This paper provides a synthesis of the 1983-98 published literature on the empirical evidence regarding the interaction between government policies and growth. Five policy areas are considered: general government consumption, tax rates, education expenditures, defense and public infrastructure. The most conclusive results in the literature relate to the positive impact of education expenditures on growth. Public infrastructure also appears important. Regression analysis remains the most commonly adopted research methodology. A better link with current theories will be obtained when parameter calibration methods for micro-foundations based models replace parameter estimation of regression models with ad hoc specifications. Nonetheless, there remain severe limitations on what can be learned for policy from highly aggregative models of endogenous growth. Better data are needed at the regional macro and meso levels to complement the currently available pooled cross-section time-series country data. The potential endogeneity of government fiscal variables can be resolved through the selection of appropriate instrumental variables, such as those that arise in cases of "natural experiments".