¹ For example: Bendix, Reinhard, “Social Stratification and Political Power,” in Class, Status, and Power, ed. Bendix, Reinhard and Lipset, Seymour (Glencoe, Ill.: Free Press, 1953), pp. 596–609Google Scholar; Bracher, Karl D., Die Auflösung der Weimarer Republik (Villingen: Schwarzwald, 1960)Google Scholar; Burnham, Walter Dean, “Political Immunization and Political Confessionalism,” delivered at the 1970 meeting of the International Political Science Association, Munich, 1970Google Scholar; Converse, Philip, “The Nature of Belief Systems in Mass Publics,” in Ideology and Discontent, ed. Apter, David (Glencoe, Ill.: Free Press, 1964), pp. 252–254Google Scholar; Heberle, Rudolf, From Democracy to Nazism (Baton Rouge, La.: Louisiana State University Press, 1945)Google Scholar; Lipset, Seymour Martin, Political Man (Garden City, New York: Doubleday, 1960), 140–154Google Scholar; Kaltefleiter, Werner, Wirtschaft und Politik in Deutschland (Köln: Westdeutscher Verlag, 1968), pp. 13–94Google Scholar; O'Lessker, Karl, “Who Voted for Hitler? A New Look at the Class Basis of Naziism,” American Journal of Sociology, 74 (07, 1968), 63–69CrossRefGoogle Scholar; Waldman, Loren K., Models of Mass Movements: The Case of the Nazis, Ph.D. thesis, University of Chicago, 1972Google Scholar.

² Dennis, Jack and McCrone, Donald J., “Preadult Development of Political Party Identification in Western Democracies,” Compartaive Political Studies, 3 (07, 1970), 247Google Scholar.

³ Converse, Philip E., “Of Time and Partisan Stability,” Comparative Political Studies, 2 (07, 1969), 141–142CrossRefGoogle Scholar.

⁴ Jaros, Dean and Mason, Gene L., “Party Choice and Support for Demagogues: An Experimental Examination,” American Political Science Review, 58 (March, 1969), 100–110CrossRefGoogle Scholar. The introduction to their paper provides an excellent review of the party identification theory of electoral stability.

⁵ Especially Campbell, Angus, Converse, Philip E., Miller, Warren E., and Stokes, Donald E., The American Voter (New York: John Wiley and Sons, 1960)Google Scholar.

⁶ See, for example, Campbell, Angus and Valen, Henry, “Party Identification in Norway and the United States,” in Elections and the Political Order, Campbell, Angus, Converse, Philip E., Miller, Warren E., and Stokes, Donald E. (New York: John Wiley and Sons, 1966), pp. 245–268Google Scholar; Butler, David and Stokes, Donald, Political Change in Britain (New York: St. Martin's Press, 1969), pp. 55–61Google Scholar; and the data on various countries (including data from the Almond and Verba study), presented in Dennis and McCrone, cited above.

⁷ Converse and Dupeux found that in France 45 per cent of respondents, as compared with 75 per cent of American respondents, expressed a sense of identification with some party. Converse, Philip E. and Dupeux, Georges, “Politicization of the Electorate in France and the United States,” in Elections and the Political Order, Campbell, et al., p. 277Google Scholar; working with the Almond and Verba data, Converse found that the average strength of party identification in Italy was well under half that in Britain and the United States. Converse, , “Of Time and Partisan Stability,” p. 151Google Scholar.

⁸ Converse, “Of Time and Partisan Stability.”

⁹ This assumption is particularly important to Converse's theory, and for what follows in this paper. There is evidence for the assumption from the United States (American Voter, p. 163) and Britain (Butler and Stokes, pp. 55–57). In both cases, the strength of an individual's identification increases as he grows older, but this is a function solely of how long he has identified with the party, not of his age per se. If the length of time that individuals have identified with the same party is held constant, the relationship between age and the strength of identification disappears. In other words, we might expect a voter aged fifty, who had initially acquired an identification with a party ten years before, to show roughly the same strength of identification as a thirty-year-old voter who had initially acquired his party identification ten years before. This, despite the fact that we would generally expect fifty-year-olds to identify more strongly with a party than thirty-year-olds.

¹⁰ Converse, , “Of Time and Partisan Stability,” pp. 161–163Google Scholar.

¹¹ The results which I discuss are reported in Kaase, Max, “Determinanten des Wahlverhaltens bei der Bundestagswahl 1969,” Politische Vierteljahresschrift” 11 (03, 1970), 46–110Google Scholar.

¹² The questions were: 1967, “Ganz allgemein gesprochen—betrachten Sie sich als CDU-Anhänger, als SPD-Anhänger, als FDP-Anhänger, als NPD-Anhänger, als Anhänger einer anderen Partei oder fühlen Sie sich keiner Partei besonders verbunden? (Do you generally think of yourself as a supporter of the CDU, the SPD, the FDP, the NPD, or some other party, or do you not feel particularly tied to any party?); 1969, “Ganz allgemein gesprochen, betrachten Sie sich als Anhänger einer bestimmten politischen Partei, oder fühlen Sie sich keiner Partei besonders verbunden?” (Do you generally think of yourself as a supporter of a particular political party, or do you not feel particularly tied to any party?).

¹³ Kaase, p. 82. The Survey Research Center's American panel from 1956 to 1960 provides two comparisons with the German figure. Of those who identified with a party in 1958, 82 per cent reported the same identification they had given in 1956; of those who identified with a party in 1960, 88 per cent reported the same identification they had given in 1958. The difference between the American and German results is a conservative one, for two reasons: (1) twenty-four months intervened between questioning in the American study, compared with twenty months in the German study; (2) the American identifiers were isolated by means of an “easier” question than the Germans (the American question cued the names of the parties), and thus should have been a less firmly partisan group than the German identifiers. Source: SRC American Panel Study: 1956, 1958, 1960 (SRC S440, ICPR 7252).

¹⁴ Butler and Stokes, pp. 40–43.

¹⁵ These figures are calculated from Butler and Stokes, Tables 2.6 and 2.7, pp. 41–42.

¹⁶ Campbell, and Valen, , “Party Identification” in Elections and the Political Order, Campbell, et al., pp. 245–268Google Scholar. The figures used here are calculated from Table 13–5.

¹⁷ Converse, , “Of Time and Partisan Stability,” p. 166Google Scholar.

¹⁸ See particularly Heberle, Lipset, and O'Lessker, cited above.

¹⁹ The relationship between nonvoting and party identification in the United States is demonstrated in The American Voter, pp. 96–101. For the lower party identification in the rural portion of the American electorate, see The American Voter, p. 409; for France, see Tarrow, Si̧dney, “The Urban-Rural Cleavage in Political Involvement: The Case of France,” American Political Science Review, 65 (06, 1971), 341–357CrossRefGoogle Scholar.

²⁰ See footnote 9 above.

²¹ The finding of Butler and Stokes that the duration of present party tie accounts for approximately all of the relationship between age and party identification in Britain (Butler and Stokes, pp. 56–58) is puzzling, and seems to weaken my interpretation of British voting. If it is true that British party identification is to a greater extent than American party identification an expression of immediate partisan choice, one would expect either that the age-party identification relationship would be weaker than in the United States (which is not the case) or that it would be less fully accounted for by the duration of the party tie. The Butler-Stokes table is not strictly comparable to the American Voter table (p. 163 of The American Voters), and it is possible that if the two tables were set up in a comparable way, the second expectation would be confirmed. But it is a puzzle.

²² “Ecological” regression is simply an interpretation placed on regression equations which use data on individuals grouped into aggregations, such as “per cent Nazi” or “per cent Catholic.” There is an increasingly large body of literature on the technique. The best presentation is probably Stokes's, Donald E. “Cross-Level Inference as a Game Against Nature,” in Mathematical Applications in Political Science, ed. Bernd, Joseph L. (Charlottesville, Va.: The University of Virginia Press, 1969), pp. 62–83Google Scholar.

²³ There are ways in which the aggregate vote in the first of two elections might serve as a cause of the aggregate vote in the second. The act of voting for a party in the first election (and the memory of having voted for it, when it came time to vote again) might lead people to repeat that vote. Repeating votes in this way might, for example, serve to maintain attitudinal balance for the voters. Again, the ardor of party workers might be affected by the results of the earlier election; this would affect the results in the second election. I have chosen to ignore these possibilities, however. The danger in doing so seems less serious than the distortion which would result from carrying the analysis only forward in time.

²⁴ Converse, Philip E., “Information Flow and the Stability of Partisan Attitudes,” in Elections and the Political Order, Campbell, et al., pp. 136–157Google Scholar. See also Shanks, J. Merrill, “The Quality of Electoral Change: 1952–1964,” paper delivered at the 1969 Annual Meeting of the American Political Science Association, New York City, 09 2–6Google Scholar.

²⁵ A third source of electoral stability which appears in the graphs does not seem to be of particular theoretical interest. For all five parties for which Catholics and Protestants are compared, Protestants are the more stable supporters. This is probably because a Catholic voting for any of these parties would have experienced a strong cross-pressure which none of his Protestant counterparts experienced—the influence of the Center Party, which was endorsed by the Catholic Church. This interpretation is strengthened by the fact that for all five parties, Catholic women, who might be expected to have felt this pull more strongly than men, were a trifle less stable relative to Catholic men that Protestant women were relative to Protestant men. (The differences are admittedly very slight.) This interpretation also is consistent with the fact that Catholic women who voted for the Center Party were so much more stable than the men. The difference between men's and women's stability is greater for the Center Party than for any other.

²⁶ Initially, the failure to generate such data was due to the economic burden of separating the sexes in the midst of the Depression. Later, it was aggravated by the Nazi scorn for elections, which left unpublished some separate tabulations.

²⁷ The figures for these cities are taken from Tingsten's, HerbertPolitical Behavior (London: P. S. King and Son, 1937), pp. 28–29 and 51–58Google Scholar. The partial data he reports on five other cities or towns indicate the same patterns as those shown by these four.

²⁸ In the case of Regensburg, it is possible to carry the comparison of men and women voters beyond the Nazi take-over. Interestingly, even given the changed nature of elections, the same process continued. Women continued to narrow the gap between themselves and the men, and finally overtook the men in 1934:

Source: Bavaria. Statistisches Landesamt. Zeitschrift, vol. 65, p. 585; vol. 66, p. 250.

²⁹ The predominantly male vote for the Nazis in the early elections has sometimes been used as evidence of the special radical nature of the Nazi voters. As Seymour Lipset puts it:

Lipset, , Political Man, p. 143Google Scholar. But the pattern as we see it here suggests that women did not really respond differently to the Nazis than did men. They lagged behind the men, but probably ended up ahead of them.

³⁰ See Duverger, Maurice, Political Parties (New York: Wiley, 1954), p. 84Google Scholar.

³¹ Allen, William S., The Nazi Seizure of Power: The Experience of a Single Town, 1930–1935 (Chicago: Quadrangle Books, 1965), p. 296Google Scholar.

³² Lipset, pp. 149–152.

³³ O'Lessker, , “Who Voted for Hitler?” pp. 66–67Google Scholar.

³⁴ Schnaiberg, Allan, “A Critique of Karl O'Lessker's ‘Who Voted for Hitler?’“, American Journal of Sociology, 74 (05, 1969), 732–735CrossRefGoogle Scholar.

³⁵ A similar analysis for nine cities between 1928 and 1930 bears out Lipset's and Schnaiberg's finding of no relationship for that period. The coefficient is −1.0 for all nine cities.

³⁶ In West Germany today, the Christian Democratic vote, which consists disproportionately of marginal voters, has acted the way I have suggested the Nazi vote should have acted. One reason the Social Democrats were able to control as many state governments as they did during the period of Adenauer's popularity was that the Christian Democratic vote dropped off sharply in state elections. Compare Preece, R. J. C., “Land” Elections in the German Federal Republic (London: Longmans, Green, 1968), p. 33Google Scholar.

³⁷ Compare Tingsten, Herbert, Political Behavior, pp. 27–29Google Scholar. Also Bremme, Gabrielle, Die Politische Rolle der Frau in Deutschland (Göttingen: Vandenhoeck and Ruprecht, 1956), pp. 35–50Google Scholar; Hartwig, , “Wie die Frauen im Deutschen Reich von ihrem politischen Wahlrecht Gebrauch machen,” Allgemeines Statistisches Archiv, 17 (1928), 497–513Google Scholar; Hartwig, , “Das Frauenwahlrecht in der Statistik,” Allgemeines Stalistiches Archiv, 21 (1931), 167–183Google Scholar. In general, women's participation lagged behind men's by several percentage points.

³⁸ Campbell, et al., The American Voter, pp. 96–101Google Scholar.

³⁹ Campbell, et al., The American Voter, pp. 110–115, 154–155Google Scholar. See also Campbell, Angus, “Surge and Decline,” in Elections and the Political Order, Campbell, et al., pp. 40–63Google Scholar. The American Voter findings deal with marginal voters in general, but they would seem to apply as well to new voters.

⁴⁰ This is illustrated below in Figure 5.

⁴¹ Statistik des Deutschen Reichs, Vol. 315, No. 6, pp. 35–40Google Scholar. The averages I give are calculated from ages broken down to eleven categories.

⁴² Tingsten, , Political Behavior, pp. 29, 56Google Scholar.

⁴³ The results of this study of Weimar do not disprove the party identification theory of electoral stability. Party identification appears to have been absent in the Weimar electorate, and Weimar voting was rather unstable. But the results do suggest that the theory's usefulness is limited. The theory presumably helps to explain why American voting (at least in Congressional elections) is more stable than Weimar elections were. But it cannot help us understand the more important questions of the development of the German electorate over time, or of variations within the German electorate, since party identification appears to have never been a factor in German voting.

₄₄ This measure would equal 100 per cent, for instance, if each party had either gained or lost all of its votes from the first election to the second.

⁴⁵ Bo Särlvick describes a similar situation in Sweden. See his “Political Stability and Change in the Swedish Electorate,” Scandinavian Political Studies (1966), pp. 188–222, esp. pp. 193–197Google Scholar.

⁴⁶ Some further support for this interpretation of Weimar voting is suggested in preliminary findings of a project in which I am seeking to measure “party distances” in voting during the German Empire, from 1871 to 1912. Two major gulfs in party distance appear to have determined voting during the Empire—that between the Center Party and all other parties, and that between the Social Democrats and all other parties. Distinctions among the remaining parties do form a third dimension, apparently along lines of the older urban-rural conflict over the constitution. But the distances separating these parties are relatively small. It may be that then, and later in the Weimar Republic, although the leaders of the various parties saw important differences among themselves, voters at the mass level saw essentially only Marxists, Catholics, and the parties of the Protestant bourgeoisie.

This interpretation fits quite well with the usual interpretation—presented in classic form by Lipset—of the Nazi Party as having been primarily a party of the Protestant bourgeoisie (Lipset, , Political Man, pp. 134–152)Google Scholar

⁴⁷ Butler, and Stokes, , Political Change in Britain p. 43Google Scholar.

⁴⁸ Downs, Anthony, An Economic Theory of Democracy (New York: Harper Row, 1957)Google Scholar, esp. chap. 12. Actually, Downs takes a dim view of the rationality of choosing a political party as one's information-crutch, but the notion stems from his general discussion of the need to delegate the jobs of information-seeking and evaluation.

⁴⁹ There is some evidence consistent with this interpretation in the Campbell/Valen and Butler/Stokes studies cited earlier. Campbell and Valen, for instance, were not certain whether party identification had an effect on Norwegians' voting independent of their social group identifications: “The Norwegian labor union member who is a member of the Labor Party may display a strong party attachment, but one wonders if this does not merely express in different form his basic identification with the working class” (Campbell and Valen, p. 268). And Butler and Stokes note the great impact that social class has in British voters' images of the parties, and in structuring switches in voting (Butler and Stokes, p. 89; pp. 298–303).

⁵⁰ The British first-vote figures are from Butler and Stokes, p. 54; the American figures are from Campbell, et al., The American Voter, p. 155Google Scholar. In both cases, the “first votes” are based on respondents' later recollections of how they had voted in the first election in which they participated. Note that the British figures are the per cent of the vote for the two main parties.

⁵¹ This term was first used by Key (Key, V. O. Jr.), “A Theory of Critical Elections,” Journal of Politics, 17 (02, 1955), pp. 1–18CrossRefGoogle Scholar. See also Campbell, et al., The American Voter, pp. 531–538Google Scholar.

⁵² Lipset, Seymour M. and Rokkan, Stein, “Cleavage Structures, Party Systems, and Voter Alignment: An Introduction,” in Party Systems and Voter Alignments: Cross-National Perspectives, ed. Lipset, S. M. and Rokkan, S. (New York: Free Press, 1967), pp. 50–56Google Scholar. Lipset and Rokkan themselves stress the development of broad party organizations, rather than individual party identification, in the process of “freezing.”

⁵³ For example, the papers in Lipset and Rokkan, Party Systems and Voter Alignments. Also, see Converse, Philip and Valen, Henry, “Dimensions of Cleavage and Perceived Party Distances in Norwegian Voting,” Scandinavian Political Studies, 6 (1971), 107–152CrossRefGoogle Scholar.

⁵⁴ MacRae, Duncan, Parliament, Parties, and Society in France, 1946–1958 (New York: St. Martin's, 1967), p. 243Google Scholar.

⁵⁵ Shively, W. Phillips, “‘Ecological’ Inference: The Use of Aggregate Data to Study Individuals,” American Political Science Review, 58 (12, 1969), pp. 1183–1196; esp., pp. 1191–1192Google Scholar.