One method of Introducing these effects Into the earnings function Is by Incorporating the work history of the Individual Into the equation. Generally, suppose there are n jobs In the Individual’s working life up to time t. Then equation (1) can be generalized as:

where e^ Is the duration of the jth job and k^ Is the Investment ratio In the t**1 year of the j**1 job. Note that the rates of return have been assumed constant within the segment, but have been allowed to vary across jobs. comments

The discontinuity In the earnings and Investment profiles Is reflected In (4) by the fact that the returns to on-the-job training have been broken up into n terms. Each of these terms depends on the investment path for the particular Job. As was argued earlier, we would expect that the investment path be declining within the job. Thus an analogue of (2) is:

*
*We also expect the level of the investment profile to be affected by the timing of the job in the life cycle. That is, more investment is likely to take place the earlier the job occurs. This prediction, of course, follows from the fact that if some of the training is general (i.e., useful in other jobs) the payoff is greater the earlier it occurs. If the training is partly specific, however, a more important prediction for the slope of the investment (and hence the earnings) profile can be derived: the level of investment in any job is likely to be positively correlated with the completed duration of the job. In other words, the earnings profile will be steeper in longer jobs. For example, suppose only general training were produced on the job. Then dollar investment costs in a given job would not be correlated to the completed duration of the job since the only factor which can diminish the value of general training is depreciation.