Although many studies indicate that both the level and composition of public spending are significant for economic growth, the results in the empirical literature are still mixed. This paper studies the importance of country sample selection and expenditure classification in explaining these conflicting results. It investigates a set of fast-growing countries versus a mix of countries with different growth patterns. The regression specifications include different components of public expenditure and total fiscal revenues, always considering the overall government budget constraint. Total public spending is first disaggregated using a definition that classifies public spending as productive versus unproductive components, an a priori criterion that is based on the expected impact of public spending items on the private sector production function. After empirically confirming the validity of this definition in the panel analysis, the authors suggest and test an alternative definition of "core" public spending that may be more appropriate for developing countries. The empirical analysis shows that the link between growth and public spending, especially the productive and "core" components, is strong only for the fast-growing group. In addition, macroeconomic stability, openness, and private sector investment are significant in the fast-growing group, which points to the existence of an economic policy environment more conducive to growth in the first group of countries. The authors conclude that public spending can be a significant determinant of growth for countries that are capable of using funds for productive purposes.