SAMPLE SELECTION AND MALE-FEMALE EARNINGS DIFFERENTIALS IN THE GRADUATE LABOUR MARKET

Abstract

FOR a large sample of 1970 graduates in the U.K., the average earnings of males in 1977 were over 27% higher than the average earnings of females. The present paper examines the extent to which this differential is due to residual factors rather than different characteristics and attributes. To this end, the earnings of men and women are compared using the standard Oaxaca (1973) measures of discrimination. The study gives a new perspective to this familiar exercise by incorporating a female labour force participation decision and analysing the possible effects of sample selection bias. The richness of the data permits new insights, not least because it contains detailed information on a number of variables which have sometimes been omitted in the past. A study of remuneration by sex in the graduate labour market may be motivated in several ways. First, the graduate labour market has an importance to the national economy which far exceeds its numerical size. Since a large majority of technical, professional and managerial workers are graduates, it is important to question whether their pay is unfairly conditioned by sex. Second, other researchers have previously investigated particular aspects of this market such as professional workers (Chiplin and Sloane (1976), Shah (1984)) or librarians (Siebert and Young (1983)). It would therefore seem appropriate to consider the market as a whole since the previous results may be specific to the occupations studied. Third, there are advantages in considering a cohort, all of whom have experienced similar labour market conditions and have a comparatively homogeneous educational background. By contrast, the study of a diverse cross section sample may encounter problems with unobserved heterogeneity, competing groups and considering group rather than workplace samples.1 Many researchers base their estimates of sex discrimination on a comparison of single men and single women. It is argued that single women are more like men than married women because they have similar motivations and work histories.2 Further it is recognised that many married women are not actually working and information is often not available