Bayesian Poisson Regression for Reconstructing Fertility Rates

Afua Durowaa-Boateng, Vienna Institute of Demography
Dilek Yildiz, Wittgenstein Centre for Demography and Global Human Capital

The impact of education on fertility has been widely studied. Many of these studies have found an association between higher educational attainment levels and decreasing fertility rates. However, access to quality historic data on fertility by level of education has been limited, especially in developing countries. The main source of detailed fertility data in many developing countries has been provided by the Demographic and Health Surveys (DHS). However due to the data collection methods employed, the quality of data is sometimes sub-optimal. Certain issues including but not limited to incomplete and missing data sets may arise from collection methods. To deal with missing data, Schoumaker, (2013) proposes a Poisson regression model for reconstructing birth histories from DHS waves. We extend the model by Schoumaker to include education differences in fertility. The number of observations decrease when broken down by age and education. To deal with the small sample size we propose a Bayesian model to estimate the true age and education speci c fertility rates and the uncertainty around them. We present preliminary results for Brazil and Turkey.

Keywords: Bayesian methods / estimation, Fertility and childbirth, Demographic and social surveys

See extended abstract.

  Presented in Session P14.