¹ Campbell, Angus, Converse, Philip E., Miller, Warren E., and Stokes, Donald E., The American Voter (New York: Wiley, 1960), pp. 161–165Google Scholar.

² Blacks have had markedly different political experiences from whites, and the assumptions I make about the development of party loyalties among whites do not apply to blacks. For a discussion of partisan change among blacks, see Abramson, Paul R., Generational Change in American Politics (Lexington, Mass.: D. C. Heath, 1975), pp. 65–68Google Scholar.

³ The measure of party identification used by the SRC classifies respondents as strong Democrats, weak Democrats, Independents who lean toward the Democrats, Independents with no party leanings, Independents who lean toward the Republicans, weak Republicans, and strong Republicans. Tables 1 and 2 follow the procedures used by Campbell et al. by “folding” this scale to combine strong partisans, and to combine strong and weak partisans. The mean score measure of strength of partisanship follows the same logic, but assumes, for purposes of illustration, that the party identification scale can be treated as interval. My measure is similar to the index of partisan strength used by Converse, Philip E. in “Of Time and Partisan Stability,” Comparative Political Studies, 2 (July, 1969), 139–171, at 149–151, although his index scores ranged from zero through twoCrossRefGoogle Scholar.

The percentages reported in Tables 1 through 3 are based upon all respondents, except those for whom party identification was not ascertained. Persons who supported minor parties and those classified as apolitical are included in the numbers upon which percentages are based. As these two groups never made up more than three per cent of the total, similar results would be obtained if they were excluded from the analysis. In computing the mean strength of partisan identification for this study, apoliticals were scored as zero, but persons who supported minor parties were excluded from the analysis. The totals upon which the mean scores were calculated are therefore slightly smaller than those upon which the percentages were based.

⁴ Duncan, Otis Dudley, “Social Stratification and Mobility: Problems in the Measurement of Trend,” in Indicators of Social Change: Concepts and Measurements, ed. Sheldon, Eleanor Bernert and Moore, Wilbert E. (New York: Russell Sage, 1968), pp. 675–719, at p. 679Google Scholar.

⁵ Hyman, Herbert H., Secondary Analysis of Sample Surveys: Principles, Procedures, and Potentialities (New York: Wiley, 1972), p. 243Google Scholar.

⁶ Throughout this analysis, respondents were divided into eight-year age cohorts, although I present data about younger cohorts before the full eight-year age group had entered the electorate. Ideally, a cohort analysis should divide respondents into the same age range as the number of years between surveys, allowing comparison of persons of the same age at different points in time by reading diagonally down and across the table. Practical considerations, however, dictate that the Ns for each cohort be substantial, and to divide respondents into two-year age groups would lead to small Ns as well as unreadable tables.

The 1952, 1968, 1970, 1972, and 1974 surveys recorded years of birth. For other surveys year of birth has been calculated from data about the respondent's age. The 1954 SRC survey employed arbitrary age categories, and could not be used in my time-series cohort analysis.

⁷ Campbell, et al. , The American Voter, p. 161Google Scholar.

⁸ In this analysis, the Michigan authors did not differentiate between Independents who leaned toward a party and those who did not.

⁹ Campbell et al., p. 161.

¹⁰ Ibid., p. 161.

¹¹ Ibid., p. 161.

¹² Ibid., p. 163. For a similar analysis examining the relationship of age and strength of partisanship in Britain, see Butler, David and Stokes, Donald, Political Change in Britain: Forces Shaping Electoral Choice (New York: St. Martin's Press, 1969), pp. 55–58Google Scholar.

¹³ The basic procedure was comparison of persons 44 years or younger with those 45 years or older. The authors, however, found only 27 persons 45 years old and older who had held their current identifications for less than four years, and only 40 who had held their current identification for between four and seven years.

¹⁴ See Dreyer, Edward C., “Change and Stability in Party Identifications,” Journal of Politics, 35 (August, 1973), 712–722CrossRefGoogle Scholar.

¹⁵ On the other hand, Independence might be a form of loyalty in and of itself. Pure Independents might become more committed to Independence as they gained political experience.

¹⁶ See Campbell, et al. , The American Voter, pp. 125–128Google Scholar, and Butler, and Stokes, , Political Change in Britain, pp. 44–64Google Scholar. Also see McPhee, William N. and Ferguson, Jack, “Political Immunization,” in Public Opinion and Congressional Elections, ed. McPhee, William N. and Glaser, William (New York: Free Press, 1962), pp. 155–179Google Scholar.

¹⁷ A three-wave panel study of fourth, sixth, and eighth graders in the San Francisco Bay area conducted between December 1968 and May 1969 showed children to have relatively unstable partisan identifications compared to those of adults. (See Vaillancourt, Pauline Marie and Niemi, Richard G., “Children's Party Choices,” in The Politics of Future Citizens, ed. Niemi, Richard G. [San Francisco: Jossey-Bass, 1974], pp. 126–148.)Google Scholar For a further discussion of the strength of party identification among pre-adults, see Sears, David O., “Political Socialization,” in Handbook of Political Science, Vol. 2: Micropolitical Theory, ed. Greenstein, Fred I. and Polsby, Nelson W. (Reading, Mass.: Addison-Wesley, 1975), pp. 93–153, at pp. 121–123Google Scholar.

¹⁸ For a discussion of party leadership during the postwar years, see Broder, David S., The Party's Over: The Failure of Politics in America (New York: Harper and Row, 1972)Google Scholar. For another provocative discussion, see Burnham, Walter Dean, “The End of American Party Politics,” Transaction, 1 (December, 1969), 12–22Google Scholar.

¹⁹ Gerald Pomper, who also discusses the weak party identifications of young Americans, argues that they result from the political events of the last decade. See Voters' Choice: Varieties of American Electoral Behavior (New York: Dodd, Mead, 1975), pp. 23–24Google Scholar.

²⁰ As Karen Oppenheim Mason and her colleagues clearly show, any cohort analysis faces a basic methodological problem: each of the three independent variables—age, period of survey, and years of birth—is a perfect function of the other two. Mason and her colleagues propose a model, employing Multiple Classification Analysis, that allows the analyst to estimate the effects of age, period, and cohort. (See Mason, Karen Oppenheim, Mason, William M., Winsborough, H. H., and Poole, W. Kenneth, “Some Methodological Issues in Cohort Analysis of Archival Data,” American Sociological Review, 38 [April, 1973], 242–258)CrossRefGoogle Scholar. David Knoke and Michael Hout, using the Michigan SRC data, have applied the Mason et al. technique to the study of party identification. (See “Social and Demographic Factors in American Political Party Affiliations, 1952–1972,” American Sociological Review, 39 [October, 1974], 700–713CrossRefGoogle Scholar.) However, both the applicability of the Mason et al. model to attitudinal data and the use of this model by Knoke and Hout have been questioned in a recent paper by Norval D. Glenn, who argues that the model necessitates unwarrented assumptions. (See “Some Cautions Concerning Statistical Attempts to Separate Age, Period, and Cohort Effects,” paper presented at the 71st Annual Meeting of the American Political Science Association, San Francisco, September, 1975)Google Scholar.

While the methodological problems of cohort analysis may be unresolved, the theoretical expectations of the life-cycle explanation are so clear that the failure of partisanship to rise among cohorts as they age provides, at the very least, a basis for seriously questioning The American Voter formulation.

²¹ All these surveys are based upon national probability samples, and, as the Ns have been reported, the reader can evaluate the probability of differences between age groups in Tables 1 through 3 resulting from sampling error. (As weighted Ns were used for 1958, 1960, and 1974, tests of significance are not valid for those years.) However, the best assurance that differences between cohorts are meaningful is the replication of age-group differences in numerous surveys based upon separately conducted samples. In this respect, too, the 1958, 1960, and 1974 data are less satisfactory. More than half the respondents in 1958 and 1960 originally were sampled as part of a panel study begun in 1956, and more than half of the respondents in 1974 originally were sampled as part of a panel study begun in 1972. Thus, the high consistency in levels of identification among cohorts in 1956, 1958, and 1960, and between cohorts in 1972 and 1974 partly results from sampling procedures.

One can also gain greater confidence that age group differences are meaningful by replicating this analysis with other data sets. To this end, I report upon a study by Norval D. Glenn who used Gallup data to conduct a similar analysis.

²² Although levels of partisanship are reported for the 1932-1939 cohort in 1956 and 1958, the full eight-year cohort was not sampled until 1960. The full eight-year cohort born between 1940 and 1947 was not sampled until 1968. In 1970, the SRC began to sample eighteen-, nineteen-, and twenty-year olds. Nevertheless, the full eight-year cohort born between 1948 and 1955 was not sampled until 1974. Respondents born in 1956 were not included in this youngest cohort, but since only ten respondents born during that year were included in the electorate sample, their levels of identification have not been reported.

²³ See Glenn, Norval D. and Hefner, Ted, “Further Evidenceon Aging and Party Identification,” Public Opinion Quarterly, 36 (Spring, 1972), 31–47CrossRefGoogle Scholar; and Glenn, , “Sources of Shift to Political Independence: Some Evidence from a Cohort Analysis,” Social Science Quarterly, 53 (December, 1972) 494–519Google Scholar.

Since women tend to outlive men, the proportion of females in a cohort will increase as it ages. Glenn introduced a weighting factor to compensate for this tendency. As he reports, sex differences in levels of identification were small, and his weighting procedure had little effect on his results. The present analysis consistently controlled for sex, but since sex differences were small, no weighting factor was used.

²⁴ Glenn usually reports a slightly higher proportion of identifiers than that reported in Table 2 of this article. Glenn's analysis includes nonwhites, but this difference does not account for his tendency to find a larger proportion of identifiers. When blacks are included in the present study, the overall percentage of identifiers is affected only marginally.

²⁵ Glenn, and Hefner, , “Further Evidence,” p. 44Google Scholar.

²⁶ Jennings, M. Kent and Niemi, Richard G., “Continuity and Change in Political Orientations: A Longitudinal Study of Two Generations,” American Political Science Review, 69 (December, 1975), 1316–1335CrossRefGoogle Scholar.

²⁷ These scores are based upon the same procedures used in Table 4, except that apoliticals were not included in these calculations.

²⁸ Jennings, and Niemi, , “Continuity and Change,” p. 1325Google Scholar.

²⁹ Niemi, Richard G. and Weisberg, Herbert F., “Is Party Identification Stable?”, in Controversies in American Voting Behavior, ed. Niemi, and Weisberg, (San Francisco: W. H. Freeman, 1976)Google Scholar.

³⁰ Shively, W. Phillips, “The Development of Aggregate Levels of Partisanship,” mimeo, University of Minnesota, January, 1975Google Scholar. Shively was discussing a mimeographed version of this article written in May, 1974.

³¹ Although Shively does not use this label, his proposed measurement technique can be considered a “correction for trend” procedure. For a more detailed discussion of reasons such “corrections” must consider compositional effects, see Cutler, Neal E, “Toward a Generational Conception of Political Socialization,” in New Directions in Political Socialization, ed. Schwartz, David C. and Schwartz, Sandra Kenyon (New York: Free Press, 1975), pp. 254–288, at pp. 272–273Google Scholar.

³² As several studies have shown, the shift toward partisan independence was most marked in the South. See Schreiber, E. M., “‘Where the Ducks Are’: Southern Strategy Versus Fourth Party,” Public Opinion Quarterly, 35 (Summer, 1971), 157–167CrossRefGoogle Scholar; Glenn, “Sources of Shift”; and Beck, Paul Allen, “Partisan Stability and Change in the American South: 1952–1972,” paper presented at the 70th Annual Meeting of the American Political Science Association, Chicago, September, 1974Google Scholar. But the shift away from partisan loyalties was also a nationwide trend. I have also avoided conducting this cohort analysis within regions for a basic methodological reason: cohort analysis assumes that when one traces an age group across time its social composition is unchanged. Migrations between regions in the United States are age-related, and the composition of a cohort within a region may change over time as a result of these migrations.

³³ Persons born before 1924 are labeled the prewar electorate, even though a small proportion entered the electorate during World War II.

³⁴ Personal communication, May 30, 1974.

³⁵ These figures include persons in cohorts older than those presented in the tables.

³⁶ The cohorts born before 1900, which had very high identification levels, made up 37 per cent of the prewar electorate in 1952, but only 15 per cent in 1974.

³⁷ Glenn reports the ages of his cohorts during the years they were surveyed. My reports about the years of birth of his cohorts are based upon my approximations. Since all persons sampled in 1945 were born before 1925, the overall identification level for that year is reported; with the 1971 data, the partisanship of persons born between 1885 and 1924 is reported.

³⁸ Jennings, and Niemi, , “Continuity, and Change,” pp. 1323–1324Google Scholar; Niemi and Weisberg, “Is Party Identification Stable?”

³⁹ Converse, “Of Time and Partisan Stability.”

⁴⁰ Inglehart, and Hochstein, , “Alignment and Dealignment of the Electorate in France and the United States, Comparative Political Studies, 5 (October, 1972), 343–372CrossRefGoogle Scholar. Inglehart and Hochstein focus on the role of issues in eroding American party loyalties.

⁴¹ See Philip E. Converse, Warren E. Miller, Jerrold G. Rusk, and Arthur C. Wolfe, "Continuity and Change in American Politics: Parties and Issues in the 1968 Election," American Political Science Review, 63 (December, 1969), 1083-1105, at 1103-1105.

⁴² Converse, et al. , “Continuity and Change,” p. 1104Google Scholar.

⁴³ These arguments are advanced in Campbell, et al. , The American Voter, pp. 120–145Google Scholar, and in several essays in another volume by these same authors, Elections and the Political Order (New York: Wiley, 1966)Google Scholar. Converse advances similar arguments in, “Of Time and Partisan Stability,” and, with his colleagues, in “Continuity and Change.”

⁴⁴ Most notably by Shively, W. Phillips, “Party Identification, Party Choice, and Voting Stability: The Weimar Case,” American Political Science Review, 66 (December, 1972), 1203–1225CrossRefGoogle Scholar.

⁴⁵ For a discussion of possible new alignments, see Abramson, , Generational Change, pp. 71–121Google Scholar.