¹ Upon being victorious in the October 1972 elections in British Columbia, the New Democratic party announced that its first bill to be submitted to the legislature will severely restrict political advertising outlays.

² This phrasing is carefully chosen not to include a vast amount of research done on the influence of mass media on intentions to vote.

³ A recent brief article calculated Spearman rank correlation coefficients between total campaign expenditures and votes cast in the 1970 British elections. It is not clear whether a distinction can be made in Britain between advertising and other electoral expenses and if the major part of such expenses is captured by the reporting requirements. See Taylor, A.H., “The Effect of Party Organization: Correlation between Campaign Expenditure and Voting in the 1970 Election,” Political Studies (September 1972), 329–31.Google Scholar

⁴ I would like to thank Professor Paltiel of Carleton University for drawing my attention to this fact.

⁵ For different views see Milbrath, Lester, Political Participation (Chicago, 1965)Google Scholar, chap. 4.

⁶ I refer here to the Dorfman-Steiner model and its later market-share versions, particularly that of Lambin. See Dorfman, Robert and Steiner, Peter O., “Optimal Advertising and Optimal Quality,” American Economic Review (December 1954), 826–36Google Scholar; Lambin, Jean-Jacques, “Optimal Allocation of Competitive Marketing Efforts,” Journal of Business (October 1970), 468–84Google Scholar. Economists working in political theory envisage the electoral process in terms of spatial competition rather than as set forth above. See Otto Davis, Melvin J. Hinich, and Ordeshook, Peter C., “An Expository Development of A Mathematical Model of the Electoral Process,” American Political Science Review (June 1970), 426–48Google Scholar. I have been unable to find anything corresponding to my approach in the latest survey of economic analysis applied to political phenomena, Attali's, JacquesAnalyse économique de la vie politique (Paris, 1972).Google Scholar

⁷ Voter turnout as related to campaign intensity is the subject of a study I am now undertaking.

⁸ The economist designates as “tastes” all demand – influencing variables except those that can be expressed in monetary terms: prices, income, and, conceivably, product quality.

⁹ The meaning of the ratios is left deliberately vague; it will be sharpened further on.

¹⁰ For scholarly analyses of the 1966 and 1970 elections see Quatre élections provinciates an Québec, ed. Lemieux, Vincent (Quebec, 1969)Google Scholar; Lemieux, V., Gilbert, M., Blais, A., Une élection de réalignement (Montreal, 1970).Google Scholar

¹¹ Loi électorate du Québec, 11–12 Elizabeth II, 1963, ch. 13 and subsequent amendments (ch. 12 Statuts du Québec, 1965 and ch. 5 Statuts, 1966).

¹² Except for category 1 and 6, the expenses are only identified by title. The examples are taken from a term paper prepared for Vincent Lemieux at Laval University in December, 1971. The student group who authored the paper obtained these examples from the chief returning officer of the province.

¹³ Unless otherwise stated all data were taken from the 1966 and 1970 Rapports du président général des élections (Québec, no dates).

¹⁴ See Mellos, Koula, “Quantitative Comparison of Party Ideology,” this Journal, (December 1970), 540–58Google Scholar. An economist would probably choose to tackle the problem of quantifying campaign issues with the help of an analytical scheme that goes under the name of hedonic price index. For its most powerful use to date see Cowling, Keith and Cubbin, John, “Price, Quality, and Advertising Competition,” Economica, (November 1970), 378–84.Google Scholar

¹⁵ Annuaires du Québec, 1966, 1970, (“Gouvernement et administration”).

¹⁶ Brome, D'Arcy McGee, Jacques Cartier, Marguerite Bourgeoys, Notre-Dame-de-Grâce, Ste-Anne, Robert Baldwin, West-mount.

¹⁷ Department of National Revenue, Taxation Statistics for 1966 and 1970. Total reported income was divided by the number of the district's electors.

¹⁸ Stigler, in a recent paper, did not, however, detect any relationship between unemployment levels and outcomes of US congressional elections. See Stigler, George J., “General Economic Conditions and National Elections,” American Economic Review, (May 1973), 160–7Google Scholar. The fragmentary evidence I have seen points equally to no relationship between party electoral performance in a district and that district's unemployment rate.

¹⁹ The definition of this variable suffers somewhat from the fact that new ridings were created. In the 1944, 1948, 1952, and 1956 elections there were 94 ridings; in 1960 and 1962 there were 96 ridings.

²⁰ Dummy variables, however, and the urban-rural variable have not been used in log form.

²¹ I am grateful to Paul H.S. Lau for the excellent computing assistance he has furnished me.

²² An F=2 value “to remove” was used in the stepwise regression. Its precise equivalent in terms of a t-statistic for regression coefficients changes with sample size and number of conditioning variables. Suffice it to say that here F=2 is the equivalent of 1.5 < t < 1.6.

²³ Durbin-Watson tables do not list values for samples larger than 100 observations. The D-W test is most commonly employed to find evidence of autocorrelation. For references on cross-sectional employment see Christ, Carl F., Econometric Models and Methods (New York, 1966), 527Google Scholar. Obviously, we are testing for homoskedasticity against the advertising variable only. An excellent summary of recent literature on this problem is Aigner, Dennis J., Basic Econometrics (Englewood Cliffs, N.J., 1971), 124–31.Google Scholar

²⁴ This statement must be interpreted with a grain of salt. Budgetary reasons prevented me from making an unexceptional comparison between linear and logarithmic R²'s, which requires that the residuals of the log regressions be first converted to anti-logs before calculating a comparable R² statistic.

²⁵ From here on the sample sizes are N = 108, corresponding to the number of electoral districts.

²⁶ At this stage it must be acknowledged, however, that these remarks are speculative. Note that in 1966 RIV A DP is significant at the 5 per cent level only if we postulate a priori that its sign is negative. If no assumption is made about its sign, the critical t-value for the 5 per cent level rises to 1.96. Ongoing research devoted to the exploration of impact of campaign expenditures on abstentionism does, however, show in preliminary results that campaign intensity influences positively voter participation.

²⁷ Schmalensee, Richard, The Economics of Advertising (Amsterdam, 1972), 104–12.Google Scholar

²⁸ I have examined this phenomenon in a commercial context in The Measurement of Cumulative Advertising Effects (Englewood Cliffs, N.J., 1964).

²⁹ The average plurality with non-augmented advertising, WAD, is 1095 votes.

³⁰ Lambin, Jean-Jacques, “Is Gasoline Advertising Justified?,” Journal of Business (October 1972), 585–619Google Scholar; Telser, Lester G., “Advertising and Cigarettes,” Journal of Political Economy (October 1962), 471–99.Google Scholar

³¹ As was mentioned, a variable called YEL, income per elector, was available in 69 of the 108 ridings. When employed in the step-wise (2a) regressions, its coefficient had no statistical significance whatever. The tentative conclusion is therefore that this “taste” characteristic did not affect the outcome of the Quebec elections.