1 See his “Institutionalization of Voting Patterns, or Is Mobilization the Source of Decay?” American Political Science Review 69 (1975), 49–67CrossRefGoogle Scholar.

2 See for example, Wilson, John, “The Canadian Political Cultures: Towards a Redefinition of the Nature of the Canadian Political Systems,” this JOURNAL 7 (1974), 438–83Google Scholar, much of which is taken up with attempts to identify crucial periods of transition from one type of party system to another in each province.

3 See Simeon, Richard and Elkins, David J., “Regional Political Cultures in Canada,” this JOURNAL 7 (1974), 397–437Google Scholar, for some observations about patterns of political culture and the likelihood of electoral change.

4 The debate began with the publication of Sniderman, Paul M., Forbes, H. D., AND Melzer, Ian, “Party Loyalty and Electoral Volatility: A Study of the Canadian Party System,” this JOURNAL 7 (1974), 266–88Google Scholar. See n. 9 below for other contributions.

5 Ibid., 270–73.

6 Illustrating this position are statements like the following: “a seven or eight percentage point gain or loss by a party is not unusual,” and “the unusual instability in the major parties’ share of the national vote, the relative openness of the voters to third party alternatives.” For sources of these statements and other examples see ibid.

7 Ibid., 271.

8 Ibid., 286.

9 The first major response to the Sniderman et al. article completely ignores the volatility part of the stability of partisanship-volatility relationship. See Jenson, Jane, “Party Loyalty in Canada: The Question of Party Identification,” this JOURNAL 8 (1975), 543–53Google Scholar.

10 Calculations based on “Institutionalization of Voting Patterns,” Table 1. The nine-country group includes (in addition to Canada), France, Belgium, Denmark, Sweden, Norway, Germany, United Kingdom, and Finland. The list expands to ten with the addition of Switzerland.

11 See his “Electoral Politics and Expansion of the Public Economy: A Comparative Analysis” (1977). The calculation of volatility scores is a by-product of Cameron's analysis. He considers 17 countries including Japan, Australia, Italy, Ireland, the Netherlands, Austria, and the United States in addition to those treated by Przeworski. Canada ranks sixth in stability in this larger grouping, the US (seen as more volatile in the Sniderman et al. study) ranks second, and the UK ranks eleventh.

12 This is obviously an unrealistic assumption in the case of Canada, especially when comparing nonconsecutive elections, and is not required for the purposes of this paper. The relationship between electoral volatility and changes in the size of the electorate is considered in n. 46 below.

For an interesting attempt to break down electoral change into its theoretical components, see Butler, David and Stokes, Donald, Political Change in Britain (College Edition; New York: St. Martin's Press, 1971)CrossRefGoogle Scholar, especially chap. 9.

13 Przeworski uses the volatility score as a dependent variable measuring deinstitutionalization of the party system in a regression model where rates of mobilization and demobilization are independent variables. The results in the case of Canada for the period 1896–1965 are suggestive of changes in the vote distribution between elections produced by the movement into and out of the electorate of voters without fixed preferences.

Lest this be thought an obvious result it should be noted that other patterns are possible and are represented in his analysis. In the UK, for example, demobilization has no effect suggesting that neither voters for particular parties nor floaters are more likely to leave or to stay. In Germany, “decay” is produced by mobilization and demobilization suggesting movement into and out of the electorate by groups with specific preferences. See his “Institutionalization of Voting Patterns,” especially 63. Przeworski's model does not fare well when applied at the subnational level in Canada. See n. 47 below.

14 In calculating the index for purposes of this paper, the Liberal, Conservative, Social Credit créditiste, CCF-NDP, Progressive, Reconstruction, and bloc populaire parties were treated as separate entities. The Union des Electeurs is treated as part of Social Credit in 1949. Prior to the formation of the CCF, votes for all socialist and labour candidates were aggregated to form a single “labour party.” All other votes were aggregated into a single “other” group. In comparisons involving 1917, Government and Opposition votes were treated as Conservative and Liberal, respectively.

Election results were drawn from Scarrow, Howard, Canada Votes (New Orleans: Hauser Press, 1962)Google Scholar for elections from 1878–1917; from Beck, J. M., Pendulum of Power (Scarborough: Prentice-Hall, 1968)Google Scholar for elections from 1921 to 1968; and from relevant reports of the Chief Electoral Officer for subsequent elections. Beck is, of course, also a possible source for the earlier elections, and his figures sometimes differ from those reported by Scarrow. These discrepancies are noted when they may make a difference to the interpretation.

15 “A Theory of Critical Elections,” Journal of Politics 17 (1955), 4Google Scholar.

16 The literature on critical elections is now too voluminous to summarize here. For a useful bibliography together with an analysis of methodological issues see Flanigan, William H. and Zingale, Nancy H., “The Measurement of Electoral Change,” Political Methodology 1 (1974), 49–82Google Scholar.

17 These have not been the only concerns. Some scholars have sought to explain the apparent periodicity of critical elections in the US (realignments occurred approximately once a generation until the mid-1960's) and to relate critical elections to major policy shifts. For a sophisticated attempt to relate the periodicity of realignments to socialization and party identification as well as an explanation for the failure of realignment to materialize in the 1960's, see Beck, Paul Allen, “A Socialization Theory of Partisan Realignment,” in Niemi, R. G. and Weisberg, H. F. (eds.), Controversies in American Voting Behavior (San Francisco: W. H. Freeman, 1970), 396–411Google Scholar. The question of a policy relationship is exhaustively considered in Ginsberg, Benjamin, “Elections and Public Policy,” American Political Science Review 70 (1976), 41–49CrossRefGoogle Scholar.

18 Other students of critical elections have made this distinction as well. See, for example, Campbell, Angus, “A Classification of Presidential Elections,” in Campbell, Angus et al. , Elections and the Political Order (New York: Wiley, 1966), 63–77Google Scholar: and Pomper, Gerald, “Classification of Presidential Elections,” Journal of Politics 29 (1967), 535–66CrossRefGoogle Scholar.

19 The US studies, forexample, examine variations overtime in the support for a single party and/or changes over time in the geographical distribution of single party support, whereas our method explores variations over time and across space in the distribution of support across all parties.

20 We do not mean to imply that the measurement of “normal vote” is a trivial problem where survey data are available. See Philip Converse, “The Concept of a Normal Vote,” in Campbell et al., Elections and the Political Order, 9–39.

21 This procedure is closest to that adopted by Pomper. “Electoral eras” is an expression he uses. Pomper examines the correlation in all pairs of elections between the distributions across states of support for the Democratic party for signs of shifts in the geographical basis of party support (“Classification of Presidential Elections”).

22 This is the case for a variety of reasons. Our measure taps net shifts in the support for parties rather than gross shifts. It is thus conceivable that wholesale exchanges of party supporters could occur without disturbing the total distribution. While logically possible, this seems unlikely. More serious is the fact that changes in the distribution between elections could be produced by the entry of new voters in the second election whose preferences differed from the (unchanged) preferences of repeating voters. The entry of new parties and the demise of existing ones also complicates the picture. In the case of the demise of parties, for example, the disappearance can force a “rearrangement” of voters among other parties or nonvoting without a genuine realignment at the individual level taking place. In the discussion which follows we have tried to keep these possibilities in mind when assessing the significance of change.

23 “Institutionalization of Voting Patterns,” 52.

24 There are a number of other reasons for the focus on 1896 in addition to its significance for the Canadian party system, including some interesting parallels with the critical election of 1896 in the United States. For example, John English sees in this election and its aftermath the emergence of significant extra-local interests and the beginning of the end of the post-Confederation party system. See his The Decline of Politics: The Conservatives and the Party System 1901–20 (Toronto: University of Toronto Press, 1977)Google Scholar, especially chap. 1. The best known exponent of the pivotal role played by 1896 in the US is Walter Dean Burnham. See his Critical Elections and the Mainsprings of American Politics (New York: Norton, 1970) for a number of similar argumentsGoogle Scholar.

25 Deaths, dual representation, and voided elections together with subsequent byelections complicate the Manitoba picture after 1896. See Dean, Edgar P., “How Canada Has Voted: 1867 to 1945,” Canadian Historical Review 30 (1949), 227–48 for detailsCrossRefGoogle Scholar.

26 Turnout figures are based on figures from Urquhart, M. C. and Buckley, K. A. H., Historical Statistics of Canada (Toronto: Macmillan, 1965), 616Google Scholar.

27 Winn, C. and McMenemy, J., Political Parties in Canada (Toronto: McGraw-Hill Ryerson, 1976), 57ffGoogle Scholar. There is inevitably an element of arbitrariness in the selection of cutting points for the designation of critical elections. Thus, while we have noted that the Conservative showing of 46 per cent of the vote in Quebec was the lowest since Confederation, Gingras and Winn argue that this “proportion exceeds the national share of the vote received by the Conservatives in almost every election in the last 74 years” (58). While strictly speaking this is true, it ignores the fact that since 1921, support at the level of 46 per cent nationally is rare for any party. In the analysis to follow, rather than focussing on absolute levels of change or volatility, we will be looking for sharp breaks with the past and with levels of volatility in intervening elections which appear to cluster together.

Ibid., 61.

29 For a summaryof these and other features of the late nineteenth century party system see Reid, Escott, “The Rise of National Parties in Canada,” in Thorbum, Hugh (ed.), Party Politics in Canada (2nd ed.; Scarborough: Prentice-Hall, 1967)Google Scholar.

30 This interpretation is also supported by Dean, “How Canada has Voted,” see especially 235 and 238.

31 On the operations of “ministerial” politics see Reid, “National Parties,” 20–21.

32 Calculations are based on Urquhart and Buckley, 14.

33 Norman Ward provides figures which suggest that the percentages of the population enfranchised in 1882, 1891, and 1900 in BC were 11.0, 13.3, and 21.9, respectively (The Canadian House of Commons: Representation [Toronto: University of Toronto Press, 1950], 214–30)Google Scholar.

34 The Conservative share of the vote in the region jumped from 40.2 per cent in 1874 to 50.7 per cent in 1878.

35 To my knowledge only Escott Reid writing in 1933 has noted the return to the prewar alignment in 1926 and 1930. See his “The Effect of the Depression on Canadian Politics,” American Political Science Review 27 (1933), 455–65Google Scholar. F. H. Underhill notes the phenomenon only to dismiss it, calling the apparent effect of 1930 “superficial.” See his “The Canadian Party System in Transition,” Canadian Journal of Economics and Political Science 9 (1943), 300–13CrossRefGoogle Scholar.

36 English, The Decline of Politics, contains an excellent analysis of disruptions in party organization and party attachments produced by early twentieth century social and economic changes as well as by the Unionist government movement. It is significant for what follows that his analysis reveals only a token and opportunistic commitment to unionism in Nova Scotia and Prince Edward Island (174–75).

37 The peculiarity of New Brunswick may be due to the greater impact of unionism in that province. See English, The Decline of Politics, 175–76.

38 Tables 4 and 5 seem to be the most economical way to compare provincial level changes with each other and with national changes given the impracticality of producing figures comparable to Figure 1 for each province, or comprehensive tables involving interelection comparisons for every pair of elections between 1921 and 1974. Copies of these results can be obtained from the author.

The choice of the four largest (and four smallest) changes for inclusion in the table, while in a sense arbitrary, is based on the fact that the four largest changes nationally identify the only critical elections at that level during this period and thus serve as an obvious reference point. It is clear from the discussion which follows whether the four largest changes so identified demarcate critical elections and whether there are other candidate elections in particular provinces.

The ranges within which volatility scores fall by province are as follows for the period 1921–1974:

It is noteworthy that small interelection changes are not restricted to provinces where two-party contests are the norm, and that even in provinces not marked by third-party activity, the range of scores is substantial. The fact that maxima are generally higher in western Canada is not particularly troubling so long as periods of relative stability can be isolated there. The figures for Newfoundland are, of course, based on post-1949 shifts.

39 These individual patterns are relatively easy to interpret and demonstrate the sensitivity of the volatility indicator. The election of 1953 saw the arrival of Social Credit on the federal scene in BC; 1945 and 1949, the rise and fall of bloc populaire in Quebec. An examination of the relationship between these elections and those surrounding them suggests that 1953 was a realigning election in BC in the sense that the previous partisan pattern was permanently disrupted, but a new pattern of stability did not emerge until 1962. In the case of Quebec, 1945 is clearly a deviating election. The Alberta and New Brunswick anomalies are dealt with below.

40 Although to save space the figures are not presented here, interelection difference scores by province for the set of elections following 1935 until 1957 are relatively low with the exceptions noted in Table 4, most of which have been discussed. According to our criteria, 1935 was a deviating election in PEI, realigning in Manitoba, Saskatchewan, Alberta, and British Columbia as well as (in combination with the election of 1940) in New Brunswick, Nova Scotia, and Ontario.

41 This, of course, is the title of Peter Regenstreifs book on the period. See The Diefenbaker Interlude: Parties and Voting in Canada (Toronto: Longmans. 1965)Google Scholar. Regenstreif is one of the strongest proponents of the “instability of partisanship” perspective. See especially his concluding chapter.

42 This expectation is based on Przeworski”s analysis (see especially p. 49). For other expressions of the view that stability of partisan ship (at the individual level and at the level of the whole electorate) is a function of the length of time voters have participated in elections, see Converse, Philip. “Of Time and Partisan Stability,” Comparative Political Studies 2 (1969), 130–71CrossRefGoogle Scholar; and McPhee, William N. and Ferguson, Jack, “Political Immunization,” in McPhee, and Glaser, William A.(eds.), Public Opinion and Congressional Elections (New York: Free Press, 1962), 155–79.Google Scholar

Curiously, Regenstreif notes the dramatic increases in turnout produced by these elections (69–70), but as noted continues to adhere to the historical instability argument. As he expresses it: “the point must be emphasized … that the changeable nature of electoral behaviour is not a creation of the Diefenbaker years but a carry-over of a pattern inherited from the past. The years of Liberal dominance, constant through they may have made Canadian political orientations appear, were essentially a façade” (169).

43 For a more detailed analysis of the 1958–62 shifts see Meisel, John, “Conclusion: An Analysis of the National (?) Results,” in his Papers on the 1962 Election (Toronto: University of Toronto Press, 1964), 272–88Google Scholar, as well as other papers in the collection.

44 Discussions of the 1962 Quebec case can be found in W. P. Irvine, “An Analysis of Voting Shifts in Quebec,” in Meisel, Papers, 129–43; and Pinard, Maurice. The Rise of a Third Party (Englewood Cliffs: Prentice-Hall, 1971)Google Scholar.

45 The scores, calculated on the same basis as those in Table 2 are .244, .217, .198, and .120 for the elections of 1965, 1968, 1972, and 1974, respectively.

46 Most correlations (Pearson r) between electoral volatility and interelection changes in the number of voters are low. The figures by province are as follows for the period beginning with the 1878–82 pair (1908–1911 in the case of Saskatchewan and Alberta):

Changes in the number of voters were calculated as the absolute value of the proportionate change in the numberof voters between the first and second election in each pair. The results (low correlations) do not differ appreciably when actual proportionate changes are used or when volatility is correlated with turnout changes (using either actual or absolute values).

47 These possibilities may help to explain the failure of the Przeworski model to account for volatility changes at the subnational level as a function of mobilization and demobilization patterns. Ourestimates of the coefficients in the Przeworski model for the period 1896–1965 at the provincial level produce contradictory results and in only three cases (Nova Scotia, New Brunswick and Saskatchewan) is explained variance comparable to that achieved by Przeworski at the national level.

While our mobilization variable (changes in the size of the electorate) differed from that used by Przeworski (changes in the proportion of the adult population voting), analysis based on a similar measure (eligibility turnout) produces even less support for his model. The results of his analysis are not presented here.

48 The straightforward interpretation of parameters in the Przeworski model requires that previous participants ”did not change their aggregate vote.” See “Institutionalization of Voting Patterns.” 54. emphasis in original.

Curiously, the major empirical study of realignment in that series of elections makes no mention of the possible significance for partisan alignment of the influx of new voters. See Jenson, Jane,“y Strategy and Party Identification: Some Patterns of Partisan Allegiance,” this JOURNAL 9 (1976). 27–48Google Scholar.