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
Thus, in all of the specifications in Table 2, the evidence indicates that longest tenure attained in the immediate post-schooling period-i.e., during the school-to-work transition-has a positive effect on adult wages. In addition, the estimated magnitude is large, with an additional year of tenure leading to adult wages that are higher by seven to 13 percent. Of course, this should not be thought of as a return to tenure per se, since attention is restricted to the five-year post-schooling period. In particular, as reported in Table 1, the standard deviation of longest tenure attained is 1.65 for men. Thus, the estimates in Table 2 imply that a one standard deviation increase in longest tenure attained results in adult wages that are higher by 11.6 to 21.5 percent, numbers that are high but perhaps not implausible. (Below, it is suggested that these estimates may be upward biased, but still reflect positive effects.)
Finally, recall that these specifications do not include contemporaneous tenure. Column (1) of Table 6 reports the results of estimating the wage equation with contemporaneous tenure controls included. The specification from columns (6) and (7) of Table 2 is used, since this specification uses the most information on early unemployment while still easily satisfying the overidentifying restrictions. As expected, the estimated effect of early tenure falls; in particular, the estimated effect of an additional year of early tenure on the adult wage, conditional on adult tenure, is five percent and no longer statistically significant. Thus, at least some of the beneficial effect of early job stability arises through its effect on adult tenure, which is not surprising.