Describing the socio-demographic structure of social transfers across Europe with regression trees

Author: Klaus Pforr

Cooperation partners: Johanna Jung, Barbara Felderer

Project description

There is considerable research on the effects of macro variables on social transfers and other welfare state characteristics (e.g. labor market rigidity). There is also a large theoretical literature describing and systematizing the general structure of welfare states. In this project, we take an intermediate approach by examining the socio-demographic structure of social transfers descriptively, without making theoretical assumptions, and using a data-driven analysis. The descriptive objective is to find out which socio-demographic groups can be delineated that receive or pay similar social transfers, and whether and how these groups differ across countries and in terms of macro indicators. More specifically, we look at social transfers at the household level, i.e. the sum of unemployment, old-age, survivors, sickness and disability benefits and education-related allowances at the individual level in a household and family-related and housing allowances and benefits for other social exclusions at the household level minus the sum of wealth and income taxes, social security contributions and inter-household cash transfers. We regress this dependent variable on several socio-demographic variables and macro-level indicators. We use a regression tree as a descriptive tool to overcome the limitations of OLS, which are that all non-linear relationships of the socio-demographic and macroeconomic level have to be specified ex-ante. Moreover, the regression tree directly separates clusters described by the relevant independent variables with similar levels of social transfers. Our preliminary results show that social transfers are mainly structured by age, household income, family structure and labor intensity. The differences in social transfers between these groups within countries are larger than the differences between countries as a whole, i.e. the country-level indicators play only a minor role for the differences in social transfers compared to the socio-demographic variables. Our approach and its concrete application are subject to some limitations. Most importantly, regression trees, similar to techniques such as cluster analysis, are notoriously unstable and prone to overinterpretation. Therefore, we perform multiple specifications to find stable patterns across different trees.

Publications:

Pforr, Klaus, Johanna Jung und Barbara Felderer (2024): "Describing the socio-demographic structure of social transfers across Europe with regression trees." ifo Conference on Understanding Socio-Economic Inequalities with Novel Data and Methods, ifo Institut – Leibniz-Institut für Wirtschaftsforschung an der Universität München e.V., Munich, 2024-02-22.

Pforr, Klaus, Johanna Jung und Barbara Felderer (2023): "Describing the socio-demographic structure of social transfers across Europe with regression trees." Frühjahrstagung der DGS-Sektion „Soziale Ungleichheit und Sozialstrukturanalyse", GESIS - Leibniz-Institut für Sozialwissenschaften, Mannheim, 2023-03-24.

Pforr, Klaus, Johanna Jung und Barbara Felderer (2023): "Describing the socio-demographic structure of social transfers across Europe with regression trees." 8th European User Conference for EU-Microdata, Universität Mannheim, Mannheim, 2023-03-17.