Analysis of fertility in metropolitan areas of Turkey: an application of structural equational modelling

Hande Tunckanat, Hacettepe University

This study is mainly an attempt for analyzing Demographic and Health Survey data from a different perspective in order to gain deeper insights about mechanisms of fertility decline in Turkey. Influenced by social psychology, it considers fertility as a phenomenon revealed as a result of a decision process instead of a concept which stands-alone in the space and affected directly by various factors. The focus is on the fertility-related decision process of women, which is modeled as one’s perception affects her attitudes which in turn affect her fertility behavior. Also influenced by social network theory and diffusion theory, it is hypothesized that a woman’s daily-life activities and the social environment a woman surrounded by have an indirect effect on her fertility-related decision process. Although this idea should, ideally, be confirmed or falsified with longitudinal data, it has been tested with cross-sectional data drawn from the Turkey Demographic and Health Survey-2003 (TNSA-2003). Because the intellectual interest in here is to see if it is possible to handle DHS data from a causal perspective, using secondary data that already obtained for other purposes is not considered as a barrier. The unit of analysis selected as ever-married women aged 15-49 living in metropolitan areas and those in rural areas as a control group in order to control the potential indirect effect of social environment on women’s decision processes about fertility. Structural Equational Modeling (SEM) that enables the examination of casual and indirect relations and to work with variables that do not directly exist in the data set (latent variables) is used. Analyses show that it is possible to catch causal relations and indirect effects of determinants which are significant in woman’s fertility-related decision process as well as the effect of the social environment that surrounds her even with the secondary cross-sectional data.