1 Kiewiet, Roderick D. and Rivers, Douglas, “A Retrospective on Retrospective Voting,” Political Behavior 6 (1984), 375.CrossRefGoogle Scholar For a comparison of results from US and Canadian aggregate analysis, see Happy, J. R., “Voter Sensitivity to Economic Conditions: A Canadian-American Comparison,” Comparative Politics 19 (1986), 45–56.CrossRefGoogle Scholar

2 See Monroe, Kristen and Erickson, Lynda, “The Economy and Political Support: The Canadian Case,” Journal of Politics 48 (1986), 622–23CrossRefGoogle Scholar; and Clarke, Harold D. and Zuk, Gary, “The Politics of Party Popularity: Canada, 1974–1979,” Comparative Politics 20 (1987), 311.Google ScholarJohnston, Richard reports that unemployment has an effect on prime ministerial approval (Public Opinion and Public Policy in Canada: Questions of Confidence, Research Studies of the Royal Commission on the Economic Union and Development Prospects for Canada, vol. 35 [Toronto: University of Toronto Press for Supply and Services Canada, 1986], 60–62Google Scholar) while Archer, Keith and Johnson, Marquis report a prospective voter orientation for this variable (“Inflation, Unemployment and Canadian Federal Voting Behaviour,” this Journal 21 [1988], 575Google Scholar).

3 Monroe and Erickson, “The Economy and Political Support,” 642; and Johnston, Public Opinion and Public Policy, 60–64.

4 Kiewiet and Rivers, “A Retrospective on Retrospective Voting,” 370.

5 Clarke, Harold D., Jenson, Jane, Leduc, Lawrence, and Pammett, Jon H., Absent Mandate: The Politics of Discontent in Canada (Toronto: Gage, 1984).Google Scholar

6 Peffley, Mark, “The Voter as Juror: Attributing Responsibility for Economic Conditions,” Political Behavior 6 (1984), 275–93.CrossRefGoogle Scholar

7 Kramer, Gerald H., “The Ecological Fallacy Revisited: Aggregate Versus Individual Level Findings of Economics and Elections and Sociotropic Voting,” American Political Science Review 77 (1983), 92–111.CrossRefGoogle Scholar

8 The data for party choice were obtained from Beck, J. M., Pendulum of Power (Englewood Cliffs: Prentice-Hall, 1968Google Scholar) for the elections of 1926 through 1968; for 1972, 1974 and 1979, the data were obtained from the relevant Report of the Chief Electoral Officer. The economic data were obtained from the following Statistics Canada sources, catalogue numbers in parentheses. Data for personal and disposable income were calculated from the National Accounts: Income and Expenditure, selected years, Cansim Matrix 000581 (13–201). The inflation series was calculated from the major cities index. For the period 1926 through 1940 it was computed from Price and Price Indexes (61–501); for 1940 through 1975 from Historical Statistics of Canada (2nd ed., 1983, CS11–516E); and for the period 1975 through 1979 from the Consumer Price Index (62–001). The conversion factors for the inflation series, as well as the provincial-level unemployment data, were obtained directly from Statistics Canada.

Disaggregation of the variables to the provincial level produces a nominal sample size in excess of 100 observations. Without disaggregation the sample size would be 15 elections, insufficient for multivariate analysis. The nominal pooled sample size is 130. However, the major cities consumer price index is not available for the four western provinces for the 1930 election, for Alberta and British Columbia for 1935 and 1940, or for Newfoundland for 1953. Unemployment figures are not available on a provincial basis before 1951. These missing observations have the effect of reducing the effective sample size to 90 for purposes of hypothesis-testing.

9 Lau, Richard R., “Two Explanations for Negativity Effects in Political Behavior,” American Journal of Political Science 29 (1985), 119.CrossRefGoogle Scholar

10 Thurow, Lester C., The Zero-Sum Society (New York: Basic Books, 1980), 49.Google Scholar

11 The appropriateness of OLS estimation procedures was examined as follows. Tests for first- and second-order auto-correlation were conducted for each of the equations reported below following Durbin's alternative procedure for the inclusion of lagged endogenous variables. See Durbin, J., “Testing for Serial Correlation in Least Squares Regression When Some of the Regressors are Lagged Dependent Variables,” Econometrica 38 (1970), 420–21.Google Scholar In all cases the hypothesis of auto-correlation could be safely rejected. The “t” values for the first- and second-order residual coefficient estimates for equation (5) are 1.12 and 0.23 and for equation (6), 1.16 and 0.23. In addition, provincial dummy variables were included to test for unit homogeneity. The estimates for the economic variables are unaffected by this procedure, and British Columbia is the only province with a significant dummy coefficient. The insignificance of the remaining provincial dummy variables is probably a consequence of using percentages throughout the analysis. The results of the regression using dummy variables in equation (5) are the following coefficients for each variable (with corresponding “t” values given in parentheses): DISPINC, .58 (3.37); (V)DISPINC,.01 (2.34); INFL, −.30 (1.89); VIL,.46 (4.71); NEW BRUNSWICK,.24 (0.08); QUEBEC, 2.55 (0.80); ONTARIO, −1.25 (0.39); MANITOBA, −4.12 (1.30); SASKATCHEWAN, −2.30 (0.62); ALBERTA, −4.57 (1.26); BRITISH COLUMBIA, −6.75 (1.94); NEWFOUNDLAND, −0.13 (0.04). Following Stimson, James A., “Regression in Space and Time: A Statistical Essay,” American Journal of Political Science 29 (1985), 921–25CrossRefGoogle Scholar, these results suggest the appropriateness of OLS estimation procedures with a lagged dependent variable and without provincial dummy variables.

12 The positive sign on the unemployment coefficient is not due to multi-collinearity. Although inflation and unemployment are slightly correlated in the sample (r = .17, significant at the .06 level), removing unemployment and inflation alternatively from the specification does not affect the results. The unemployment and nominal variables are uncorrelated in the sample.