By constructing and analyzing time-series cross-section data of 21 countries from 1998 to 2018, I show that long-term associations between employment rates of different socio-demographic groups and public pension spending per older person systematically vary across pension regimes.

In "Beveridgean" pension systems, female employment rates and poverty risks are more sensitive to changes in public pension spending, whereas in the "Bismarckian" regime, educational differences in the effects on male employment are particularly larger