¹ On party identification as a long-term force, see Converse, P. E., ‘The Concept of the Normal Vote’, in Campbell, A., Converse, P. E., Miller, W. E. and Stokes, D. E., Elections and the Political Order (New York: John Wiley & Sons, 1968).Google Scholar

² Key, V. O. Jr., and Munger, F., ‘Social Determinism and Electoral Decision: The Case of Indiana’, in Burdick, E. and Brodbeck, A. J., eds, American Voting Behavior (New York: Free Press of Glencoe, 1959).Google Scholar

³ See Burnham, W. D., Critical Elections and the Mainsprings of American Democracy (New York: W. W. Norton, 1970).Google Scholar

⁴ Campbell, A., Gurin, G. and Miller, W. E., The Voter Decides (Evanston, Ill.: Row-Peterson, 1954). p. 89.Google Scholar

⁵ Converse, P. E., ‘Public Opinion and Voting Behavior’, in Greenstein, F. I. and Polsby, N. W., eds. Handbook of Political Science, Volume 4 (Reading, Mass.: Addison-Wesley, 1975), p. 117.Google Scholar

⁶ The requirement of stability in the strength component has always appeared somewhat paradoxical, since the creators of this measure lead us to expect that individuals' identification will strengthen over their life-cycles. Converse resolves this apparent contradiction by showing that the expected increment to the strength component over a brief period – such as that covered by the extant national panel data sets – will be indistinguishable from sampling and measurement error. In other words, we can ignore life-cycle increases in strength when considering whether party identification exhibits the required individual-level stability. See Campbell, A., Converse, P. E., Miller, W. E. and Stokes, D. E., The American Voter (New York: John Wiley & Sons, 1960)Google Scholar; Converse, P. E., The Dynamics of Party Support (Beverly Hills, Calif: Sage Publications, 1976), p. 46.Google Scholar

⁷ Converse, P. E. and Markus, G. B., ‘Plus ça change…: The New CPS Election Study Panel’, American Political Science Review, 73 (1979), 32–49, p. 46CrossRefGoogle Scholar; Converse, P. E., ‘The Nature of Belief Systems in Mass Publics’, in Apter, D., ed., Ideology and Discontent (New York: Free Press of Glencoe. 1964).Google Scholar

⁸ Dreyer, E. C., ‘Change and Stability in Party Identification’, Journal of Politics, 35 (1973), 712–22.CrossRefGoogle Scholar

⁹ Dobson, D., and Meeter, D. A., ‘Alternative Markov Models for Describing Change in Party Identification’, American Political Science Review, 18 (1974), 487–500CrossRefGoogle Scholar; Dobson, D., and Angelo, D. St., ‘Party Identification and the Floating Vote: Some Dynamics’, American Journal of Political Science, 69 (1973), 481–90CrossRefGoogle Scholar; Brody, R. A., ‘Stability and Change in Party Identification: Presidential to Off-Years’, paper delivered at the annual meeting of the American Political Science Association (1977).Google Scholar

¹⁰ These calculations do not include people classified as pure independents in the initial wave of the panel.

¹¹ Brody, R. A., ‘Change and Stability in the Components of Party Identification’, DAE News, APSA, 13 (1977), 13–18.Google Scholar

¹² See Markus, G. B. and Converse, P. E., ‘A Dynamic Simultaneous Equation Model of Electoral Choice’, American Political Science Review, 73 (1979), 1055–70, p. 1060CrossRefGoogle Scholar; see also Page, B. I. and Jones, C. C., ‘Reciprocal Effects of Policy Preferences, Party Loyalties and the Vote’, American Political Science Review, 73 (1979), 1071–89CrossRefGoogle Scholar; Fiorina, M. P., Retrospective Voting in American National Elections (New Haven, Conn.: Yale University Press, 1981).Google Scholar

¹³ Franklin, Charles H. and Jackson, John E., ‘The Dynamics of Party Identification’, American Political Science Review, 77 (1983), 957–73CrossRefGoogle Scholar; Franklin, Charles H., ‘Party Identification and Party Realignment’, paper delivered at the annual meeting of the American Political Science Association, 1986.Google Scholar

¹⁴ Franklin, and Jackson, , ‘The Dynamics of Party Identification’, p. 969.Google Scholar

¹⁵ The data used in this study were gathered by the National Election Study and the Center for Political Studies at the University of Michigan and were made available through the Inter-University Consortium for Political and Social Research. Neither NES/CPS nor the ICPSR are responsible for our use and interpretation of this information.

¹⁶ Even this figure is inflated, since we do not distinguish between those who say that they have no preference for any of the parties and those who describe themselves as pure independents.

¹⁷ See Zeller, R. A. and Carmines, E. G., Measurement in the Social Sciences: The Link Between Theory and Data (Cambridge: Cambridge University Press, 1980), pp. 63–7.Google Scholar

¹⁸ By contrast, Lazarsfeld et al. hypothesized that only political opinions change during an election season. See Lazarsfeld, P., Berelson, B. and Gaudet, H., The People's Choice, 3rd edn (New York: Columbia University Press, 1968).Google Scholar

¹⁹ Unfortunately, we cannot distinguish between these two possibilities with a single intraelection panel. We would need to follow our respondents for a period after the election long enough for any ephemeral activation effects to wear off.

²⁰ Converse, , ‘The Nature of Belief Systems’Google Scholar; Converse, P. E., ‘Attitudes and Non-Attitudes: Continuation of a Dialogue’, in Tufte, E. R., ed., The Quantitative Analysis of Social Problems (Reading, Mass: Addison-Wesley, 1970).Google Scholar

²¹ Converse, P. E., ‘Comment: The Status of Non-Attitudes’, American Political Science Review, 68 (1974), 650–60.Google Scholar

²² Converse, unlike Dreyer, would not treat party identification as a belief that is highly subject to the problem of non-attitudes. Quite the contrary – he considers it the best example of a real political attitude. Nevertheless, since he does not offer a theory of partisan change in the short-run, there is no reason not to use his model to check for random response. See Dreyer, , ‘Change and Stability in Party Identification’Google Scholar; Converse, and Markus, , ‘Plus ça change…: The New CPS Election Study Panel’.Google Scholar

²³ Converse, , ‘Comment: The Status of Non-Attitudes’, pp. 650–60.Google Scholar

²⁴ Our measures of directional stability exclude leaning independents; only strong and weak identifiers are analysed.

²⁵ See Brody, , ‘Stability and Change in Party Identification: Presidential to Off-Years’Google Scholar; Franklin, and Jackson, , ‘The Dynamics of Party Identification’.Google Scholar

²⁶ See Fiorina, , Retrospective Voting in American National Elections, pp. 84–105.Google Scholar

²⁷ In decomposing the dependent variable, i.e. the measure of partisanship, we depart from the practice followed by Franklin and Jackson, who dismiss intransitivities in the seven-point, party identification scale as inconsequential. We are more impressed than they with the potential for heterogeneity, that is, for a different intra-election dynamic for different party-strength groups. Accordingly, and to explore potential heterogeneity, we have decomposed the measure of partisanship. See Franklin, and Jackson, , ‘The Dynamics of Party Identification’, p. 962.Google Scholar

²⁸ In adopting this coding scheme, we assume that leaning independents are partisans and that the party identification scale is monotonic. While we do the former with little hesitation, we adopt the latter assumption with a great deal of trepidation. Although it is necessary to assume monotonicity in our analysis, we recognize that our measure will be unreliable and our findings will tend toward statistical insignificance to the extent that this assumption is violated. On classifying leaning independents as partisans, see Keith, B., Magleby, D., Nelson, C., Orr, E., Westlye, M. and Wolfinger, R., ‘The Myth of the Independent Voter’, paper delivered at the annual meeting of the American Political Science Association, 1977Google Scholar; and Brody, , ‘Stability and Change in Party Identification: Presidential to Off-Years’.Google Scholar

²⁹ We recognize that we do not distinguish between personal and sociotropic evaluation of economic conditions. In the present study, we are less concerned about the etiology of system blame than about its effects. We believe that it is possible to assess the effects of economic satisfaction or dissatisfaction on partisanship irrespective of whether the judgement is personally or socially located. On the distinction between personal and sociotropic evaluations, see Sniderman, P. M. and Brody, R. A., ‘Coping: The Ethic of Self-Reliance’, American Journal of Political Science, 21 (1977), 501–21CrossRefGoogle Scholar; Kinder, D. R. and Kiewiet, R. D., ‘Sociotropic Polities’, British Journal of Political Science, 11 (1981), 129–61.CrossRefGoogle Scholar

³⁰ See Hertel, B. R., ‘Minimizing Error Variance Introduced by Missing Data Routines in Survey Analysis’, Sociological Methods and Research, 4 (1976), 459–74.CrossRefGoogle Scholar

³¹ See McKelvey, R. D. and Zavonia, W., ‘A Statistical Model for the Analysis of Ordinal Level Dependent Variables’, Journal of Mathematical Sociology, 4 (1975), 103–20.CrossRefGoogle Scholar

³² The potential existence of some (although not all) individuals in the electorate who randomly change their partisan identification makes estimating a simultaneous recursive structure unfeasible: voters making random changes might alter their party identifications and then re-evaluate their political world to make it more consistent. Consequently, specifying a simultaneous recursive system would tend to yield artificially strong evidence that short-term electoral forces influence partisan change; instead we choose a conservative approach that reduces the probability of finding evidence for rejecting the null hypotheses.

³³ See Weisberg, H., ‘Towards a Reconceptualization of Party Identification’, paper delivered at the annual meeting of the Midwest Political Science Association, 1978Google Scholar; Brody, R. A., ‘Measures of Party Identification in the 1979 Pilot Study’, a Report to the Board of Overseers of the National Elections Project, 1979.Google Scholar

³⁴ Because of estimation problems with our probit analysis, the interaction term for nomination preference was subsequently dropped.

³⁵ We had no a priori expectations about the directions the estimated coefficients for our interaction terms should take. We assume that estimates in the same direction as that predicted for the whole sample signify that Republicans are influenced more heavily than Democrats by the factor in question (since scores for Republicans were multiplied by one to form the interaction term). Similarly, estimates in the opposite direction suggest that Democrats are more strongly affected.

³⁶ Our sample includes only forty-six strong identifiers who changed their identification in the P2 to P3 period. Consequently, the finding that party evaluations alone are significant is not much of a surprise.

³⁷ It has been suggested that party assessments may be also measuring party identification. But if this is conceptually true, then party evaluations at P1 and P2 should not have any influence on changes in party identification in P2 and P3 respectively, i.e., change in party identification cannot be a function of party identification at a prior period. Thus, it is reasonable for both conceptual and empirical reasons to treat party evaluations as reflections of substantive forces and not as party identification.

³⁸ When comparative party assessments are evaluated for the entire sample (not shown), even clearer evidence is found for the hypothesis that they are surrogates for other substantive forces. However, the results in Table 6 also demonstrate that if our other independent variables were used to form an instrument for party evaluations, the resulting measure would be highly unreliable because the level of explained variation ranges only between 10 and 30 per cent. These low R² coefficients are not surprising given the small sample sizes and our approach to handling missing values. None the less, because of them we eschew an instrumental variable strategy.

³⁹ Ideally, we would employ a multiple indicator approach to measure the underlying, short-term, substantive electoral forces. As it is not yet feasible to do so with limited dependent variables, we must leave this task to future analyses. On multiple indicators, see Joreskog, K. G. and Sorbom, D., LISREL V: Analysis of Linear Structural Relationships in Maximum Likelihood and Least Squared Methods (Uppsala, Sweden: University of Uppsala, Department of Statistics, 1981).Google Scholar

⁴⁰ Of the strong partisans, 81.7 per cent and 82.7 per cent claim that they are concerned with the election's outcome in the P1 and P2 periods respectively; for other identifiers, these figures are 40.9 per cent and 46.0 per cent.

⁴¹ On the normal vote model see Converse, , ‘The Nature of Belief Systems in Mass Publics’Google Scholar; Miller, A., ‘Normal Vote Analysis: Sensitivity to Change Over Time’, American Journal of Political Science, 23 (1979), 406–25.CrossRefGoogle Scholar

Even if we follow Achen's advice and replace normal vote analysis with regression-based approaches, we still need to know the best way to incorporate the long-term component into the model. See Achen, C., ‘The Bias in Normal Vote Estimates’, Political Methodology, 6 (1979), 343–55.Google Scholar