¹ See, especially, Sears, David O. and Chaffee, Steven H., ‘Uses and Effects of the 1976.Debates: An Overview of Empirical Studies’, in Kraus, Sidney, ed., The Great Debates: Carter vs. Ford, 1976 (Bloomington: Indiana University Press, 1979)Google Scholar; Sigelman, Lee and Sigelman, Carol K., ‘Judgements of the Carter–Reagan Debate: The Eyes of the Beholders’, Public Opinion Quarterly, 48 (1984), 624–8CrossRefGoogle Scholar; LeDuc, Lawrence and Price, Richard, ‘Great Debates: The Televised Leadership Debates in Canada’, Canadian Journal of Political Science, 18 (1985), 135–53.CrossRefGoogle Scholar

² See, respectively, Lanoue, David J., ‘One That Made a Difference: Cognitive Consistency, Political Knowledge, and the 1980 Presidential Debate’, Public Opinion Quarterly, 56 (1992), 168–84CrossRefGoogle Scholar; Lanoue, David J., ‘Debates That Mattered: Viewers' Reactions to the 1984 Canadian Leadership Debates’, Canadian Journal of Political Science, 24 (1991), 51–66CrossRefGoogle Scholar; Schrott, Peter R., ‘Electoral Consequences of “Winning” Televised Campaign Debates’, Public Opinion Quarterly, 54 (1990), 567–85.CrossRefGoogle Scholar

³ See Lanoue, , ‘One That Made a Difference’.Google Scholar

⁴ For more details on the organization and the content of the debates, see Monière, Denis, Le Combat des chefs: Analyse des débats télévisés au Canada (Montréal: Québec-Amérique, 1992).Google Scholar

⁵ The principal investigators were Richard Johnston, André Blais, Henry Brady and Jean Crête. The fieldwork was carried out by the Institute for Social Research, York University.

⁶ All interviewing was completed from centralized telephone facilities using Computer Assisted Telephone Interviewing (CATI) technology. The response rate for the campaign-period survey was 57 per cent and the re-interview rates for the post-election survey and the mailback questionnaire were 81 per cent and 80 per cent respectively. Further details on the sample and the survey can be found in Northrop, David and Oram, Anne, The 1988 National Election Study: Technical Documentation (Toronto: Institute for Social Research, York University, 1989).Google Scholar

⁷ The survey started on 4 October, three days after the election had been called. Interviews were conducted every day except Thanksgiving Day (10 October) until the very last day of the campaign, 20 November. Seventy-seven interviews were completed per day, on average.

⁸ See Lang, G. E. and Lang, Kurt, ‘The Formation of Public Opinion: Direct and Mediated Effects of the First Debate’, in Bishop, G. et al. , eds, The Presidential Debates (New York: Praeger, 1978)Google Scholar; Steeper, Frederick, ‘Public Responses to Gerald Ford's Statements on Eastern Europe in the Second Debate’Google Scholar, in Bishop, et al. , The Presidential DebatesGoogle Scholar; Patterson, Thomas E., The Mass Media Election (New York: Praeger, 1980), p. 123Google Scholar; Lanoue, David and Schrott, Peter R., The Joint Press Conference: The History, Impact, and Prospects of American Presidential Debates (New York: Greenwood Press, 1991), p. 135.Google Scholar

⁹ See Spector, Paul E., Research Designs (Beverly Hills, Calif.: Sage, 1981)CrossRefGoogle Scholar; Cook, Thomas D. and Campbell, Donald T., Quasi-Experimentation: Design and Analysis Issues for Field Settings (Boston, Mass.: Houghton Mifflin, 1979).Google Scholar

¹⁰ The NDP seems to increase its share of voting intentions from week 1 to week 3, but this is only because the very first poll, conducted on 1–2 October, put them at 22 per cent. All subsequent polls put the NDP at between 26 and 30 per cent.

¹¹ We start with 5 October, which is the first full interviewing day, but the responses obtained on 4 October are included in the computation of moving averages. Note that the moving average for 5 October includes only one day before, and those for the penultimate and last days of the campaign include one and no day after respectively.

¹² Sixty-three per cent of our respondents indicated, in the post-election questionnaire, that they had seen the debate. Note that these figures, especially for non-watchers, are based on very small numbers (the average daily number of respondents indicating a vote intention among non-watchers is eighteen) and must be interpreted cautiously.

¹³ The data also seem to indicate that watchers depart from non-watchers in the last week of the campaign, the former being more inclined to vote Conservative and less likely to vote NDP. We do not have a ready explanation for this divergence. It is the only week of the campaign for which we observe a substantial difference between watchers and non-watchers. The most plausible explanation, we believe, is sampling error, as differences in the distribution of vote intentions between watchers and non-watchers in the last week are not significant at the 0.05 level (p = 0.08) and as we find no difference in the voting behaviour of watchers and non-watchers in our post-election survey.

¹⁴ The group that saw the debates is affected by both the direct and the indirect effects. The group that did not see them is affected only by the latter. The difference is the direct effect. We refer to the net direct effect because the indirect effect could be greater among non-watchers (as they do not have first-hand information to countervail that coming from friends or the media) than among watchers. The difference between watchers and non-watchers thus corresponds to the direct effect minus any differential between the two groups in the magnitude of the indirect effects.

¹⁵ All regressions reported in this article are ordinary least squares (OLS) regressions. In Tables 2,3 and 4, the dependent variable is a dummy variable; we also ran LOGIT regressions, and the results proved to be very similar. We report the OLS regressions, which are easier to interpret.

¹⁶ Lanoue, ‘Debates That Mattered’.

¹⁷ Leduc and Price, ‘Great Debates’.

¹⁸ Schrott, ‘Electoral Consequences of “Winning” Televised Campaign Debates’.

¹⁹ Lanoue, ‘One That Made a Difference’.

²⁰ Fifty-three per cent said Turner had performed the best in the debates, 19 per cent chose Mulroney, and 6 percent Broadbent. The analysis includes both watchers and non-watchers. We also ran regressions including an interactive Winner-Watch variable, which came out significant in the PC-Lib equation, but not in the Lib-NDP one.

²¹ See McDowall, David, McCleary, Richard, Meidinger, Errol E. and Hay, Richard A., Interrupted Time Series Analysis (Beverly Hills, Calif.: Sage, 1980).CrossRefGoogle Scholar

²² See Price, Simon and Sanders, David, ‘Modeling Government Popularity in Postwar Britain: A Methodological Example’, American Journal of Political Science, 37 (1993), 317–34.CrossRefGoogle Scholar

²³ Price and Sanders, for instance, had a coefficient of 0.86 for the lagged dependent variable, indicating that 86 per cent of the initial impact remains after one period (one quarter in their case), 74 per cent (0.86 × 0.86) after two periods, and so on. See Price and Sanders, ‘Modeling Government Popularity in Postwar Britain’.

²⁴ We thank Marcel Dagenais for suggesting that possibility. An alternative way of allowing the debate variables more chance of having an effect would be to have more than one lag for the dependent variable and lagged debate variables. This would create degrees of freedom problems, given the small number of observations. For this reason, we believe our method is more suitable.

²⁵ We found no autocorrelation in any of the equations shown in Table 5.

²⁶ More precisely, according to our estimates, the Conservatives lost 7 points (0.22 * 0.33) on the first day after the debate, 16 points after two days (0.22 * 0.67 + 0.16 * 0.07), 25 points after three days (0.22 * 1 + 0.16 * 0.16), 17 points after four days (0.22 * 0.6 + 0.16 * 0.25), and 16 points after five days (0.22 * 0.6 + 0.16 * 0.17). The loss remains at that level until election day.

²⁷ More precisely, the Liberals gained 3 points (0.08 * 0.33) on the first day after the debates, 6 points after two days (0.08 * 0.67 + 0.22 * 0.03), 9 points (0.08 + 0.22 * 0.06) after three days, and 10 points (0.08 + 0.22 * 0.09) after four days. The gain remains at that level until election day.

²⁸ It could be argued that the differences between some of the scenarios are extremely small, and that it is not possible to tell which one is really most plausible. The results are unequivocal, however, in showing that the least plausible scenarios, for the two races, are those that assume most of the initial impact to vanish. Furthermore, though the data cannot tell us the exact rate of decay, they converge on what the final impact is. All the scenarios which assume the final impact to be at least half the initial one (final = 0.5 or more) indicate a final impact of between 15 or 17 points in the case of the PC–Liberal contest and of between 9 and 11 points with respect to the Liberal–NDP contest.

²⁹ We take the average vote in the first three weeks of the campaign, as there is no indication of movement before the debates. This gives the Conservatives 43 per cent, the Liberals 27 per cent, and the NDP 28 per cent.

³⁰ With respect to the 1988 Canadian election, there is no evidence that any other specific campaign event played a major role; see Johnston, Richard, Blais, André, Brady, Henry and Crête, Jean, Letting the People Decide (Montreal: McGill-Queen's University Press, 1992).Google Scholar

³¹ The lowest estimate, a 6 point impact, is provided by the ‘reactions’ analysis; the highest estimate, a 12 point difference, comes from the simple comparison of vote intentions in the public polls before the debates and in the final week of the campaign. The time-series estimate is a 10 point effect.

³² We deal here only with the debates' impact on the vote. In 1988, the debates necessitated major changes in party strategies. We are unable, within the confines of this article, to determine whether the debates' impact is partially imputable to campaign strategies. Neither do we explore the psychological process through which debates affect the vote. It is quite possible, for instance, that strategic voting benefited the Liberals at the expense of the NDP. Such strategic voting, however, was prompted by the debates, which substantially increased perceived probabilities of a Liberal victory (see Johnston, et al. , Letting the People Decide, p. 201).Google Scholar