Multilevel Determinants of Household Poverty: A Mixed Effects Logistic Regression Analysis

Abstract

This study investigates the socioeconomic determinants of poverty in Turkey by incorporating household characteristics with regional indicators using 2024 data using a multilevel mixed-effects model. Relative poverty, defined as 60% of the median equivalised household disposable income, serves as the dependent variable. Considering the hierarchical structure of the data, the determinants of poverty status were examined employing a two-level generalised mixed-effect logistic regression model, accounting for the nesting of households within NUTS-2 regions. The findings reveal that poverty cannot be explained solely by household-level vulnerabilities; the regional economic environment also significantly shapes poverty outcomes. Also, education is determined as the most security measure; having a university degree decreases poverty risk by 42.6%. Across regions, resources must be directed towards GDP growth and infrastructure to mitigate spatial disparities by integrating household-level microdata with regional macro indicators. This study provides a comprehensive understanding of the spatial and key structural factors behind poverty in Turkey. Policymakers should prioritise opportunities to pursue higher education and vocational training at the household level.