The segmented earnings function (with and without interaction terms) is presented in Table 4 for the pooled sample. As can be seen, the interaction terms are mostly negative, and in fact the addition of the interaction terms to the simpler segmentation in Column 1 significantly Increases the explanatory power of the equation (the F statistic is 2.21, significant at the 10 percent level). Note also that the coefficients of previous experience are significantly weaker than the effect of current experience and that the coefficient of the longest job prior to the current job is by far the largest and most significant of all the previous job coefficients.
By estimating the segmented earnings function within each mobility pattern, it is possible to calculate the investment ratios for the different jobs in each mobility pattern. These regressions are shown in Table A-l of the Appendix. As can be seen, the estimates are generally not very significant but this is mainly due to the large amount of multicollinearity among the variables. Table 5 presents the initial investment ratios for several values of rg/2 (assumed to be the same across all jobs in the individual’s life cycle):21 .0010, .0015, and .0020. The estimates are presented assuming r ■ .10, since varying the rate of return did not affect the qualitative results of the analysis. Table 5 also presents estimates of the “projected” investment ratio, к ., defined as what investment would have been if the particular job had been the first Job in the life cycle.
Segmented Earnings Functions Pooled Sample, Dependent = Ln(RATE)
TABLE 5 Investment Ratios
|Segment||гв/2к . ol||-.0010к . ol||rB/2′к . ol||-.0015кol||rB/2к . ol||-.0020к . ol|