1 For a dispute on this question, see Cairns, Alan, “The Electoral System and the Party System in Canada, 1921–1965,” this Journal 1 (1968), 55–80Google Scholar; Lovink, J. A. A., “On Analyzing the Impact of the Electoral System on the Party System in Canada,” this Journal 3 (1970), 497–516Google Scholar; and Cairns's reply to Lovink, ibid., 517–20. Cairns's original article is, of course, the point of departure for the arguments in this paper.

2 Cairns, “The Electoral System,” 60–63. The logic of exaggerated regional differences is exactly the same as the logic of two other effects of single-member plurality systems: the disproportionate number of seats given parties with relatively large proportions of the popular vote and the exaggeration of swings from election to election. For the parameters of the first effect, see Rae, Douglas, The Political Consequences of Electoral Laws (New Haven and London: Yale University Press, 1967)Google Scholar, chaps. 4 and 5. For the second effect, see ibid., Table 5.2, and Tufte, Edward R., “The Relationship Between Seats and Votes in Two-Party Systems,” American Political Science Review 67 (1973), 540–54.CrossRefGoogle Scholar

3 Cairns, “The Electoral System,” 64–72.

4 Ibid., 59–60.

5 Ibid., 65. See as well Meisel, John, The Canadian General Election of 1957 (Toronto: University of Toronto Press, 1962), 167–68.Google Scholar

6 The variance and its square root, the standard deviation, are both quite sensitive to the party's size. The smaller a party, the smaller will be its standard deviation, almost regardless of the degree to which its support is sectionally concentrated. The coefficient of variation corrects for this problem as it is calculated as the ratio of the standard deviation (in this case, of the party's vote across geographic units) over the size of the party. The coefficient expresses the variation as a proportion of the party's size. Some may feel uncomfortable with a variance measure taken as one of “concentration” rather than of “dispersion.” Usually, one thinks of the variance as an indicator of the dispersion of a distribution across the range of possible values. The larger the variance, the more dispersed across the range the distribution is. The usage with geographic data is almost exactly opposed to the more common usage. The more dispersed over space (rather than over the range of possible values) is some attribute, such as a party's vote, the smaller are the differences between geographic units and, thus, the smaller is the geographic variance. Conversely, the more concentrated in one or a few geographic units the attribute is, the larger must be the difference between units and, thus, the larger must be the geographic variance. On the vagaries of areal measures and statistics, see Duncan, Otis Dudley, Cuzzort, Ray P., and Duncan, Beverly, Statistical Geography (Glencoe: Free Press, 1961).Google Scholar

7 This is the measure used in Cairns. “The Electoral System,” Table 11.

8 For elegant examples of the use of regression to explore aspects of the electoral system, see Rae, The Political Consequences of Electoral Laws, chap. 4; Tufte, “The Relationship Between Seats and Votes”; and Spafford, Duff, “The Electoral System of Canada,” American Political Science Review 64 (1970), 168–76.CrossRefGoogle Scholar Spafford did not consider, or at least did not report the consideration of, the distribution of the vote across provinces. He did, however, find a significant effect on the party's seat total of the distribution of its vote across constituencies of different sizes.

9 For the elections from 1921 to 1958 inclusive our data are taken from Scarrow, Howard A., Canada Votes (New Orleans: Hauser, 1962).Google Scholar The data for elections from 1962 to 1974 inclusive come from the Chief Electoral Officer's reports. From 1921 to 1974 inclusive there have been seventeen general elections. Of course, the Liberal and Conservative parties have contested all seventeen. The Progressive party contested the four elections from 1921 to 1930 inclusive. The CCF-NDP and some version of Social Credit have contested all thirteen elections from 1935 to 1974 inclusive. We have treated Social Credit and the Ralliement des Créditistes as a single party. Both Cairns and Spafford begin their studies with the 1921 election. That election produced a veritable quantum jump in the fractionalization of the Canadian party system. Estimates of the parameters of effects on the seat-to-vote ratio are probably not transferable from the post-1921 party system to the system in place before 1921.

10 The regression coefficient on X is unstandardized. The bracketed figure below the coefficient is the standard error of the coefficient. As a rule of thumb, a coefficient cannot be considered significantly different from zero unless it is at least twice its standard error.

11 The argument breaks down as the party's popular vote total exceeds that necessary to win a bare plurality in each and every constituency.

12 The point defined by b₁ / 2b₂Z_ij is that at which Y is a maximum. The point is found by the partial differentiation of the seat-vote function with respect to X_ij, holding Z_ij constant:

setting the differential equation equal to zero, and solving for X_ij. This yields:

13 The coefficients are significant in spite of the considerable multicollinearity among the independent variables. The correlations between the independent variables are:

The national popular vote and the geographic concentration of the vote are related to each other more closely than each is to the interaction term. On the question of multicollinearity, see Johnston, J., Econometric Methods (New York: McGraw-Hill, 1972), 159ff.Google Scholar

14 Toronto, Globe, June 3, 1882.Google Scholar This is taken from the epigraph to Dawson, R. MacGregor, “The Gerrymander of 1882,” Canadian Journal of Economics and Political Science 1 (1935), 197–221.CrossRefGoogle Scholar For the pattern of redistribution in the 1882 case, see especially 210 and Appendix B.

15 Morton, Desmond, “Polling the Soldier Vote: The Overseas Campaign in the Canadian General Election of 1917,” Journal of Canadian Studies 10 (1975), 43–44.CrossRefGoogle Scholar The number of soldiers actually eligible to assign their votes at will was not large. As it happened, many reassigned their votes illegally, but had their votes discarded as a result. See ibid., 48, 51.

16 See Munro, John M., “Highways in British Columbia: Economics and Politics,” Canadian Journal of Economics 8 (1975), 192–204CrossRefGoogle Scholar, and Blake, Donald E., “LIP and Partisanship: An Analysis of the Local Initiatives Program,” Canadian Public Policy 2 (1976), 17–32.CrossRefGoogle Scholar For an argument in the same spirit, but with respect to policy benefits distributed in response to the slightly more complex incentives embodied in the American Electoral College, see Wright, Gavin, “The Political Economy of New Deal Spending: An Econometric Analysis,” The Review of Economics and Statistics 56 (1974), 30–38.CrossRefGoogle Scholar The strategy described in the text and illustrated by Munro, Blake, and Wright corresponds to the “economic” approach described in Lovink, “On Analyzing the Impact,” 502–04. Lovink suggests that the strategy of reinforcing strength be called the “patronage” strategy.

17 However much Churchill may have urged the party to cut its losses, the 1953–1957 shift in popular response to the party reflected no such strategy. As it grew from 1953 to 1957, the party's popular vote became less differentiated geographically. The party's vote declined in no region and displayed the greatest proportionate growth where the party had hitherto been weakest—on the prairies. In any case, according to Meisel, Churchill's advice to his party was more ambiguous than Cairns argued it to be (General Election, 167–68).

18 The distinction between functional and territorial cleavages is made in Lipset, Seymour Martin and Rokkan, Stein, “Cleavage Structures, Party Systems, and Voter Alignments: An Introduction,” in Lipset, and Rokkan, (eds.), Party Systems and Voter Alignments (New York: Free Press, 1967), 1–64.Google Scholar

19 For recent examples, see Stewart, Walter, Divide and Con: Canadian Politics at Work (Toronto: New Press, 1973).Google Scholar The classic instance of mutually contradictory appeals must surely be those made by the Conservative party in 1911. The party's vote gained in English Canada on the basis of an Imperialist appeal. In French Canada, the party gained on the basis of an anti-Imperialist one. The irony is bitter even today.

20 Electoral system incentives may, however, explain some of the departure of the NDP from class-based appeals. See, for example, Stewart, Divide and Con, chap. 4, especially 89–90, and chap. 11.