COUPLES’ DECISIONS ABOUT WORK: Evidence from the British Household Panel Survey

When talking about couples, it is common to hear the saying ‘opposites attract’. But evidence suggests that people with similar characteristics are more likely to be together: for example, couples are likely to share similar levels of educational achievement. Is the same true when it comes to a couple’s employment decisions?

New research by Melisa Sayli, to be presented at the Royal Economic Society's annual conference at the University of Warwick in April 2019, shows that while a male partner’s non-employment seems to have a negative effect on a woman’s labour force participation, this effect disappears when taking account of unobserved permanent characteristics of the couple.

To understand the relationship between partners’ labour market outcomes, the study focuses on cohabiting and married couples, using couples’ monthly labour market histories from the British Household Panel Survey 1991-2009 (BHPS). The author explores the effect of a male partner’s non-employment on a woman’s labour force participation among couples in which the partners are not in full-time education and are in a continuous relationship.

Assuming that a partner’s labour market activity is exogenous to a woman’s decision to be active in the labour market, previous research on the UK finds a strong negative correlation between a woman’s labour force participation and her partner’s unemployment (non-employment).

But it is not clear whether the negative relationship arises as a causal effect of the partner’s non-employment or due to the partners’ tastes, preferences or coordination in labour market decisions. And if the latter, whether the source of endogeneity is due to idiosyncratic shocks or due to unobserved permanent characteristics of a couple?

Due to the lack of natural experiments that would serve for identification during the study period, the research uses instrumental variables approach. One of the instruments for the male partner’s non-employment the study evaluates is the male partner’s self-assessed health and other objective and other measures of health, e.g. limitations to work, number of GP visits.One of the instruments evaluated for a partner’s non-employment is his self-assessed health and other measures of health

The idea behind the instrument is that a male partner’s poor health would lower his probability of being employed, which would either increase the need for additional income and/or lowers a woman’s opportunity cost for market work. Hence, the woman would be expected to participate in the labour force.

A potential concern for the validity of this instrument is that the woman’s labour force participation may be influenced by her partner’s poor health directly as she may care for her partner due to his poor health. Thus her participation may be directly affected by his health and not, and not through his non-employment. The study addresses this concern by controlling for whether the including a dummy variable for a  woman cares for her partner.

What is the observed sorting pattern among couples? Are ‘opposites’ or ‘likes’ more likely to be together in the data? Comparing the couples to a counterfactual case where individuals select their partner randomly, couples with the same education levels are more likely to be together.

This positive ‘assortative mating’ is strongest on the basis of educational homogamy (46% of couples), especially among couples with first degrees or higher who are observed three times more compared to random matching. But there is no clear sorting pattern based on parental employment backgrounds and their health.

While people with similar non-market demographic groups characteristics form couples, after controlling for observed characteristics there is negative assortative mating based on unobservable factors common to both partners’ labour market outcomes, such as preferences and tastes towards paid work. The unobserved factors and characteristics that make a woman more likely to be active in the labour market are positively correlated with the factors that make the male partner more likely to be non-employed. This correlation ranges from 25% to 10% and presents an upper bound as it does not differentiate between couples’ permanent unobserved characteristics and idiosyncratic shocks.

As expected, the study The finding from the pooled model suggests that there is a finds a monotonic relationship between with a partner’s health and his non-employment – that is, if a partner has poor health, he is more likely to be not working.

The male partner’sHis non-employment reduces the participation probability of an average woman by 31 percentage points. This effect is potentially overestimated and inconsistent as the partner’s non-employment may be correlated with some unobserved factors that played a role in couple formation (couple fixed effects).

Controlling for couple fixed effects, the research finds that this negative effect drops to 13 percentage points,. Instrumenting partner’s non-employment when fixed effects are controlled for yields an effect of the same magnitude, yet the significance of partner’s non-employment disappears.  once instrumented, The study shows that suggesting that a selection effect rather than a causal effect of partner’s non-employment, is the source of endogeneity in the couple’s labour market outcomes relationship.

Partners Match and then the Couple Decides: An empirical analysis of couples’ labour supply in the UK by Melisa Sayli

Melisa Sayli

Postdoctoral Research Associate at University of York