Protective Effect of Marriage on Health: Instant or Cumulative, Short- or Long-Term?

Wednesday, 18 July 2018: 10:30
Oral Presentation
Malgorzata MIKUCKA, MZES, Mannheim University, Germany, Université catholique de Louvain, Belgium, Belgium
Oliver ARRANZ BECKER, Martin Luther University Halle - Wittenberg, Germany
Christof WOLF, GESIS Leibniz-Institute for the Social Sciences, Germany, Mannheim University, Germany
Although research on the protective effect of marriage for health has yielded vast empirical evidence, the dynamics of this effect is not always clearly conceptualized. Moreover, the previous literature is dominated by U.S. studies, which creates the need to analyze different social contexts.

This paper contributes to the field by explicitly distinguishing between an instant vs. cumulative effect of marriage on health, as well as between a short- vs. long-term effect of the transition into first marriage. We analyze mental and physical health, functional limitations, self-rated health, and health satisfaction. Our contribution extends the existing empirical evidence to German panel data, which allow observing long-term health trajectories.

We use German SOEP data covering the period 1984-2015 and fixed effects regression for panel data to model the dynamics of the effects of marriage on health, separately for men and women.

Our results show a cumulative protective effect of first marriage on self-rated health among men, which, however, seems to disappear after controlling for selection into marriage. Among women, self-rated health decreases after marriage and remains on a low level in the long run. Marriage effects on the other health measures are rather inconsistent and weak. The only consistent effect is the positive short-term effect of the transition into first marriage on various health outcomes.

In contrast to most previous studies, our analysis explicitly models various types of dynamics in the protective effect of marriage on health. Our results suggest that the protective effects of marriage for health are generally overstated. This shows how important it is to use FE models to control for time-invariant covariates, to model simultaneously the various types of dynamics, and to attempt to control for selection on factors correlated with health trajectories.