¹ On spatial voting and its relationship to prospective voting, see Downs, Anthony, An Economic Theory of Democracy (New York: Harper and Row, 1957)Google Scholar; Enelow, James M. and Hinich, Melvin J., The Spatial Theory of Voting (New York: Cambridge University Press, 1984)Google Scholar; and Fiorina, Morris P., Retrospective Voting in American National Elections (New Haven, Conn.: Yale University Press, 1981), esp. pp. 8–9.Google Scholar On the distinction between retrospective and prospective voting, see Fiorina, , Retrospective VotingGoogle Scholar; and Miller, Arthur H. and Wattenberg, Martin P., ‘Throwing the Rascals Out: Policy and Performance Evaluations of Presidential Candidates, 1952–1980’, American Political Science Review, 79 (1985), 359–72.CrossRefGoogle Scholar

² Fiorina, , Retrospective VotingGoogle Scholar, Chap. 1; Downs, , An Economic Theory of Democracy, Chaps. 3, 7, and p. 298Google Scholar; and Key, V. O. Jr, The Responsible Electorate (Cambridge, Mass.: Harvard University Press, 1966).CrossRefGoogle Scholar

³ Downs, , An Economic Theory of Democracy, p. 298.Google Scholar

⁴ On the two-step flow, see Lazarsfeld, Paul F., Berelson, Bernard and Gaudet, Hazel, The People's Choice (New York: Duell, Sloan, and Pearce, 1948), pp. 151–2.Google Scholar The re-alignment studies are: Carmines, Edward G. and Stimson, James A., ‘On the Structure and Sequence of Issue Evolution’, American Political Science Review, 80 (1986), 901–20CrossRefGoogle Scholar; and Macdonald, Stuart Elaine and Rabinowitz, George, ‘The Dynamics of Structural Realignment’, American Political Science Review, 81 (1987), 775–96.CrossRefGoogle Scholar

⁵ Downs, , An Economic Theory of Democracy, Chap. 6.Google Scholar

⁶ The data in the first two columns of Table 1 are reported in LeDuc, Lawrence and Murray, J. Alex, ‘A Resurgence of Canadian Nationalism’, in Kornberg, Allan and Clarke, Harold D., eds., Political Support in Canada: The Crisis Years (Durham, N.C.: Duke University Press, 1983), pp. 270–90, at p. 289.Google Scholar Data in the third column are reported in Irvine, William, ‘Epilogue: The 1980 Election’, in Penniman, Howard R., ed., Canada at the Polls, 1979 and 1980 (Washington, DC: American Enter prise Institute, 1981), pp. 337–98, at p. 374.Google Scholar Data for the fourth column are reported in McDougall, Gordon H. G. and Keller, Gerald, Energy: Canadians' Attitudes and Reactions (1975–1980) (Ottawa: Consumer and Corporate Affairs Canada), p. 17.Google Scholar Data for the final column are reported in Pammett, Jon H., ‘Elections’, in Whittington, Michael S. and Williams, Glen, eds, Canadian Politics in the 1980s, 2nd edn (Toronto: Methuen, 1984), pp. 271–86, at p. 273.Google Scholar

⁷ The percentage of respondents who mentioned energy as the most important issue in 1980 rose by a factor of 3.3 from 1979. See Clarke, Harold D., Jenson, Jane, LeDuc, Lawrence, and Pammett, Jon H., Absent Mandate: The Politics of Discontent in Canada (Toronto: Gage, 1984), p. 81, Table 4–1.Google Scholar

⁸ Clarke, et al. , Absent Mandate.Google Scholar

⁹ On Canadian parties as brokerage institutions, see Clarke, et al. , Absent Mandate, pp. 11, 13–15Google Scholar; on the link between partisanship and issue preferences, see Clarke, and Stewart, Marianne, ‘Short-Term Forces and Partisan Change in Canada: 1974–1980’, Electoral Studies, 4 (1985), 15–35.CrossRefGoogle Scholar For diverse findings on the impact of retrospective factors on party evaluations, see Clarke, Harold D. and Zuk, Gary, ‘The Politics of Party Popularity’, Comparative Politics, 19 (1987), 299–315CrossRefGoogle Scholar; Clark, Harold D., Stewart, Marianne and Zuk, Gary, ‘The Political Economy of Party Support in Canada, 1980– 1984’, European Journal of Political Economy, 2 (1986), 25–45CrossRefGoogle Scholar; and Monroe, Kristen and Erickson, Lynda, ‘The Economy and Political Support: The Canadian Case’, Journal of Politics, 48 (1986), 616–47.CrossRefGoogle Scholar

¹⁰ The simple correlation between the feeling thermometer for Clarke and a dummy variable for region was just 0.12. For studies on the effects of key variables on vote choice, see Clarke, et al. , Absent MandateGoogle Scholar; Clarke, Harold D. et al. , Political Choice in Canada (Toronto: McGraw-Hill Ryerson, 1978)Google Scholar; LeDuc, Lawrence, ‘Canada: The Politics of Stable Dealignment’, in Dalton, Russell J. et al. , eds, Electoral Change in Advanced Industrial Democracies (Princeton, NJ.: Princeton University Press, 1984), pp. 404–24Google Scholar; and Pammett, Jon H., ‘Class Voting and Class Consciousness in Canada’, Canadian Review of Sociology and Anthropology, 24 (1987), 269–90.CrossRefGoogle Scholar

¹¹ Berry, Glyn R., ‘The Oil Lobby and the Energy Crisis’, Canadian Public Administration, 17 (1974), 600–35CrossRefGoogle Scholar; and Toner, Glen and Doern, G. Bruce, ‘The Two Energy Crises and Canadian Oil and Gas Interest Groups’, Canadian Journal of Political Science, 19 (1986), 467–93.CrossRefGoogle Scholar

¹² Brady, David W., ‘Critical Elections, Congressional Parties and Clusters of Policy Change’, British Journal of Political Science, 8 (1978), 79–100.CrossRefGoogle Scholar

¹³ Sec Uslaner, Eric M. and Conway, M. Margaret, ‘The Responsible Congressional Electorate: Watergate, the Economy, and Vote Choice in 1974’, American Political Science Review, 79 (1985), 788–803.CrossRefGoogle Scholar

¹⁴ Clarke, et al. , Absent Mandate, p. 90.Google ScholarIrvine, , ‘Epilogue’Google Scholar, reports a Toronto Star poll which gave the Liberals a 3–2 advantage on energy issues.

¹⁵ The index was computed as follows. For both the most important and second most important issue in the election, a dummy variable was created for voters who cited the energy issue (coded ‘1’) and those who did not (coded ‘0’). Citizens who were scored ‘1’ were weighted by how important they said the issue was in their voting decisions (by three for those who said very important, by two for those who said somewhat important, and by one for those who said not very important). Respondents who did not give one of these responses were recoded as zero. These scores (ranging from one to three) were weighted by the party preferred on the issue. For the Liberal analysis, citizens preferring the Liberals were weighted by + 1 and those preferring other parties were weighted by − 1 (again with those who had no party preference re-coded to zero). A similar weighting was done for the Progressive Conservatives (+ 1 if the PCs were preferred, − 1 if some other party was preferred). Finally, the scores for each respondent on the two most important issues were summed, producing a range of − 6 to + 6 for each measure. The direct summation yielded superior results to alternative weighting schemes for the two most important issues. Fiorina, , Retrospective Voting, p. 70Google Scholar, also employs the most important issue as a measure of prospective voting. Such a measure is prospective in that voters have preferences about policy areas and can state the party with which they most agree on an issue. Such voters thus fulfil the key conditions for spatial voting, which is posited to be prospective, by Downs in An Economic Theory.

¹⁶ Technically, the problem is that instrumental variable estimators are neither unbiased nor efficient. However, they are consistent and asymptotically efficient. Thus, it is hardly surprising that there would be some difference in mean scores and that the standard errors would be larger than for the measured variables. For the ‘full sample’ analysed below (N = 1,051), the measured attitudes were employed for the first half sample and the estimated attitudes were employed for the second half sample. A critical test of the instrumental variable technique, but not the only one performed, compared the probit coefficients for the vote choice model for two half-samples (one with measured attitudes and the other with estimated attitudes on budget disapproval). For the Liberal vote choice equations, the probit coefficients estimated were 0.472 and 0.498 respectively. For the Progressive Conservative vote choice equations, the coefficients were −0.751 and −0.741 respectively. Clearly these results are more than satisfactory. The probit R² for the instrumental variable estimation was 0.405. For a similar approach, see Jackson, John E., ‘Issues and Party Alignment’, in Maisei, L. and Sacks, P. M., eds, The Future of Political Parties (Beverly Hills, Calif.: Sage Publications, 1975), pp. 101–21.Google Scholar A more detailed appendix on the methodology employed is available from the author.

¹⁷ See Doern, G. Bruce, ‘The Mega-Project Episode and the Formulation of Canadian Economic Development Policy’, Canadian Public Administration, 26 (1983), 191–238, p. 223.CrossRefGoogle Scholar

¹⁸ By the West, I include all four Western provinces (Alberta, British Columbia, Manitoba and Saskatchewan). Removing British Columbia from the West does not affect the results.

¹⁹ Both variables, of course, are highly correlated with the Quebec dummy (since Quebec has most of the French-speaking voters in Canada and is the most heavily Catholic province). The modest impact of Quebec residence in most of the equations to follow, however, provides further evidence that in these models of prospective and retrospective vote choice, the effects of language and religion are very modest.

²⁰ See McKelvey, Richard D. and Zavoina, William, ‘Statistical Analysis of Ordinal Level Dependent Variables’, Journal of Mathematical Sociology, 4 (1975), 103–30.CrossRefGoogle Scholar

²¹ Specifically, the predicted values for the dependent variable in a probit analysis are z (standardized) scores. Thus, each score can be placed on the cumulative normal distribution and associated with a probability value. The technique operates by assigning to the predictor in question its minimum value and estimating a mean probability for the sample from an equation in which all other variables vary freely. Then the same procedure is followed for the maximum value taken by that predictor. The mean probability for the variable at its minimum is subtracted from that for the predictor's maximum to give the marginal impact of the variable. The procedure is discussed in Raymond E. Wolfinger and Steven J. Rosenstone, Who Votes? (New Haven, Conn.: Yale University Press, 1981), and the specific algorithm for computing the probabilities was graciously provided by Steven J. Rosenstone.

²² All tests of significance reported herein are one-tailed. The addition of these prospective and retrospective variables does make a statistically significant difference for both the Liberal and PC equations. I also estimated probit equations for these two dependent variables without either energy-related variable. To test whether the addition of these variables is statistically significant, one simply subtracts – 2 Log Likelihood Ratio for the restricted model (with seven predictors) from that for the full model (with nine predictors). The resulting figure is distributed as x² with two degrees of freedom. For the PC equation, the difference in –2 Log Likelihood Ratios is 50.08; for the Liberals, the figure is 32.49. The critical value of x² at the 0.001 level is 13.82.

²³ In regression analysis, this is the familiar t-ratio.

²⁴ For a similar proportionate reduction in error measure, see Frohlich, Norman, Oppenheimer, Joe A., Smith, Jeffrey and Young, Oran R., ‘A Test of Downsian Voter Rationality’, American Political Science Review, 72 (1978), 178–97.CrossRefGoogle Scholar

²⁵ For similar findings in American Congressional elections, see Uslaner, and Conway, , ‘The Responsible Congressional Electorate’.Google Scholar

²⁶ Clarke, et al. , Absent Mandate, Chap. 5.Google Scholar

²⁷ Clarke, et al. , Politicai dioice in Canada, p. 81.Google Scholar

²⁸ In each case there were very few voters at the extreme (+ 6, −6) values for energy salience. The boundaries were thus adjusted to reflect cases with more than just one or two voters at these extremes. The relevant boundaries become (− 6, 4) for the Liberals, (− 4, 6) for the PCs, and (− 3, 3) for switchers. For estimating the joint effects of both energy salience and budget disapproval, the boundaries (− 3, 3) were employed.

²⁹ Many observers believed that the activist bent of the NEP was designed by the Liberals to attract NDP supporters. See, inter alia, Doern, Bruce G. and Toner, Glen, The Politics of Energy (Toronto: Methuen, 1985), p. 33.Google Scholar

³⁰ Is the budget-disapproval measure a valid one? The National Election Study asked voters what the buget meant to them. Only 12 per cent of the 725 respondents who gave answers to this question specifically mentioned something related to energy. When more general responses about effects on prices, business and low-income people are considered, the figure rises only to 23.1 per cent. However, virtually all other answers are general comments in favour of, or opposed to, the budget. No other specific comments were made in response to this question. Thus, budget evaluations do relate largely to the anticipated effects of the 18 cent tax to the extent that they reflect anything of substance. To the extent that the responses do not provide specific responses, the problem may lie more with the question asked than with voters' lack of knowledge about the 18 cent tax. The question about budget meaning was particularly vague.

³¹ Clarke, et al. , Absent Mandate, Chap. 6.Google Scholar

³² While the correlation was considerably stronger for PC party identification for all voters (−0.60), this is not surprising. As noted above, retrospective evaluations tend to punish the governing party more than they reward the party out of power.

³³ Clarke, et al. , Political Choice in CanadaGoogle Scholar, esp. Chap. 7, also examine the role of political interest, but in a very different context from that employed here.

³⁴ Clarke, et al. , Absent Mandate, p. 90.Google Scholar

³⁵ For an important statement of such results in an American context, see Kernell, Samuel, ‘Presidential Popularity and Negative Voting’, American Political Science Review, 71 (1977), 44–66.CrossRefGoogle Scholar

³⁶ In contrast, both parties mobilized virtually all of their supporters on the energy salience measures.

³⁷ Doern, and Toner, , The Politics of Energy, p. 33.Google Scholar

³⁸ Of the total of 33.8 per cent, 0.9 per cent cited energy responses as both the first and second most important issues. Virtually identical percentages of the voters analysed above and of the full weighted sample of 1,770 (24.4, 9.2 per cent) cited energy as the first and second most important issue.

³⁹ Miller, and Wattenberg, , ‘Throwing the Rascals Out’, p. 370.Google Scholar

⁴⁰ Fiorina, , Retrospective Voting, pp. 5–6.Google Scholar