Social Norms, Labor Market Opportunities, and the Marriage Gap for Skilled Women

Marianne Bertrand, Patricia Cortes, Claudia Olivetti, Jessica Pan
2016 Social Science Research Network  
In most of the developed world, skilled women marry at a lower rate than unskilled women. We document heterogeneity across countries in how the marriage gap for skilled women has evolved over time. As labor market opportunities for women have improved, the marriage gap has been growing in some countries but shrinking in others. We discuss a theoretical model in which the (negative) social attitudes towards working women might contribute towards the lower marriage rate of skilled women, and
more » ... also induce a non-linear relationship between their labor market prospects and their marriage outcomes. The model is suited to understand the dynamics of the marriage gap for skilled women over time within a country with set social attitudes towards working women. The model also delivers predictions about how the marriage gap for skilled women should react to changes in their labor market opportunities across countries with more or less conservative attitudes towards working women. We test the key predictions of this model in a panel of 23 developed countries, as well as in a panel of US states. To explain these patterns, we propose a theoretical model in which (negative) social attitudes 1 http://ec.europa.eu/eurostat/statistics-explained/index.php/Marriage_and_divorce_statistics 2 In 2012, the share of births outside of marriage was just over 2% in Korea and Japan, compared to the OECD average of 39%. The share of out-of-wedlock births in the US is about 40% (OECD Family Database, 2012) 4 See Appendix Table 1 for the average response to this question for each country with ISSP data. Hong Kong, Greece, and Italy are excluded from the 2002 and 2012 ISSP samples. 5 Larger gaps within countries are observed between skilled and unskilled men: educated men tend to be less conservative. 6 The earlier years of the surveys cover a relatively small number of countries, making it di cult to distinguish actual di↵erences in responses across countries from di↵erences due to the di↵erent survey years. Therefore, rather
doi:10.2139/ssrn.2728392 fatcat:ooofvrs6fvdmtewosuusyp24na