The most recent literature on aid effectiveness finds a positive effect of aid on growth. To the extent that aid goes through the budget, this either reflects an aid-financed increase in government expenditures (quantity effect) or an improvement in the use of government resources as a result of donor involvement and lower taxes (quality effect). This study investigates the causal link between on-budget aid and government expenditures using a large cross-country panel data set for 53 countries and recent methodology to test Granger causality in heterogeneous panels. I find that in most countries donors do not change aid in response to changes in the level of government expenditures and that governments react to aid by changing the way they use their own resources rather than by increasing spending. There is little support for a quantity effect: aid Granger causes government expenditures in only eight countries. This suggests that aid substitutes for domestic government revenue and that aid is effective largely through the quality effect.