¹ The Winnipeg electoral constituency represented the city of Winnipeg proper rather than the whole greater Winnipeg area. The constituency included about three-quarters of the total number of voters in the greater Winnipeg area.

² Winnipeg Free Press and Winnipeg Tribune, 24 September 1945.

³ lpp platform positions are to be found in the Tribune, 15 and 29 September 1945 and the Free Press, 10 October 1945. The spc's platform called for “the establishment of a system of society in which all the means of production and distribution will be owned and controlled by all the people and operated for no other purpose than to satisfy human needs.” Quoted in the Tribune, 18 September 1945.

⁴ Free Press, 10 October 1945. Stubbs, 12 years earlier, was an unsuccessful ccf candidate in Saskatchewan's federal Mackenzieriding. In 1921 he ran as a federal Liberal in Marquette (Manitoba) against Progressive leader T.A. Crerar. These inconsistent political patterns paralleled themselves in Stubb's transferable ballots as Table iv shows.

⁵ Free Press, 8 September 1945.

⁶ The Tribune's error and the Free Press's accuracy is confirmed by independent and separate data dealing with the numbers of eligible voters in each poll. The Manitoba Election Record, 1923–57 (Provincial Library of Manitoba) published by the chief returning officer provides official poll-by-poll figures for every constituency in 1945 except Winnipeg. Reliance on newspaper figures, therefore, was unavoidable. We believe, however, that the Free Press figures are accurate since the total vote figure and the total votes for each candidate coincide with the official Result Sheet, a summary of results certified correct by the returning officer. This Result Sheet is reproduced, in percentage terms, in Table iv below.

⁷ Ottawa, 1949.

⁸ Wage earners are “that sector of the gainfully occupied population working for salary, wages, commission, or piece-rate forms of payment” (ii, xiv). Earnings include the above listed forms of payment but not “income from pensions, investments, workman's compensation, relief and other sources of this kind” (ii, xix). Of Winnipeg's 103,804 residents who were 14 years of age and over and gainfully employed 91,113 were wage-earners (ii, 44 and 214).

⁹ Years of schooling “is the total number of school years attended by the individual” (i, xx). This is not a reference to level of education but, rather, to the duration of time of going to school.

¹⁰ Mother tongue is “the first language learned in childhood if still understood by the person” (i, xx).

¹¹ Census of Canada, 1941 (iv, 266–7). The raw figures are: 148,899 English mother tongue, 129,478 “British Isles Races”; 15,340 Yiddish mother tongue, 17,027 Jewish; 21,027 Ukrainian mother tongue, 22,578 Ukrainian.

¹² Winnipeg's population grew steadily and slowly before and after the 1945–6 period. The 1941 census gave a population figure of 221,960. In 1946 this rose to 229,045 and by 1951 it was 235,710.

¹³ Census of the Prairie Provinces, 1946. Detailed information on the methods of handling the armed service personnel can be found in the Administrative Report of the Dominion Statistician (Ottawa, 1945), 93–4.

¹⁴ If other class indicators, such as the number of wage earners, are added, class variables can explain virtually all the variation in the Left vote. Other class indicators were excluded here mainly for simplicity of exposition, and because their effects are marginal.

¹⁵ Since the units of analysis in this study are areal units rather than individuals, our correlations are what Robinson, W.S., in “Ecological Correlations and the Behavior of Individuals,” American Sociological Review, 15 (June, 1950) 351–7CrossRefGoogle Scholar, has referred to as ecological correlations. In other words our correlations are between attributes aggregated within the areal units, rather than correlations between the attributes of individuals themselves. Since the population of voters within any social area is smaller than the total population measured by the census, the possibility of artifactual relationships arises. Close study of the data has convinced us that no important artifacts of this sort have occurred. We have measured a sufficient number of variables so that their effects will be minimal in the multiple regressions. The remarkable consistency of the relationship between the class, ethnic, and voting variables is sufficient evidence against artifactuality in these relationships. For extended methodological discussion and evaluation of survey versus ecological techniques see Dogan, Mattei and Rokkan, Stein, Quantitative Ecological Analysis in the Social Sciences (London, 1969Google Scholar), especially articles 5 and 8.

¹⁶ For stylistic reasons the prefix “percentage” is omitted, but understood, in subsequent references.

¹⁷ Revised Statutes of Manitoba, 1940, i, chap. 57.

¹⁸ The quota for election was the total number of valid ballots cast divided by a number exceeding by one the number of members to be elected and that result being increased by one. In others words: (79433/11) + 1 = 7222.

¹⁹ Only 3396 of 79,433 ballots were non-transferable, about 4.3 per cent.

²⁰ London, 1964.

²¹ Ibid., 109.

²² Ibid.

²³ Ibid., 250.

²⁴ Ibid., 257.

²⁵ Ibid., Table 9–1, 264.

²⁶ Ibid., Table 9–3, 268.

²⁷ Ibid., 79–80.

²⁸ Ibid., Table 3–2, 70, and n. 11, 72.

²⁹ “Some Problems in the Measurement of Class Voting,” in American Journal of Sociology, 78, no. 3 (November, 1972), 627–42.

³⁰ party and Society, Table 9–3, 268.

³¹ Census of the Prairie Provinces, 1946, ii, 3.

³² Ibid., ii, 317 and 603.

³³ The regression and correlation computations were done with Nie, Norman H., et al., Statistical Package for the Social Sciences (New York, 1970Google Scholar). The regression routine in spss utilizes a stepwise procedure that allows variables to be “forced” into the regression equation for control purposes. We did not attempt to change the two default criteria for inclusion of variables in the regression equation employed by the spss routine, an 0.05 F-test criterion and a tolerance-level criterion.