Trends in educational homogamy amongst couples, and its effects
Malcolm Brynin and Marco Francesconi, University of Essex, and Tim Liao, University of Urbana
[Project number 30028]
The project hitherto has aimed to increase our understanding of the relationship between the marriage and labour markets. We have produced one paper, based on the British Household Panel Study, in which we explore the impact of each partners human capital on the others labour market success, but also the probability that this human capital draws people together in the first place, so that we can look at how closely human capital in partnerships is matched as a predictor of labour market success. Such factors are closely bound up with homogamy.
Our objective in the second phase of this project is to analyse the extent of homogamy in British society and homogamy trends, as we believe this has a significant effect on the association between marriage and labour markets. This is an important area of research in its own right, which has also produced conflicting results. Some research suggests that homogamy is increasing, some decreasing. Important arguments as to the nature of modern society depend on settling this empirical issue - in particular, whether marriage and partnership trends are making society more or less open. Further understanding is, however, hampered by the lack of datasets with sufficient observations over time
We have two subsidiary objectives: First, to produce reliable trends in homogamy over time. Here we hope to innovate methodologically by using different indices of homogamy and different indicators of homogamy. Our expectation is that this will produce the definitive account of the extent of and trends in homogamy in the UK. Second, we will look at the effects of homogamy on trends in labour-market outcomes over time, using social class indicators.
The LS has three major advantages over other available datasets such as the LFS, GHS, or BHPS. First, it is much bigger than any of these, which greatly increases cell sizes and therefore reliability. For certain forms of homogamy, for instance by ethnicity, it is the only possible source. Second, its cross-sectional base goes back to 1971, which is further than other datasets can take us. Third, it includes a longitudinal component, so we can see how the effects of homogamy have changed over time for the same individuals. (The BHPS can provide this too, but with much smaller numbers and only for a select group of individuals.)
Our aim is to match spouses and partners by educational level and ethnicity, to construct various statistical measures of homgamy based on these, and then to produce trends. We will also test how far both the level of each partners education and the closeness of this match is associated with their social class. The dataset does present some problems in doing this. For some analysis we will have to restrict ourselves to graduates (or equivalent) only as in 1981 and 1991 only post-18 education is available. We can, however, compare 1971 and 2001 using a fuller set of educational measures. Finally, for a subsample we can use the panel element to match back education known in 2001 to 1991, and perhaps to 1981.
Finally, we will use this last sample to test the longitudinal effect of educational matching within marriage and partnership over the lifecycle.