Over the past decades, there has been considerable interest in how the ethnic, racial, and class composition of a community affects its behavior. There have been numerous discussions of how the ethnic background of immigrants interacted with conditions they faced in urban or rural America to produce group behavior and community choices that contributed to the relative performance of those groups (e.g., see Sowell, 1981; Glazer and Moynihan, 1963; Glazer, 1983; Lieberson, 1980). One choice has received particular emphasis: a group's commitment to increased education for its own members and others. Most of this scholarly work has relied on historical and survey methods to analyze differences in various groups' commitments to educational improvement. In this article, we use a new technique and Michigan schooldistrict data from 1970 to identify how the ethnic, racial, andclass composition of heterogeneous communities systematically affects levels of local public educational expenditures. The literature on school spending contains three types of hypotheses that need to be distinguished. The first hypothesis is that the demand for school spending varies with income and that the relationship is probably not monotonic but instead U-shaped-i.e., the rich and the poor favor larger expenditures. While that sounds straightforward, it is in fact one of the most confusing areas of debate. One favored explanation is that the poor want large school budgets because they do not pay very much of the cost and the rich want large budgets because they are public-regarding persons who are glad to pay for the betterment of the community (Wilson and Banfield, 1964). This explanation, however, illustrates a possible confusion between a price effect and an income effect. According to this logic, if the poor pay a smaller share of.a publicly provided good, they will demand more because they face a lower price. But we also expect that if prices are held constant while income is varied, expenditures on any commodity will begin at some subsistence or threshold level, rise as income increases, and eventually level off as demand for the commodity is sated.' Therefore, the rich may demand more despite having to pay a higher share of the cost (or price). It is also possible that the very poor, seeing education as a path to a better life for their children, are willing to pay a disproportionate share of their income for schooling (again, holding price constant). If this hypothesis about spending is correct, a typical family's demand function for schooling will have a nonmonotonic relation to income. As Figure 1 shows, demand as a function of family income (Yf) (other things, including price, held constant) falls to point A then rises to a peak. at B. If public-regardingness is extremely important, it will push the point where that curve reaches a maximum out. to extremely high income levels. A second set of hypotheses concerns the effects of race and ethnicity on public expenditures. In the early versions of their theory, Wilson and Banfield (1964) suggested that Northern and Southern Europeans brought quite different values with them to America and that these imported tastes determined the political choices such groups made. In later versions (Wilson and Banfield, 1971) they retreated from identifying a particular ethos with a particular country of origin. But they still maintained that members of an ethnic or racial group fit predominantly into one of two opposing world views about 'the role of government: a public-regarding, or unitary, ethos and a private-regarding, or individualistic, ethos. These views could manifest themselves either directly in attitudinal survey data or indirectly through ethnic variations in voting patterns or (with long and variable lags) through variations in political institutions. The local school system is an especially good place The authors contributed equally to this article. They wish to thank several anonymous referees for Sociology of Education for helpful comments on an earlier draft. Address all correspondence to Professor Daniel H. Saks, Peabody College, Vanderbilt University, Nashville, TN 37203. I There have been many studies of these relationships, called Engel curves. For a discussion, see Ramsey (1974).