At the margin: By how much do social transfers reduce material deprivation in Europe ?

Since the adoption of the Europe 2020 social inclusion target, the population at risk of poverty or social exclusion has not decreased sufficiently to meet the target in EU countries. It is therefore important to evaluate the role and effectiveness of the policies adopted to combat income poverty and social exclusion in Europe. This paper takes a regression-based approach to estimating the effects of transfers on material deprivation, using the 32 countries covered by the EU Statistics on Income and Living Conditions (EU-SILC) and the new indicator of material and social deprivation agreed at EU level in 2017 (see Guio et al., 2017). It thereby complements established methods using income to evaluate policy impacts. The approach has broader applicability, suited to social indicators whose scaling has similar properties, such as housing deprivation indicators. This paper is the first to estimate and compare the average marginal effect (AME) of a small income transfer across the 32 EU-SILC countries. It finds that the impact of social transfers on material deprivation is higher at lower living standards, an effect that is present both within and across countries and underlines the importance of a progressive social transfer system. It further finds that households receiving non-pension transfers would experience, on average, the largest decrease in material deprivation. In many Eastern European countries, receivers of pension transfers would also see above population average reductions in material deprivation. A comparison of the Baltic States illustrates how factors other than the national social transfer system and living standards studied here play a role in explaining cross-national differences in AME. The paper further calculates the predicted effects of such increases in social transfers on the EU’s official material deprivation rate and social spending levels. From an econometric point of view, this paper offers new methodological insights, by systematically comparing the performance of count data models (Poisson, negative binomial and zero-inflated negative binomial) and ordered regression models (ordered logit and generalized ordered logit) to predict the material deprivation distribution. It finds that ordered logit models systematically outperform the count models. [+]