Despite the explosive increase in the research program on government growth in recent years, little work has been done on government growth disaggregated to the subnational level. I examine the empirical validity of five competing models of government growth for the fifty U.S. states from 1945 to 1984: Wagner's Law, fiscal illusion, party control, bureau voting, and intergovernmental grant. Government size is defined in terms of state government spending as a proportion of total state economic output, with separate implicit price deflators being employed for the public and private sectors. Based on a longitudinal test of these competing models, the analysis uncovers strong empirical support for the bureau voting and intergovernmental grant models, moderately weak support for the Wagner's Law model, and virtually no support for the fiscal illusion and party control explanations. These findings have important implications for the study of government growth in general and, more specifically, in the states.