Panel data analysis and machine learning practice for the factors that affect the birth rate: The Turkiye sample

This research aims to examine the factors that affect the fertility rate in Turkiye for the years 2009-2021, utilizing panel data analyses and machine learning (ML). To achieve this, the relationship between the fertility rate at the provincial and yearly levels and various factors such as the rate of people with undergraduate or higher education, women's employment rate, women's population rate, gross domestic product (GDP) level, married population ratio, COVID-19, and immigrant dummy variables has been investigated. Three main models, namely Pooled Ordinary Least Squares (OLS), Fixed Effects Model, and Random Effects Model, have been employed for panel data analyses. For ML, different classification models have been implemented to solve this problem, and the model that best learns the change of pattern in birth rate has been classified. The results obtained from ML models have been compared with panel data analysis results, and differences in terms of model variables have been examined. The findings; it has been observed that conducting panel data analysis with fixed effects model yields meaningful results. In Turkiye, the birth rate is found to be negatively correlated with the level of undergraduate and higher education, female employment, COVID-19, and the immigrant dummy variable, while it is positively correlated with the female population rate, married ratio, and GDP level. In ML, focused on transforming the change in birth rate into a classification problem, examining which model performed better and the levels of importance for the variables. Both models used have demonstrated effective performance in identifying factors that affect the fertility rate.