Last updated 09/07/24: Online ordering is currently unavailable due to technical issues. We apologise for any delays responding to customers while we resolve this. For further updates please visit our website: https://www.cambridge.org/news-and-insights/technical-incident
We use cookies to distinguish you from other users and to provide you with a better experience on our websites. Close this message to accept cookies or find out how to manage your cookie settings.
Published online by Cambridge University Press: 04 January 2016
Since the publication of V.O. Key’s seminal article, “A Theory of Critical Elections,” numerous scholars have worked to add theoretical depth and analytical sophistication to the study of critical elections. On the latter count, efforts have been ambitious and illuminating; on the former they largely have not. Typologies of elections and descriptions of electoral change have emerged from empirical analyses of aggregate vote data. A variety of techniques ranging from straight forwara t-tests to correlation matrices to a particularly rich analysis of variance approach have been employed to uncover patterns of electoral change. But while a good deal of descriptive, analytical headway has been made, the development of a theory of electoral change has lagged behind.
What theoretical overtures have been made, generally regard critical elections both as the products of prolonged malfunctions in the political system and as the means by which the system begins to correct these malfunctions. According to this view, American electoral politics follows a cyclic dynamic. Critical elections mark the beginnings and the ends of adjoining periods (i.e., party systems) in a repeating cycle of stability, disequilibrium, and adjustment.
1 Testing for the statistical significance of differences between average votes for adjacent periods of electoral history has been used often to identify critical elections. See Burnham, Walter Dean, Critical Elections and the Mainsprings of American Politics (New York, 1970)Google Scholar; and Pomper, Gerald, “Classification of Presidential Elections,” Journal of Politics, 29 (August 1967)CrossRefGoogle Scholar.
2 The use of correlation matrices of time series voting data to analyze electoral trends is illustrated by Pomper, Gerald, “Classification of Presidential Elections”; and Shover, John L., “Was 1928 a Critical Election in California” in Clubb, Jerome and Allen, Howard, eds., Electoral Change and Stability in American Political History (New York, 1971)Google Scholar.
3 The analysis of variance technique for identifying types of electoral change was developed by Zingale, Nancy H. in Electoral Stability and Change: The Case of Minnesota, 1857-1966 (Ph.D. Dissertation, University of Minnesota, 1971)Google Scholar. A concise explication of this approach appears in Flanigan, William H. and Zingale, Nancy H., “The Measurement of Electoral Stability and Change,” Political Methodology, 1 (Summer 1974)Google Scholar.
4 This interpretation of critical realignments is explicit in Burnham, Walter Dean, Clubb, Jerome M., and Flanigan, William H., “Partisan Realignment: A Systemic Perspective,” in Silbey, Joel, Bogue, Allen, and Flanigan, William H., eds., The History of American Electoral Behavior (Princeton, New Jersey: Forthcoming)Google Scholar.
5 Paul Allen Beck, for example, suggests that as a critical election fades into history, and along with it its symbols and emotions, the “efficiency” with which parental partisanship is transmitted to the younger generation declines. The “realignment generation,” i.e., those persons acquiring partisanship during the crises of the realignment, wilt hold their loyalties intensely. However, subsequent generations who did not experience the crises acquire partisanship with less of this intensity. The partisanship basis of a party system therefore weakens over time and the electorate becomes “ripe” for realignment. See Beck, , “A Socialization Theory of Realignment,” in Neimi, Richard, ed., Politics of Future Citizens (San Francisco, 1974)Google Scholar. James Sundquist argues along analogous lines that the parties lose their distinctiveness as the realignment fades into history, and consequently, that new voters will fail to see the parties as relevant or as real alternatives. These processes eventually ripen the electorate for realignment. Sundquist, James L., Dynamics of the Party System (Washington, D.C., 1973), 33–36.Google Scholar
6 Burnham, Clubb, and Flanigan, “Partisan Realignment: A Systemic Perspective,” 34. (The pagination reported for this article is taken from the unpublished manuscript of the same name prepared for presentation at the Conference on Popular Voting Behavior sponsored by the Mathematical Social Science Board at Cornell University, June 1973.)
7 Ibid., 35-36.
8 Ibid., 29-36.
9 A summary and analysis of the major hypotheses dealing with future party systems is offered in Clubb, Jerome M., Flanigan, William H., and Zingale, Nancy H., “Partisan Realignment Since 1960: (Paper prepared for delivery at the Annual Meeting of the American Political Science Association, Chicago, Illinois, September 2-5, 1976)Google Scholar.
10 Walter Dean Burnham has argued in numerous articles that the partisanship basis of party systems has been decomposing since 1900 and will continue to do so in the future. See Burnham, , “The Changing Shape of the American Political Universe,” American Political Science Review, 59 (March 1965), 1–25CrossRefGoogle Scholar; Burnham, , Critical Elections, 71–91, 175-95Google Scholar; and Burnham, , “Theory and Voting Research: Some Reflections on Converse’s ‘Change in the American Electorate’,” American Political Science Review, 68 (September 1974), 1002–23.CrossRefGoogle Scholar
11 Several scenarios for traditional partisan realignments have been developed. Cf: Phillips, Keven, The Emerging Republican Majority (New Rochelle, N.Y., 1969)Google Scholar; Scammon, Richard M. and Wattenberg, Ben J., The Real Majority (New York, 1970)Google Scholar; Clubb, Flanigan, and Zingale, “Partisan Realignment Since 1960,” 23-32.
12 Ibid., 24.
13 Diachronic change is used throughout this paper to denote a change in systems—in the conceptual as well as in the formal modeling sense. For an introduction to the complementary concepts of diachronic and synchronic change, see Cortes, Fernando, Przeworski, Adam, and Sprague, John, Systems Analysis for Social Scientists (New York, 1974), 3–25.Google Scholar
14 Converse’s normal vote model provides an expected value for a party’s proportion of the vote based on the distribution of party identification. In contrast to the dynamic model offered here, the normal vote model is static, and inapplicable to party systems predating opinion surveys. See Converse, Philip E., “The Concept of a Normal Vote,” in Campbell, Angus, et. al., Elections and the Political Order (New York, 1966), ch. 14.Google Scholar
15 The classic formalization of partisanship development, of course, is due to Philip Converse. His model, however, does not accommodate realignments. See Converse, , “Of Time and Partisan Stability,” Comparative Political Studies, 2 (July 1969), 139–71.CrossRefGoogle Scholar
16 The tendency of American party competition to remain within a narrow equilibrium range has been commented upon frequently. Cf: Sellers, Charles, “The Equilibrium Cycle in Two-Party Politics,” Public Opinion Quarterly, 30 (1965), 16–38CrossRefGoogle Scholar; and Stokes, Donald E. and Iverson, Gudmund R., “On the Existence of Forces Restoring Two-Party Competition,” in Campbell, Elections and the Political Order, 180–93.Google Scholar
17 Cortes, et. al., Systems Analysis for Social Scientists, 3-21.
18 Ibid., 25-47.
19 It is important to recognize that the structure of a system in the context of a systems analysis is quite unlike the concept of a structure in traditional structural-functionalism. In the latter structures are abstractions of reality while in systems analysis they are models of the relationships underlying observable phenomena.
20 For a complete development of the concept of system function see Cortes, et. al., Systems Analysis for Social Scientists, 48-80.
21 The concept of synchronic time is analogous to Claude Levi-Strauss’s concept of “microtime.” For an explanation of synchronic change see ibid., 5-16, 271-92.
22 Przeworski argues that systems persist because they “(re)produce their synchrony.” But if a system contains internal “contradictions” (e.g., the fundamental conflicts which Marx saw in capitalist systems), the system will not (re)produce its synchrony, and will perish. The important epistemological issues of system change are discussed by Przeworski in the “Epilogue” of ibid., 271-92.
23 Przeworski argues that the interdependence of ultimately incompatible systems is a likely cause of diachronic change. Ibid., 289.
24 The conventional graph algebra approach to systems modeling is developed at length in ibid., 1-121.
25 A close inspection of this mobilization model will reveal a misspecification problem. The strong and weak partisan pools are not partitioned to the mutual exclusion of each other. This problem can be rectified by including another parameter, f3, to account for those voters who fail to repeat a Democratic vote at a second point in time. This correctly specified model is: 
Unfortunately, this reduces to a system of equations that is underdetermined by second order autoregression. It can be shown, however, that the misspecification in equation (1) does not affect the estimates of f2 and g, and produces what is usually only a small positive bias in f1. Consequently, the specification problem in (1) is ignored safely.
26 Theoretically this category includes non-voters as well as Republicans and Independents. However, the empirical analysis uses proportions of the vote to measure Dt, and not Democratic proportions of the adult population. The latter data would accommodate changes in turnout.
27 An easily understood introduction to difference equations and their solutions is given by Samuel Goldberg in Difference Equations (New York, 1958).
28 The derivation of this solution form is given in ibid., 138-39.
29 These very appropriate labels for this type of system are suggested in Cortes, et. al., Systems Analysis for Social Scientists, 157-63.
30 Ibid., 157-63.
31 The modulus, r, must be positive.
32 For discussions of the ecological fallacy and precautionary measures see Goodman, Leo A., “Some Alternatives to Ecological Inference: The Use of Aggregate Data to Study Individuals,” American Political Science Review, 63 (December 1969)Google Scholar. [Also of interest is the article following in this issue of SSH by Allen J. Lichtman and Laura Irwin Langbein, Editor’s Note.]
33 Where data were missing, average votes were still calculated if at least two of the pertinent elections were found; otherwise, the observation was coded missing.
34 These problems are demonstrated clearly by Hibbs, Douglas A., “Problems of Statistical Estimation and Causal Inference in Time-Series Regression Models,” in Costner, Herbert L., ed., Sociological Methodology 1974 (San Francisco, 1974).Google Scholar
35 The alternative estimation procedures that demand consideration are generalized least squares, and maximum likelihood estimation of a Box-Jenkins ARIMA model. Generalized least squares is not superior to OLS because it requires that the structure of the process generating the error pattern be assumed. If this assumption is wrong (and chances are good it will be), the GLS estimates may be more biased than the OLS estimates.
The Box-Jenkins techniques (see Box, George E.P. and Jenkins, Gwilym, Time Series Analysis [San Francisco, 1976]Google Scholar) are inappropriate because the systems theory and its formalization fix the underlying process as a second order autoregressive one. Box-Jenkins techniques are designed to fit, in the absence of substantive theory, an ARIMA model to the data. Since other forms of autodependence cannot be admitted, Box-Jenkins techniques offer no advantages. Moreover, the Box-Jenkins maximum likelihood algorithm will not yield estimates significantly different from least squares. See Box and Jenkins, Time Series Analysis, 213.
36 Although OLS residuals from purely dynamic models are biased guides to true error patterns, and generally will fail to reflect serial error correlations of the same order as lags specified in the regression model, they can be used in an approximate or “portmanteau” test of the hypothesis of model adequacy. A statistic Q (see Box and Jenkins, Time Series Analysis, 289-93), based on the series of possible autocorrelations among residuals, can be tested for significance against chi square. The Q statistic was calculated for sixty sets of residuals randomly selected from the 335 time series utilized in this study. Q was non-significant in every case, suggesting no grounds for questioning the adequacy of the model, nor for worrying about correlated errors at least beyond lag two.
While this test still leaves first and second lag error correlations suspect, substantive theory supports the assumption of independent errors. Serially correlated errors will result to the extent that the partisanship component extracted from the voting data by autoregression is not the only long-term factor underpinning the vote. But, except for partisanship, the other most prominent components of voting choices can be viewed as short-term. For example, important factors such as candidates’ personalities and issues should be fairly independent over time, and consequently not produce bias in the OLS estimates.
37 The mean change for the thirty-seven counties in the sample was -8.62 percent. Hence the sample is closely representative of the whole state.
38 Urban areas are defined here as they are in the United States’ census as locations with more than 2500 persons.
39 See Burnham, Critical Elections, 38-55, 71-90.
40 The relative isolation of rural areas from the mainstream of political information has been suggested by Converse as an explanation for damped partisan change in rural areas. Converse, Philip E., “Information Flow and Stability of Partisan Attitudes,” in Campbell, , et. al., Elections and the Political Order, 136–57.Google Scholar
41 Burnham, Critical Elections, 50-53.
42 See the number of iterations required for convergence as listed in Table 1.
43 Cf: Burnham, , Critical Elections, 91–134Google Scholar; Converse, Philip E., “Change in the American Electorate,” in Campbell, Angus and Converse, Philip, eds., The Human Meaning of Social Change (New York, 1972)Google Scholar; Rusk, Jerrold G., “The Effects of the Australian Ballot Reform on Split Ticket Voting: 1876-1908,” American Political Science Review, 64 (December 1970), 1220–38CrossRefGoogle Scholar; Burnham, Walter Dean, “Theory and Voting Research: Some Reflections on Converse’s ‘Change in the American Electorate’,” American Political Science Review, 68 (September 1974), 1002–23CrossRefGoogle Scholar; Rusk, Jerrold G., “Comment: The American Electoral Universe, Speculation and Evidence,” American Political Science Review, 68 (September 1974), 1028–49CrossRefGoogle Scholar; and Converse, Philip E., “Comment on Bumham’s ‘Theory and Voting Research’,” American Political Science Review, 68 (September 1974), 1024–25.CrossRefGoogle Scholar
44 The Democratic ranks were also buoyed by a large influx of new voters, particularly immigrants, who were not mobilized until the 1930s. See Campbell, Angus, et. al., “Development of Party Identification,” in The American Voter (New York, 1960), 146–68Google Scholar. A systems analysis employing a non-constant input function would detect the contribution of new voters to realignment. However, data limitations precluded the use of non-constant inputs in this analysis.
45 The studies pointing to the weakening relationship between experience and partisanship strength are quite numerous. For example, Cf: Abramson, Paul R., “Generational Change and the Decline of Party Identification in America: 1952-1974,” American Political Science Review, 70 (June 1976), 469–78CrossRefGoogle Scholar; Glenn, Norval D., “Sources of Shift to Political Independence: Some Evidence From a Cohort Analysis,” Social Science Quarterly, 53 (1972), 494–519Google Scholar; and Mr.Jennings, Kent and Niemi, Richard G., “Continuity and Change in Political Orientations: A Longitudinal Study of Two Generations,” American Political Science Review, 69 (December 1975), 1316–35.CrossRefGoogle Scholar
46 The life-cycle theory, which dates to Campbell, et. al., The American Voter, 161-65, views the strengthening of partisanship as a social psychological process dependent upon the duration of the attachment. Data presented by Miller and Levitan, however, show clearly that the percentage of independents among most cohorts rose between 1952 and 1974. See Miller, Warren E. and Levitan, Teresa E., Leadership and Change: The New Politics and the American Electorate (Cambridge, Mass., 1976), 195.Google Scholar
47 The defense of the life-cycle theory has been carried largely by Philip E. Converse. A comprehensive analysis of the issue is contained in his monograph, The Dynamics of Party Support: Cohort Analyzing Party Identification (Beverly Hills, Calif., 1976).
48 Converse argues in his monograph that the pre- and post-1964 periods must be examined separately. Life-cycle effects, though present in both periods, definitely are damped in the latter period by generational factors affecting the young, and period factors slowing the habituation process throughout the electorate.
