Expected Rates of Return and the Demand for Education : Some Empirical Evidence

The great flowering of research in the economies of education over the past dozen years has been based essentially upon one observation : that individuals with more education tend to earn higher incomes. The major aim of research in the field has been to attach figures to this relation and to assess the rates of return realized by students on their investments in education. This work on the ultimate economic effects of investment decisions in education has provided a much better understandingof the significance of these decisions — once taken. But it has done little to shed light on the importance of these economic aspects of education for students at the moment when they make the key investment choices. How aware are students of the financial costs and possible returns on their education ? Do these financial aspects actually affect enrolment choices ? The object of this paper is to shed some new light on these questions. It begins by arguing that one should adopt an explicit model of student expectations in the investigation of student decision processes, for the type of realized rate of return normally presented in the literature almost certainly provides poor estimates of the rates expected by students on enrolment The paper then presents some new evidence on expected rates of return which indicates that students do in fact seem to be influenced in their educational decisions by the economic aspects of the investment.

The usual starting point for studies of the value of educational investments, as with most predictive models of the economic behaviour of individuals, is the theory of the optimizing consumer. This starting point provides the approach with considerable intuitive appeal. It calls for careful estimates to be made of the discounted value of the stream of lifetime financial returns to educational investments. Particular care will be taken to isolate the effects of education from the impact of other associated factors such as intelligence and family background. Attention will be given to the actual movement of returns over time. The ultimate object of the approach is to obtain an accurate estimate of the returns realized by students on a given educational investment.

In examining studies made under this approach, one often encounters a strange and disquieting difficulty in assessing just where the focus of a particular study is directed : whether it falls on the determinants of students' decisions, or whether it is concerned with the ultimate effects of these decisions. The ambiguity is curious, for the difference in approach is methodologically quite distinct. The source of the confusion, however, is not difficult to trace. It lies in the particularly accute problems of measurement and hypothesis testing in the economies of education. Being very difficult to test, the picture of the student as an efficient investment planner becomes built in at the foundation of the analysis. Moreover, once the student can be granted some measure of accuracy in his forecasts of the returns to investment, the distinction between expected and realized rates of return become very shadowy indeed. It is for this reason that few authors feel it necessary to indicate whether their studies are intended to provide estimates of expected or of realized rates of return.

The lack of concern in the literature for the distinction betweenex ante andex post rates is very disquieting. There are at least two major reasons for suspecting that most of the rates presented in the literature, calculated as they are under the inflexible logic of theex post approach, give a very biased measure of the rates of return expected by students at the moment they make their educational decisions. First, most of these estimates are measures of the rates of return on investments made by those students who have successfully managed to reach graduation. They thus ignore the (usually lower) rates of return on the investments made by students who fail to complete the course. The possibility of being a drop-out is a legitimate concern for the prospective student. A student cannot « invest in a graduation » ; he can only invest in the years leading up to graduation. The gap between rates ofreturn obtained by successful graduates and those received by drop-outs can be substantial, particularly in the case of certain types of professional training where the risk of not completing the course is high, and where the usefulness of the training of those who fail to obtain certification is limited. The gap can still be significant for other more general types of training. Becker (1964) has estimated, for example, that the rate of return to the 1949 cohort of white male college graduates stood at around 13 percent, while the comparative rate on the investment of those who dropped out of the program stood at 8 percent. At the same time, the probability of the student dropping out was almost one in two. The rate of return to graduates thus substantially overestimates the rates of return to initial enrolment. Viewed in this light, the fears expressed in many quarters about the possibility of substantial underinvestment in college education appear to be rather exaggerated.

The second major source of bias in the use of realized rates of return as a proxy for expected rates can be traced to the lack of perfect foresight on the part of students. Where conditions affecting the scale of returns change radically and unpredictably at some point after the period of investment has been terminated, a substantial gap will appear between any reasonable estimate of expected returns and the returns ultimately realized by students. A good example of such divergence can be found with the cohorts investing in education during the 1930's in the United States. During those years, there was little reason to believe that a general world war would shortly break out, and that as a result personal income taxes in the U.S. would rise substantially. The new tax rates applied during the 1940's substantially lowered the returns realized by individuals on their educational investment. Becker (1964) has estimated that the application of the 1949 tax rates reduced the rate of return realized by the 1939 cohort of college graduates by 1.5 percentage points, from a level of around 16 percent to approximately 14.5 percent. In other words, the gap between expected and realized rates of return to the 1939 cohort of male college graduates appears to represent over 10 percent of the realized rate of return. The difference is substantial. Moreover the gap implies that the expected rates of return fell significantly between 1939 and 1959, a conclusion very different from the one drawn by Becker that theex post rates of return actually rose between 1939 and 1959.

In sum, it would seem that the type of rate of return generally presented in the literature will frequently provide very poor estimates of the rates expected by students at the moment of their decision to enrol in a particular program. This strongly suggests that if one wishes to investigate student decision processes, it would be preferable to utilize an explicit model of expectations.

Table I (p. 395) presents estimates of the expected rates of return for students enrolling in the final years of secondary school by province and by sex in Canada during the early 1960's. These rates are explicitly intended as estimates of the financial rates of return that the reasonably perspecacious student might have expected in the early 1960's. The object of the approach is to see whether such estimates of expected returns might explain enrolment patterns which would other-wise be difficult to understand. The rates are therefore calculated on the assumption that student expectations of returns are formed by observation of current income differentials obtaining between individuals of different educational attainment. No attempt has been made to correct these estimates of returns for the actual evolutionin differentials since the early 1960's ; and no attempt has been made to forecast what returns will be actually realized by these individuals during the possible thirty or forty years of working lifetime that still remain to them.

Two features of these results are striking. First, the rates of return tend to be highest in the poorest and least developed provinces. This pattern is not as surprising as it may seem at first glance. It can be traced to the response of individuals to economic incentives — specifically, to the geographical relocation of labour in response to income differentials. To the extent that such migration tends to equalize salary incomes across the country, the absolute levels of incomes of people remaining within a given province will be influenced by incomes in more distant regions. As educated people tend to be more mobile than their less qualified colleagues, migration will lead to a relative shrinking of the supply of well-educated manpower in the poorer regions and to an increase in the differentials obtaining! between given age-education categories. While the size of these differentials in the poorer regions may not reach the levels of other areas in the country, the costs of obtaining educational qualifications still remain substantially lower in these poorer regions due to lower foregone earnings. The net result is a tendency for rates of return to be highest in the less developed regions of the country.

The second striking feature of these results is the uneven relative standing of the expected rates of return for the two sexes within each of the provinces. The rates for females are higher than for males in some provinces and not in others. Various possible factors can be ruled out a explanations of these differences. Psychic costs and benefits of secondary education are presumably no higher for one sex than for the other ; or if they are then the difference is likely to repeat itself systematically in all provinces. Since approximately the same proportions of children of each sex are enrolled in secondary school in each province, factors such as differences in family background, intelligence, or differences in the cost of capital are similarly unlikely to produce these patterns of intra-provincial differences. One would there-fore conclude that if these rates indicate real differences in the net attractiveness of investments, and if students tend to invest more heavily in investments offering higher rates of return, then the sex expecting the highest rate of return should send a higher proportion of its members to enrol in secondary school. This in fact seems to be the case. The final two columns of Table I show the participation rates for each of the two sexes by province for 1961-62. Except in one province where the participation rates for the two sexes are equal, the sex expecting the highest rate of return tends to enrol a higher proportion of its population at the matriculation level. This evidence gives tentative support to the major hypothesis that expected rates of return influence students' enrolment decisions.