Income predictions that provide some indication of the potential value of attending college are considered. Standard multiple regression analysis of data describing the income experiences of men 25 years old and older were used to determine differences in incomes of high school and college graduates. Information on the gross national product was then added to the income prediction equation for the period 1956 through 1975, and the resulting predictions were improved. Freeman's (1976) three-equation recursive adjustment model is presented and assessed. Based on further studies, it is suggested that the number of high school graduates is the most useful variable, particularly for predicting the income of men, and the second best predictor is birth rate. The economic measures of personal income and disposable personal income are the most useful predictors of the incomes of women high school graduates who do not graduate from college. Other economic and labor market variables such as gross national product per person, percent of the labor force that is unemployed or professional workers are also useful, as are share of population living in urban households, household average size, and higher education income and expenditure statistics. The findings indicate that aptitude and ability scores, enrollment and retention rates, and grades are not useful income predictors. The marginal productivity theory of wages, based on the work of Adam Smith, and Thurow's (1975) job competition theory of wages are addressed. A bibliography is appended. (SW)