¹ Median family income under $3,000 per year. Needless to say, most voters in this category are Negroes.

² The cities and elections examined are:

Cleveland-Cuyahoga County: Nov., 1956; Nov., 1959; May, 1960; Nov., 1960.

Chicago-Cook County: June, 1957; Nov., 1958; Nov., 1959; April, 1962.

Detroit-Wayne County: August, 1960; Feb., 1961; April, 1961; April, 1963.

Kansas City: Nov., 1960; March, 1962.

Los Angeles: Nov., 1962.

Miami: Nov., 1956; May, 1960.

St. Louis: March, 1962; Nov., 1962; March, 1963.

³ The degree of association was also calculated using a nonparametric statistic (Kendall's tau). The relationship persists but at lower values. Since we are satisfied that the relationship found by r is not spurious, we have relied on it for the balance of the analysis because of its capacity to produce partial correlation coefficients.

⁴ Only two measures of tax liability can be got from the Census: median home value and median family income. We have used the latter for the most part. The Census classifies all homes valued at over $25,000 together, thereby collapsing distinctions that are important for us. We think, too, that people are more likely to know their incomes than to know the current market value of their homes, and that therefore the Census information on incomes is more reliable. Finally, in neighborhoods populated mostly by renters, median home values are likely to be unrepresentative of the class character of the neighborhood: this is so, for example, where a few owner-occupied slums exist in a district of luxury apartments.

⁵ Other studies which suggest that upper-income groups may have a greater preference for public expenditures than middle-income groups include Williams, Oliver P. and Adrian, Charles R., Four Cities: A Study in Comparative Policy Making (Philadelphia: University of Pennsylvania Press, 1963)CrossRefGoogle Scholar, ch. v; Boskoff, Alvin and Zeigler, Harmon, Voting Patterns in a Local Election (Philadelphia: J. B. Lippincott Co., 1964)Google Scholar, ch. iii; Watson, Richard A., The Politics of Urban Change (Kansas City, Mo.: Community Studies, Inc., 1963)Google Scholar, ch. iv; and Salisbury, Robert H. and Black, Gordon, “Class and Party in Non-Partisan Elections: The Case of Des Moines,” this Review, Vol. 57 (September, 1963), p. 591Google Scholar. The Williams-Adrian and Salisbury-Black studies use electoral data; the Boskoff-Zeigler and Watson studies use survey data. See also Otto A. Davis, “Empirical Evidence of ‘Political’ Influences Upon the Expenditure and Taxation Policies of Public Schools,” Graduate School of Industrial Administration of the Carnegie Institute of Technology, January, 1964 (mimeo), and Birdsall, William C., “Public Finance Allocation Decisions and the Preferences of Citizens: Some Theoretical and Empirical Considerations,” unpublished Ph.D. thesis, Department of Economics, Johns Hopkins University, 1963Google Scholar. A difficulty with the Davis and Birdsall studies is the size (and thus the heterogeneity) of the units of analysis—entire school districts in one case, entire cities in the other.

⁶ Michigan is one of the few states which restricts the right to vote on expenditures to property owners and their spouses. Because the Flint returns were tabulated on a precinct basis, demographic data had to be obtained from block rather than tract statistics; since median family income is given only for tracts, median value of owner-occupied homes had to be used.

Possibly the flood control benefits would be distributed roughly in proportion to the value of properties; about this we cannot say. However, it is worth noting that the vote in Flint on other expenditures which presumably would not distribute benefits in proportion to the value of properties (e.g., parks) followed the same pattern.

⁷ We isolated all precincts in Census tracts having median family incomes of at least $10,000 a year, with at least 70 percent home ownership (the central city of Chicago was excepted here), and at least 70 percent of the population third (or more) generation native born.

⁸ A person is of “foreign stock” if he was born abroad or if one or both of his parents was born abroad. We believe that the reason why a significant relationship does not appear for the suburbs is that there is a considerable number of Jews among the foreign stock of the suburbs. In the central city, there are practically no Jews. Like other Jews, Jews of Eastern European origin tend to favor expenditures proposals of all kinds. Their presence in the suburbs, therefore, offsets the “No” vote of the non-Jews of foreign stock.

⁹ Since no home-owning ward or town in Cuyahoga County is more than 25 percent Polish-Czech according to the 1960 Census, it may be that no inferences can be drawn from the voting data about Polish-Czech behavior. Three considerations increase our confidence in the possibility of drawing inferences, however. (1) Only first- and second-generation Poles and Czechs are counted as such by the Census, but third- and fourth-generation Poles and Czechs tend to live in the same wards and towns; thus the proportion of the electorate sharing Polish-Czech cultural values (the relevant thing from our standpoint) is considerably larger than the Census figures suggest. (2) When other factors are held constant, even small increases in the number of Poles and Czechs are accompanied by increases in the “No” vote; nothing “explains” this except the hypothesis that the Poles and Czechs make the difference. (3) When we take as the unit for analysis not wards but precincts of a few hundred persons that are known to contain very high proportions of Poles and Czechs, we get the same results. Because we are using ecological, not individual, data, we are perforce analyzing the behavior of ethnic “ghettos” where ethnic identification and attitudes are probably reinforced. Poles in non-Polish wards, for example, may behave quite differently.

¹⁰ The method by which these precincts were selected is given in the Appendix. Unfortunately, it proved impossible to identify relatively homogeneous precincts typical of other ethnic groups at various income levels and degrees of home-ownership. For example, middle-income Jews tend to be renters, not home owners, and there are practically no low-income Jewish precincts in either city. A complete list of these precincts is available from the authors.

¹¹ Cf. Downs, Anthony, “The Public Interest: Its Meaning in a Democracy,” Social Research, Vol. 29 (Spring 1962), pp. 28–29Google Scholar.

¹² The proposition that “subculture” can be defined in ethnic and income terms is highly provisional. We are looking for other and better criteria and we think we may find some. But so far as the present data are concerned, ethnic and income status are all we have.

¹³ Two other explanations are possible and, in our opinion, plausible. One is that the low-income renters may have taken into account costs to them other than taxes—e.g., the cost, (perhaps monetary) of changes in the neighborhood that would ensue from expenditures. (Irish objections to urban renewal in Chicago may have been due, not to a fear of higher taxes, but to fear of neighborhood “invasion” by Negroes displaced from land clearance projects.) The other is that in these precincts a much higher proportion of renters than of home-owners may have stayed away from the polls. In Cleveland (though not, interestingly, in Chicago) voter turnout is highly correlated with home ownership and almost all white renter precincts have at least a few homeowners in them. Conceivably—we think it unlikely—all those who voted in some “renter” precincts were actually owners.