YOUTH LABOR MARKETS IN THE U.S: Results 5

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.

OLS First stage IV First stage IV First stage IV First stage IV
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Longest tenure attained, five-year post-schooling period .02(.01) .24(.08) .16(.04) .12(04) -.06(.15)
Schooling .05 .13 .02 .10 .03 .10 .04 .14 .06
(01) (04) (.02) (.03) (.01) (.03) (01) (03) (02)
Experience .02 .04 -.01 -.03 .00 -.02 .00 .07 .02
(.02) (.07) (.03) (.06) (02) (06) (-02) (-06) (.02)
Experience squared * 102 .16 1.7 -.10 2.2 -.01 2.2 .04 1.4 .25
(.12) (.40) (.17) (.39) (.14) (.39) (.13) (.38) (.23)
Currently married .02 .12 .00 .11 .01 .10 .01 .06 .02
(.03) (09) (.04) (.09) (-03) (.09) (.03) (09) (.03)
Non-white .03 .13 .00 .15 .01 .18 .02 .13 .05
(.04) (.13) (.05) (-12) (.04) (.12) (.04) (.12) (.04)
Minimum unemployment rate -.07 .03 -.08 .05 -.08 -.02 -.08 -.01 -.07
on current job (.01) (.05) (02) (04) (-01) (.05) (01) (-04) (01)
Current unemployment rate .03 -.06 .04 -.04 .03 -.05 .03 -.04 .02
(.01) (.03) (.01) (-03) (.01) (.03) (.01) (.03) (.01)
uii, .01(.03)
Ц2 .08(.02)
u8J -.01(.02)
Ц* -.02(-02)
U,5 -.01(.02)
Uj. .34(.12) .34(.12)
.02(.04) .03(.04)
Up -.02(.08) -.03(08)
U, -.24(.19)
(Uiji-Uj.) -.02(.03)
(Uj2-U.j2) -.01(.03)
(U1J3-U.J3) .00(.03)
(Ц-Г^) .02(.03)
(Uu5-Ujs) .05(.03)
(u,-iu .04(.03)
F-statistic for instruments in first stage 5.3 24.9 10.8 2.4
P-value for Hausman exogeneity test .00 .00 .00 .65
P-value for test of over identifying restrictions .48 .89 .16 .46
P-value for test of overidentifying restrictions when minimum unemploymen rate on current job is excluded It .00 .97 .00 .09

The estimates in columns (4) and (5), which use the cross-cohort variation in early unemployment rates, are preferable on a priori grounds to those in columns (2) and (3), while in columns (6) and (7), in which the individual deviations are also used, there is some weak evidence against the overidentifying restrictions (the p-value is .16) even when the minimum unemployment rate on the current job is included as a control variable, suggesting that the specification in columns (4) and (5) is preferred overall. The estimated effect of .16 in column (5) implies that a one standard deviation increase in longest tenure attained increases adult wages of women by 25.4 percent. As for men (and perhaps more so), in the first-stage estimates the overall effect of early unemployment on longest tenure attained is positive. Thus, the estimates in columns (l)-(7) are qualitatively similar to those for men, although the beneficial effects of early tenure are stronger for women. Here