Education, Vintage, and Learning by Doing

There is now a considerable literature in capital theory dealing with "learning by doing" and with the question of the "embodiment" of technical knowledge of a given period in the capital goods produced in that period. There have, however, been relatively few attempts to examine these questions empirically, and hardly any work has been done on the application of these approaches to skill formation and education. This paper is an attempt to begin this work by drawing some empirical implications and then testing them with data from the U. S. Census of 1960 and the Canadian Census of 1961. It is clear that if workers in different age groups, but with the same number of years of schooling, are examined, an older worker will have had a longer opportunity to "learn by doing," but probably will have received a formal education inferior to that of a younger one. By assuming that formal education and learning by doing are complementary, and also that there are diminishing returns, it is possible to draw some firm analytical conclusions about the age-productivity profiles of people with the same number of years of formal education. But productivity as such is difficult to measure, and the translation of the results to age-earning profiles, by assuming that marginal productivity equals earning rates, requires the fulfillment of some strong conditions, for example, perfect competition. Instead, it was assumed that earnings were a linear function of productivity, which covers a wide variety of cases, including uniform degrees of imperfection. With this background, two results were anticipated in the age-earning profiles of each educational bracket, viz., (1) each would have a single peak, i.e., earnings would uniformly rise or uniformly fall, or rise up to a point and then fall; and (2) the rate of rise in earnings must fall uniformly over time (or the rate of fall in earnings rise uniformly over time). The U. S. Census of 1960 was used to calculate the age-earning profile for each educational group; the first result was verified without exception and the second with one minor exception. The Canadian Census of 1961 also confirmed the results without exception. However, some limitations of the study are also pointed out, and too much significance should not be attached to the empirical verification. The paper is mainly an attempt to marry two branches of economic thinking, and the empirical exercises illustrate one, but not the only, possible marriage. An incidental by-product of the study is the set of age-earning profiles from the 1960 U. S. Census, and some of the contrasts to be drawn between this and those of 1940 and 1950 are noted.