The extensive literature on the size and growth of government attests to the long-standing interest of social scientists in the interrelations of economic development, income distribution, political processes, bureaucracy, and tax rates. Recent surveys of parts of this literature (Peacock, 1979;Cameron, 1978;and Larkey, Stolp and Winer, 1981) show that neither theoretical nor empirical work has resolved the main issues. (See also Peltzman, 1980.) There is little agreement about a common model or framework for predicting the size of government or discussing the causes of government growth or decline. And there is no consensus about the empirical evidence on the reasons for changes in the size of government.The disputes are not about the relevant facts for the twentieth century. Nutter (1978) shows that if the average size of government is measured by either the share of income taken in taxes or spent by government, the size of government has increased in all developed, market economies during the past quarter century. Growth is not a recent phenomenon. Data for the average tax rate or government's share of spending suggest that, judged by these measures, the size of government increased also during the first half of the century in several countries.Many explanations of the increased size of government emphasize the role of 'suppliers' of government services. Several recent studies (Romer and Rosenthal, 1978;Niskanen, 1971;Fiorina and Noll, 1978) suggest that there is an element of monopoly power on the supply side.1 Congress, bureaucrats, or 'interest groups' are able to raise government spending above the level that utility maximizing households or voters would choose in the absence of this monopoly power. Models of this kind posit some explicit or implicit cost of acquiring information or taking action and reject the stylized model of the fully informed maximizing voter. Models of 'supply' usually exclude the combined effects of taxes and spending on voters' choices by omitting the * We wish to acknowledge the issistance of Joshua Angrist in compiling the data used in our tests and the helpful advice of Dennis Epple. Thanks also to Peter Aranson for helpful comments in the editing process. GSIA, Carnegie-Mellon University, Pittsburgh, PA 15213.