1 Dawson, R. MacGregor, The Government of Canada (Toronto: University of Toronto, 1970), 326Google Scholar.

2 Since 1940, on only two occasions were more than 6 by-elections held simultaneously, 8 on October 24, 1949 and 15 on October 16, 1978.

3 During the investigation period, of the 57 dates on which federal by-elections were contested only 3 times were as many as 4 provinces represented, and only on October 16, 1978 were more than 4 provinces included.

4 Schwartz, Mildred A., Politics and Territory (Montreal: McGill-Queen's, 1974), 139–65CrossRefGoogle Scholar.

5 Harold Clarke, Jane Jenson, Lawrence Leduc and Jon Pammett suggest that 77 per cent of the electorate is “permanent” in the 1974 general election, but this figure might be questionable in by-elections (Political Choice in Canada [Toronto: McGraw-Hill Ryerson, 1979], 363)Google Scholar.

6 Scarrow, Howard A., “By-Elections and Public Opinion in Canada” in Public Opinion Quarterly 29 (1961), 79–91CrossRefGoogle Scholar.

7 Ibid., 84–86. It might be mentioned that Scarrow's definition of a “favourable” performance included vote declines of up to 5 per cent.

8 Ibid., 89–90.

9 The 1940 election also marked the virtual end of uncontested by-elections which had previously been commonplace.

10 Several Quebec by-elections were included in which only Liberals and representatives of minor parties or independents ran as candidates, because the latter won at least 20 per cent of the total vote. In a few 1940s contests, candidates running under the banner of United Reform, New Democracy, Bloc Populaire and Union des Electeurs were included in the Social Credit totals, where there appeared to be an accommodation between those movements.

11 For example, the Globe and Mail was filled with articles and columns devoted to this topic from October 17 to October 28, following the by-elections of October 16,1978.

12 This question is raised in Pollock, J. K., “British By-Elections Between the Wars,” American Political Science Review 35 (1941), 519–28CrossRefGoogle Scholar.

13 Morton, Desmond, Social Democracy in Canada (Toronto: Samuel, Stevens Hakkert, 1977), 54Google Scholar.

14 Wilson, John, “The Canadian Political Cultures: Towards a Redefinition of the Nature of the Canadian Political System,” this JOURNAL 7 (1974), 463Google Scholar.

15 Leonard, R. L., Elections in Britain (London: D. Van Nostrand, 1968), 119–26Google Scholar.

16 Kernell, Samuel, “Presidential Popularity and Negative Voting: An Alternative Explanation of the Midterm Congressional Decline of the President's Party,” American Political Science Review 71 (1977), 52CrossRefGoogle Scholar.

17 The turnout rates over the 147 by-elections varied from 89 per cent to 23 per cent, a range much broader than occurred in the corresponding previous general elections, when the figures were 90 per cent to 46 per cent. The overall average participation rate among the by-elections was 60 per cent, considerably below the general election average of 72 per cent for the constituencies where by-elections were held. A comparable 12 percentage point mean decline in by-election turnout was also found in Great Britain as reported by Taylor, Stan and Payne, Clive, “Features of Electoral Behaviour at By-Elections” in Cook, Chris and Ramsden, John (eds.), By-Elections in British Politics (London: MacMillan, 1973), 337Google Scholar.

18 Mishler, William, Political Participation in Canada (Toronto: Macmillan, 1979), 40Google Scholar.

19 Converse, Philip E., “The Nature of Belief Systems in Mass Publics,” in Apter, David (ed.), Ideology and Discontent (Glencoe: The Free Press, 1964), 242Google Scholar.

20 This matching of constituency boundaries for the 35 additional by-elections was performed by Dawne Rennie and Isabelle Kuczinski on a grant provided by Wilfrid Laurier University's Research Fellowship programme.

21 Vote totals for the CCF are combined with the New Democratic party, prior to the establishment of the latter in 1961.

22 Moreover, when the Atlantic region was excluded from consideration, the NDP rises to an average gain of 5.8 points in Ontario and the West.

23 The relatively strong by-election showings of other parties in Quebec tended to occur between 1940 and 1957, when Independents and Independent Liberals frequently constituted the strongest alternative to the Liberal candidates.

24 Another relevant consideration is that the CCF-NDP vote at T3 is in part a reflection of the party's gradual growth in support during the period of the study. However, other data, not shown here, suggest that this effect has only minimal impact.

25 Since 1940, 10, of the 15 new CCF and NDP seats won were retained at the following general election. However, since the founding of the New Democratic party in 1961, the rate of subsequent success is 6 out of 7.

26 This pattern of regional variation appears unique to by-elections, since relatively less incidence of interregional disparity is displayed in the constituencies studied between successive general elections.

27 The British Columbia totals include one by-election held in the Yukon Territory.

28 Scarrow, “By-Elections and Public Opinion,” 84.

29 This figure is an estimate based upon data presented in Leonard, Elections in Britain, 126.

30 Between Tl and T2 the similar turnout category included those constituencies with an absolute change in turnout of less than 11 points. Between T2 and T3 the criterion was 17 points.

31 This general finding was also corroborated in shifts between the preceding and following general elections, Tl and T3.

32 Scarrow, “By-Elections and Public Opinion,” 87.

33 Although the regression coefficients of shifts in turnout upon Conservative change between T2 and T3 was .190, most beta weights were below .10.

34 The timing variable was dichotomized such that between Tl and T2 the by-elections categorized as close to the previous election were those held within 25 months of Tl. Between T2 and T3, contests defined as close to the following general election occurred within 22 months of T3.

35 Scarrow's findings (“By-Elections and Public Opinion,” 88) were inconsistent with the hypothesis, but were based upon more crudely derived indicators than used here.

36 The single highest regression coefficient was .276 predicting Social Credit change between T2 and T3, but most were no higher than .15.

37 The strongest simple correlation coefficient occurred when change in turnout provided a beta weight of .484 in predicting variation in the Conservative vote between T2 and T3. The multiple correlation coefficient in that particular equation was .601.