ABSTRACT
The empirical basis of a rate of return calculation is largely a series of age-earnings profiles, i.e. a series of earnings figures for various age intervals throughout the prospective working life of an individual. Separate earnings data are given for each educational level under consideration. The source of this data is invariably census or other cross-section surveys. Thus data which are derived from observations about a given point in time are used to construct a time-series which is intended as a guide to prospective life-time earnings. Broadly speaking, the conversion of cross-sectional into time-series data is only valid when the relevant variables are stable over time, and in this case the time period is forty years or more (the working life-span). Considering the social and economic changes which are likely to take place over such a period, is the assumption of such stability unwarranted? Several answers have been given to this. Firstly, it may be argued that the objection is irrelevant,1 that choices are made using observations of the contemporary scene as a guide to the future, and that for individual choice at least, cross-section data
provide as good a basis as any. But the objection can also be met by stating that stability in earnings differentials in the past has been sufficient to lead one to expect (in the absence of contrary evidence) stability in the future. Furthermore, in the long run, the increased probability of changes in earnings differentials is heavily discounted — earnings say thirty years hence are discounted to such an extent that their impact on the rate of return calculation is relatively minor. Blaug2 holds the view that cross-section data are unaffected by cyclical factors or other movements in the economy, and thus provide
a 'truer' estimate of returns. This is to some extent the case, but on the other hand there is evidence that earnings differentials at any one time are affected by cyclical factors. Differentials between skills tend to narrow during booms, and widen during slumps.3 Consequently,
differentials at any time will reflect cyclical forces, unless they happen to be for a cyclically 'neutral' year. Reder's observations relate to cyclical movements in skill differentials, and while there is no direct evidence of educational differentials showing cyclical variation, there must at least be a presumption that similar forces will be at work.
