Socioeconomic Background, Schooling and Monetary Rewards in the United Kingdom

Abstract

It is a commonplace observation that those coming from favourable social origins earn more in the labour market. But this may be due in part to differences in the level of educational attainment. In this paper we examine the way in which education interacts with family background in the determination of earnings. We begin by considering the empirical relationship between family background and earnings using data from the 1972 General Household Survey.' We then discuss possible explanations of the results along with their policy implications. I. THE INTERACTION BETWEEN EDUCATION AND FAMILY BACKGROUND The best way of investigating the interaction effect is to fit a Mincerian human capital earnings function to groups of workers with the same family background. This function has the form lnY=f(S, EX, EX2) where ln Y stands for the natural logarithm of the individual's annual earnings, S is the number of years of schooling completed, and EX is the number of years of labour force experience of the individual (defined as age - years of schooling - 5). The EX2 term is introduced to take care of the observed parabolic age-earnings profile.2 The results of fitting this function are shown in Table 1, where nine groups of workers are distinguished according to the occupation of their father. Concentrating on the coefficient on S (which in this specification measures the private rate of return to schooling), we observe that it falls as the status of the father's occupation rises. At the two extremes, the sons of agricultural workers have an average 13 per cent rate of return on their investment to schooling, whereas the sons of professional workers make 8 per cent. This suggests a negative interaction between education and the status of one's social origins in determining earnings.3 The hypothesis that the coefficient on schooling is different across occupational groups has been investigated by means of a Chow, or structural break, test in the following way. The observations were grouped and two regressions were run for the whole sample. First, an unrestricted regression of the form