As shown earlier, a more direct test of the specific training hypothesis can be obtained if tn, the total duration of the current Job, Is observed.
The results using equation (11) are shown in Table 6, using a small sample of individuals who left the Job they held in 1966 before 1969. The coefficient of the Interaction term (REM x CURRENT) is positive and significant as expected, thus rn pn « .0087. The coefficient of time remaining (REM) is negative and approaching statistical significance, thus pn – .0620. The implied estimate of rn about 14 percent. The results unambiguously suggest that the correlation between Job duration and Investment is a significant determinant of the distribution of earnings. The addition of these two variables Increases the explanatory power of the equation (the F statistic is 2.55, significant at the 10 percent level).
The negative effect of REM on current earnings can only be explained by referring т t° the effects of specific training and Job mobility on the Investment profile.
Earnings Functions When Completed Duration of Current Job is Known
Dependent m Ln (Rate), n *» 350
If training were totally general, then time remaining in the current Job would have no effect on the investment profile or on the earnings distribution. Once specific training is introduced, the payoff period to that portion of the training which is firm-specific become the time remaining in the current job.
The longer the time remaining, the higher the marginal revenue of Investment, so that more investment takes place in the current period. This, of course, implies lower earnings now. Similarly, the positive Interaction term says that the negative effect of REM on current earnings is balanced by the fact that longer REM and longer CURRENT would imply a longer job, so that more would have been invested on the job prior to the current time period, leading to higher current earnings.EDUC