Arend Lijphart's dimensions of democracy: Logical connections and institutional design

Abstract

Lijphart's (1999) analysis maps countries along two dimensions of democratic institutions: 'executives-parties' or 'joint-power', and 'federal-unitary' or 'divided-power'. My 'meta-study' maps the methodology of Lijphart's mapping: the nature of indices (inputs or outputs), their logical interconnections, their susceptibility to institutional design ('constitutional engineering'), and their suitability for expressing the intended underlying concepts. Strikingly, the joint-power indicators are highly correlated and mostly logically connected output measures, which are not susceptible to institutional design, while the opposite is true for the divided-power dimension. For this dimension most indices are expert estimates of inputs, marginally correlated, yet subject to institutional design, limited by size dependence. Surprisingly, the parliamentary-presidential aspect of institutional design does not affect the picture. The connection between cabinet life and the number of parties is even stronger than found by Lijphart. Interest groups and central bank independence fit his dimensions empirically but less so logically. In sum, institutional design may be more difficult than sometimes assumed, but offers hope.