Abstract. This article analyzes various pitfalls that arise in the application of panel data methods in comparative political economy. Empirically, we refer to the debate on the globalization-welfare state nexus by re-assessing a study by Garrett and Mitchell (‘Globalization, Government Spending and Taxation in the OECD’, European Journal of Political Research 39(1) (2001): 145–177). We discuss the properties of specifications with time invariate political variables, dynamic models with nonstationary data, and autocorrelated residuals. We demonstrate that the findings of previous empirical studies are often driven by mis-specifications. Presenting a statistically well-behaved model, we find evidence that government spending is primarily driven by the state of the domestic economy. Neither partisan effects nor the international economic environment have affected public expenditure considerably.