Skip to main content

What a difference a term makes: the effect of educational attainment on marital outcomes in the UK


In the past, students in England and Wales born within the first 5 months of the academic year could leave school one term earlier than those born later in the year. Focusing on women, those who were required to stay on an extra term more frequently hold some academic qualification. Using having been required to stay on as an exogenous factor affecting academic attainment, we find that holding a low-level academic qualification has no effect on the probability of being currently married for women aged 25 or above, but increases the probability of the husband holding some academic qualification and being economically active.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6


  1. Moreover, using household surveys from 34 countries, Fernandez et al. (2005) find strong empirical evidence of a positive and significant relationship between several measures of the skill premium and of the degree of correlation of spouses’ education (marital sorting).

  2. There are two key reasons why we focus our analysis specifically on women’s marital outcomes. The first reason is a statistical one. Below, we argue that there is no evidence of any impact of holding an academic qualification on women’s probability of being married, thus allowing us to argue that we can identify the effect of women’s education on the economic properties of their husbands. A corresponding analysis of the impact of holding a degree on men’s probability of being married leads to less conclusive results, allowing us neither to rule out a positive effect nor verify it. Given that possible effect of academic qualification on selection into marriage, we then cannot identify the effect of men’s education on the economic properties of their wives. The second reason relates to the outcome variable studied. For women, we consider whether holding an academic qualification increases the probability of the husband being economically active, and we perform this analysis under the interpretation that the husband being economically active is a favourable outcome for the wife. While a corresponding analysis could be done for men, it is less clear that the wife working would indicate a favourable outcome as it more likely would reflect specialization.

  3. Del Bono and Galindo-Rueda (2007), focusing on the wage returns to education, present similar finding using, in part, the same data.

  4. For early studies of marital patterns, see Hunt (1940), Burgess and Wallin (1943) and Rockwell (1976).

  5. See also Peters and Siow (2002).

  6. The current model draws in part on the model by Konrad and Lommerud (2010).

  7. It can be shown that the benefit takes the form

    $$B( \boldsymbol{\alpha}_{i}) =\Delta y+\left( \theta_{s}-\theta_{m}\right) \left\{ p( \boldsymbol{\alpha}_{i},1) -\left[ 1-p\left( \boldsymbol{\alpha}_{i},0\right) \right] \right\} -\gamma \pi_{m}\Delta y\left\{ [ 1-p( \boldsymbol{\alpha}_{i},1) ] +p\left( \boldsymbol{\alpha}_{i},0\right) \right\} , $$

    where \(\theta _{s}\equiv E[ \theta I_{\theta \geq 0}]\) and \(\theta _{m}\equiv E[\theta I_{\theta \geq \gamma \Delta y}]\) (where \(I_{S}\) is the indicator function which is unity if S is true and zero otherwise), with \(\theta _{s}>\theta _{m}\).

  8. For the current purposes, we do not need to consider the equilibrium distribution in education choice in the population. It suffices to say that a complete description of the equilibrium would need to specify the matching technology and then a characterization of two mutually consistent functions \(x(\boldsymbol {\alpha },z)\) and \(p(\boldsymbol {\alpha },x)\) as either depends on the other.

  9. Formally, we define the effect of education on the partner’s skill level as

    $$\Delta x_{-i}( \boldsymbol{\alpha}_{i}) \equiv \Pr \left( x_{-i}=1|m_{i}=1,\boldsymbol{\alpha}_{i},x_{i}=1\right) -\Pr \left( x_{-i}=1|m_{i}=1,\boldsymbol{\alpha}_{i},x_{i}=0\right). $$

    Expression (3) follows from applying Bayes’ rule.

  10. Note that even using date of birth relative to the cut-off date between academic cohorts (in our case, the August–September threshold) could fail the identifying assumption in so far as an individual’s academic cohort constitutes a marriage-market-relevant social grouping. This is suggested by the finding of McCrary and Royer (2011) who use date of birth relative to school entry cut-off as instrument for educational attainment for mothers in Texas and California and find that mothers born after the school entry cut-off date have younger partners.

  11. Note also that the analysis of identification assumes that education precedes marriage. In the analysis below, we will use as measure of educational attainment an indicator for whether an individual holds any academic qualification. Since this is typically determined by exams taken at the age of 16 in the UK, we perceive this to be a negligible issue.

  12. The education system in Northern Ireland differs slightly from that in England and Wales. For instance, the cut-off date between academic cohorts is July 1 in Northern Ireland as opposed to September 1 in England and Wales. For this reason, we will not include Northern Ireland in the analysis below.

  13. The justification for dual exit dates seems to have been the belief that a common exit date, given the share of students leaving school at the minimum age, would negatively affect the functioning of the labour market.

  14. Even when the minimum school-leaving age was 16, students leaving at Easter had the option of returning for exams and evidence suggests that a substantial fraction of students did so (Del Bono and Galindo-Rueda 2007).

  15. Indeed, with the restructuring of the LFS in 1992, the survey was transformed into a “rotating panel”. Each quarter’s LFS sample is made up of five “waves”. Each wave is interviewed in five successive quarters, such that in any one quarter, one wave will be receiving their first interview, one wave their second and so on, with one wave receiving their fifth and final interview. However, since we are not interested in time-varying characteristics or outcomes, we will not be making use of the panel structure of the LFS. Instead, we will only be using information provided by individuals in their first interview.

  16. Prior to 2001, there is no information about in which country of the UK individuals were born. We then keep those born in the UK and currently living in England and Wales. Hence, for earlier survey years, there is some unavoidable degree of noise due to migration from Scotland and Northern Ireland.

  17. We do not impose any explicit upper limit on age. However, per construction, the oldest individual who will be included in the data will be someone born in the autumn of 1957 and observed in the autumn of 2006. Hence, no one in the main sample will be aged above 50.

  18. We focus on whether the respondent is currently married. Hence, we do not measure “partnership status” which would include cohabitation. Unfortunately, cohabitation can only be consistently identified in the data from 1991 at which point the cohabitation rate was 6 % while the married rate was 76 %. More generally, about 40 % of women in the main cohorts of interest had ever cohabited at some point in their lives by the age of 40 (Beaujouan and Bhrolchïn 2011). However, in most cases, this constituted premarital cohabitation.

  19. Formally, unbiasedness obtains in the limit as the window size is reduced to zero. See, for instance, Hahn et al. (2001) for a theoretical discussion and McCrary and Royer (2011) for an application.

  20. The use of quarter of birth as an instrument for educational attainment in the US context has recently been criticized by Buckles and Hungerman (2008). They highlight, for instance, that women giving birth in the winter months are more often teenagers, less frequently married, less frequently white, less educated and younger.

  21. In contrast, academic attainment does not change monotonically at the threshold between academic cohorts: while those born before this (August–September) threshold are more likely to hold some low-level qualification, those born after the threshold are more likely to hold some higher-level qualification.

  22. A detailed analysis of the impact of the school-leaving rule for actual school-leaving behaviour is presented in Del Bono and Galindo-Rueda (2007). Some of their main findings on this are summarized below.

  23. The finding that the main effect of having been required to stay on was an increase in the probability of holding a low-level academic qualification is in line with that of Del Bono and Galindo-Rueda (2007).

  24. It is also possible that, in the pre-RoSLA period in particular, the estimated effect may due to students completing the academic year at age 15, in some cases receiving a “school-leaving certificate” which in some cases are likely to have been recorded as a CSE academic qualification (Dickson and Smith 2011).

  25. The presented specifications for the pre-RoSLA cohorts and the post-GCSE cohorts use the narrow December-to-March window and no weighting. Other specifications are available upon request.

  26. Specifically, the figure illustrates the coefficients from a set of regressions, one for each age, of the outcome variable “currently married” on the various levels of academic attainment where the regressions also include controls for academic cohort, survey year and ethnicity.

  27. Specifications 1–4 are basic 2SLS IV models where the outcome variable is a dummy for the individual being currently married and the endogenous variable—the dummy variable indicating whether the individual holds an academic qualification—is instrumented for using \(z_{i}\), the dummy indicator for whether the individual was, due to her month of birth, required to stay on. Specification 5 is the Wald IV estimator formed by taking the ratio of the estimated gap in the outcome variables at the threshold point to the estimated gap in the endogenous variable. See Imbens and Lemieux (2008) and Lee and Lemieux (2010).

  28. The effect of holding an academic qualification on “ever being married” could be different from the effect of “being currently married” if there was an effect on divorce probability. However, we find no evidence to suggest an impact of holding an academic qualification on the probability of being divorced. In the age group \(25+\), the reduced-form estimates of the effect of being born after the January–February threshold on the probability of being currently divorced range from \(-0.008\) to \(0.002\) and are never statistically significant. The corresponding IV estimates are centered on zero and never statistically significant. (Details are available upon request.) Nevertheless, it should be noted that our results apply to “current” marriages.

  29. More generally, it is also true that there is marital sorting by qualification level. For instance, for any academic qualification level j (including no qualification), a woman with qualification level j is more likely to be married to a qualification level j male than any other women and vice versa.

  30. The effect of the wife holding an academic qualification on the husband’s economic activity rate persists, with nearly identical point estimates, also if we control for the husband himself holding some academic qualification. Moreover, this is true whether or not we instrument for the husband’s holding of an academic qualification using whether or not he, due to his month of birth, would have been required to stay on.

  31. Results are available upon request from the authors.

  32. For example, their preferred IV specification suggests that having academic qualification increases the probability of labour force participation by 32 percentage points (\(p<0.01\)).


  • Angrist J, Imbens G, Rubin D (1996) Identification of causal effects using instrumental variables. J Am Stat Assoc 91:444–455

    Article  Google Scholar 

  • Angrist JD, Krueger AB (1991) Does compulsory school attendance affect schooling and earnings? Q J Econ 106:979–1014

    Article  Google Scholar 

  • Beaujouan E, Bhrolchïn MN (2011) Cohabitation and marriage in Britain since the 1970s. Popul Trends 145:1–25

    Google Scholar 

  • Becker GS (1973) A theory of marriage: part I. J Polit Econ 81:813–846

    Article  Google Scholar 

  • Becker GS (1993) A treatise on the family. Harvard University Press, Cambridge. Enlarged edition

    Google Scholar 

  • Behrman JR, Rosenzweig MR (2002) Does increasing womens schooling raise the schooling of the next generation? Am Econ Rev 92:323–334

    Article  Google Scholar 

  • Breierova L, Duflo E (2004) The impact of education on fertility and child mortality: do fathers really matter less than mothers? NBER working paper no. 10513

  • Buckles K, Hungerman DM (2008) Season of birth and later outcomes: old questions, new answers. NBER working paper no. 14573

  • Burgess EW, Wallin P (1943) Homogamy in social characteristics. Am J Sociol 49:109–124

    Article  Google Scholar 

  • Chiappori PA, Iyigun M, Weiss Y (2009) Investment in schooling and the marriage market. Am Econ Rev 99:1689–1713

    Article  Google Scholar 

  • Crawford C, Dearden L, Meghir C (2007) When you are born matters: the impact of date of birth on child cognitive outcomes in England. CEE DP 93, Centre for the Economics of Education, London School of Economics, London

  • Del Bono E, Galindo-Rueda F (2007) The long term impacts of compulsory schooling: evidence from a natural experiment in school leaving dates. CEE DP 74, Centre for the Economics of Education, London School of Economics, London

  • Dickson M, Smith S (2011) What determines the return to education: an extra year or a hurdle cleared. Econ Educ Rev 30:1167–1176

    Article  Google Scholar 

  • Duflo E, Dupas P, Kremer M (2010) Education and fertility: experimental evidence from Kenya. Mimeo, Massachusetts Institute of Technology, Cambridge

    Google Scholar 

  • Fernandez R, Rogerson R (2001) Sorting and long-run inequality. Q J Econ 116:1305–41

    Article  Google Scholar 

  • Fernandez R, Guner N, Knowles J (2005) Love and money: a theoretical and empirical analysis of household sorting and inequality. Q J Econ 120:273–341

    Google Scholar 

  • Fort M (2007) Just a matter of time: empirical evidence on the causal effect of education on fertility in Italy. Mimeo, European University Institute, Florence

    Google Scholar 

  • Goldin C (1992) The meaning of college in the lives of american women: the past one-hundred years. NBER working paper no. 4099

  • Hahn J, Todd P, van der Klaauw W (2001) Identification and estimation of treatment effects with a regression discontinuity design. Econometrica 69:201–209

    Article  Google Scholar 

  • Hunt TC (1940) Occupational status and marriage selection. Am Social Rev 5:495–505

    Article  Google Scholar 

  • Imbens G, Angrist J (1994) Identification and estimation of local average treatment effects. Econometrica 62:467–475

    Article  Google Scholar 

  • Imbens GW, Lemieux T (2008) Regression discontinuity designs: a guide to practice. J Econometrics 142:615–635

    Article  Google Scholar 

  • Kirdar MG, Tayfur MD, Koç I (2010) The impact of schooling on the timing of marriage and fertility: evidence from a change in compulsory schooling law. MPRA paper number 13410

  • Konrad K, Lommerud KE (2010) Love and taxes—and matching institutions. Can J Econ 43:919–940

    Google Scholar 

  • Lee D, Lemieux T (2010) Regression discontinuity designs in economics. J Econ Lit 48:281–355

    Article  Google Scholar 

  • Lefgren L, McIntyre F (2006) The relationship between women’s education and marriage outcomes. J Labor Econ 24:787–830

    Article  Google Scholar 

  • McCrary J, Royer H (2011) The effect of female education on fertility and infant health: evidence from school entry laws using exact date of birth. Am Econ Rev 101:158–195

    Article  Google Scholar 

  • Oreopoulos P, Salvanes KG (2009) How large are returns to schooling? Hint: money isn’t everything. NBER working paper no. 15339

  • Peters M, Siow A (2002) Competing premarital investments. J Polit Econ 110:592–608

    Article  Google Scholar 

  • Rockwell R (1976) Historical trends and variations in educational homogamy. J Marriage Fam 38:83–96

    Article  Google Scholar 

Download references


The authors would like to thank Stacey Chen, Arnaud Chevalier, John Knowles, Costas Meghir, Jonathan Wadsworth and seminar participants at Kent, Sheffield, CESifo, the Max-Planck Institute, WPEG, ESPE, EEA, CUHK, RES and the GEARY Institute for the helpful comments and Vikesh Amin for the excellent research assistance. Financial support from the Nuffield Foundation (grant no. CPF/37724) is gratefully acknowledged.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Dan Anderberg.

Additional information

Responsible editor: Erdal Tekin



In this appendix, we explore whether individuals who, due to their month of birth, were required to stay on had parents with different economic characteristics than those individual who were not required to stay. In order to do this, we assemble a sample of youth from the LFS for whom we can observe also their parents as they are in the same household. The sample consists of individuals born towards the later stages of our main sample period, specifically between September 1967 and August 1971, who are observed in 1985 through to 1987, and for whom we have information about parents’ characteristics. For this sample, we estimate regression models of the same type used in the main body of the paper in order to explore whether the parents of those individuals born after the January–February threshold had different characteristics to the parents of those individuals born before the January–February threshold. The outcome variables used directly correspond to those used in the analysis of partner characteristics in the Table 6 of the paper, that is, whether the parent holds an academic qualification and whether the parent is economically active. The results are provided in Table 8 and reveal no systematic association between the individual’s requirement to stay on and parental characteristics.

Table 8 Estimates of discontinuity in parental characteristics at the January–February threshold

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Anderberg, D., Zhu, Y. What a difference a term makes: the effect of educational attainment on marital outcomes in the UK. J Popul Econ 27, 387–419 (2014).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:


  • Education
  • Marriage
  • Assortative mating

JEL Classifications

  • J12
  • J24