Until recently, there has been virtually no discussion among professional economists of the impact of government expenditures on the distribution of income.1Neoclassical economics has traditionally shown little interest in distributional issues. Little is said beyond the assumption that factors are paid their marginal products. Micro economics is said to take a “neutral” stance with regard to distributional issues. Static efficiency of allocation is attainable for any income distribution, and consequently, so the parable goes, no income distribution is superior on purely economic grounds to any other. Macro economics also purports to be neutral with respect to distribution. Government expenditures in Keynes' model appear as an undifferentiated blob called “G”. The only interest macro economics takes in distribution issues is concerned with the marginal effect of redistribution on the marginal propensity to consume out of income. Keynesian economics, therefore, is unable to say whether one form of government expenditure is superior to another so long as both accomplish macro objectives.When orthodox economists have approached the issue of the government's distributional impact, they have until recently focused solely on its use of taxes and transfer payments. Public finance has traditionally ignored the expenditure side of state activity since, after all, government activity was a necessary evil, benefiting no one. Gillespie's path‐breaking study in 1965 finally acknowledged the utility of government spending, but his analysis and those that have followed in the orthodox tradition have been hampered by a number of awkward premises.First, the orthodox studies of fiscal incidence implicitly accept the view of the government as a neutral arbiter rather than a protagonist of the dominant classes in society. Second, benefits of government services are assumed to be accurately measured by outlays. Thus, if we find that the government spends four times as much on highways as on police, it is assumed that the utility of highways is four times that of police even though one cannot even imagine the continuity of the status quo without the police while many responsible citizens argue that we should drastically curtail outlays on roads. Obviously, the utility of the police in terms of system maintenance exceeds that of the more expensive highway expenditures. Third, it is assumed that for each dollar spent by the government, only one person will benefit when, in fact, many disparate groups can benefit from the same expenditure. A dollar spent on education benefits the student as well as hislher employer. Fourth, Gillespie and his orthodox followers ignore any effect of the government on the pre‐tax, pre‐transfer distribution of income which they take as given. A hypothesis which we examine in this paper is that the government has an enormous influence over the shape of the pre‐tax, pre‐transfer income distribution.A more general criticism of previous studies of fiscal incidence is that they suffer from a poorly defined theory of the state. This assertion is most clzarly illustrated by the categorization in previous studies of a wide variety of public exp‐enditures as “public goods” (such as national military expenditures). The benefits of these “public goods” are allocated among various income groups in several ways, for example on the basis of wealth ownership (both productive and consumptive) or on a per capita basis. The method of allocation chosen has enormous consequences for one's estimate of overall fiscal incidence. According to Herriot and Miller, those with incomes over $50,000 either receive a net benefit of 4.5 percent of their total income from the government or lose 42.1 percent, depending upon the allocation formula chosen for public goods. Previous studies have taken an agnostic position with respect to the appropriateness of the several allocative assumptions. But this is merely simple empiricism without theoreticai foundation, and thus the formulation of specific hypotheses which employ scientific procedures is impossible. What is needed to provide an interpretation of the data is a well‐articulated theory of the state‐an area to which we turn our attention in