Causal analysis of time to first birth and first marriage in Turkey

Elif Yiğit, Hacettepe University

The fertility research in Turkey is mainly concentrated on static analysis of fertility models. However in order to understand the determinants of time interval passed until the event experienced, it is rather important to use dynamic approaches. In a methodological view, the main purpose of this study is to model the age at first marriage and first birth by using event history techniques with demographic, social, economic and contextual characteristics of women and to estimate their effects on the transitions to first union and first birth. In the data analysis of this study, the 2003 Turkey Demographic and Health Survey (TDHS-2003) is used as the principle data source.As a result of applying all the event history methods to the transition to first birth, it is found that according to the AIC values for the time to first birth Log-logistic parametric method is the best one to explain the time to event with selected covariates. Moreover, the log-logistic model has the largest log likelihood which indicates the best-fitting model is the log-logistic model. For age at first marriage Gamma Generalized model is selected according to AIC although this model has not the largest log likelihood which indicates the best-fitting model. The cohort effect is important in the hazard of first birth. Being in an older cohort reduces the hazard of first birth compared to being in a younger cohort. The effect of living in Central and Eastern regions shows an increasing hazard of ten percent higher than living in the Western region. Kurdish speaking women are significantly different from Turkish speaking women with higher hazards of first birth. Both in age at first marriage and age at first birth analysis the most important covariates are age, education of women and ethnicity.