¹ Variation in the amount of voter turn out among the American states has been diminishing in recent elections. But substantial differences still exist, especially between the South and non-South as has been the case for the last century. See Burnham, Walter Dean, “The Changing Shape of the American Political Universe,” The American Political Science Review, 59 (March, 1965), 7–28CrossRefGoogle Scholar; Flanigan, William H., Political Behavior of the American Electorate (Boston: Allyn and Bacon, 1968), p. 16Google Scholar; Milbrath, Lester W., “Political Participation in the States,” in Politics in the American States, ed. Jacob, Herbert and Vines, Kenneth (Boston: Little, Brown, 1965), pp. 25–60Google Scholar.

² Milbrath, , “Political Participation in the States,” pp. 45–50Google Scholar.

³ Kelley, Stanley Jr., Ayres, Richard and Bowen, William G., “Registration and Voting: Putting First Things First,” American Political Science Review, 61 (June, 1967), 359–377CrossRefGoogle Scholar.

⁴ Campbell, Angus, Converse, Philip E., Miller, Warren E. and Stokes, Donald E., The American Voter (New York: John Wiley, 1960), Chapter 11, esp. pp. 276–286Google Scholar.

⁵ Report of the President's Commission on Registration and Voting Participation (Washington, D.C., Government Printing Office, 1963).

⁶ Among others whose studies we do not specifically comment on are Campbell et al., American Voter, Chapter 11; Milbrath, “Political Participation in the States”; Dawson, Richard E., “Social Development, Party Composition and Policy,” in The American Party Systems: Stages of Political Development, ed. Chambers, William Nisbet and Burnham, Walter Dean (New York: Oxford University Press, 1967), pp. 203–237Google Scholar.

⁷ Downs, Anthony, Economic Theory of Democracy (New York: Harper, 1957)Google Scholar. See Chapter 14 in particular.

⁸ Kelley, et al., “Registration and Voting …”, p. 360Google Scholar.

⁹ Key, V. O., Southern Politics in State and Nation (New York: Knopf, 1951), Chapter 29Google Scholar.

¹⁰ Burnham, , “Changing Shape of the American Political Universe,” and his “End of Party Politics,” Transaction, 7 (December, 1969), 12–22Google Scholar.

¹¹ Rusk, Gerald F., “The Effect of the Australian Ballot Reform on Voting: 1876–1908,” American Political Science Review, 64 (December, 1970), 1220–1238CrossRefGoogle Scholar, for a counter argument; and, for further elaboration on this counter argument, see Converse, Phillip E., “Change in the American Electorate” in The Human Meaning of Social Change, ed. Campbell, Angus and Converse, Phillip (New York: Russell Sage Foundation, 1972)Google Scholar.

¹² We will cite only a few: Campbell et al., The American Voter; Almond, Gabriel A. and Verba, Sidney, The Civic Culture: Political Attitudes and Democracy in Five Nations (Princeton, N.J.: Princeton University Press, 1963), Chapters 11 and 12CrossRefGoogle Scholar; Lazarsfeld, Paul F., Berelson, Bernard, and Gaudet, Hazel, The People's Choice (New York: Columbia University Press, 1964)Google Scholar; Lipset, Seymour, Political Man, Garden City, N.Y.: (Doubleday, Anchor Books Edition 1963), Chapter 3Google Scholar; Verba, Sidney and Nie, Norman, Participation in America: Political Democracy and Social Equality (New York: Harper, 1972)Google Scholar.

¹³ Lipset, , Political Man, p. 187Google Scholar.

¹⁴ Milbrath, “Participation in the American States.”

¹⁵ Kelley et al., “Registration and Voting,” and Matthews, Donald R. and Prothro, James W., “Political Factors and Negro Voter Registration in the South,” American Political Science Review, 57 (June, 1963), 355–367CrossRefGoogle Scholar.

¹⁶ For a discussion of ecological fallacies, see Robinson, W. S., “Ecological Correlation and the Behavior of Individuals,” American Sociological Review, 15 (June, 1950), 351–357CrossRefGoogle Scholar; also Alker, Hayward R. Jr., “A Typology of Ecological Fallacies,” in Quantitative Ecological Analysis in the Social Sciences, ed. Dogan, Mattei and Rokkan, Stein (Cambridge, Mass.: M.I.T. Press, 1969), pp. 69–86Google Scholar.

¹⁷ Most studies based on aggregate data analysis have a tendency to mix the two meanings of socioeconomic factors. For example, see Kelley et al., “Registration and Voting,” and Matthew and Prothro, “Political Factors and Negro Voter Registration.”

¹⁸ For a discussion of several types of variables, see Lazarsfeld, Paul F. and Menzel, Herbert, “On the Relation between Individual and Collective Properties,” in Complex Organizations, ed. Etzioni, Amitai (New York: Holt, Rinehart and Winston, 1964), pp. 422–440Google Scholar.

¹⁹ We are not saying that socioeconomic factors can only be used at the individual level. In studies of the policy output of states or the stability of the system as a whole, the level of economic development may be a significant variable apart from the relationships at the individual level.

²⁰ The ideal design would replicate this analysis for several elections. Longitudinal design would permit a more complete test of these components. Changes instituted by the Voting Rights Act of 1965, the 1970 amendments to it and the 26th Amendment along with the Republican effort to mobilize the South will be a pertinent subject of analysis and may provide some keys to the controversy over the relative importance of party competitiveness and legal structures for voter turnout and electoral systems in general. See Burnham, “Changing Shape of the American Political Universe” on the one side, and Converse, “Change in the American Electorate” and Rusk, “The Effect of the Australian Ballot” on the other side.

²¹ Key, V. O., Politics, Parties and Pressure Groups (New York: Crowell, 1956, Third Ed.), p. 305Google Scholar.

²² Such a hierarchical assumption is often made in analyzing the effects of correlated factors even if the full data set is available. See the section on the decomposition of two systemic factors in Table 5. One of the readers suggested the possibility of coding systemic factors on the survey data. The resultant analysis at the individual level would eliminate the problems of committing the ecological fallacy and allow the examination of interaction effect between the systemic (contextual) variables and other individual characteristics. Furthermore, it would eliminate the problems of weighting (see footnote 34). For our analyses, the liabilities of such a procedure seem to outweigh its advantages. First, the SRC sample does not contain every state; second, even for the states that are included, the sample for any particular state is too small to be a reliable representation of the state; third, the SRC sample design is not meant to be representative samples of each state, but rather of the U.S. as a whole; fourth, different sample sizes for different states, coupled with an overall small sample size, would make variance decomposition very unreliable; finally, the focus of this study is not the probability of voting of the individuals, but rather the variation in voter turnout among the states.

²³ For an excellent discussion of comparative studies, see Przeworski, Adam and Teune, Henry, The Logic of Comparative Social Inquiry (New York: Wiley, 1970)Google Scholar.

²⁴ Examples of such relationships are found in Campbell et al., American Voter, and Verba and Nie, Participation in American Political Life. Also, see Burnham, Walter Dean and Sprague, John, “Additive and Multiplicative Models of the Voting Universe: The Case of Pennsylvania: 1960–1968,” American Political Science Review, 64 (June, 1970), 471–490CrossRefGoogle Scholar. For an interesting discussion of comparability and hidden uniformities in the relationships among variables across the nations, see also Verba, Sidney, “Cross National Survey Research: The Problem of Credibility,” in Comparative Methodologies: Changes and Continuities, ed. Vallier, Ivan (Berkeley: University, of California Press, 1972)Google Scholar. More important, lack of additivity does not automatically invalidate the current procedure, however. The examination of residuals would indicate any gross deviations from the additive model, and such indirect information may be of some value.

²⁵ Gross categorization—professionals, white collar, skilled worker, farmer, etc.—can be made transferable between the two different data sets. Such categorizations of occupation, however, do not add much to the predictability of individual voting when education and income of the individual are already known. See footnote 26.

²⁶ The multiple correlation between voting and the four variables—education, income, age and race— is .350. The addition of religion, residency, occupation, and sex increases the multiple r to .369. It may also be noted that most survey reports do not employ truly multivariate analysis. In particular, many comments in Lipset's Political Man with respect to factors affecting voting are based on serial examination of each rather than on multivariate control (Chapter 6). For a multivariate analysis of the effect of urban residency, for example, see Nie, Norman H., Powell, G. Bingham, and Prewitt, Kenneth, “Social Structure and Political Participation: Developmental Relationships, Part I, II,” American Political Science Review, 63 (June and September, 1969), 361–378 and 808–832CrossRefGoogle Scholar.

²⁷ The national average in the last row is the turnout for the voting age population of the U.S. in 1960 and it is not the population who reported voting in the 1960 survey, which is about 74 per cent. It has been consistently higher than voting rates based on election statistics. The discrepancy runs between 5 and 10 per cent. See U.S. Bureau of the Census, Current Population Reports, Series P-20, No. 143, “Voter Participation in the National Election: November, 1964,” U.S. Government Printing Office, Washington, DC, 1965. Also Parry, Hugh J. and Crossley, Helen M., “Validity of Responses to Survey Questions,” Public Opinion Quarterly, 14 (Spring, 1960), 61–80CrossRefGoogle Scholar.

²⁸ Johnson, J., Econometric Methods (New York: McGraw Hill, 1972, 2nd Ed.), Chapter 6Google Scholar. For discussion of dummy regression, see Suits, Daniel B., “The Uses of Dummy Variables in Regression Equations,” Journal of the American Statistical Association, 52 (December, 1957), 548–551CrossRefGoogle Scholar; Melichar, Emmanuel, “Least Squares Analysis of Economic Survey Data,” Proceedings of the Business and Economic Statistics Section, American Statistical Association (1965), pp. 373–375Google Scholar. For approaches based on the quantification perspective, see Hayashi, Chikio, “On the Quantification of Qualitative Data from the Mathematical-Statistical Point of View,” Annals of the Institute of Statistical Mathematics (Japan) 2 (1950)CrossRefGoogle Scholar; and “On the Prediction of Phenomena from Qualitative Data and the Quantification of Qualitative Data from the Mathematical-Statistical Point of View, Ibid., Vol. 3 (1951–1952). For a discussion of the MCA technique, see Andrews, Frank, Morgan, James, and Sonquist, John, Multiple Classification Analysis: A Report on a Computer Program for Multiple Regression Using Categorical Predictors (Ann Arbor, Michigan: Survey Research Center, Institute for Social Research, University of Michigan, 1967).Google Scholar

²⁹ Mostly from the Statistical Abstract of the U.S., 1962, 83rd Ed., U.S. Bureau of the Census, Washington, D.C. We have used exactly corresponding categories in age, education and race—both in survey and census data—but there was a slight discrepancy between the categories of income in survey data and the categories of income in census data. Since, however, the discrepancy was slight, we proceeded as if there were no discrepancies.

³⁰ If the population distribution pattern found in the sample survey is exactly the same as that of the entire voting age population of the U.S., the expected voter turnout calculated in this manner for the nation should be 0.0. Because of the slight discrepancies between the distribution of the sample data and the population distribution, the actual calculation does not exactly turn out that way. Since, however, the discrepancy is minor (at most, to the magnitude of 0.06 per cent), we do not think it necessary to reestimate the coefficients using some iterative procedure. We simply adjusted the coefficients in such a way that the national mean based on the population became zero. The final coefficients used are given in Column C of Table 2.

³¹ Our estimate of individual components contain some of the effects of legal and political systems, since there is likely to be some interaction effect.

³² To cite only a few examples: Franklin, Billy J., “Urbanization and Party Competition: A Note on Shifting Conceptualization and a Report of Further Data,” Social Forces, 49 (June, 1971), 544–549CrossRefGoogle Scholar; Dawson, Richard E. and Robinson, James A., “Inter-Party Competition, Economic Variables, and Welfare Policies in the American State,” Journal of Politics, 25 (May, 1963), 265–289CrossRefGoogle Scholar; Golembiewski, Robert T., “A Taxonomic Approach to State Political Party Strength,” Western Political Quarterly, 11 (September 1958), 494–513CrossRefGoogle Scholar; Pfeiffer, David F., “The Measurement of Inter-Party Competition and Systemic Stability,” American Political Science Review, 61 (June, 1967), 457–467CrossRefGoogle Scholar; Ranney, Austin, “Parties in State Politics,” in Politics in the American States, ed. Jacob, Herbert and Vines, Kenneth (Boston, 1965)Google Scholar; Ranney, Austin and Kendall, Willmoore, “The American Party System,” American Political Science Review, 48 (June, 1954), 477–485CrossRefGoogle Scholar; Riley, Dennis D., “Party Competition and State Policy Making: The Need for a Re-examination,” The Western Political Quarterly, 24 (September, 1971), 510–513CrossRefGoogle Scholar; Schlesinger, Joseph A., “A Two-Dimensional Scheme for Classifying the States According to Degree of Inter-Party Competition,” American Political Science Review, 49 (December, 1955), 1120–1128CrossRefGoogle Scholar.

³³ Milbrath, “Participation in the American States”; for quantification techniques, see footnote 28.

³⁴ The readjustment amounted to adding .9 per cent to each value. It would have been neater statistically to weight each state by its voting age population so that the resulting mean turnout for the nation would remain the same whether one uses survey data or aggregate statistics. But it is neater conceptually to treat each state as an equal unit. For instance, to examine the impact of a regulation in effect, say, both in Texas and North Dakota, there is no reason to give more weight to the results of Texas than that of North Dakota.

³⁵ This difficulty stems from the fact that the variation of Y explained by, for example, two independent variables X₁, X₂, when the relationship is Y = b₁ X₁ + b₂X₂ is given by times (the variance of X₁ + times (the variance of + times (the covariance between X₁ and Y₂). The last term is often interpreted as the amount of variance explained jointly by X₁ and X₂. Note, however, that these components can be negative if one of the b's or the covariance between X₁ and X₂ is negative.

³⁶ The last strategy is plausible if one is not willing to assign causal priority to either factor. See footnote 35.

³⁷ For discussion of multicollinearity, see Blalock, Hubert H., “Correlated Independent Variables. The Problems of Multicollinearity,” Social Forces, 41 (December, 1963) 233–237CrossRefGoogle Scholar.

³⁸ For an example of this kind of deviant case analysis, see Johnson, Gerald W., “Research Note on Political Correlates of Voter Participation: A Deviant Case Analysis,” American Political Science Review, 65 (September, 1971), 768–776CrossRefGoogle Scholar.

³⁹ In examining the relationship between voting and legal factors, Campbell, et al., American Voter, p. 152Google Scholar, pointed out that it is the informal and extralegal barriers, not state legislation, that accounts for much of the variability in the turnout of the Southern Negro.

⁴⁰ Out of Milbrath's 12 items, we created one additional one: Existence or lack of registration system.

⁴¹ The statistical procedure used here is the same as the one used in estimating coefficients in Table 3. See footnote 28 for reference.

⁴² The correlation (multiple correlation with dummy variables) between our measure of electoral competitiveness and Golembiewski's typology—with categories such as one-party, weak minority party, two-party states, and others—is .778. See Golembiewski, “A Taxonomic Approach …”

⁴³ These eight elections are used on the ground that party alignment during this period has been relatively stable. For such rationale, see Key, V. O. Jr., “A Theory of Critical Elections,” Journal of Politics, 17 (February, 1955), 3–18CrossRefGoogle Scholar. However, the years in which a particular state had a substantial third-party vote were eliminated from consideration. Consequently, only seven elections were used for Alabama, Louisiana, and South Carolina, and only six for Mississippi.

⁴⁴ These figures are based on unweighted average. See footnote 34.