Hostname: page-component-848d4c4894-mwx4w Total loading time: 0 Render date: 2024-06-15T17:42:39.166Z Has data issue: false hasContentIssue false

Post-dicting Electoral Cleavages in Canadian Federal Elections, 1949–68: Material for a Footnote

Published online by Cambridge University Press:  10 November 2009

J.A. Laponce
Affiliation:
University of British Columbia

Extract

The first systematic correlation of socio-demographic factors and party preference in Canadian federal elections was made by Escott Reid in 1933. No expansion or even replication followed until the mid-1950s when John Meisel, using data obtained by sample survey of individuals, continued the enquiry begun by Reid on the basis of aggregate statistics. Since then – especially after 1960 – many data have been collected and some studies made on the links between party preference and factors such as religion, language, region, social class, and age. But these studies, even for the variables that have received most attention – religion, region, and class, for example – give us findings which remain scattered over time and space, like building materials on different construction sites. We need an overall view of the relationship between socio-demographic factors and party preference – if only to facilitate the interpretation of more refined but narrowly focussed studies; and we need such a view to extend over more than one or two elections.

Type
Notes
Copyright
Copyright © Canadian Political Science Association (l'Association canadienne de science politique) and/et la Société québécoise de science politique 1972

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 “Canadian Political Parties: A Study of the Economic and Racial Bases of Conservatism and Liberalism in 1930,” Contributions to Canadian Economics, VI (1933), 7–39.

2 “Religious Affiliation and Electoral Behaviour: A Case Study,” Canadian Journal of Economics and Political Science, XXII (1956), 481–96.

3 For a bibliography of the literature on parties and electoral behaviour, see Van Loon, Richard J. and Whittington, Michael S., The Canadian Political System (Toronto, 1970).Google Scholar

4 This paper is the result of a classroom exercise. I am grateful to the students who were involved in its preparation and in particular to I. Bernstein, A. Katona, R. Keenlyside, and J. Twigg. For her help with programming and the checking of the data, we are indebted to Pamela Smortchevsky.

5 For an electoral analysis using subprovincial units of analysis, see Copes, P., “The Fishermen's Vote in Newfoundland,” this JOURNAL, III, no. 4 (Dec. 1970), 579604.Google Scholar

6 For studies using aggregate statistics applied to variables similar to those analysed here, see Wilson, J.M.The Use of Aggregate Data in the Analysis of Canadian Electoral Behaviour,” paper presented at the CPSA Conference on Statistics, Ottawa, 1967Google Scholar; Laponce, J.A., “Ethnicity, Religion, and Politics in Canada: A Comparative Analysis of Survey and Census Data,” in Dogan, M. and Rokkan, S., eds., Quantitative Ecological Analysis in the Social Sciences (Boston, 1969)Google Scholar; and Blake, D. E., “The Measurement of Regionalism in Canadian Voting Patterns,” this JOURNAL, V, no. 1 (March 1972), 5581.Google Scholar

7 The CIPO surveys analysed here are those which have been deposited at the Roper Center in Williamstown. Description of this data set is in Schwartz, Mildred A., Public Opinion and Canadian Identity (Berkeley and Los Angeles, 1967), 53 ff.Google Scholar This work offers a rare systematic study of consensus/dissensus in the Canadian electorate. 8 Because of the small number of cases and because of the party split of 1963 between Social Credit and Créditistes, the cleavages between that movement and the other parties will not be systematically reported in this paper. Whenever the term Social Credit appears in the tables or footnotes, it refers to the combined Social Credit/Créditiste electorate.

9 The fir st small scale election survey of individuals made in Canada by an academic researcher (Meisel, “Religious Affiliation and Electoral Behaviour”) followed the first commercial survey by more than ten years. In the United States the Survey Research Center has surveyed each presidential election since 1948 and has made the data available through the Inter-University Consortium for Political Research.

10 These figures, as well as those measuring party support by province, are based on Beck, M., Pendulum of Power (Scarborough, Ont., 1968).Google Scholar

11 If Independents and splinter parties are included, the grand total is 56,806,884 votes cast, of which 41.8 per cent are Liberal, 36.1 per cent Conservative, and 13.5 per cent CCF/NDP.

12 When two parties appear in a single cell, each scores one-half.

13 This criteria remains hard to meet. It eliminates variables which random variations or exceptional elections would have pushed temporarily under the party's average; it eliminates Newfoundland from the Liberal column, for example. The choice of criteria was determined by our overall tactic which is to be left, at the end of our search, with the sole variables which remained stable over the twenty years considered.

14 For the interpretation of this and subsequent findings, it is important to know that the SES of the respondent is determined subjectively by the interviewer and that the category SES A describes the top 3 to 4 per cent of the electorate. Using a more liberal definition of upper class by adding the wealthy to the very wealthy would likely shift the political preference of the group from Conservative to Liberal. In his survey of the 1968 election, John Meisel finds that the group he defines as upper class and which comprises the top 8 per cent of the social hierarchy is more likely to vote Liberal and less likely to vote Conservative than the electorate as a whole. See Meisel, “The Basis of Support in the 1968 election,” unpublished.

15 According to our criteria of stability, the only predictor to survive the Social Credit shift from an English to a French base is “primary education.” However, the small number of Social Credit/Créditiste respondents makes the observation subject to confirmation.

16 See Sonquist, John A., Multivariate Model Building: The Validation of a Search Strategy (Ann Arbor, Mich., 1970).Google Scholar

17 The measuring of the independent effect of the variables studied here is the object of a forthcoming study by the author and Russell Uhler.

18 In order to relate the position of all three parties on all the “stable” factors we need to reintroduce some of the non-stable party positions which we had decided to ignore. Only four of those, however (those indicated by a parenthesis on Figure 4) have been subject to wide fluctuations over the eight elections. The reader may wish to ignore these deviant cases or may prefer to locate them at the “o” neutral point. A wide fluctuation is defined as a crossing of the dividing line on Figure 4 by more than ten points to the side other than that of the average in at least one election.

19 Overall indices are most useful when they identify a balance point around which known and recognizable factors are grouped; they are most disturbing when they mash and hide these factors beyond recognition. Mashing variables is dangerous and amusing.

20 For descriptions of this technique, see Guttman, L., “A General Nonmetric Technique for Finding the Smallest Coordinate Space for a Configuration of Points,” Psychometrika, XXXIII (1968), 469506;CrossRefGoogle ScholarLingoes, J.C., “New Computer Developments in Pattern Analysis and Non-metric Techniques,” in Uses of Computers in Psychological Research (Paris, 1966), 122;Google ScholarLingoes, J.C., “The Multivariate Analysis of Qualitative Data,” Multivariate Behavioral Research, III (1968), 6194CrossRefGoogle Scholar; and Bloom-Baum, M., “Doing Smallest Space Analysis,” Journal of Conflict Resolution, III (1970), 409–16.CrossRefGoogle Scholar

21 The missing data on religion for the years 1949, 1953, and 1958 has not been replaced. In those years religion is thus not a factor. I tested two alternatives: (a) replacing the missing data by an average trend value, and (b) replacing it by values obtained from the average ratio of French/Catholic and English/Protestant indices in the other elections. The results were basically similar to those presented here, the difference being in greater stability and closer clustering. For a computer simulation of the effects of alternative ways of dealing with missing data, see Mackelrang, A., Rodgers, H., and Klassen, D., “Missing Data: Some Implications for Multivariate Analysis,” University of Iowa, Laboratory for Political Research, no 27, 1968.Google Scholar

22 The difficulty of reproducing three dimensions on paper explains our using only two; we are further justified in doing so by the fact that the coefficients of alienation are always under .10 in two dimensions (.15 is considered an acceptable level of error) and by the fact that the three dimensional data, though of course more precise, would not have led us to different conclusions.

23 This can be verified by measuring the distance between the three electorates at each election time. For example, the distances between NDP, Liberals, and Conservatives measured in centimetres on the original print out of Figure 5(e) are as follows (in chronological order): between Conservatives and Liberals 10, 6, 14.5, 17, 18.5, 19.5, 18.5, 17.5; between Liberal and NDP 22.5, 23, 22.5, 20, 24, 23.5, 22.5, 24.5; between NDP and Conservatives 23, 23, 22, 22, 15.5, 17.5, 19.5, 20.5. Obviously there are deviant cases (Conservative-Liberal in 1949; NDP-Conservative in 1953) but no emergence of stable new relationships over time. Added together these three distances give an overall measure of stability which shows a very slight tendency for the electorates to be more contrasted at the end of the period. The total distances are: 55.5, 52, 59, 59, 58, 60.5, 60.5, 62.5.

24 Admittedly the fourteen structures presented on Figure 4 and the five presented on Figure 5 are only a small proportion of the 16.852 possible combinations offered by fourteen variables. The five solutions of Figure 5 were not, however, selected arbitrarily; we were led to them by the logic of our preceding analysis which consisted in eliminating the less stable single discriminators.

25 For the greater importance of the factors “religion,” “class,” and “urban/rural,” especially the first two, when compared to sex and age, see Arendt Lijphart, “Class Voting and Religious Voting in the European Democracies,” University of Strathclyde, occasional paper no 8: see in particular Table I which gives indices for Great Britain, Italy, West Germany, the Netherlands, Belgium, France, Austria, Sweden, and Norway.

26 See Lipset, S.M. and Rokkan, S., Party Systems and Voter Alignments (New York, 1967).Google Scholar