentification tests are higher, and the restrictions are not rejected whether or not the minimum unemployment rate is excluded from the wage equation. In addition, the exogeneity of longest tenure attained is rejected, with a p-value of .02. review