1 Irvine, W. P., “Explaining the Religious Basis of the Canadian Partisan Identity: Success on the Third Try,” this JOURNAL 7 (1974), 560–63Google Scholar; and Irvine, W. P. and Gold, H., “Do Frozen Cleavages Ever Go Stale? The Bases of the Canadian and Australian Party Systems,” British Journal of Political Science 10 (1980), 187–218.CrossRefGoogle Scholar

2 The census and electoral data are from the Blake file and were analysed by myself. In 1972, a unit increment in the percentage Catholic in a constituency produced a 0.209 point increment in the Liberal share and a 0.386 point decrement in the Conservative share. The 1972 impact on the Conservative vote was the greatest for all elections in. the Blake file except 1917. The 1972 impact on the Liberal vote was about average. Estimates are OLS from bivariate regressions. Attempts to separate religious from linguistic effects were bedevilled by multicollinearity; religion almost always dominated language as a predictor, however. On the relative power of religion and language, see also footnote 4.

3 For the 1950s and 1960s, see Laponce, J. A., “Postdating Electoral Cleavages in Canadian Federal Elections, 1949-68: Material for a Footnote,” this JOURNAL 5 (1972), 270–86.Google Scholar For 1965, 1968, 1974, and 1979, the generalizations are based on analyses by myself with national election study data. Copies of the analyses are available on request.

4 The statements in this paragraph are based on my own analyses with national election study data and on the following sources: Irvine, “Explaining the Religious Basis”; and H. D. Forbes and P. M. Sniderman, “The Statistical Relation Between Religion and the Vote in Canada,” unpublished manuscript, University of Toronto, 1976. The latter is by far the most subtle account of the Canadian religious cleavage to date.

5 See above, footnote 1.

6 See above, footnote 1.

7 I attempted to reproduce the coding scheme of Irvine and Gold, “Frozen Cleavages,” but was unable to do so exactly. The minor differences between my estimates and theirs do not affect any conclusions here or below.

8 The empirical statements in this paragraph are based on Irvine and Gold, “Frozen Cleavages,” Table 1. Their findings with 1974 data are substantially confirmed with data from other years.

9 Irvine, “Explaining the Religious Basis,” 563.

10 A “family” party measure would shift estimates in Table 1 only slightly, as most respondents impute the same party loyalty to their mothers as to their fathers.

11 On self-directed bias in recall of parents'preferences, see Niemi, R. G., How Family Members Perceive Each Other (New Haven: Yale University Press, 1974).Google Scholar

12 Classic sources on political socialization include Hyman, H., Political Socialization (Glencoe: Free Press, 1959)Google Scholar; Hess, R. D. and Torney, J. V., The Development of Political Attitudes in Children (Chicago: Aldine, 1967)Google Scholar; Easton, D. and Dennis, J., Children in the Political System: Origins of Political Legitimacy (New York: McGraw-Hill, 1969)Google Scholar; and Jennings, M. K. and Niemi, R. G., The Political Character of Adolescence (Princeton: Princeton University Press, 1974).Google Scholar For a comparative evidence on the “heritability” of party loyalty see R. Johnston, “Families and the Fate of Party Systems,” paper presented to the Canadian Political Science Association Annual Meeting, Halifax, 1981.

13 The equilibrium distribution implied in Table 2 is 50.3 per cent Liberal, 18.7 per cent Conservative, 10.5 per cent NDP, 3.1 per cent Social Credit, and 16.9 per cent nonpartisan. To calculate the equilibrium, let entries in the equilibrium probability vector, p*, be p₁*, with the parties numbered in the row and column order that they appear in Table 2. From Table 2:

1 = P₁* + p₂* + P₃ + P₄* + P₅*

P₁* = .652p₁* + .335p₂* + .307p₃* + .286p₄* + .422p₅*

p₂* = .105p₁* + .398p₂* + .159p₃* + .229p₄* + .211p₅*

p₃* = .056P_l* + .067p₂* + .409p₃* + .171p₄* + .094p₅*

p₄* = .026p₁* + .019p₂* + .011p₃* + ,257p₄* + .030p₅*

p₅* = .161p₁* + .182p₂* + .114p₃* + .057p₄* + .243p₅*

With six equations and five unknowns, solving for p₁* is straightforward.

The calculations in the preceding paragraph and the argument in the body of the text assume that the intergenerational transmission of party loyalty is some kind of Markov process. Although party loyalty transmission does not meet all of the assumptions of the simplest'kind of Markov process, the stationary, first-order, homogeneous-population process, the Markov image is still apt. It is especially so as it dramatizes the dependence of equilibrium on interparty exchanges. For simple examples of Markov chains, see Kemeny, J. G., Snell, J. L., and Thompson, G. L., Introduction to Finite Mathematics (2nd ed.; Englewood Cliffs: Prentice-Hall, 1966), 194–201, 271–89.Google Scholar For a more formal exposition see Kemeny, J. G. and Snell, J. L., Finite Markov Chains (New York: Van Nostrand, 1960).Google Scholar

The bulk of the difference between row and column margins comes from the bloated “other-none” share on the row margin. This probably reflects the large percentage of immigrants in the population. The standing of parties relative to each other differs only a little between margins. The differences that do appear reflect the historical record: third-party growth relative to the major parties.

A more vexing problem with Table 2 is the overestimation of the Liberal share. This overestimation is endemic to Canadian national election study surveys and merits investigation.

14 Clearly other factors are at work as well as the simple collective inertia described in the text. The parties' relative standing shifts from the fathers' margin to the respondents' margin, as noted above. This realignment may help explain some of the strength of the NDP inheritance rate in Table 2. The strong NDP rate may also reflect relatively intense influence within the family, as conventional wisdom leads us to expect. The difference between the NDP inheritance rate and the random likelihood of becoming a New Democrat is much greater than the corresponding difference for any other party.

15 Within father's party groups, the Catholic–Non-Catholic difference in current Liberal or Conservative preference is always significant. The bulk of the religious difference is mapped by recruitment from the children of nonpartisans. This is so for two reasons: first, the sharp Catholic–Non-Catholic differences in Liberal and Conservative recruitment; and, second, the large proportion of respondents unable to impute a party loyalty to their fathers.

16 The implied “equilibrium” Catholic-Non-Catholic difference is +18.8 for the Liberals and –22.1 for the Conservatives. This implied equilibrium is only notional, of course. Survey evidence, cited in footnote 3, suggests that the religious cleavage has shrunk since 1965. But the shrinkage itself reflects a contemporary reorientation of political forces. The cleavage that remains is a product of the Catholic–Non-Catholic transition rate differences that endure. The remaining cleavage cannot be simply an artifact of family socialization.

17 Wald, K. D., Crosses on the Ballot: Patterns of British Voter Alignment Since 1885 (Princeton: Princeton University Press, 1983), chap. 5.CrossRefGoogle Scholar

18 At present, the government of Ontario gives province-wide support to separate schools only through the elementary grades; in some school districts, support extends to grades 9 and 10. In 1971, the Ontario Liberal party campaigned pointedly on full financial support for the Catholic system. Although the disaster which befell the Liberals in the 1971 election deterred them from emphasizing the schools issue in more recent elections, partisan differences on the schools issue remained much as before (see the Toronto Globe and Mail, November 26,1983,11). It may be no accident, then, that the religious cleavage as measured in 1974 is wider in Ontario than in any other province.

On June 12, 1984, the provincial government announced full support for the Catholic system, to begin in September 1985. This may take some of the partisan edge off the issue, although the traditional lines of controversy revealed themselves the moment the announcement of support was made. For immediate comment and a telling editorial, see the Toronto Globe and Mail, June 13, 1984.

19 For grades one to ten in Ontario, the ratio of Catholic system enrolment is very close to the whole-population Catholic–on-Catholic ratio (Globe and Mail, November 26, 1983). Novelists are well in advance of political scientists in seeing the social separation of Catholics and non-Catholics. Here, for example, is Marian Engel: “Now I think of it, there was a profound division all along that shore. Tess's church, white board-and-batten with a rounded apse that glimmered like a ghost on summer nights, was bigger than ours. Its churchyard bore witness to many more generations of settlement than ours did. The names were French and Irish. And we didn't know any of the French or the Irish except the McCrorys because my father worked with Tess's sometimes; otherwise, Catholics were not available to us. We were like the French and English in Montreal, looming invisibly over each others' shoulders” (The Glassy Sea [Toronto: McClelland and Stewart, 1978], 42).Google ScholarPubMed

20 See footnote 4.