Esping-Andersen developed a typology of welfare regimes: conservative, liberal and social democratic, which are measured on the basis of seven indicators. We re-examine Esping-Andersen's data as well as replication data compiled by Scruggs and Allan to show that the seven indicators do not form valid measures of these welfare regimes. In addition to divergences in their measurement, the seven indicators are a mixture of institutional characteristics of welfare systems and outcome measures of social stratification. A measurement model based on the five institutional characteristics of welfare regimes that pertain to social insurance, however, does fit both the original and replication data. This article therefore proposes a three-dimensional model of conservative and liberal social insurance, which treats universal insurance coverage as the third dimension, instead of Esping-Andersen's ‘socialist’ regime. Although this does not fundamentally alter the typology of countries, it has implications for previous studies that employ country scores based on Esping-Andersen's method as independent variables in causal models. To illustrate these implications, this article re-examines a study by Noël and Thérien and calls into question their conclusions on the causal connections between the social democratic welfare state and levels of foreign aid.