Hostname: page-component-7c8c6479df-24hb2 Total loading time: 0 Render date: 2024-03-28T18:41:58.084Z Has data issue: false hasContentIssue false

The Measurement of Regionalism in Canadian Voting Patterns*

Published online by Cambridge University Press:  10 November 2009

Donald E. Blake
Affiliation:
University of British Columbia

Abstract

Image of the first page of this content. For PDF version, please use the ‘Save PDF’ preceeding this image.'
Type
Articles
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 See, for example, Engelmann, F. C. and Schwartz, M., Political Parties and the Canadian Social Structure (Scarborough, Ont., 1967), esp. pp. 52–3.Google Scholar

2 See Alford, Robert, Party and Society (Chicago, 1963Google Scholar), chap. 9, and “Class Voting in the Anglo-American Political Systems,” in Lipset, S. M. and Rokkan, S., eds., Party Systems and Voter Alignments (New York, 1967), 6593Google Scholar; and Schwartz, M., “Canadian Voting Behaviour,” a paper presented to the Conference on Comparability in Voting Studies, Loch Lomond, Scotland, 1968, p. 105Google Scholar and passim.

3 See, for example, Pinard, Maurice, “One-Party Dominance and Third Parties,” Canadian Journal of Economics and Political Science, XXXIII, no 3 (Aug. 1967), 358–73CrossRefGoogle Scholar; Dean, E. P., “How Canada Has Voted: 1867 to 1945,” Canadian Historical Review, XXX, no 3 (Sept. 1949), 227–48.CrossRefGoogle Scholar

4 See Schwartz, M., Public Opinion and Canadian Identity (Berkeley, 1967, p. 146Google Scholar ff.). The nature of the data used in her study, Canadian Gallup surveys from 1941, prevented the identification of “regional subcultures”; however, Schwartz reports that “region” is the prime differentiator of opinions (cf. Engelmann and Schwartz, Political Parties, 43), and that regions differ in the degree of “within-region” homogeneity of opinions across party lines, with the Maritimes exhibiting greatest homogeneity followed by, in order of decreasing homogeneity, the Prairies, Quebec, British Columbia, and Ontario.

5 “The Electoral System and the Party System in Canada, 1921–1965,” this JOURNAL, I, no 1 (March 1968), 55–80. Cairns demonstrates among other things that studying the proportion of seats won by a party in a given region may give a misleading picture of the degree of homogeneity within the region because of the “bias” of the single-member, plurality electoral system.

6 “Conclusion: An Analysis of the National (?) Results,” Papers on the 1962 Election (Toronto, 1964), p. 286.

7 His measure of class voting – the percentage of those in “manual” occupations supporting “left” parties minus the percentage of those in “non-manual” occupations – is not concerned with the electoral strength of a party or the homogeneity of its support. Thus, for example, a situation in which 75 per cent of manuals and 50 per cent of non-manuals supported a party would produce the same level of class voting (+25) as a situation in which a party receives the support of 25 per cent of manuals and 0 per cent of non-manuals.

8 Alford combines support for the Liberals and the CCF/NDP as support for parties of the left. Some quarrel with the combination and others with his assumption that the Liberals are more left wing than the Conservatives, the claim being made that such a distinction exaggerates ideological differences between the parties. Alford meets the former criticism in an updated version of his study by separating the CCF/NDP from the Liberals. He finds that such a separation makes little difference to his generalizations about the low level of class voting in Canada. See his “Class Voting,” 84–5. For an attempt to distinguish ideological differences between the parties, see Meisel, John, “Recent Changes in Canadian Parties,” in Thorburn, Hugh, ed., Party Politics in Canada (2nd ed., Scarborough, 1967), 3354.Google Scholar

9 See Schwartz, “Canadian Voting Behaviour,” 105.

10 party and Society, 261.

11 See Rokkan, Stein and Valen, H., “The Mobilization of the Periphery,” in Rokkan, Stein, ed., Approaches to the Study of Political Participation (Bergen, 1962Google Scholar), and “Regional Contrasts in Norwegian Politics,” in Allardt, E. and Littunen, Y., eds., Cleavages, Ideologies and Party Systems (Helsinki, 1964Google Scholar); Rokkan, S., “Electoral Mobilization, Party Competition, and National Integration,” in LaPalombara, J. and Weiner, M., eds., Political Parties and Political Development (Princeton, 1966Google Scholar), chap. 9, and “Geography, Religion, and Social Class: Cross-cutting Cleavages in Norwegian Politics,” in Lipset and Rokkan, Party Systems.

It is, of course, often hazardous to make judgments about large bodies of literature, but it does seem that North American political scientists have been less concerned with the evolution of cleavages than have Europeans. For an excellent summary statement of differences in approaches, see Linz, Juan, “Ecological Analysis and Survey Research,” in Dogan, M. and Rokkan, S., eds., Quantitative Ecological Analysis in the Social Sciences (Cambridge, Mass., 1969), 92.Google Scholar The collection containing the Linz essay contains a number of excellent studies using social structure as an independent variable in analysis.

The political geographers are represented by Kevin Cox, “The Spatial Evolution of National Voting Response Surfaces: Theory and Measurement,” and David Reynolds, R., “Spatial Dimensions of Electoral Behavior: Some Theoretical and Methodological Considerations,” both papers presented at the 1969 Annual Meeting of the American Political Science Association, New York.Google Scholar

12 Cox, “Spatial Evolution,” 6–9.

14 Converse, Philip, “On the Possibility of Major Political Realignment in the South,” in Campbell, Anguset al., Elections and the Political Order (New York, 1966Google Scholar), chap. 12.

15 See Rokkan and Valen, “The Mobilization of the Periphery.”

16 See Cox, “Spatial Evolution”; Stokes, Donald E., “A Variance Components Model of Political Effects,” in Claunch, John M., ed., Mathematical Applications in Political Science (Dallas, 1965Google Scholar), and Stokes “Parties and the Nationalization of Electoral Forces,” in Chambers, William N. and Burnham, Walter D., The American Party Systems: Stages of Political Development (New York, 1967Google Scholar).

Some less methodologically sophisticated attempts to come to grips with the same phenomenon, and also based to some extent on measuring between-region variance in voting patterns, have been undertaken for the Canadian case. Thus see Beck, J. M., “The Democratic Process at Work in Canadian Elections,” in Courtney, J. C., ed., Voting in Canada (Scarborough, 1967Google Scholar); and Robert Alford, Party and Society. Beck's analysis, based on “swings” away from the “governing party” when power changed hands, finds few differences in the direction of swing by region until after 1958. Alford looks at swings in all elections from 1921–58, and by comparing the variance of the swing across regions (of course, ignoring direction) concludes that the range of the swing has been becoming smaller, as would be expected from the evolutionary perspective. However, this writer extended his analysis to the four elections following that of 1958 and found that the pattern did not persist.

17 He uses measures of these variables as independent variables in a regression model in which the dependent variable is the positive square root of the squared deviation of each state's Democratic presidential vote from the national mean in each election from 1920–56. See “Spatial Evolution,” 16–18.

18 A more complete analysis of these approaches is contained in this writer's doctoral dissertation, “Regionalism in Canadian Voting Behaviour, 1908–1968,” chap. 2, in preparation.

19 For a critique of political development studies from this perspective, see Huntington, Samuel P., Political Order in Changing Societies (New Haven, Conn., 1968Google Scholar), chap. 1. A similar criticism of assumptions underlying studies of “political integration” has been made by Lijphart, Arend, “Cultural Diversity and Theories of Political Integration,” this JOURNAL, IV, no 1 (March 1971), 114.Google Scholar

20 See Pinard, Maurice, “Working Class Politics: An Interpretation of the Quebec Case,” Canadian Review of Sociology and Anthropology, VII, no 2 (May 1970), 87109.Google Scholar

21 See Martin Robin, “The Social Basis of Party Politics in British Columbia,” in Thorburn, Party Politics in Canada, 201–11.

22 See Wilson, G. W., Gordon, S., and Judek, S., Canada: An Appraisal of its Needs and Resources (New York, 1965), 150–1.Google Scholar

23 See Stone, Leroy O., Urban Development in Canada: An Introduction to the Demographic Aspects (Ottawa, 1967Google Scholar), Table 2–2.

24 Economic Council of Canada, Interregional Disparities in Income, Staff Study no 14 (Ottawa, 1966), esp. p. 65.Google Scholar

25 A slightly different use of multiple regression to measure regional effects has been recently suggested by William Irvine. See his “Assessing Regional Effects in Data Analysis,” this JOURNAL, IV, no 1 (March 1971), 21–4.

26 For a more detailed mathematical treatment of the use of dummy variables, see Suits, Daniel B., “The Use of Dummy Variables in Regression Equations,” Journal of the American Statistical Association, LII (1957), 548–51.CrossRefGoogle Scholar

27 These are not the only kinds of interactions which could have been tested. For example, interactions between the background variables themselves might have been investigated to see whether the assumption of “additivity” of these characteristics is justified. Some preliminary work along this line has been done by the author which suggests that the assumption of additivity is not violated. For an interesting discussion of the concept of interaction and its importance in the analysis of survey data, see Pool, Ithiel de Solaet al., Candidates, Issues and Strategies (Cambridge, Mass., 1965), 127–43.Google Scholar

The presence of interaction in analysis using aggregate data has been investigated by Claudio Soares and Hamblin, Robert, “Socio-Economic Variables and Voting for the Radical Left: Chile, 1952,” American Political Science Review, LXI (1967), 1053–65Google Scholar; and Burnham, Walter D. and Sprague, John, “Additive and Multiplicative Models of the Voting Universe: The Case of Pennsylvania, 1960–1968,” American Political Science Review, LXIV (1970), 471–90.CrossRefGoogle Scholar As a result of the latter study, the authors suggest as a general proposition that interactions are more important for “stable clientele” parties than for “portmanteau” or “catch-all” parties (p. 489). Our analysis suggests that this proposition is not true for regional interactions in the Canadian case.

28 Porter, John, The Vertical Mosaic: An Analysis of Social Class and Power in Canada (Toronto, 1965), 87.CrossRefGoogle Scholar

29 I am indebted to John Meisel of Queen's University for supplying me with the constituency level census data for 1961. Professor Meisel also supplied census subdivision information on ethnic origin not previously published which was used for estimating constituency characteristics in 1949.

30 See his “Ethnicity, Religion and Politics in Canada: A Comparative Analysis of Survey and Census Data,” in Dogan and Rokkan, Quantitative Ecological Analysis, 187–216.

31 A more extensive discussion of the problem of multicollinearity can be found in Blalock, H. M., “Correlated Independent Variables: The Problem of Multicollinearity,” in Tufte, E., ed., The Quantitative Analysis of Social Problems (Reading, Mass., 1970), 418–25.Google Scholar See also the matrix of correlations in the Appendix.

32 Census of Canada, 1961.

33 One cannot use all categories of the religious, ethnic, and occupational variables simultaneously in the analysis for mathematical reasons. The size of each residual category is 72 per cent for ethnic origin (consisting mainly of British and French); 16 per cent for religion (mainly Lutherans and Eastern Orthodox religions); and 36 per cent for occupation (mainly unskilled labour, recreation, and protective occupations). It would have been useful to use unskilled labour as a separate variable, but is could not be extracted from our data.

34 Party and Society, 263–7. Ontario obtained the highest score in six of ten surveys and second highest score in the others.

35 For a discussion of specification error, see Draper, N. and Smith, N., Applied Regression Analysis (New York, 1968), 81–5.Google Scholar

36 A non-significant coefficient indicates that the presence of the group so measured in a constituency is unrelated to the party's level of support. This could occur if a group does not disproportionately support a given party or if that group's effect is more adequately measured by some other characteristic. There is of course a further possibility which we have taken pains to guard against, namely that correlations between independent variables have masked their separate effects.

Significant coefficients indicate that the presence of the group so measured is related to the party's expected vote. The size of the coefficient indicates what increment we expect in a party's vote with a unit increase in the percentage of that group in the constituency. Differences in the sizes of coefficients indicate differences in the degree to which group characteristics are related to a given partisan preference. However, these coefficients do not measure the degree to which each group contributes to a party's support because they do not take into account the relative sizes and dispersions of groups.

At this point some attention should be paid to the question of the “ecological fallacy,” since we are dealing with aggregate data and making inferences about the relationship between the presence of certain groups in constituencies and their party support. This can be construed as an inference about individuals.

Our analysis could have been couched in terms of “behavior of electorates,” an “out” suggested by Austin Ranney (see his “The Utility and Limitations of Aggregate Data in the Study of Electoral Behavior,” in Ranney, , ed., Essays on the Behavioral Study of Politics [Urbana, Ill., 1962], 91102Google Scholar). Given, however, the similarity between many of our results and results from studies using survey data, and because of our concern with the “net” effect of region once characteristics of the electorate have been taken into account, we prefer to accept the advantages claimed for ecological “regression” as opposed to ecological “correlation.” For an extended discussion of the evolution of the controversy and an examination of various situations in which knowledge of the marginal characteristics of a joint distribution can give unbiased estimates of conditional relationships, see Shively, W. Phillips, “‘Ecological’ Inference: The Use of Aggregate Data to Study Individuals,” American Political Science Review, LXIII (1969), 1183–96.Google Scholar One of Shively's findings is that estimates of conditional relationships will be unbiased if “individuals have been grouped in such a way that their scores on the dependent variable are unrelated to the aggregation in which they fall, except indirectly through their scores in the independent variable” (p. 1186, emphasis in original). This would seem to be a justifiable assumption when trying to explain the relationship between party preference and demographic variables.

37 That is, in calculating the contribution of the percentage of Germans to a party's support in a prairie constituency, one would first multiply the percentage German by .34 and subtract from that the same percentage multiplied by .37.

38 However, the surveys conducted by Meisel and associates of the 1965 and 1968 elections, the former now publicly available, will allow more sophisticated analysis. Comparison of our results with results using the 1965 Meisel survey is in progress.

39 “Ethnicity, Religion and Politics,” 213–14. The major differences are that Laponce used stepwise regression, a different set of variables, some of which were the same as used in this study, and did the analysis separately for rural and metropolitan constituencies. This study also examined rural-urban differences, although the results are not presented here, by including an “urban” dummy variable in the party support models. However, once religion, ethnic origin, and occupation are incorporated in the models, urban-rural differences were not significant.

40 See Meisel, John, “Religious Affiliation and Electoral Behaviour: A Case Study,” Canadian Journal of Economics and Political Science, XXII, no 4 (Nov. 1956), 481–96CrossRefGoogle Scholar; and Anderson, Grace M., “Voting Behaviour and the Ethnic-Religious Variable: A Study of a Federal Election in Hamilton, Ontario,” ibid., XXXII, no 1 (Feb. 1966Google Scholar). Laponce also reports the expected Anglican-Conservative relationships found by Regenstreif, S. Peter, “Some Aspects of National Party Support in Canada,” ibid., XXIX, no 1 (Feb. 1963Google Scholar).

41 See Regenstreif, P., The Diefenbaker Interlude: Parties and Voting in Canada (Toronto, 1965Google Scholar).

42 Ibid., 94.

43 Ibid., 92.

44 Papers on the 1962 Election, 281–5.

45 These figures were calculated from Table 126 in Dominion Bureau of Statistics, “Characteristics of Immigrants,” Bulletin 1.3–11 (Ottawa, 1961Google Scholar).

46 See Wagenheim, Elizabeth, “The Ukrainians: A Case Study of the ‘Third Force.’” in Blishen, Bernardet al., eds., Canadian Society: Sociological Perspectives (3rd ed., Toronto, 1968), 648–65.CrossRefGoogle Scholar

47 Robert R. Alford, “The Social Bases of Political Cleavage in 1962,” in Meisel, , Papers on the 1962Google Scholar Election, 215, Table XI.

48 This does not contradict the earlier comment about the relationship between the presence of Catholics and Conservative support in certain regions since we are measuring “conditional” relationships, and, in addition, since we have a multiparty situation, we do not expect Liberal party support to be the mirror image of Conservative support.

49 Cited in Alford, “Class Voting,” 88.

50 “The Social Bases of Political Cleavage, Table XI.

52 See Ryder, N. B., “The Interpretation of Origin Statistics,” Canadian Journal of Economics and Political Science, XXI, no 4 (Nov. 1955), 466–79CrossRefGoogle Scholar, for an extended discussion of problems, particularly in earlier censuses, with census classifications of “ethnic origin,” “birthplace,” and “mother tongue.” For example, Ryder finds a surprising number of “Germans” disappearing after the First World War in the census of 1921 and re-emerging in 1931.

53 Again a debt to John Meisel must be acknowledged for his provision of unpublished subunit information on ethnic origin for the census of 1951.

54 See “The Measurement and Impact of Regionalism” presented at St John's, June 1971, Appendix. Briefly, characteristics of census subdivisions within electoral districts were aggregated to arrive at constituency characteristics. In cases where electoral district boundaries cut through census subdivisions, characteristics of the latter were apportioned between electoral districts.

55 We used all background variables rather than only the most “statistically significant” because we wanted our provincial dummy variables to measure the “net” effect of region rather than have them include the effects of “unmeasured” variables. This, of course, does not resolve the problem completely since we have no measure of class for this period.

56 “‘Limited Identites’ in Canada,” Canadian Historical Review, L, no 1 (March 1969), 1–10.

57 Ibid., 7.