Electoral Bias and the New Democratic Party: Provincial Competition in Manitoba, Saskatchewan, and British Columbia

Transcription

1 Electoral Bias and the New Democratic Party: Provincial Competition in Manitoba, Saskatchewan, and British Columbia Tony L. Hill Department of Political Science University of Minnesota 1414 Social Sciences Tower th Ave. S. Minneapolis, Minnesota Internet: Presented at the annual convention of the Society for Socialist Studies, June 3-6, 2004, Winnipeg, Man.

2 The New Democratic Party (NDP) has proposed proportional representation (PR) on the national level in Canada. The party believes its voters would fare better under PR than under the country's single member plurality (SMP) electoral system, in which each constituency elects a single member on the basis of a plurality election. The party's belief is understandable in light of the last three elections. In 1993, the party received seven percent of the popular vote but only three percent of parliamentary seats. In 1997, 11 percent of Canadian voters chose the NDP, but the party received only seven percent of seats in the House of Commons. In 2000, the party received nine percent of the vote and only four percent of seats. By contrast, the Liberal Party received 41, 38, and 41 percent of the vote in the three elections, respectively, but won 60, 51, and 57 percent of seats. This reflects a long-noted trend in SMP elections for the biggest winners to take extra-large shares of seats. 1 Such a move to PR might cure the party's ills on the national scene; a previous study by the author 2 found that PR with a 5 percent viability threshold would have yielded the NDP (among other parties) approximately the share of seats as the popular vote would indicate. But this begs the question as to the impact of proportional representation in the three provinces in which the party is not only not a minor party but also routinely forms the 1 Tufte, p. 540; Rossiter, Johnston & Pattie, p. 466; M.G. Kendall and A. Stuart, "The Law of Cubic Proportion in Election Results," British Journal of Sociology 1: (1950). 2 Tony L. Hill, A Simulation of Proportional Representation and Instant Runoff Voting in the 1993, 1997, and 2000 Canadian General Elections, paper presented at the 17th biennial conference of the Association for Canadian Studies in the United States (ACSUS), Portland, Ore., November 17-23,

3 government: Manitoba, Saskatchewan, and British Columbia. 3 What is the impact of SMP on the NDP in those three provinces, and what aspects of the electoral system there, if any, disadvantage the party? Gudgin and Taylor define electoral bias as the difference between the proportion of votes a party receives in an election and the proportion of seats it obtains. 4 They note a paradox that befalls small parties under SMP: The party is organized on the basis of a national electorate, with a party like the NDP having support not only in concentrated pockets across the country but also at some level in every riding, yet at the same time, the elections are centred on geographically compact ridings. 5 Beginning in 1993, two regionally-based parties flourished in Canada: the western-based Reform Party (renamed the Canadian Alliance in 2000 and merged into the Conservative Party in 2004) and the Quebec nationalist Bloc Québécois. By being regionally rather than nationally based, they have a direct territorial meaning, in the words of Gudgin & Taylor. 6 On the other hand, they would classify the NDP (and also the Progressive Conservative Party) in this time period as a functional cleavage based on class interests having no direct territorial 3 The party also formed the government in Ontario from 1990 to 1995, but this is believed to have been a flash in the pan, and the party has since been relegated to its traditional third place in the province, failing to even make official party status in the 2003 election. Although the party also forms the government in the Yukon Territory occasionally, there are serious problems with analyzing electoral bias in the territory due to the vast differences in size of the provincial ridings and their mercurial population shifts caused by changing economic conditions. 4 Gudgin & Taylor, 1973, p Gudgin & Taylor, 1973, p

4 meaning. 7 Thus, the pitfall the regional parties avoided, but the NDP as a nationally based minor party could not is having substantial support, but not enough of it concentrated in local constituencies to win an equally substantial number of ridings. Rossiter, Johnston & Pattie describe this as the superimposition of a geography of constituency boundaries on the geographies of party support. 8 They put forth three broad generalizations, including that large parties benefit the most, small parties with widespread support suffer the most, and the large party with the widest spatial spread of support tends to benefit more than that with a greater concentration of its votes in certain regions and constituency types, such as inner cities. 9 This question lends the greatest theoretical support to this paper. The authors take exception to Gudgin & Taylor s definition of bias, which centres on the national share of the vote, preferring the alternative approach of Brookes, whose definition of bias is in terms of the additional number of seats the object party would have won had it won the same share of the vote as its competing party. 10 They lay out a series of equations for calculating bias, modified from Brookes by Mortimore as the Full Brookes Method. 11 Siaroff further modified this set of 6 Ibid. 7 Ibid. 8 Rossiter, Johnston & Pattie, p Ibid. 10 Rossiter, Johnston & Pattie, p Rossiter, Johnston & Pattie, pp R. Mortimore s treatment is found in his Constituency Structure and the Boundary Commission, University of Oxford, doctoral dissertation,

5 equations 12 to analyze electoral bias in Quebec, and the Siaroff set of equations are used here. In Manitoba, the Progressive Conservative Party dominates the agrarian south of the province and the NDP dominates the northern part, where fishing, mining, tourism, and forestry are more important than agriculture. Most of the voters live in Winnipeg, and the north-south demarcation has some sway within the city too. The NDP dominates the northern half of Winnipeg (including the North End, the Kildonans, and Transcona), and the older part of St. Boniface, St. Vital, and Fort Garry. The Progressive Conservative Party dominates the outer ring along the perimeter of the southern half of the city, including Charleswood, Assiniboia, Tuxedo, St. Norbert, and the newer part of St. Boniface. The party is almost irrelevant in the northern half of the city. In the 2003 election, the party won a small number of polls in North Kildonan, in the extreme northeastern corner of Winnipeg. The Liberal Party of Manitoba was competitive in only three ridings in One of these, Inkster, is in the NDP s stronghold in Northwest Winnipeg. The other, River Heights, home of the party s leader, former MP Jon Gerrard, MD, is a marginal area that often goes Tory in provincial elections. The party was also strong in a riding in Charleswood, although the Tories won it. The party won only River Heights in 1999, and only three ridings in This 12 As described in Johnston, et al, 2001, pp

6 leads to the conclusion that Manitoba is principally a two-party province, in which the Liberal Party is frequently irrelevant. Data for electoral bias show a total bias in favour of the NDP in the elections of 1999 and 2003, which the party won, of 3.76 and 7.6 seats, respectively. For 1995, which the party lost, the data show a bias against the NDP of 5.8 seats. This is consistent with the idea that electoral biases augur most heavily in favour of the biggest winner. Despite the similar direction of the total bias for the most recent two elections, there is great dissimilarity in the composition of the bias through the five separate elements. The concept of gerrymandering is perhaps inapt in the analysis of bias, since what is being measured is not principally the intentional manipulation of electoral district boundaries to favour the in-party, but rather the natural clustering of the electorate. Siaroff demonstrated a consistent pattern of bias against the Liberal Party in his research on Quebec, but the chief cause of this was the pattern whereby most Liberal voters (who are not coincidentally disproportionately Anglophone) live in Montreal s West Island, the Outaouais, or the Eastern Townships. While they can capture ridings there with phenomenal percentages, Liberals are frequently unable to make inroads in most of the province s largely Francophone ridings. This measure was the chief component of the bias in Manitoba in 1995 and 2003 but unimportant in Unlike Quebec, it cannot be said that 5

7 whether the NDP or the Tories win an election in Manitoba is due to a historic pattern of maldistribution of the electorate. Constituency size variations, however, are a substantial bias that in this three-election sample is consistently in favour of the NDP. This is due in part to the tendency of redistricters in the western provinces to make the vast ridings that occupy the north of the province much smaller in terms of population than typical ridings. (This is also true to some extent in federal redistricting.) These ridings tend to be the domain of the NDP. While the average riding in Manitoba had 6,966 actual voters at the 2003 election, Rupertsland and Flin Flon had only 2,534 and 3,300, respectively, much less than half the provincial average. The average Saskatchewan riding in 2003 had 7,347 actual voters, but Athabasca riding had 3,453. All of these are reliable NDP ridings. (In fact, the paradigm in rural areas of both provinces is that the NDP is strongest in the north, away from the U.S. border.) Redistricting in British Columbia for the 2001 election lessened the degree of these disparities relative to the other two provinces. But in the 1991 election, the average riding had 19,500 voters, and all of the ridings at the low end on this measure (including two in the 10,000 range) were in remote parts of the interior and dominated by the NDP: Peace River North, North Coast, Bulkley Valley-Stikine, Prince George-Mount Robson, and Skeena. Thus, the NDP enjoys an additional advantage if this practice is to continue. As Gudgin & Taylor put it, Clearly a party has an advantage if its vote is located in small 6

8 constituencies with low turnout and low two-party voting since it is in such places that seats can be won with relatively few votes. 13 Indeed, the NDP won Athabasca in 2003 with 2,407 votes, fewer than the Saskatchewan Party lost with in nine ridings, and even than the hapless Liberals took in two. The NDP won with only 2,203 votes in Rupertsland in 2003, and the Tories lost with more votes in 13 ridings and the Liberals failed to take a third riding with more than that number. The third element of bias, abstentions, consistently works in favour of the NDP in all three provinces for the period covered herein, with the exception of Saskatchewan in This is somewhat contrary to expectation, since the NDP has traditionally been the party of the lower class, and it is an axiom of political analysis that the lower class normally turns out to vote less consistently than the upper class, who have more typically been supporters of the opponents of the NDP. The model makes the assumption that those who don t vote are apt to vote the same as those who do, but this might be a flaw in the model. High residential turnover might also be a factor in skewing these results, as there will be more new voters in parts of the province with high turnover than low. Minor party votes (which in Manitoba principally means those for the Liberal Party) are consistently a bias against the NDP, although actual wins are a mixed bag. It isn t entirely true that the Liberals and NDP split the so-called left of 13 Gudgin & Taylor, 1980, p

9 centre vote and let the Tories win by default. Since support for Liberals is so sporadic in Manitoba, depending almost entirely on where the party can make a serious run at a small number of seats, it should be taken into consideration that there is a small N problem analyzing the Liberal vote in Manitoba. For example, the Liberal win in 1999 shows up as a bias in favour of the NDP, because it was the seat in River Heights in which the Tories were otherwise leading. Had Gerrard lost and the party s candidate in Inkster instead won, it would have shown up here as a bias against the NDP. Since the two Liberal wins in 2003 were in seats where one of each party was otherwise leading, these Liberal wins appear to cancel each other and appear as no bias in this analysis. On the other hand, all three of the Liberal wins in 1995 are counted as biases against the NDP, suggesting that vote splitting is bad for the NDP when Tories have a serious chance of forming the government but is almost irrelevant to NDP fortunes when the Tories are not strong. More data might be needed to examine how bias affected the elections of the 1980s in which the Tories won a minority government. However, the bias literature is clear that the methodology is most appropriate for two-party systems, or those where third parties do quite poorly, which is the situation in Manitoba right now. This might encapsulate the NDP s situation with regard to bias in Manitoba as a whole. Most elements of it seem to work in their favour most of the time, but critical elements work against them when Tory fortunes are rising. It is worth noting that PR 8

10 in Manitoba would energize the Liberal Party greatly, giving them access to voters provincewide instead of those in the handful of ridings they are now able to target. We can conclude that replacing SMP with PR in Manitoba would be a bad move for the NDP. Saskatchewan, The Land of Living Skies, is where the NDP (then the CCF) formed the first democratically elected socialist government in North America in the 1940s. It has been a twoparty system since the slow decline of the Liberal Party began in the late 1960s. Like Manitoba, the Liberals are frequently irrelevant in Saskatchewan, except when the election breaks so closely between the NDP and the Saskatchewan Party that Liberal seats become pivotal, as was the case in 1999 when the NDP under Roy Romanow won exactly half the seats and managed to stay in power by forming a coalition with the Liberals, the first coalition in Canada in 75 years. This happened in spite of an electoral inversion, when the Saskatchewan Party actually won the popular vote. The Liberals won no seats in 2003, and the NDP won a narrow majority of both seats and votes. Total electoral bias in Saskatchewan has been lower recently than in the other two provinces. The 1995 election is problematical for the bias analysis due to the strong three-party showing. (The bias method used herein is most effective when analyzing a two way race.) Almost all of the bias in 1995 is accounted for by what the model calls gerrymandering, or maldistribution of the voters. Interestingly, operationalizing bias for 1995 using the Progressive Conservative Party (now the 9

11 Saskatchewan Party) as the second party instead of the Liberals (which is not an orthodox application of the methodology) results in a high bias of in favour of the NDP (as opposed to against the Liberals). Although gerrymandering-maldistributive bias is an even higher in this scenario, the third party win [Tory] using the Liberals as the second party results in a bias of one seat in favour of the NDP, while the third party win [Liberal] using the Tories as the second party results in a bias of nine seats against the NDP. Effectively, this means that of the eleven seats the Liberals won in 1995, ten were in ridings where the NDP was otherwise leading, and only one was in a riding where a Tory was leading. The one-seat bias in favour of the NDP the other way is due to three of the five Tory wins being in ridings where the Liberals were leading and two being in ridings where the NDP was leading. This suggests that vote splitting between the NDP and the moribund Liberals in Saskatchewan is more salient than in Manitoba. The NDP bias was much smaller in the more recent elections. Strangely enough, the gerrymandering-maldistribution effect which was so strong for the NDP in the 1995 election turned against them in the 1999 and 2003 elections. In part, this is due to the tendency of the NDP to do well in Regina and Saskatoon and the boreal north and rather poorly in the rural south. This is a pattern very similar to neighbouring Manitoba. Federally, redistricters in Saskatchewan in the 1990s and 2000s have exploited this pattern by drawing all Regina and Saskatoon 10

12 districts outward to include large swaths of rural areas that are usually friendly to right-of-centre parties; there is not a purely urban federal riding in Saskatchewan despite the substantial share of the province s population living in the two cities. The abstention factor was key in 1999; its bias in favour of the NDP was more than enough to wipe out the bias against the party due to gerrymandering. In 2003, it swung to a small bias against the party. Overall, electoral bias has been positive for the NDP in the three elections surveyed here. The SMP system works to the party s advantage in Saskatchewan. Additionally, considering the boost PR would give to the moribund Liberals here, it would be a mistake for the NDP to change the electoral system here. British Columbia frequently defies rational political analysis. At the same time it had a socialist government running the province from 1991 to 2001, it elected mostly members of the right-wing Reform Party/Canadian Alliance to represent it in Ottawa. The 2001 election was such a massive blowout for the NDP the premier conceded the election a week early that it is a serious question how much information can be gleaned from the election results. The bias calculations show the NDP should have won 28 seats more than it did based on its vote. Thus, while the Liberals still would have won, the NDP would have been a strong opposition with 30 seats to the Liberals 49. The gerrymandering-maldistribution effect accounts for nearly all of the bias against the NDP in 2001, and nearly all of the bias in favour of 11

13 the party in Thus, the paradigm that the party that loses didn t have its votes in the right place holds true in B.C. too. As noted, the variation in constituency size built into the northern districts in the Prairie Provinces has largely been eliminated from B.C., and while this measure gave the NDP a bias of more than three seats in both of the elections it won in the 1990s, it gave them an advantage of less than a single seat in In fact, the ridings in B.C. are now much closer to each other in population than those of Manitoba and Saskatchewan, so the likelihood of constituency size making a difference for any party there are small. The abstentions factor, even in the 2001 Liberal landslide, came down consistently in favour of the NDP. The third party vote has been a source of bias against the NDP. While there were no third party wins in B.C. in either 1996 or 2001, the now-defunct Social Credit Party took seven seats in 1991, but three were in Liberal areas, and this resulted in a bias of only one seat against the NDP. Overall, the SMP system helped the NDP in the recent elections it won and wouldn t have helped the NDP avoid defeat in the election it lost. While some may argue that the party taking only two seats in 2001 is evidence of the unfairness of the electoral system, SMP in the long run helps an established party like the NDP and hurts would-be rivals for the left-of-centre vote, of which there are many in B.C. In fact, one of the reasons Premier Ujjal Dosanjh conceded defeat a week before election day in 2001 is that he was afraid the NDP would take 12

14 fewer seats than the Green Party. His strategy can be seen as a partial success, since the Green Party took no seats. While the NDPs in Manitoba and Saskatchewan were formed by farmers, workers, and northerners, in B.C. the party is frequently an uneasy association of various sympathizers to left wing causes who don t share the traditional ties to the party found on the prairies. As such, they could be more easily attracted to some other protest party than their copartisans in Manitoba and Saskatchewan could be. B.C. is the kind of place where one can envisage an Israel-like din of small parties competing for seats under PR. The NDP (although not currently in a position to do anything about it) would be foolhardy to support changes to the electoral system that would threaten its role as the province s protest party in a province where protest rivals soccer as a major sport. In conclusion, while PR would give the NDP a boost in Ottawa were it installed for national elections, no one believes that the NDP would ever form the national government under PR. Thus, while PR might make the NDP a bigger opposition than it is now, it would still be merely an opposition. However, the NDP governs three provinces in western Canada more than half the time. Tinkering with the electoral systems there is apt to not only give an edge to the party s usual opponent, it might also boost nonviable rivals like the Liberals in Manitoba and Saskatchewan and the Green Party in British Columbia to the point where they could displace the NDP. Many observers would find it 13

15 hypocritical for the NDP to advocate sweeping reform of the federal electoral system while simultaneously discouraging such reform in its own houses. It would be rational for the NDP to cease advocating electoral reform for the national Parliament. REFERENCES: R. H. Brookes, Electoral Distortion in New Zealand, Australian Journal of Politics and History 5: (1959).. The Analysis of Distorted Representation in Two Party, Single Member Elections, Political Science 12: (1960). Graham Gudgin & Peter J. Taylor, Electoral Bias and the Distribution of Party Voters, Transactions on the Institute of British Geographers 63: (1974).. The Decomposition of Electoral Bias in a Plurality Election, British Journal of Political Science 10: (1980). Tony L. Hill, Canadian Politics, Riding by Riding: An In-Depth Analysis of Canada's 301 Federal Electoral Districts, Minneapolis: Prospect Park Press, ISBN Ron J. Johnston, Spatial Structure, Plurality Systems and Electoral Bias, Canadian Geographer 20: (1976). Ron J. Johnston, Charles J. Pattie, Danny F.L. Dorling and David J. Rossiter, From Votes to Seats: The United Kingdom's Electoral System in Operation Since Manchester, U.K.: Manchester University Press, David Rossiter, Ron J. Johnston and Charles J. Pattie, Redistricting and Electoral Bias in Great Britain, British Journal of Political Science 27: (1997). Alan Siaroff, Electoral Bias in Quebec Since 1936, paper presented at the 17th biennial conference of the Association for Canadian Studies in the United States (ACSUS), Portland, Ore., November 19-23, Edward R. Tufte, The Relationship Between Seats and Votes in Two-Party Systems, American Political Science Review 67: (1973). The author acknowledges the assistance of Polly Hrenyk of Elections Saskatchewan in providing election data in a timely manner. *** 14

16 Table 1. Electoral bias in Manitoba, VAR DESCRIPTION x NUM SEATS WON BY PARTY A (NDP) y NUM SEATS WON BY PARTY B (PC) b NUM OF SEATS WHERE A LEADS B f NUM OF SEATS WHERE B LEADS A P AVG NO OF VOTES WHERE A LEADS B Q AVG NO OF VOTES WHERE B LEADS A R AVG REGISTERED ELECTORATE IN SEATS WHERE A>B S AVG REGISTERED ELECTORATE IN SEATS WHERE B>A C AVG NO OF ABSTENTIONS WHERE A>B D AVG NO OF ABSTENTIONS WHERE B>A U AVG NO MINOR PARTY VOTES WHERE A>B V AVG NO MINOR PARTY VOTES WHERE B>A G SEATS pos is bias in favour of NDP; neg is bias against CSV SEATS pos is bias in favour of NDP; neg is bias against A SEATS pos is bias in favour of NDP; neg is bias against TPV SEATS pos is bias in favour of NDP; neg is bias against TPW SEATS pos is bias in favour of NDP; neg is bias against Bias SEATS pos is bias in favour of NDP; neg is bias against Table 2. Electoral bias in Saskatchewan, VAR DESCRIPTION x NUM SEATS WON BY PARTY A (NDP) y NUM SEATS WON BY PARTY B (Lib 95; Sask Pty 99/03) b NUM OF SEATS WHERE A LEADS B f NUM OF SEATS WHERE B LEADS A P AVG NO OF VOTES WHERE A LEADS B Q AVG NO OF VOTES WHERE B LEADS A R AVG REGISTERED ELECTORATE IN SEATS WHERE A>B S AVG REGISTERED ELECTORATE IN SEATS WHERE B>A C AVG NO OF ABSTENTIONS WHERE A>B D AVG NO OF ABSTENTIONS WHERE B>A U AVG NO MINOR PARTY VOTES WHERE A>B V AVG NO MINOR PARTY VOTES WHERE B>A G SEATS pos is bias in favour of NDP; neg is bias against CSV SEATS pos is bias in favour of NDP; neg is bias against A SEATS pos is bias in favour of NDP; neg is bias against TPV SEATS pos is bias in favour of NDP; neg is bias against TPW SEATS pos is bias in favour of NDP; neg is bias against Bias SEATS pos is bias in favour of NDP; neg is bias against

17 Table 3. Electoral bias in British Columbia, VAR DESCRIPTION x NUM SEATS WON BY PARTY A (NDP) y NUM SEATS WON BY PARTY B (Liberal) b NUM OF SEATS WHERE A LEADS B f NUM OF SEATS WHERE B LEADS A P AVG NO OF VOTES WHERE A LEADS B Q AVG NO OF VOTES WHERE B LEADS A R AVG REGISTERED ELECTORA TE IN SEATS WHERE A>B S AVG REGISTERED ELECTORATE IN SEATS WHERE B>A C AVG NO OF ABSTENTIONS WHERE A>B D AVG NO OF ABSTENTIONS WHERE B>A U AVG NO MINOR PARTY VOTES WHERE A>B V AVG NO MINOR PARTY VOTES WHERE B>A G SEATS pos is bias in favour of NDP; neg is bias against CSV SEATS pos is bias in favour of NDP; neg is bias against A SEATS pos is bias in favour of NDP; neg is bias against TPV SEATS pos is bias in favour of NDP; neg is bias against TPW SEATS pos is bias in favour of NDP; neg is bias against Bias SEATS pos is bias in favour of NDP; neg is bias against

18 Appendix. Equations of bias components x = number of seats won by NDP y = number of seats won by other major party (Progressive Conservative Party in Manitoba; Liberal Party in British Colunbia and Saskatchewan 1995; Saskatchewan Party in Saskatchewan 1999 and 2003) b = number of seats where NDP leads other major party f = number of seats where other major party leads NDP P = average number of combined votes for two major parties where NDP leads other party Q = average number of combined votes for two major parties where other party leads NDP R = average registered electorate in seats where NDP leads S = average registered electorate in seats where other party leads C = average number of abstentions in seats where NDP leads D = average number of abstentions in seats where other party leads U = average number of minor party votes in seats where NDP leads V = average number of minor party votes in seats where other party leads G = gerrymander effect CSV = constituency size variations (malapportionment effect) A = abstentions effect TPV = third party votes effect TPW = third party wins effect G = {[f(pb/qf-1)] [b(qf/pb-1)]} / 2 CSV = {[f(s/r-1)] [b(r/s-1)]} / 2 A = {f*[(r/(r-c))*[(c/r)-(d/s)]]-b*[(s/(s-d))*[(d/s)-(c/r)]]} / 2 TPV = {f*[(r/(r-u))*[(u/r)-(v/s)]]-b*[(s/(s-v))*[(v/s)-(u/r)]]} / 2 TPW = (x-b)-(y-f) Source: Alan Siaroff, "Electoral Bias in Quebec Since 1936," paper presented at the 17th biennial conference of the Association for Canadian Studies in the United States (ACSUS), Portland, Ore., November 19-23, 2003, p. 17, adapted from Johnston, et. al., From Votes to Seats: The United Kingdom's Electoral System in Operation Since Manchester, U.K.: Manchester University Press, 2001, pp