² Merriam, Charles E. and Gosnell, Harold G., Non-Voting: Causes and Methods of Control (Chicago: University of Chicago Press, 1924), p. 251Google Scholar.

³ Ibid., p. 232.

⁴ Campbell, Angus and his associates at the Survey Research Center of the University of Michigan discuss the relation of voting to certain of the legal arrangements governing suffrage in The American Voter (New York: John Wiley and Sons, 1960), 276–286Google Scholar, and Warren Miller has reported additional findings on the subject in a memorandum to the President's Commission on Registration and Voting Participation. Key, V. O. examined the effect of poll taxes on voting in his Southern Politics (New York: Alfred A. Knopf, 1950), 599–618Google Scholar, and Howard Freeman, Arnold Simmel, and Murray Gendell discuss the turnout of registered voters in McPhee, William N. and Glaser, William A., Public Opinion and Congressional Elections (Glencoe: The Free Press, 1962), 240–250Google Scholar. Matthews, Donald R. and Prothro, James W. report findings about rates of registration among Negroes in the South in two articles: “Social and Economic Factors and Negro Voter Registration in the South,” this Review, 57 (03, 1963), 24–44Google Scholar, and “Political Factors and Negro Voter Registration in the South,” ibid. (June 1963), 355–367.

Much earlier, Gosnell, presented some interesting information on the relation of registration procedures to turnout in elections in Getting Out the Vote (Chicago: University of Chicago Press, 1927)Google Scholar and Why Europe Votes (Chicago: University of Chicago Press, 1930), as did Harris, Joseph P. at about the same time in his Registration of Voters in the United States (Washington: The Brookings Institution, 1929), 106–108Google Scholar. There are of course comments on the relation of registration to voting in works not cited here, but very few additional sources of systematic information about that relationship.

⁵ For a list of the cities included in our sample, see p. 373 below. Originally we planned to study factors affecting rates of registration in all cities in the United States with populations above 100,000; we were unable to proceed as planned, however, because we were unable to secure information about registration or registration procedures in some of these cities. Most commonly we could not secure accurate data regarding the percentage of the population of voting age registered to vote or verify the accuracy of such data as we could get. This was the case for 16 cities: Mobile, Montgomery, Tucson, Savannah, Chicago, Evansville, Indianapolis, Shreveport, Jackson, Albuquerque, Knoxville, Amarillo, Beaumont, El Paso, Lubbock, and Wichita Palls. We could not get accurate information regarding times and physical arrangements for registration in eight additional cities: Phoenix, Anaheim, Long Beach, San Jose, Denver, Providence, Madison, and Milwaukee. A figure for the percentage of recent migrants from out of state was missing for Yonkers, New York. Washington, D. C. had no procedures for registering voters in 1960, since, at that time, it had no voters.

⁶ Downs, Anthony, An Economic Theory of Democracy (New York: Harper and Brothers, 1927), p. 260Google Scholar.

⁷ Merriam and Gosnell reported that a sub-stantial number of the non-voters they studied had given fear of loss of income through loss of time away from work as an important reason for not voting; see Non-Voting, 86–95.

⁸ Literacy tests tend to lower registration, of course, not only because they are a bother but also because they disqualify a certain number of voters.

⁹ The most important have been race, place of residence, education, age, income, marital status, sex, ethnic group affiliation, occupation, geographical mobility, intensity of partisan preferences, perceived closeness of the election, interest in the campaign, concern with the outcome of the election, sense of the efficacy of voting, sense of civic duty, attitudes with respect to issues of public policy, attitudes toward candidates, and level of political information.

¹⁰ Nor could we study the relationship of urban and rural residence to rates of registration, since we limited ourselves to the study of registration in large cities.

¹¹ See Anthony Downs, op. cit., 232–236, 260–273.

¹² Ibid., p. 235, and Lipset, Seymour Martin, Political Man (New York: Doubleday, 1960), p. 205Google Scholar.

¹³ Campbell, Angus, Converse, Philip, Miller, Warren and Stokes, Donald, The American Voter, p. 476Google Scholar.

¹⁴ See particularly Seymour Martin Lipset, op. cit., 202, 209–211; Lane, Robert, Political Life (Glencoe: The Free Press, 1959), 48–49Google Scholar; Campbell, Anguset al., The American Voter, 496–497Google Scholar; and Milbrath, Lester W., Political Participation (Chicago: Rand McNally and Company, 1965), p. 135Google Scholar.

¹⁵ See Seymour Martin Lipset, op. cit., 202–208

¹⁶ The F-ratio for the regression was 361.39.

¹⁷ The regression coefficient of the rate of registration in this regression was .96, its standard error was .05, and its “t” value was 19.01.

¹⁸ The results in all of our text tables are reported in terms of partial regression coefficients, standard errors of these coefficients, and t-values. Appendix III contains partial correlation coefficients and Beta-weights.

¹⁹ The R²'s reported in Table I are not corrected for degrees of freedom; the R²'s for Regression I.1 must therefore be larger than the R² for Regression 1.2, since the latter contains only six of the twelve independent variables included in the former. Actually, as the F-ratios indicate, Regression I.2 gets higher marks on the statistical significance scale than does Regression I.1, although it is important to remember that we already knew something about the behavior of the variables included in Regression I.2 before it was run. A “t” value significant at the 95 percent level in Regression I.1 was the criterion for the inclusion of variables in Regression I.2, and in such circumstances the meaningfulness of standard significance tests is open to some question.

²⁰ Precise definitions of these and other variables are given in Appendix I in order of which the variables are discussed below.

²¹ See Appendix II for the matrix of simple correlations.

²² Campbell, Angus, et al., The American Voter, 475–476Google Scholar.

²³ Ibid., p. 487.

²⁴ Because of the way this variable was measured (See Appendix I), low values imply a high level of competition in a state's party politics; hence, the correlation between the variable and rates of registration is negative.

²⁵ See especially Gosnell, Harold F., Why Europe Votes, 182–183, 199–203Google Scholar; Milbrath, Lester W., “Political Participation in the States,” p. 43Google Scholar in Jacob, Herbert and Vines, Kenneth N. (eds.), Politics in the American States (Boston: Little, Brown and Company, 1965)Google Scholar; and Tingsten, Herbert, Political Behaviour (London: P. S. King and Son, 1937), p. 216Google Scholar.

²⁶ Robert Lane, op. cit., p. 310.

²⁷ Lane's data for the election of 1950 are as reported in Nicholas, H. G., The British General Election of 1960 (London: Macmillan, 1951), p. 318Google Scholar; Gosnell's data were for the election of 1924.

²⁸ Robert Lane, op. cit., p. 309.

²⁹ It does have meaning for the value of the vote in county elections, of course; and if county elections were held in years when major state and national offices were not at stake, one would expect to find a relationship between the competitiveness of a county's politics and the turnout of voters.

³⁰ Campbell, Angus, et al., The American Voter, p. 100Google Scholarf.n.

³¹ Campbell, Angus, Gurin, Gerald, and Miller, Warren E., The Voter Decides (Evanston: Row, Peterson and Company, 1954), p. 193Google Scholar.

A re-arrangement of data presented in The American Voter, Table 17.4, p. 479, yields the following results:

³² The assignment of values in these indices was arbitrary, as has been indicated in the notes to Table 1, but a few words about the kind of arbitrariness involved are in order here. This can best be done, perhaps, by indicating how we arrive at the scale of values for one of these indices, that for the variable we have called “registration system.” The four positions on that index were assigned values of 10, 25, 34, and 100. This set of values was chosen because it was our guess that a system of permanent registration (valued at 100) was ten times less demanding on voters than a system which combines a cumulative poll tax with annual registration (valued at 10), four times less demanding than a system of annual registration coupled with a non-cumulative poll tax (valued at 25), and three times less demanding than a system of annual personal registration (valued at 34).

In a moment of less boldness, we eschewed estimates of how much more restrictive one system was than another and constructed a simple ordinal scale which involved only guesses as to which systems had a greater, and which had a lesser, tendency to restrict registration. The substitution of the ordinal indices had little effect on our results. The cardinal indices showed somewhat stronger relationships to registration than those with ordinal scales, but if one of our cardinal indices was significantly related to registration, so was its ordinal equivalent, and if one of our cardinal indices was not significantly related to registration, neither was its ordinal equivalent.

³³ Cf. Milbrath, Lester, “Political Participation in the States,” p. 48Google Scholar in Jacob and Vines, op. cit., and Matthews, Donald R. and Prothro, James W., “Political Factors and Negro Voter Registration in the South,” this Review, 57 (06, 1963), at p. 358Google Scholar.

³⁴ Riker, William H., Democracy in the United States (New York: Macmillan, 1953), p. 66Google Scholar.

³⁵ To the best of our knowledge only Warren Miller has previously presented systematic evidence tending to show non-voting to be related to an early closing date for registration. In his memorandum to the President's Commission on Registration and Voting Participation, Miller reported that “One out of every four or five citizens lives in a state where registration closes in September, before an election campaign is well under way…. Non-voting by two or three percent of the population is associated with the September closing dates, both North and South.”

³⁶ Gosnell, Harold F., Getting Old the Vote, p. 104Google Scholar.

³⁷ See Joseph P. Harris, op. cit., 106–108, and Robert Lane, op. cit., p. 315. Warren Miller reported to the President's Commission on Registration and Voting Participation that “The absence of any prevoting registration requirement and the provision for permanent registration are clearly associated with higher turnout at the polls.” Key, V. O. has shown the poll tax to depress turnout in Southern Politics, 599–618Google Scholar.

³⁸ This seems to be the sense of Lipset's statement that he does not wish to consider the effects of “legal and technical restrictions like residence requirements, poll taxes and property qualifications, literacy tests (often used as a cover for racial discrimination) and burdensome registration requirements” on the ground that his interest is in voluntary (his emphasis) non-voting. (See Seymour Martin Lipset, op. cit., p. 181 f.n.) But for the greatest numbers of those who fail to register, non-registration is just as voluntary or non-voluntary as is failure to vote.

³⁹ See Campbell, Angus, et al., The Voter Decides, p. 37Google Scholar; Charles E. Merriam and Harold F. Gosnell, op. cit., 79–86; and Bradley, Phillips and Cope, Alfred H., “A Community Registration Survey,” this Review, 45 (09, 1951), at p. 777Google Scholar.

⁴⁰ It is possible, of course, that the index we constructed did not reflect with sufficient exactness the actual differences in purging practices that exist from city to city. The procedures for purging were difficult to document.

⁴¹ These “t” values were obtained by running five versions of Regression III.1, dropping a different regional dummy each time, and using the t's for the other dummies to find whether or not they are significantly different from the omitted dummy.

⁴² Actually, for cities in counties with large percentages of Catholics in their populations. Our source was National Council of the Churches of Christ in the United State of America, Churches and Church Membership in the United States, New York, 1956–1958Google Scholar.

⁴³ Our source for percentages of aliens in the population of our cities was U. S. Department of Commerce, Bureau of the Census, U. S. Census of Population I960: Characteristics of the Population, Table 55.

⁴⁴ To avoid forcing a linear pattern on this variable, we used a set of dummy variables; however, none had a “t” value as high as 1.00.

⁴⁵ Cf. Milbrath, Lester W., “Political Participation in the States,” p. 47Google Scholar in Jacob and Vines, op. cit.

⁴⁶ This variable had the expected negative sign, but the “t” value of its regression coefficient was only 1.19.

⁴⁷ Cf. Matthews, Donald R. and Prothro, James W., “Social and Economic Factors and Negro Voter Registration in the South,” this Review, 57 (03, 1963), 28–32Google Scholar.

⁴⁸ Our source for this variable (actually the percentage of those over 25 years of age who are college graduates) was U. S. Department of Commerce, Bureau of the Census, County and City Data Book 1963, Table 6.

⁴⁹ Stability of population was measured by the percentage of persons in a city's population over five years of age that were living in the same house in 1960 as they had in 1955. The source for these figures was U. S. Department of Commerce, Bureau of the Census, County and City Data Book 1962, Table 6.

⁵⁰ Gosnell, Harold F., Why Europe Votes, p. 185Google Scholar. That differences in turnout for elections in Europe and America may be largely a function of differences in systems of registration has also been suggested by Seymour Martin Lipset, op. cit., p. 181 f.n., and Converse, Philip E. and Dupeux, George, “Politicization of the Electorate in France and the United States,” Public Opinion Quarterly (Spring, 1962), 8–9Google Scholarf.n.

⁵¹ These averages were for six post-war Canadian elections (1945, 1949, 1953, 1957, 1958, 1963), five post-war elections for the French National Assembly (1946, 1951, 1956, 1958, 1962), and six post-war elections in Great Britain (1945, 1950, 1951, 1955, 1959, 1964). The average rate of turnout of registered voters in the cities of our full sample was 81.6 percent, and in Northern cities only it was 84.6 percent.

⁵² Key, V. O., Politics, Parties, and Pressure Groups (New York: Thomas Y. Crowell, 1964)Google Scholar, fifth edition, p. 578.

⁵³ Joseph P. Harris, op. cit., p. 89.

⁵⁴ Lazarsfeld, Paul F., Berelson, Bernard, and Gaudet, Helen, The People's Choice (New York: Columbia University Press, 1948), p. 47Google Scholar. This conclusion is not well supported even by the data adduced as evidence for it. Some 27 percent of the voters who said they had no interest at all in the campaign nonetheless voted. See Ibid., p. 46.

⁵⁵ Berelson, Bernard, Lazarsfeld, Paul P., and McPhee, William N., Voting (Chicago: University of Chicago Press, 1954), p. 32Google Scholar.

⁵⁶ One of the authors of the present study found, for instance, that there was a very high positive correlation between an index of the convenience of registering in Chicago wards and differences among wards in the Democratic percentage of the two-party vote, a f act which suggests that Mayor Richard J. Daley's organization is alert to the possibilities of manipulating the composition of Chicago's registered voters: See Ayres, Richard E., Registration 1960: Key to Democratic Victory? (Princeton University senior thesis, 1964), 34–35Google Scholar.

⁵⁷ New York Times, September 6, 1964.