Ethnicity, education, and the non-proportional hazard of first marriage in Turkey

Abstract

This study uses the 1998 Turkish Demographic and Health Survey to estimate non-proportional piecewise-constant hazards for first marriage among women in Turkey by education and ethnicity, with controls for region of residence and rural–urban migration. At low education levels Kurdish speakers married earlier than women who spoke Turkish or other languages, but at high education levels Kurdish women delayed marriage more than other women. This reversal across education groups furnishes a new illustration of the minority-group-status hypothesis specifically focused on marriage as the first step in the family formation process. The ethnic contrast concerned only marriage timing in Turkey, not proportions ever marrying. Eventual marriage remained nearly universal for all groups of women. This means that an assumption of proportional duration hazards (widespread in contemporary research) across the whole range of marriage-forming ages should be replaced by models with non-proportional duration hazards.