Poverty in Germany : Statistical Inference and Decomposition

Based on six harmonized cross-sections of the German Sample Survey of Income and Expenditure, we study inter-temporal changes in poverty from year 1978 to 2003. Results are decomposed by region and household types, and the bootstrap method is applied to test for the statistical significance of all our findings. Across household types, single parents with children have the highest poverty risk. Most striking is a huge regional divide in poverty which only narrows slightly over the period under investigation: the incidence and the intensity of poverty are substantially higher in the New states. A nonlinear Oaxaca-Blinder decomposition is conducted to quantify the separate contribution of regional differences in households' characteristics to the likelihood of being poor. Estimates from the decomposition indicate that differences in the distributions of socioeconomic characteristics play a negligible role for the 1993 poverty divide. Already in year 2003, however, differences in the distributions of characteristics explain more than fifty percent of the poverty divide, indicating that the poverty divide is likely to become a persistent phenomenon