¹ See Key, V. O. Jr, The Responsible Electorate (New York: Vintage Books, 1966).CrossRefGoogle Scholar

² White, Theodore H., The Making of the President, 1968 (New York: Simon & Schuster, 1970).Google Scholar

³ Witcover, Julius, Marathon: The Pursuit of the Presidency, 1972–1976 (New York: Viking Press, 1977).Google Scholar

⁴ Wilson, Major L., The Presidency of Martin Van Buren (Lawrence: University of Kansas Press, 1984).Google Scholar

⁵ Muted succession effects might also be found when a cabinet official is nominated to succeed the president, but there is no appropriate test case during the period of the NES surveys.

⁶ Key, , The Responsible ElectorateGoogle Scholar; Fiorina, Morris P., Retrospective Voting in American National Elections (New Haven, Conn.: Yale University Press, 1981).Google Scholar

⁷ Shanks, J. Merrill and Miller, Warren E., ‘Policy Direction and Performance Evaluation: Complementary Explanations of the Reagan Elections’, British Journal of Political Science, 20 (1990), 143–235CrossRefGoogle Scholar; Shanks, J. Merrill and Miller, Warren E., ‘Partisanship, Policy and Performance: The Reagan Legacy in the 1988 Election’, British Journal of Political Science, 21 (1991), 129–97.CrossRefGoogle Scholar

⁸ Shanks and Miller do find a ‘strong impact’ of evaluations of Reagan's performance as president, even within their system of extensive controls on causally prior variables. Their results are perhaps dependent on their exact assumptions as to the ordering of variables in a causal chain, even though there is no consensus in the literature on the proper ordering of variables in such a chain (see, for instance, Wattenberg, Martin, The Rise of Candidate-Centered Politics (Cambridge, Mass.: Harvard University Press, 1991) pp. 141–3).CrossRefGoogle Scholar Indeed, the ordering they used for 1988 is not fully identical to that they had used for 1984. We would not necessarily question the ordering they employ for 1988, but we prefer a model which makes fewer assumptions as to the causal order, to test whether succession effects exist outside of a single assumed sequence of causally prior variables. We note also that emotions appear to affect the approval of presidential performance more strongly and consistently than do issues or prospective and retrospective views of personal finances and the economy. See Ragsdale, Lyn, ‘Strong Feelings: Emotional Responses to Presidents’, Political Behavior, 13 (1991), 33–65.CrossRefGoogle Scholar

⁹ Kelley, Stanley Jr and Mirer, Thad W., ‘The Simple Act of Voting’, American Political Science Review, 68 (1974), 572–91CrossRefGoogle Scholar; see also Kelley, Stanley Jr, Interpreting Elections (Princeton, NJ: Princeton University Press, 1983).CrossRefGoogle Scholar

¹⁰ Miller, Arthur H. and Miller, Warren E., ‘Partisanship and Performance: “Rational” Choice in the 1976 Presidential Election’ (paper presented at the Annual Meeting of the American Political Science Association, Washington, DC, 1977)Google Scholar; Shanks, and Miller, , ‘Policy Direction and Performance Evaluation’Google Scholar; Shanks, and Miller, , ‘Partisanship, Policy and Performance’.Google Scholar

¹¹ Kelley, and Mirer, , ‘The Simple Act of Voting’Google Scholar; see also Brody, Richard A. and Page, Benjamin I., ‘Indifference, Alienation and Rational Decision: The Effects of Candidate Evaluations on Turnout and the Vote’, Public Choice, 15 (1973), 1–17.CrossRefGoogle Scholar

¹² Markus, Gregory B., ‘Political Attitudes During an Election Year: A Report on the 1980 NES Panel Study’, American Political Science Review, 76 (1982), 538–60CrossRefGoogle Scholar; see also Markus, Gregory B. and Converse, Philip E., ‘A Dynamic Simultaneous Equation Model of Electoral Choice’, American Political Science Review, 73 (1979), 1055–70.CrossRefGoogle Scholar

¹³ Fiorina, , Retrospective Voting in American National Elections.Google Scholar

¹⁴ Candidate electability has been used in this analysis rather than candidate viability (the respondent's estimate of the candidate's chance of winning the nomination), since the candidate's chance of winning the general election would theoretically affect the rational voter's utility calculations.

¹³ Bartels, Larry M., ‘Expectations and Preferences in Presidential Nominating Campaigns’, American Political Science Review, 79 (1985), 804–15CrossRefGoogle Scholar; Bartels, Larry M., Presidential Primaries and the Dynamics of Public Choice (Princeton, NJ: Princeton University Press, 1988).Google Scholar

¹⁶ The 0.58 is obtained by inserting the variable means shown in the text into Equation 1.

¹⁷ Given the 0.064 coefficient for the net candidate score, it takes slightly more than 7 thermometer points to match the estimated 0.474 impact of approval of Reagan's performance.

¹⁸ To compute this change of probability Δp we use the formula Δp = exp(L₁)/[1 + exp(L₁)] -exp(L₀)/[1 +exp(L₀)], where L₀ is the value of the logit before the change of the independent variable and L₁ is the value of the logit after the change in the independent variable. Given our baseline (P = 0.5), L₀ = 0.

¹⁹ The 1988 NES study actually included nine one-year retrospective orientation questions plus five eight-year items. In addition to the analysis in Table 2, we examined each of these items alone and in various combinations; the results reported in Table 2 are fully representative of those larger results.

²⁰ The estimate of the comparative assessment of candidate attributes is based on questions asking how well a trait describes Bush or Dukakis. The estimate of issue effects is based on the distance between the respondents' positions and the candidates' perceived positions. In both instances, missing data are scored as zero (the neutral point).

²¹ These probability estimates are the result of changes moving a respondent from the neutral point (scored as 0) on partisanship, assessment of candidate traits, evaluations of Reagan's performance, issue and ideological orientation to a ‘typical’ value' – the variables' sample means. Consider, for instance, the direct and indirect influence of the level of approval of Reagan on the probability of voting for Bush. The sample mean of Reagan's performance is 0.357; according to the coefficient shown in Table 3, this value translates into a 1.6 ( = 0.357 × 4.465) pro-Bush advantage on the net candidate score. This score, along with the mean level of approval of Reagan, is used to derive the average probability that an otherwise undecided voter has to cast her preference for the vice-president. Given the estimated effect of net candidate score and Reagan's performance obtained in Table 1b, the probability that such a voter may support Bush rises from 0.50 to 0.57. The same procedure is used to compute the other probabilities presented in the text.

²² The estimation of the net candidate score carried out in Table 3 yields a sizeable pro-Bush residual, which points to an inability to account fully for the advantage enjoyed by the vice-president on the thermometer. We have attempted to test this by adding additional variables related to important 1988 campaign themes. The 1988 NES questionnaire did not ask respondents to rate candidates according to their experience, so we relied on the open-ended questions (master codes 211, 212, 217–21) to measure the impact of experience. We relied on the same questions to estimate the influence of the law and order theme (master codes 968–78, 988–90,1041, 1042) and abortion (codes 985–7). Finally, two retrospective economic evaluation questions and the question about the country's current conditions (see Table 2) were also included. Our analysis shows that only perceived experience and the abortion issue had significant effects on the net candidate score. Both variables worked to Bush's advantage in the election, and, as a consequence, the size of the regression intercept decreases (though it still does not disappear). On the other hand, the parameter estimates for the variables of the original model are only slightly affected by the new inclusions, and the greater probability of voting for Bush on the basis of succession effects than other factors remains largely unchanged.

²³ The analysis is based on telephone interviews conducted in the sixteen states (all but two in the South) that held primaries on 8 March 1988. According to the Super Tuesday Study, Bush received 57.3 per cent and Dole 23.2 per cent of the votes cast by the respondents. The percentages based on the respondents' preferences for the Republican nomination are 57.4 and 23.5, respectively. Given the low number of reported voters and the close correspondence between votes and preferences, we decided to use the latter for the present study.

²⁴ To parallel the analysis in Table 1, we tested a model including a direct succession effect on the primary season preferences. The Reagan approval rating was not significant; instead, as shown below, it had an indirect effect on candidate choice. The Super Tuesday Study does not allow controls for retrospective orientations.

²⁵ Obviously the two-stage least-squares results are limited by our ability to find instruments to estimate the relationship between candidate evaluations and electability. There is agreement that the two affect each other (Bartels, , Presidential Primaries and the Dynamics of Public Choice)Google Scholar, but inevitably there will be disagreement as to the choice of instruments. We assume that electability is directly affected by the election results and by the respondents' political knowledge and involvement (here measured by recognition of Bush's name and by the respondent's educational level, respectively). We also assume that overall candidate evaluations are shaped by specific candidate traits, succession effects and the candidates' issue positions. The available data do not permit the estimation of the impact of the candidates' issue stands.

²⁶ Bartels, , ‘Expectations and Preferences in Presidential Nominating Campaigns’Google Scholar; Brady, Henry E. and Johnston, Richard, ‘What's the Primary Message: Horse Race or Issue journalism?’ in Orren, Gary R. and Polsby, Nelson W., eds, Media and Momentum (Chatham, NJ: Chatham House, 1987).Google Scholar

²⁷ The estimated net candidate score based only on Reagan's performance evaluation is obtained by writing the equation in its reduced form, with the dependent variable a function of the exogenous variables. Separate analysis of the Bush and Dole thermometers (not shown here) indicates that evaluation of Reagan's performance influenced the net candidate score by affecting Bush's rating but not Dole's.

²⁸ See fn. 21 above.

²⁹ These correlations are not simply due to partisanship: with party identification controlled, the 1988 and 1968 correlations are both 0.63, compared with 0.44, for 1984 and 0.34 for 1976.

³⁰ The 1968 election was also unusual in that Nixon was a former vice-president, but the NES asked no questions that would permit a test of whether that service affected the 1968 race.

³¹ The higher value of the logit coefficient for the 1968 net candidate score may be due, to a large extent, to the timing of the thermometer questions that in 1968 were asked only in the survey's post-election wave. When the net candidate score between Bush and Dukakis is based upon the thermometer scores obtained after the election, the corresponding logit coefficient is 0.1 and the parameter estimate for Reagan's affect becomes 0.042. The two surveys contain a similar question on the president's performance asked before the vote. Adding the post-election net candidate score and party identification, the parameter estimates (standard errors in parenthesis) for the model are as follows: 1968; intercept – 0.673 (0.214), net candidate score 0.127 (0.012), party identification –0.911 (0.162), incumbent's performance 0.579 (0.204); and in 1988, intercept – 1.031 (0.175), net candidate score 0.100 (0.009), party identification 0.750 (0.120), incumbent's performance 0.618 (0.104).

³² The 1968 national election study does not contain questions about candidate traits.

³³ Fiorina, , Retrospective Voting in American National Elections, pp. 167–75.Google Scholar

³⁴ An alternative interpretation of the 1976 and 1984 results is that bad memories do not damage vice-presidents in the same way as good memories help them.

³⁵ Campbell, Angus E. et al. , The American Voter (New York: Wiley, 1960).Google Scholar

³⁶ Shanks, and Miller, , ‘Partisanship, Policy and Performance’.Google Scholar

³⁷ Kessel, John H., ‘Comment: The Issues in Issue Voting’, American Political Science Review, 66 (1972), 459–65.CrossRefGoogle Scholar