Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Social cleavage and the party pie : the relationships between social heterogeneity and party systems… Tanaka, Kashi 2005

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_2006-0114.pdf [ 5.86MB ]
JSON: 831-1.0092493.json
JSON-LD: 831-1.0092493-ld.json
RDF/XML (Pretty): 831-1.0092493-rdf.xml
RDF/JSON: 831-1.0092493-rdf.json
Turtle: 831-1.0092493-turtle.txt
N-Triples: 831-1.0092493-rdf-ntriples.txt
Original Record: 831-1.0092493-source.json
Full Text

Full Text

Social Cleavage and the Party Pie: The Relationships between Social Heterogeneity and Party Systems in Canadian Provinces by Kashi Tanaka B .A . (Honours) University of British Columbia, 2003 THESIS SUBMITTED IN P A R T I A L F U L F I L L M E N T OF T H E R E Q U I R E M E N T FOR THE D E G R E E OF M A S T E R OF ARTS in The Faculty of Graduate Studies (Political Science) UN1VERISTY OF BRITISH C O L U M B I A December 2005 © Kashi Tanaka 2005 Abst rac t One of the purposes of political parties is to reduce a heterogeneous polity into a few political elements. This thesis determines i f there is a relationship between political parties and social cleavages in Canada. I have used provincial election results and census data from 1956 to 1991. Electoral results are converted into two measures of party system size, the effective number of parties (the number of significant parties in a legislature) and the competitive number of parties (the number of relevant parties in an election). Social heterogeneity is measured by converting census data into a series of indexes that measure the ethnic, religious, and linguistic diversity. 1 also examine the affect of rural/urban and centre periphery cleavages in provincial politics. I have found that there is a significant relationship between social heterogeneity and party system size in Canada. Of the cleavage structure examined, ethnicity is positively correlated with party system size and the size of a province's rural population is negatively correlated with party system size. Curiously, religion and language have mixed affects; religion is positively correlated with the number of parties that get elected but negatively correlated with the number of parties that win seats. Similarly, the size of a provinces French speaking population has a positive relationship with the number of parties that win seats but a negative relationship with the vote distribution among parties. There are two important conclusions in this thesis. First, there is substantial evidence that social heterogeneity influences party systems size in Canadian provinces. This result challenges institutional explanations which suggest that party systems in polities that use ii plurality electoral systems which elect single members will not be affected by social diversity. M y second conclusion is the identification of a largely untouched area of research on provincial party systems. European theorists have used social structural approaches for fifty years to explain how societies and political parties co-evolve. This thesis proves that this approach has an important role to play on this side of the Atlantic. iii Table of Contents Abstract Table of Contents Lists of Tables Lists of Figures Acknowledgements Dedication Chapter I - The shape of provincial party systems 1 Measuring parties 4 The unit of analysis - Canadian Provinces (1951-1996) 6 Determinants of ENP and CNP 11 Institutional effects - Electoral systems 11 Institutional effects - District magnitude 15 Chapter II - The shoulders of social structural giants 19 Social structural models 20 Critiques of social explanations 26 The value cleavage 28 Explanations of party system development in Canada 31 Summary 36 Chapter III - The building blocks of party systems 39 Building an index 40 Ethnic diversity 43 Religion 49 Language 53 The proximity cleavage in Canada 56 Rural/Urban 60 Methodology 63 Chapter IV - Social heterogeneity and party systems 66 Social heterogeneity and ENP 67 Social heterogeneity and C N P 71 Comparing ENP and C N P models 75 Testing the models 79 Summary 84 Works Cited «7 ii iv v vi vii iv Lists of Tables Table 1.1 ENP and CNP for Provinces 1951 -1996 (CNP in italics) Table 1.2 District magnitude for Provinces with a history of employing multi-member ridings (1951-1996) Table 3.1 Average size of ethnic population for single respondents by province as a percentage (1951-1996) Table 4.1 Estimates of the Effective Number of Parties (ENP) in Canadian Provinces (1951-1996) Table 4.2 Estimates of the Competitive Number of Parties (CNP) in Canadian Provinces (1951-1996) Table 4.3 Estimated ENP for selected provincial cases - with comparisons between early period and late period in study Table 4.4 Estimated CNP for selected provincial cases - with comparisons between early period and late period in study Lists of Figures Figure 1.1 ENP and CNP for Provinces (1951-1996) 9 Figure 2.1 Lipset and Rokkan's model of social cleavage structures 22 Figure 3.1 Percentage of Catholic population by region 50 Figure 3.2 Percentage of Protestant population by region 51 Figure 3.3 Estimated number of religious groups by region 53 Figure 3.4 Adjusted linguistic index for Quebec and New Brunswick 55 Figure 3.5 Approximated magnitude of centre-periphery cleavage in provinces 58 Figure 3.6 Rural population as a percentage 61 vi Acknowledgements I have had four wonderful years at U B C and a large part of that is due to the excellent teaching and mentorship I received from the faculty members I have worked with. The department continually attracts tremendously bright people from around the word to learn and to be challenged. The quality of the students in the Honours and Graduate programs has always impressed and humbled me. I am grateful for the friendship and support my peers gave me. In addition, there are several people I owe particular thanks. I would like to thank my thesis supervisor, Dr. Fred Cutler, whose insight and soccer abilities were equally inspirational. I owe thanks to two of my past teachers. Mrs Beverly Reid, who convinced me I was smart and stimulated a life of curiosity and Mr Raj Tour whose passion and integrity will continue to inspire me. I must thank Cam Gross, Don Colangelo, E l i Walker, Jeff Bell, Kristen Stevenson, Lindsey Galvin, Paul Allanson, Tim Bottomer, and Scott Graham whose camaraderie, sober and not so sober judgement, and above all else, love and friendship for the past two years I will spend a lifetime repaying. Finally, I would like to thank my family which never wavered in their love and support. This thesis is dedicated to them. vii To my sister, Nikko and my parents, Lynn and Yosh viii Chapter I - The shape of provincial party systems Generating a list o f divisive topics in any polity is rarely difficult. In Canada the hardest part would be to know where to begin, or perhaps more accurately where to end. A t any point in time there are thousands o f potential points o f conflict that have the potential to drive people apart. Some of the most basic divisions in Canada are social cleavages such as race, religion, and language. Effective governance requires reducing heterogeneous polity's to a few political elements of which the most salient are political parties. The central question o f this thesis is to determine i f there is a relationship between political parties and social cleavages. The reduction o f political views into a handful of political parties is well documented by spatial theories.1 Spatial models elegantly demonstrate the mechanism with which the complex calculation o f individual preferences is translated into partisanship, but this does not eliminate the impact of social diversity. This thesis w i l l demonstrate that "the relationship between electoral rules and party systems is not mechanical and automatic: A particular electoral regime does not necessarily produce a particular party system; it merely exerts pressure in the direction o f this system." 2 Research on party systems indicates that the number of political parties that successfully participate in a democracy is a product o f both institutional and cultural elements: in particular electoral system design, district magnitude, and societal cleavages. 3 1 The roots of spatial theories of party systems are with Downs. Variations and critiques of his work are found in Hinich and Enelow (1984), and Rabinowitz and MacDonald (1989). 2 Duverger (1964), pg. 40. 3 See Cox and Neto (1997), Rae (1971), Lijphart (1993), and Taagepera and Shugart (1993). 1 The connection between social cleavages and political parties is a relatively unexplored area o f Canadian politics. The absence of literature would suggest there is very little connection at all . Are societal cleavages reflected in party systems? Or are parties aggregates o f a multitude of constituencies concerns? Unt i l now a common convention in Canadian politics has been that, at a federal level, the major political parties act as omnibuses aggregating political concerns. 4 This is a persuasive argument. However, this thesis w i l l argue that the omnibus theory misses out on part o f the explanation because it is arguable that parties are also a product o f social cleavages. None o f the major parties in Canada claim direct connections religious or ethnic groups. However, parties like these have emerged in several democracies; the governing Christian Democrats in Germany and BJP in India are examples. Moreover, while the major Canadian parties do not advertise connections to specific populations, students of Canadian politics undoubtedly recognize that these connections exist. The relationship between Catholics and the federal Liberal party is an example which w i l l be looked later on. This thesis w i l l explore these connections and other to determine the extent to which societal cleavages influence party systems in provinces. The defining characteristics o f party systems are the participants, the nature o f interaction between these participants and the voting population, and how both o f these change over time. Two measures o f party and voter participation are employed in this paper, the effective number of parties (ENP) and the competitive number of parties (CNP) . E N P measures the number of significant parties in a legislature, whereas C N P measures the number o f relevant parties in an election. 4 See Meisel (1992), 328-350; and Carty (1995) 187-202. 2 Current literature argues that the degree of variation caused by social heterogeneity is limited when party systems are constrained by electoral ridings with small district magnitudes and the mechanical effects of plurality electoral systems. In the summary o f their work, Ordeshook and Shvetsova argue that, " i f district magnitude equals one, then the party system is relatively 'impervious' to ethnic and linguistic heterogeneity."5 Combined with the restrictive effects of the plurality electoral system, their findings suggest that the effects o f social heterogeneity on E N P and C N P may be overwhelmed by the constricting effects o f institutional variables. This paper looks at two interrelated questions. First, do E N P and C N P vary in systems with low district magnitudes and plurality electoral formats? A n d second, i f there is evidence of variation, which variables explain this variation? Canadian provinces provide a unique opportunity for studies o f party systems because it is possible, for the most part, to control for the effects o f electoral systems and district magnitude. A t present every Canadian province uses the plurality formula and single-member districts to elects its representatives. A t various times, provinces have experimented with different electoral arrangements such as multi-member districts and majoritarian electoral formulas. However, there are few instances o f these elections in the latter half o f the last century. A t the same time the provinces are not demographically uniform. The ethnic, linguistic and religious make-up o f provinces varies considerably. The central hypothesis is that social cleavages influence party systems. Therefore, the institutional similarity and social diversity o f provinces make them an ideal crucible to test to this hypothesis. Ordeshook and Shvetsova (1994), pg. 122. 3 Measuring Parties Explaining the connection between social heterogeneity and party systems first requires determining how party systems can be measured. One way to is to establish who is participating and the power the each party has in a legislature. To this end counting the number o f parties that operate in a political system would be the simplest place to start.. Two basic methods for counting parties would be to sum all the parties that run in elections or to sum all the parties that win seats. In both cases the power o f these measures to explain party system size is questionable. A n example o f an electoral result from Quebec provides a reasonable justification for this hesitation. In the 1989 provincial election sixteen parties contested the election and three won seats in the legislature. These figures provide some idea of the range o f competition but they are not indicative o f the actual election or the resultant legislature. Nearly 99% of the votes were cast for the three parties that won seats and, o f those parties; one received almost 75% of the seats. The standard alternative is to use some form of index or approximation derived from Douglas Rae's measure of fractionalization. His measure, presented mathematically is: F. =1 ( " \ • I? V >=i J Where: T, is group i's decimal share o f the total population. 4 The index has several advantages, "first, it is sensitive to both the number and the relative equality of the party shares. Second, it w i l l allow us to consider systems with any number of parties, and to compare the resulting states o f fractionalization . . . and third, it may prove to be a convenient device for . . . [examining] the general concept o f fractionalization in other concepts."6 The first two advantages apply directly to the question at hand while the last advantage w i l l be effectively demonstrated in Chapter 3. In this later chapter fractionalization indexes are used as an elegantly simple way of empirically examining changes in patterns of social cleavage. This paper uses a variation of Rae's index proposed by Taagepera and Shugart who manipulate the formula to produce a number which is more intuitively appealing. 7 Their formula for determining the E N P is: E N P = n HP? 1=1 Where: p is the percentage of the seats a party receives In this paper, this formula is also used to determine the C N P , where p becomes the percentage of votes that a party receives province-wide. 8 6Rae(1971),pg. 56. 7 Taagepera and Shugart (1989), and Rae (1971), pgs. 54-58. For example in a hypothetical political system with 100 seats i f two parties split the seats 50-50, the ENP is 2. Similarly, i f three parties split the seats 33-33-34, the ENP is nearly 3. Even i f seat allocation is not symmetrical, the formula still produces useful results. In an election where the seat allocation is 60-30-10, the ENP is 2.18, indicating that the combined effectiveness of the second and third party is, as would be expected, less than two. 8 For this project, only parties that obtained at least one-percent of the popular vote were included in calculations. 5 Consideration o f the C N P has generally been neglected in research because it includes non-elected parties that have an arguably limited degree of influence in party systems. In addition, the majority of studies of party systems are comparative works that include proportional electoral systems in which E N P is, by and large, equivalent to C N P . 9 Arend Lijphart argues that looking at the number of parties that receive votes "is the better indicator o f the long-term nature o f the party system" because party popularity best represents the concerns o f a pol i ty . 1 0 This thesis considers both measurements because provincial party systems are shaped by the plurality electoral system which disproportionately allocates seats to votes. Using two measures o f a party system—one that includes all parties and another that indicates which parties are succeeding in finding seats in legislatures—provides a more complete account o f Canadian party politics. The Unit of Analysis - Canadian Provinces (1951-1996) The data employed to answer the questions of this thesis are provincial election results from 1951 to 1996. The elections are paired with census enumerations which, in Canada, occur on the first and sixth year o f every decade. In periods where an election does not occur in a census year the next chronological election is used. In situations where two or more elections fall within a census period all elections after the first election are dropped from the study. Because of this treatment it is important to note that any effects attributable to time are measured in periods between censuses and not 9 Proportional systems where ENP and CNP diverge significantly are those with higher electoral thresholds (>3%). 1 0 Lijphart (1993), pg. 483. 6 elections. Wi th this data it is possible to consider how the evolution o f provincial societies affects party systems. Table 1.1 - E N P and C N P for Provinces 1951-1996 {CNP in italics) Census B C m § K M g Q N Q C N B N S N F year ^ 1996 1991 1986 1981 1976 1971 1966 1961 1956 1951 2.151 1.603 2.204 2.047 2.216 1.970 1.485 2.445 1.882 2.935 2.618 2.840 2.581 2.636 2.545 2.348 3.011 2.412 1.914 1.900 1.393 2.167 2.135 1.920 1.536 1.600 1.825 2.985 2.725 2.627 2.846 2.910 2.485 3.113 2.710 2.373 1.768 1.724 1.979 2.060 1.753 1.676 1.000 2.198 1.929 2.326 2.762 2.410 2.856 2.849 2.408 2.228 2.698 2.208 1.967 1.108 1.319 1.928 2.405 1.823 1.822 1.756 1.352 2.205 2.228 2.281 2.382 2.830 2.206 2.506 2.849 2.015 1.985 1.136 1.672 2.007 2.786 2.072 1.998 2.136 1.865 2.250 2.537 2.535 2.459 2.934 3.135 2.534 2.642 2.356 1.923 1.858 1.600 2.277 2.000 1.159 1.991 1.899 2.095 3.349 2.552 2.050 2.859 2.826 2.469 2.254 2.504 2.202 2.190 1.376 1.933 2.540 2.303 2.068 1.979 1.293 1.153 2.723 3.245 2.408 3.092 2.887 2.538 1.993 2.175 2.131 2.064 1.101 2.039 2.145 1.780 1.830 1.899 1.203 1.463 3.349 2.704 2.808 2.807 2.642 1.994 2.039 2.104 2.086 1.663 1.134 1.871 2.803 1.730 1.549 1.696 2.052 1.246 2.946 2.546 2.967 3.010 2.650 2.111 2.073 2.121 1.814 2.309 - 1.490 2.290 1.284 1.642 1.742 2.022 1.324 3.313 - 2.244 3.506 2.710 2.116 2.082 2.326 1.766 1 There are clear trends in the E N P and C N P in provinces. The average E N P across time and provinces is 1.83 with a standard deviation o f 0.36. C N P across time and provinces is 2.56 with a standard deviation of 0.36. Regression analysis o f both sets of data on time indicates that the average E N P and C N P increased between 1951 and 1996; E N P has increased by 0.20 and C N P by 0.15. The E N P figures in Table 1.1 provide an overview o f the size of Canadian provincial party systems between 1951 and 1996. 1 1 Party systems range from systems dominated by single parties to systems with consistent multi-party competition. Evidence of single-party domination in Alberta is clear with an average E N P of less than 1.5. The province has been governed by parties with overwhelming electoral majorities since 1930. A part o f this dominance is explained by the somewhat higher average C N P for the province, indicating a fractionalization o f support for the opposition parties. 1 2 Single-party dominance of a lesser magnitude is also found in Newfoundland and New Brunswick. In Newfoundland the party system has typically been dominated by a single party, beginning with the Liberals in 1949, shifting to the Conservative party in 1972, and back to the Liberals in 1989. In New Brunswick a legitimate two party system that spans much o f the sixties and seventies is flanked by a period o f power by the Conservative prior to this period and even stronger period of dominance by the Liberals after. A clear example of this dominance was manifested in 1987 when Frank 1 1 The province of Prince Edward Island has been excluded from this thesis due to variances in size and electoral systems. The population of PEI only recently exceeded 100,000 which have meant that provincial politics bears more resemblance to large municipalities than most other provinces. Indeed, a long-standing anecdote characterizing PEI politics goes that when the local Member wants to know what his constituents think he simply opens his back door. The province also employed a unique electoral system from the beginning of this study until 1996 where the province was divided into 13 districts with each riding electing a Member and a Councilman. 1 2 In every election considered in this study at least two parties competing against the Social Credit pre-1967 or the Progressive Parties after 1967 won at least 10% of the popular vote, in two elections three parties other than the eventual winner managed to winner this amount. 8 McKenna's party swept every seat in the province. For both of these provinces the narrower range of ENP is mirrored by a narrow range of CNP. As has been the case since Confederation, party systems in the Maritime region have predominantly been a two-horse race between the Liberals and the Conservatives. A multi-party landscape has been the norm in Manitoba and Ontario. Both had high ENP and CNP averages between 1951 and 1996 which reflects the competition among the Conservatives, Liberals, and New Democrats in both provinces. In these provinces there are only a handful of cases where all three parties did not win at least 15% of the popular vote and at least some seats. In additions, all three parties have enjoyed power at least once during the past fifty years. Figure 1.1 - ENP and CNP for Provinces (1951-1996) High BC MB ON CNP AB SK NS Low NB NF QC Low E N P High For ENP the designation of high, medium, and low were determined as: above 2.0, between 1.99-1.75, and below 1.74. For CNP the designation of high, medium, and low were determined as: above 2.75; between 2.74-2.50, and below 2.49. 9 The remaining four provinces fall into the middle range in E N P but are spread across the C N P category. Figure 1.1 demonstrates the variation in these provinces. In Saskatchewan and Nova Scotia the difference between E N P and C N P is fairly consistent with E N P typically 20-25% smaller than the C N P . There are different reasons for this in each province, the figures for Saskatchewan are the result of a volatile political arena. The party system has shifted several times with four different parties achieving power over the past fifty years. Generally it could be characterized as a party system with two strong parties and a weak party. Nova Scotia's results are a product o f static electoral results. The party system has been dominated by the Conservatives and the Liberals while the N D P has tenaciously been able to hold on to a small but significant wedge o f the electorate. E N P for the province hovers just below 2 while the C N P sits above 2.5. British Columbia and Quebec find themselves on opposite ends o f the C N P spectrum even though they have similar average E N P . Both provinces have seen parties emerge and fade and both have a variety of parties assume power. The main difference is the distribution of votes and how these are translated to seats in each province. In British Columbia elections have been a battle between two main parties with one or two other minor competitors. In all but three elections the two large parties have failed to win 80% of the combined popular vote. In Quebec the battle between two main parties is also present; however, these two parties typically win over 90% o f the popular vote. The key difference between the two provinces is that with a far smaller piece o f the electoral pie smaller parties in Quebec have found as much success winning seats as smaller parties in British Columbia. 1 3 This figure includes both the Cooperative Commonwealth Federation and the New Democratic Party which are arguably too similar in political ideology and support to differentiate. 10 Determinants of ENP and CNP There is evidence o f variation over time and space in E N P and C N P . It is clear that there are variations across provinces. Where does the variation come from? This section looks at how institutional factors shape party systems. The list o f independent variables in the literature regarding the causes for the proliferation or contraction of the number o f parties that operate in a polity is fairly narrow. Douglas Rae suggests that, we can generally limit our focus to those factors that are "social, economic, legal, and pol i t ical ." 1 4 Past research on the size and competitiveness of party systems point to three types o f variables that are influential: type o f electoral system, district magnitude, and societal cleavages, the last o f which can be either territorial or socia l . 1 5 Institutional Effects - Electoral Systems Elections serve as a mediating institution between individual voters and political parties. N o two states use exactly the same electoral formula; each system takes into account local characteristics and idiosyncrasies. A l l electoral systems have at least some distorting effect on the translation of votes to seats. Highly proportional systems such as those employed the Netherlands and Israel seek to minimize this distortion to the point where the level is negligible, while majoritarian or plurality systems such o f those widely used in Anglo-American democracies tend to produce significantly disproportional 1 4 Rae, pg 141. 1 5 Taagepera and Shugart (1989 and 1993), Lijphart (1993), Ordeshook and Shvetsova (1994), and Cox and Neto (1997). It has been pointed out by Lijphart (1990) that ballot structure can play a significant role in the proportionality of an electoral outcome; his analysis is largely limited to proportional electoral formulas which have minimal application to this study. 11 results. In these states the electoral systems tend to reward the largest vote getter in an election with the majority of seats and punish parties that receive fewer votes and that are not regionally concentrated. This "mechanical" effect has an obvious influence on party systems. In his seminal work, Political Parties, Duverger argues that, "the simple-majority single-ballot system favours the two-party system."16 Duverger goes on to suggest that this is nearly "a true sociological law." His hypothesis has been subject to criticism; however, a general aspect of his assertion remains true. Rae's text was among the first to provide empirical evidence that there is a negative correlation between plurality electoral systems and the ENP. However, some of Rae's work calls into question the universality of Duverger's law. Rae points out that it is not hard to find exceptions to the plurality/two-party relationship as demonstrated by the multi-party electoral competitions in India and Canada. These exceptions do not undo Duverger's work; undoubtedly it would be unjust to throw the baby out with the bathwater. It is clear that majoritarian and plurality electoral systems, in contrast to proportional systems, involve strong forces that result in the reduction of the number of political parties. Multiple studies provide evidence that there is a negative correlation between plurality electoral systems and the number of parties that get elected and vice versa for proportional systems. Further support for this hypothesis is provided by Lijphart and Cox whose work tempers the absoluteness of Duverger's assertion but nevertheless reinforces the direction of the causal relationship. More parties gain legislative representation in proportional electoral systems than they do in plurality electoral systems. 1 6 Duverger (1964), pg. 217. 1 7 Rae (1971), pgs. 133-140; Lijphart (1993), pgs. 484-485; and Gox (1999), pg. 15. 12 The explanation provided by Duverger of why ceteris paribus, party systems experience constraint in majoritarian systems and not proportional systems is a result of two factors, one he terms "mechanical" and the other "psychological." The mechanical factor consists o f the mathematical effects of the application o f an electoral system that rewards the largest party in a constituency at the expense of all others. The psychological factor consists o f the reactions, adaptations, and strategic reasoning o f both voters and parties to the aforementioned mechanical suppression. It is important to distinguish that the effects identified by Duverger differ between the national and constituency level. The constraining factors generally have the most significant effect at a local level but considerations o f party systems refer to results at a national level. Duverger himself was conscious o f this, stating that "the true effect o f the simple-majority system is limited to local bipartism." 1 8 For the purpose o f this thesis there are few examples o f deviation from the plurality, first-past-the-post electoral formula among provinces. 1 9 In two instances covered in this study a completely different electoral system was employed: In Alberta from 1921 to 1959, the province employed a mixed electoral formula that used a multi-member proportional system in urban areas and a preferential system in rural areas. In British Columbia the elections in 1952 and the following year were conducted using an alternative vote system across the province. To address this inconsistency the 1951 election for Alberta has been excluded. For the British Columbian case the results o f the first two elections after 1951 have been skipped, so the next election that uses the single-member plurality formula is the starting point in B C . 1 8 Duverger (1964), pg. 223; and Chhibber and Kollman (2004). 1 9 The notable exception here is PEI which was excluded for reasons outlined previously. 13 The last deviation is found in Manitoba which used a mixed electoral system until 1958. In elections up until this point the province elected members from rural ridings using the single-member plurality formula and used an alternative vote system in the urban regions of Winnipeg and St Boniface. In the two elections used in this data set there were three ridings in Winnipeg each electing four members and a single riding in St Boniface electing two members. Given that in the two elections considered the overall number o f seats that are not decided using the single-member plurality formula does not exceed 25% the results from the two elections have been included. A n area of some concern is the dynamics o f party system realignment and political socialization after electoral system change. In Manitoba as previously mentioned the number of seats affected by this change is not large. In Brit ish Columbia the use o f the alternative vote system was relatively short lived, so voters would have been quite familiar with the plurality formula when the system was reintroduced even i f they were not as familiar with the parties. 2 0 Alberta poses a slightly more complicated problem because o f the long period in which the entire province used the alternative-vote/single-transferable-vote system. Previous studies suggest that effect of the outcomes of the alternative vote system is not too dissimilar from plurality elections. 2 1 The effect of realignment o f the voting population is not directly considered in this paper and as such this is a potential point o f error for this study. The Alternative Vote system was originally brought in as a way to consolidate support around the two right-leaning parties in the province, the Conservatives and the Liberals. The system was meant to shut-out the New Democrats. Things did not go according to plan as the Social Credit Party emerged and won a minority government. A quick re-election brought about a majority for the new party simultaneously ushering in a 30 year period of government dominance by the Social Credits and banishing the Liberals and Conservative to the political hinterland. 2 1 Koop and Tanaka (2003). 14 Institutional Effects - District Magnitude The second institutional factor that has an influential impact on party systems is district magnitude ( D M ) , that is, the number of seats allocated to an individual electoral district. States use a variety of districting policies that range from the exclusive use o f single-member districts to systems where the entire country operates as a single district. Anglo-American states sit at one end of this spectrum, where for the most part each legislator represents a single district. 2 2 A t the other end o f the spectrum, as previously mentioned, are countries such as Israel and the Netherlands, in which the entire country is considered a single district. In these cases D M is equal to the total number o f members of the legislature. Countries fall in between these two extremes and it is possible for there to variation in D M within a country as is seen in Irish national and Australian senate elections under single-transferable vote formulae, and in the list component o f German elections for the Bundestag. While there is debate within the literature with regard to the scale o f influence that district magnitude has on party systems, the correlation between the two variables is positive. 2 3 The larger the D M the more proportional the outcome o f the election is likely to be. That is, the more seats per district, the closer the approximation between a political party's percentage o f the vote and the number o f seats that party receives in the legislature. There are two reasons for this effect: First, the vote quota that candidates must obtain is lower and, therefore, parties with smaller popular vote shares have a greater chance o f electoral success. However, success in these cases is dependent upon 2 2 Exceptions to this rule have occurred in Canada since the time of Confederations at both a national and sub-national level. At present though there are no multi-member ridings employed in Canadian elections. 2 3 Ordeshook and Shvetsova (1994), Rae (1971), Sartori (1986), and Taagepera and Shugart (1993). 1 5 the use o f a proportional electoral system within the district. Some polities have employed two- and three-member districts that were decided on the basis o f plurality. In these ridings, a plurality of votes would win all the seats in the riding, thereby negating any proportionalizing influence that increased D M would supposedly have had. The second reason is an extension of the first, much like the way in which Duverger explained that the psychological effects o f an electoral system are dependent on the mechanical effects. Because electoral systems with larger D M s tend to be more proportional, this gives ideological and social groups with smaller, diffused popular support more incentive to organize politically. It is interesting to note the connections that can be drawn between institutional influences o f party systems and social forces. More proportional systems or systems that are designed to allow smaller parties a greater chance of electoral success have a permissiveness that opens up possibilities for niche parties. Parties that represent minority groups, singular interests, or geographically dispersed populations have found success in the substantially more proportional electoral systems o f Western Europe than in Anglo-American polities. These types of parties have not been shut out in systems that employ less proportional systems but where they have found success this has been through geographic concentration as seen with the Scottish National Party and Plaid Cymru in British elections. For the most part D M in provinces examined in this thesis is constant. Table 1.2 reports the average D M for the four provinces whose electoral systems have included ridings with more than one elected representative. British Columbia was the most adventurous o f the four provinces employing a variety of multi-member districts from 16 1871 to 1986. During this period, two- and three-member districts were used in several urban ridings. The high-water mark for use of multi-member districts was 1986 where 17 seats returned two members. Manitoba's use o f multi-member districts corresponds with the use o f proportional electoral formulas in Winnipeg and St. Boniface. Nova Scotia and Newfoundland's use of multi-member districts since 1951 has been limited to handful of urban two-member ridings. Table 1 .2- District magnitude for Provinces with a history o f employing multi-member ridings(1951-1996) Census British . , Nova , T ~ , , , i u - Manitoba 0 .. Newfoundland Year Columbia Scotia 1996 1 1 1 1 1991 1 1 1 1 1986 1.327 (17,0) a 1 1 1 1981 1.140 (7,0) a 1 1 1 1976 1.140 (7,0) a 1 1.040 (2) c 1 1971 1.146 (7,0) a 1 1.069 (3) c 1.024 ( l ) c 1966 1.146 (7,0) a 1 1.069 (3) c 1.024 ( l ) c 1961 1.238 (6,2) a 1 1.075 (3) c 1.029 ( l ) c 1956 1.238 (6,2) a 1.213 (3 , l ) b 1.075 (3) c 1.029 ( l ) c 1951 1.171 (3,2) a 1.213 ( 3 , l ) b 1.156 (5) c 1.120 (3) c a. Numbers in parentheses indicate number of two- and three-member ridings. b. Numbers in parentheses indicate number of four- and two-member ridings c. Number in parentheses indicates number of two-member ridings Nova Scotia elections used two-member districts in two to five ridings until 1976. Similarly, elections in Newfoundland also used two-member districts until 1972. 17 Manitoba and British Columbia have experimented more liberally with multi-member districts. Manitoba has mixed variations in electoral formulas with multi-member districts. The average D M for the first two elections with the multimember urban ridings is 1.21. In British Columbia the average D M has fluctuated between 1 and 1.32. The difficulty o f comparing elections with variations in D M is not new. Is it possible to compare elections with different average D M ' s ? Lijphart suggests that electoral systems with an overall D M average o f 1.1 or less can, ceteris paribus, be considered to have D M ' s of one. 2 4 This rule of thumb permits considering al l but one o f the elections in Nova Scotia and Newfoundland as being conducted with an overall D M of one. This neatly eliminates nine of the 21 elections involving multi-member electoral districts reducing the affected elections from twenty percent o f the total sample to only 12. O f these none o f the multi-member districts used in the elections in question are at a higher-tier. That is to say the number of constituents that each member represents is relatively constant regardless of their being elected from a single- or multi-member district. To address concerns a D M variable w i l l be included in the regression analysis described in Chapter 3. 2 4 Lijphart (1993), pg. 486. 18 Chapter II - The shoulders of social structural giants Social structural theories o f party systems have had a short but tumultuous life. It has been less than half a century since Lipset and Rokkan described European party systems as being shaped by underlying social structures. Over the latter half o f the twentieth-century political theorists have taken turns refuting and defending the social structural hypotheses. These debates and entirely new theories o f party system development have shaken up the Lipset and Rokkan hypothesis but they have not irrevocably changed the original foundations. The purpose of this chapter is to examine social structural explanations of party system development and to argue for the applicability o f social structural models to Canada. The chapter is divided into five sections: the first examines the social structural approach outlined by Lipset and Rokkan and discusses their "frozen cleavages" hypothesis; the second reviews the social psychological literature that has followed. The second discusses several critiques which argue that many Western democracies are undergoing, or have gone through, a period of party system realignment or dealignment. The third looks at Inglehart's critique which suggests that traditional social cleavage structures are being replaced in importance by value cleavages. The fourth section takes stock o f the realignment/dealignment critiques and suggests a framework for approaching party systems in Canada. The last section reviews the current literature on social approaches to explaining Canadian party systems. 19 Social structural models In Contemporary Democracies Powell emphasizes the role o f social structures in party politics. He asserts that, "virtually all o f the multiparty systems have clear linkages to social groups." 2 5 The connection between specific populations and political parties was initially identified and categorized by Lipset and Rokkan. The introductory essay to Party Systems and Voter Alignment answers two inter-related sets o f questions within the field of comparative political sociology. One set focuses on the genesis o f political cleavages that engender different political system alignments. What are the influences o f social forces on party systems in post-industrial and post-revolutionary democracies? When are regional, linguistic, or ethnic cleavages most likely to prove polarizing? Which of these are most salient or persistent? The second set o f questions looks at the process o f transformation of social cleavages into political divisions. Which cleavages develop into stable societal conflicts that are politically represented? Which of the existing conflicts have been most salient? How resilient are cleavages over time? A s cleavages evolve or emerge how are they incorporated into pre-existing party systems? Lipset and Rokkan tackle these questions in two steps. In the first step they develop a framework that encompasses all the salient social cleavage structures that could exist in any particular society. This model has two dimensions, one which addresses functional matters and another that addresses the geographic proximity o f political actors. The functional dimension encompasses individual concerns; on one side are material concerns such as obtaining food and shelter, on the other side are non-material concerns such as political emancipation or eternal salvation. On this dimension interest-specific 2 5 Powell (1982), pg. 77. 20 concerns and ideology are polar opposites. The proximity dimension covers a range o f conflicts from local quarrels to debates in national arenas. On this dimension regional conflicts sit in opposition to deliberations between established national organizations. B y locating existing societal cleavages along these two dimensions, Lipset and Rokkan explain variation in party systems across countries. The model identifies four cleavage structures that characterize most European states. They are the conflicts between: subject and dominant culture, church and state, primary and secondary economy, and workers and employers. Lipset and Rokkan argue that two o f these cleavages are tied to the French Revolution and two are connected to the Industrial Revolution. In both revolutions one cleavage gains salience during the event itself and another cleavage rises after the revolutionary dust has settled. The conflict between subject and dominant cultures began the French Revolution as the masses battled the existing government power. The removal of the papal yoke from government led to the institutionalizing o f nation-states in the place o f feudal political structures. In this new political environment the clash between secular and ecclesiastical powers began to take hold and continues to dominate European politics to this day. In England the clash between the landed gentry and the newly arising entrepreneurial industrialists initiated the Industrial Revolution. A s the industrialists began to take over economic and political control o f the state the conflict against the landed gentry gave way to a conflict with workers. The clash between factory workers and the owners o f the means o f production exacerbated tensions between stratums of social classes. Figure 2.1 demonstrates the two-dimensional model and locates the four major cleavages Lipset and Rokkan identify in European democracies. 21 Figure 2.1 - Lipset and Rokkan's model of social cleavages structures In the second step Lipset and Rokkan develop a model that describes the transformation o f cleavage structures into party systems. This is a complex model which Lipset and Rokkan describe in four phases. The first explains how changes in social structures impact party systems. The analysis moves onto considerations o f voter and party interactions within social structures. The third phase describes how the four established cleavage structures cause variations in party systems. Finally, they produce a multi-level model for explaining how alliance-opposition structures are created. Each of these elements w i l l be briefly examined. Changes in cleavage structures at a demographic level may have no impact on the politics of a society; they may cause the decline o f a current political cleavage, or they 22 may create a new dimension for opposition. In both of the cases involving change, it is possible that this could cause the expansion or contraction of a party system. Intuitively, it is not difficult to understand how a society that undergoes a transformation which decreases the salience of a political cleavage would see a decrease in parties. This assumes that the removal of a cleavage would remove the raison d 'etre for parties that rallied around that division. However, an argument could also be made that removing a cleavage shifts focus, and that new cleavages may come into play which could increase the number of parties. Similarly, the development of a new cleavage may trump all other considerations and could, in effect, remove minor parties from the electoral competition thereby acting as restricting force on a party system. Using a rational-choice analysis of the decision making processes that voters go through, Lipset and Rokkan argue, Cleavages do not translate themselves into party opposition as a matter of course: there are considerations of organizational and electoral strategy; there is the weighing of pay-offs of alliances against losses through split-offs; and there is the successive narrowing of the 'mobilization market' through the time sequences or organizational efforts.26 There are limits to the political power that a social cleavage can have. The cleavage needs to have both salience and a critical mass of support on one side or the other to develop political momentum. Cleavages that lack importance or backing are not feasible vehicles for coordinating electoral support. A population that gravitates to a new banner in the political landscape does so through a process of rational choices. The voters need to perceive that the benefits they will derive from the new political party will outweigh the benefits they would realize if they maintained the status quo. In periods where Lipset and Rokkan (1967). 23 restructuring o f the electorate is imminent parties work to narrow the support market by making broad appeals to the electorate. Lipset and Rokkan argue that the conditions for the expression o f protest and the representation o f interests play a significant role in part system formation. Specifically, the freedom to mobilize political ideas is a powerful determinant into the likelihood o f cleavages influencing party systems. Political mobility includes decision making traditions within a polity; the channels for expressing and mobilizing protests; the payoffs, opportunities, and costs o f alliances; the ease for new movements to gain representation; and the types o f alliances that are likely to bring about the majority rule. In liberal systems party structures emerge as stable systems of party politics with a clear and distinctive pattern o f citizen alignments behind each party. Party systems evolve in patterns that reflect the four traditional cleavages previously described. The interactions between these cleavages in each state play a significant role in shaping the developmental course o f political structures. The differences between states in the strength and magnitude o f cleavages play a critical role in how parties develop. This is a powerful descriptive model for Europe because it adequately explains why an overtly religious party would emerge to a position o f strength in Germany while a fiercely nationalistic party came to occupy a similarly dominant position in France. Interestingly, Lipset and Rokkan argue that the class-based worker-employer cleavage does not actually contribute to variations in party structures between European states. Instead, the cleavage between worker and employers in European states has led to commonalities between these countries in terms of party system structure. 2 7 Classic examples of restructured electorates are democracies that have changed who are eligible to vote. The enfranchisement in many states during the latter half of the 19 t h Century and the suffrage movements in the early 20 t h Century are examples. 24 Lipset and Rokkan explain that cleavage structures are translated into party systems through a multi-level model where government alliances and rival oppositions are produced. The models are built on three dichotomies. The use o f the three dichotomies is designed to, .. .reduce the variety of empirical party systems to a set of ordered consequences of decisions and developments at three crucial juncture of history of each nation: ... first, during the Reformation—the struggle for control of the ecclesiastical organizations within the national territory; second in the wake of the "Democratic Revolution " after 1789—the conflict over the control of the vast machineries of mass education to be built up by the mobilizing nation-states; finally, during the early phases of the Industrial Revolution—the opposition between landed interests and the claims of the rising commercial and industrial leadership in cities and towns.28 (emphasis by the authors) A s indicated by the name and description within each level there are two alternatives at each juncture. The combination o f these factors leads to eight different possible patterns of party alignments. The shape o f party systems and party alignments is dictated by how the incumbent regime decides with whom they should ally with at decision making junctures. These choices place constrains on types o f alliances that the opposition can form. This process roughly describes how party systems develop. Overall, Lipset and Rokkan's framework does a good job o f describing the linkages between cleavage structures and party systems. The linking o f party systems to social heterogeneity has led to an oft-debated conclusion. The model suggests that party system transformation only arises during periods o f change in the electorate. This suggests that absent revolutionary impulses, party systems should remain stable. The term introduced by Lipset and Rokkan that has now become synonymous with their theory is "frozen." Lipset and Rokkan (1967), pg. 38. 25 Critiques of social explanations A central thread to the social structural paradigm is party system stability. Lipset and Rokkan argue the last large-scale shuffling of the political card deck occurred with suffrage. The cleavage structures o f the 1920s begat a party system that had remained largely unchanged over the fifty years leading up to the time of Lipset and Rokkan's research. They argue that, "the party systems of the 1960s reflect, with few but significant exceptions, the cleavage structures of the 1920s." 2 9 Aggregate evidence of the freezing hypothesis was provided by Rose and Urwin in their cross-national analysis of seventeen party systems between the mid-1940s and 1960s. 3 0 Their work examined the social cohesion o f 76 different parties on the basis of their support base. They focused on five types o f social groups: religious, ethnic, social class, regional groupings, and groups that were created by urban-rural divisions. This research indicated that party systems were largely stable and the social groups that contributed most to this stability were those o f religion and class. The freezing hypothesis has not been without its critics which have necessitated a fairly extensive overhauling of the social structural argument. Two significant papers refute Rose and Urwin 's aggregate analysis. 3 1 Ersson and Lane suggest that after a period of relative stability through the 1960s and early 1970s European party systems went through a period of destabilization that continued through to the 1980s. Shamir's account o f this period is quite similar although he additionally suggests that party systems were never frozen. Shamir's key piece of evidence is the significant shifts in policy 2 9 Ibid. pg. 50. 3 0 Rose and Irwin (1969). 3 1 Ersson and Lane (1982) and Shamir (1984). 26 considerations. In both cases, the conclusions they arrive at are that party systems are not stable and are no longer bound by the cleavage structures which originally shaped the European political theater in the 1920s. More criticism of the freezing hypothesis is found in Electoral Change in Advanced Industrial Societies edited by Dalton, Flanagan, and Beck and Electoral Change in Western Democracies, edited by Crewe and Denver, two cross-national collections of essays. Both books were the product of academic conferences that sought to explain why, if the party systems were supposed to be frozen, was there so much volatility in electoral results through the 1970s and early 1980s? In the view of Dalton, Flanagan, and Beck, Electoral alignments are weakening, and party systems are experiencing increased fragmentation and electoral volatility. Moreover, the evidence suggests that the changes in all of these nations reflect more than short-term oscillations in party fortunes. This decomposition of electoral alignments often can be traced to shifts in the long-term bases of partisan support—party identification and social cleavages. Virtually everywhere among the industrial democracies, the old order is changing.32 The contributors to both texts proceed from the hypothesis that the relationships identified by Lipset and Rokkan in the late 1960s are undergoing a period of realignment or dealignment. Social cleavages that motivated political actions a generation before had begun to weaken in the post-War period. In their place, new non-traditional cleavages gained prominence. Dalton, Flanagan and Beck (1984), pg. 451. 27 The value cleavage Inglehart suggests that in the latter half o f the Twentieth Century Europe has seen the development o f post-industrial, post-materialist cleavages that sits apart from traditional social structural cleavage structures. His research argues that the explanatory power that traditional socio-economic cleavages have over political behavior has diminished. Arguments on dealignment and fragmentation o f party systems existed prior to Inglehart's work but there was very little in the way o f systematic explanations o f why this was occurring. 3 3 Inglehart's post-materialist perspective argues that increasing wealth, the absence o f major wars, and increasingly accessible education has turned an unprecedented number o f people towards post-materialist values. 3 4 This value change has had a profound impact on European democracies. Increased government intervention in social and economic matters, highly developed post-secondary education systems, and the exponential growth o f information technology has led to three developments. First, a new cohort of political actors has emerged that is highly skilled and has access to an unprecedented range o f media, information, and capital resources. Second, new issues, such as the environment, global responsibility, and social justice have emerged to positions of considerable national importance. Finally, non-conventional modes o f political participation have developed, allowing an increasingly diverse range o f political actors to participate at national and sub-national levels. See for example Franklin and Jackson (1983). 3 4 Inglehart (1971, 1977, and 1990). 28 In Citizen Politics in Western Democracies, Dalton describes these phenomena as "new politics." In new politics citizens give increasing priority to global responsibility, social justice, and the environment. These newly politicized citizens challenge an older guard who emphasize political ideals like economic growth (or at least stability) and physical security. Because value orientations are both persistent and formed early in an individual's life this conflict between new politics and old politics is generally considered a conflict between new generations of political actors and the older generation.36 This is a pivotal claim in Inglehart's theory of value change. He argues that the value orientations that were founded on the traditional cleavage structure identified by Lipset and Rokkan have lost much of their significance in the second half of the twentieth century. The older voting cohorts in contemporary polities experienced first hand the economic crises of the 1930s and the physical threat of the Second World War. These experiences instilled a belief that economic well being and physical security are both scarce goods. Post war generations have grown up in far different and arguably far less arduous circumstances. Because of this bifurcation of experiences Inglehart predicts that there will be a quiet revolution in which materialist generations will be gradually replaced by generations with postmaterialist value structures. This generational shift wil l lead to a realignment of party systems where maturing voters wil l base a significant part their party allegiances on values as opposed to traditional issues. Adding to the value cleavage hypothesis, Flanagan has stated that Inglehart's conceptualization of values is actually a combination of two dimensions. First, there is the materialist/postmaterialist dimension that is central to Inglehart's work; and second, 3 5 Dalton (1988). 3 6 Inglehart (1977). 29 there is a libertarian/authoritarian dimension. The libertarian value orientation is tied to the concept o f self-actualization. The central facets to the libertarian value structure are individuality, liberty, and self-improvement. On the other end o f the dimension the authoritarian value orientation is based on concerns for security. These concerns are manifested in value structures that are based on deference to authority, respect for law and order, patriotism (to the point of jingoism at times), conformity to cultural norms, and support for traditional religious and moral values. 3 8 The libertarian/authoritarian value dimension is also seen in Kitschelt's work, where it plays a central role in explaining changes in party systems in Western democracies. 3 9 Kitschelt argues that explanations o f party systems that focus on traditional class cleavage structures underestimate the magnitude of change. This is because these explanations do not take into consideration the abilities o f party elites to forge new electoral coalitions on issues quite separate from the cleavages identified by Lipset and Rokkan. In Kitschelt 's study o f why socialist parties have slipped in popularity through the 1970s and 1980s, The Transformation of European Social Democracy, he states that: The future of social democracy to a large extent lies in the hands of party leaders and activists. External social, economic and institutional settings within which parties operate are less important for determining a party's fortunes than its own choice of objectives and strategies in the arena of party competition.40 Even though Kitschelt 's work plays down the importance o f social cleavages in contemporary party politics his work moves the social structural.approach a considerable way forward. His theory introduces a compelling rational choice argument: where political elites make strategic choice about the cleavages upon which to focus. 3 7 Flanagan and Jackson (1983), pg. 1301. 3 8 Ibid. pg. 1305. 3 9 Kitschelt (1994 and 1995). 4 0 Kitschelt (1994), pg. 4. 30 It seems undeniable that traditional social cleavages are eroding and facing heavier competition from ideological cleavages. These critiques have required rethinking how social conditions influence party systems; however, they have not toppled the social structural model. Instead, in some respects the theory seems to have come full circle, albeit as the theory returns to its origins it is a significantly modified version of its former self. Recent work from Europe, most notably by Kriesi has focused on developing a third way in the social structural approach that navigates between the frozen party systems hypothesis on the one hand, and the realigning/dealigning hypothesis on the other. Kriesi suggests that "the decline of traditional cleavages does not necessarily signify the end of structuration of politics by social divisions."41 Instead he argues that class, religion, ethnicity, and regionalism still play an important role in structuring party systems. However, these cleavages now compete with value cleavages that at time are reinforcing, and at other times cross-cutting. The following section will attempt to build evidence of these cleavages in Canada. Explanations of party system development in Canada There are virtually no social structural explanations of Canadian party systems. Crewe and Denver argue that this is because the precursors of modern mass parties in North America existed prior to the settling of the demographic landscape. They argue that this has led to an aggregation of interests in a few parties. Party systems are therefore not shaped by social cleavages but rather by enduring partisan loyalty. In The 4 1 Kriesi (1998), pg. 181. 31 American Voter, Campbell et al suggest that voters acquire enduring allegiances or identifications with an established party. The act o f party identification is an affective or emotional psychological attachment to a partisan reference group. This model has been used by several students o f Canadian party systems. Before examining these works a brief introduction o f social psychological models is necessary. Social psychological explanations of party system development suggest that voter's develop long-standing loyalties with established parties. Party identification is an internalized personal attachment an individual feels towards the partisan group o f their own choice. Mi l l e r and Shanks suggests that "party identification is a concep t . . . positing that one's sense o f self may include a feeling o f personal identity with . . . a political party." 4 2 This allegiance not only has a major role in determining vote choice, but also alters the perceptions of voters. From Campbell et al, Converse and Pierce, Green and Palmquist, and Blais et al, it is understood that party identification has two critical facets. 4 3 The first is identification, which Campbell et al describe as, ... an attachment to a party that helps the citizen locate him/herself and others on the political landscape. As thus conceived, partisans are partisan because they think they are partisan. They are not necessarily partisan because they vote like a partisan, or think like a partisan, or register like a partisan, or because someone else thinks they are a partisan. In the strict sense, they are not even partisan because they like one party more than another. Partisanship as party identification is entirely a matter of self-definition.44 The second facet is time. Party identification is considered to be a part o f an enduring belief structure, it is self-reinforcing over time, and it w i l l persist after the inspiring event that originally inspired the loyalty has passed. A common mechanism for acquiring 4 2 Campbell et al (1960), pg. 120. 4 3 Campbell et al (1986), Converse and Pierce (1987), Green and Palmquist (1990) and Blais et al (2001). 4 4 Campbell et al (1986), pg. 100. 32 partisan loyalties is through generational inheritance. Partisan belief structures are just one of several identifies that can be inherited. Palmquist explains: Like identification with a socioeconomic class, party identification reflects an awareness that one belongs to a social group. As with ethnic identification, the individual must decide whether to appropriate the group label as part of his or her self-description. And like religious identification, one's sense of group membership and attachment may develop for reasons that have to do with a person's social location (for example, the family in which one was raised, the person one chooses to marry).45 These stable and generally enduring affective attachments to a party have an obvious and powerful influence on voting decisions which in turn have an influence on party systems. Voting populations professing moderate to strong partisan loyalties are relatively stable over time. The predictability of partisan transference identified in earlier studies o f voter behaviour led to a compelling conclusion that voting populations in most Western democracies were quite stable. 4 6 Although the path to this conclusion is quite different from the social structural model the end result is quite similar to the frozen hypothesis forwarded by Lipset and Rokkan. Another similarity between the two is addressed in the following section. The connection between social heterogeneity and party identification has a modest literature. One area where there has been some active debate is the connection between religion and parties. 4 7 Most studies indicate that at a federal level there is a correlation between church affiliation and party is a strong factor (Johnston, Irvine, and Lijphart). Given the nature of Canadian politics, this is somewhat surprising. The relative strength o f religion as an important dimension o f party choice is perplexing because the more salient dimension of political conflict in Canada has traditionally been 4 5 Palmquist (2002), pg. 204. 4 6 See for example Campbell et al 1960; Butler and Stokes 1969, and Budge, Crewe, and Farlie 1974 4 7 See Stevenson (1987), Irvine (1974) and (1985), and Johnston (1977) and (1987). 33 the linguistic-ethnic cleavage. A s Meisel states religion is " o f virtually no political importance in contemporary politics; in Canada, whereas, o f course, the conflict between francophone and anglophone Canada is of the greater political significance." 4 8 Nevertheless, one of the strongest correlations in federal elections is that between Catholic voters and supporters o f the Liberal party. 4 9 Language is a complex factor in Canadian politics as it is inextricably tied to other variables. Lijphart has drawn connections between language and religion. He explains that, Religion and language are mutually reinforcing determinants of party choice in Belgium, Canada, and Switzerland. In terms of their relative strength ... the Canadian percentages are difficult to interpret because there are hardly any French speaking Protestants in the country. Nevertheless, religion emerges as the more important variable because there is relatively less difference between francophone and anglophone Catholics than between Catholic and Protestant anglophones.50 The linguistic variable is additionally limited by geographic specificity. Outside o f Quebec and New Brunswick the level of linguistic diversity is so low that it could be considered a non-factor in political concerns. There have been very few studies into the impact of ethnicity on party systems as such it might be tempting to dismiss its relevance. However, much like language, ethnic concerns are often tied to a myriad of other variables. A t a provincial level many ethnic movements have been tied with movements of gender equality. 5 1 Combining these interests increases the profile of the overall movement but obscures the influence o f ethnicity. 4 8 Meisel (1974), pg. 9. 4 9 Blais (2005). 5 0 Lijphart (1993), pg. 449. 5 1 Stasiulis and Abu-Laban (2000), pgs. 327-49. 34 The impact o f social class is oft-debated. Swartz argues that "class-based voting exists; it is consistent class-based parties that are missing." 5 2 Wi th the exception o f the N D P , which has a traditional connection to big labour, the effect o f social class is typically filtered through other variables. Class typically needs to be activated by the presence or absence of other variables. One example of this is the connection between Catholicism and class. Using union membership as a proxy for class Johnston argues that that in Canada social class can be a powerful indicator o f political partisanship. He explains "part o f the geographic story is that Catholics much more than union families are distributed unevenly over the landscape. This allows them to control the electoral agenda, so to speak, where their numbers are relatively large." 5 3 The underlying sentiments of the centre-periphery cleavage identified by Lipset and Rokkan are present in the regional argument made by Ricketts and Waltzer. The expression of regional interests by provincial government may occur either through a party bearing a traditional label or a third party. Indeed, provincial governments offer the wherewithal—powers, prestige, patronage, experience— for the political survival of third as well as traditional parties. And in a number of provinces, Canadians' perceptions of what best serves their interests have included the idea that a third-party provincial government is the more effective regional instrument. Besides, i f such a party enters national contest, it may succeed in attaining a balance-of-power position in the House of Commons and thus give more effective representation to regional interests.54 Ricketts and Waltzer effectively argue that the nature o f the federal relationship between provinces and the central government provide an adequate incentive from party to rally around this cleavage. The existence of the Bloc Quebecois, Social Credit, and Reform parties provide solid evidence. The federal question is an interesting one for Canada. Unlike the United States there is a disconnection between federal and provincial parties. 5 2 Swartz (1974), pg. 589. 5 3 Johnston (1985), pg. 128. 5 4 Ricketts and Waltzer (1970), pg. 713. 35 Blake et al suggests that this is evidence of weak party systems; he suggests that "one key indicator o f the weakness o f partisanship in Canada is identification with different parties at the federal and provincial levels, reflecting two political words that largely do not intersect." 5 5 The reasons for this disconnection is complex and beyond the scope o f this thesis. What is important to acknowledge is that the connection between federal and provincial politics is not particularly strong and therefore the influence that the federal party system has on provincial parties is not great. It is a causal variable, but for the purposes of this thesis it is assumed to be negligible. Summary Social structural perspectives explain how social situation influences voting decisions. It captures how embedded political cleavages—religion, ethnicity, regionalism, and class—form the foundation o f electoral preferences. This is not a simple relationship to explain because while the act o f voting is simple, the decision making process is not. The approach of the majority o f Canadian party system literature begs the question o f suitability o f social structuralism. A s previously stated the majority o f literature on Canadian party systems has employed social psychological approaches. Crewe and Denver offer an explanation which suggests why this might not be surprising. The 'social structural' paradigm was particularly apt for much of continental Europe, where organized communities often pre-dated the mass franchise and where proportional representation encouraged the separate party representation of distinct religious and economic groups. The second paradigm, anchored in the concept of party identification, was developed in the United States. Here organised mass parties pre-dated the class and 5 5 Blake (1985). 3 6 religious communities of modern America and were encouraged by the simple-plurality, presidential electoral system to aggregate interests and groups into a broad coalition rather than represent them exclusively (1985:2). However, it would be wrong to classify all party systems in North America as identical in their origins. The unique pathway that led to confederation and the formation o f Canada in addition to the significant volume of social structural literature suggests that societal cleavages should not be disregarded when considering party systems. There is a great deal that is unknown about Canadian party systems and employing a fresh approach deserves consideration. Provinces are socially diverse polities and it stands to reason that party systems may be influenced by demographic shifts. This thesis accepts that there are limits to social structural models. A comprehensive model o f party system development would have to include consideration of a myriad of institutional and social variables. However, a study o f that magnitude is beyond the scope of this paper. Instead this paper intends to examine a heretofore unexplored section of this puzzle in detail. Kries i has argued that social structural models can and should be viewed flexibly. Social cleavages work in concert and in opposition to value cleavages, partisan loyalties, and institutional factors. The hypothesis o f this paper is that social heterogeneity and party system size are connected. If they are, what is the nature o f these relationships? H o w does social diversity affect party systems? The hypothesis forwarded here is that growing social diversity increases the chances o f parties emerging and being electable. In provinces where there has been an increase in the number o f ethnic, religious and linguistic groups there should be a corresponding increase in the number of parties receiving votes and being represented in legislatures. How are party systems affected by social cleavages 37 where the magnitude o f the division is based on the size o f one population relative to another such as rural versus urban populations? In these cases the hypothesis tested in this thesis is that there is a negative relationship between binary cleavages and party system size. The larger the gap between the two populations the less l ikely the cleavage w i l l foster political competition. Staying with the example of the urban-rural cleavage in a province with identically sized urban and rural populations and no other political cleavages we would expect a balanced two party system to emerge centred on this division. In this isolated system we would expect the party system to contract i f the urban-rural population became unbalanced. Additionally, in a polity with multiple cleavage structures we would expect the effects of the cleavage to become negligible as the balance between urban and rural populations decreases. Wi th both of these hypotheses there are obvious difficulties with trying to prove them. Changes in cleavage structures can be glacial. This makes it difficult to draw connections between party systems and changes in the population. Being able to use a continuous run of five decades of elections and census data improves the likelihood o f identifying these linkages. Another difficulty is determining how to measure the magnitude o f social cleavages. This problem is addressed in the next chapter. The chapter identifies cleavage structures that may have influence on provincial party systems and employs a variety o f methods for attempting to quantify social diversity in a way that allows comparisons between provinces. 38 Chapter III - The building blocks of party systems Provinces are a good place to test social structural approaches to party system development. A s discussed in the previous chapter, a modest literature has emerged in Canada that draws links between social diversity and political parties. However, few efforts have been made to quantitatively explain the significance or magnitude o f these relationships. This chapter w i l l begin by looking at five cleavage structures. Building on the social structural framework this chapter w i l l focus on specific areas o f social division in Canadian provinces and w i l l present a variety o f methods for considering each cleavage. The five vectors that are examined here are ethnicity, religion, language, the centre-periphery cleavage, and urban/rural dispersion. These first four divisions are encapsulated in national cleavage structures, while the latter measure addresses the economic-class cleavage. Admittedly, this is not the optimal method for addressing the class cleavage. A s Johnston et al explain, "as a practical matter, the operative definition o f the class cleavage is the union movement versus the rest." 5 6 Unfortunately, accurate union membership figures are not obtainable for the majority o f the period under consideration. Therefore, an urban-rural variable has been developed that attempts to capture some o f the political tensions that exist between class stratums in Canada. 5 6 Johnston (1992), pg. 37. 39 Building an Index The general formula for constructing the heterogeneity indexes, unless otherwise stated, is a modified version of Rae's fractionalization index. The model is derived by summing the squared percentages o f a group's population relative to the population as a whole. n Heterogeneity Index = ^ Pf 1=1 Where P is the percentage a group's size relative to the total population So the higher the number, the less heterogeneous is the society. Unl ike Taagepera and Shugart, instead o f inverting this result we have left the result in the form o f an index. This facilitates easier comparison across social groups and provides a more elegant model. In general, the primary criteria for inclusion of individual social groups were size and consistency. A n y population that at a national level comprised at least 1% o f the population for at least five of the ten census periods was included. In addition, for certain variables additional populations were included when they exceeded 5% o f a province's population in at least five census periods. Given the design of the index, over-estimating the variables by including too many sub-groups is not an issue as the impact that small populations have on the index is minimal. Some words o f caution should be offered before addressing the variables. A n y measure of social heterogeneity poses potential problems. First, this study has divided 40 consideration of social cleavages into three individual variables to avoid over-generalization. A s Ordeshook and Shvetsova point out, "[t]he problem with any fractionalization index . . . is that it obscures the motives and actions o f voters and political elite so that it becomes difficult or impossible to discern the effect o f institutional structure on these separate motives." 5 7 Choosing to convert each variable into an index w i l l not eliminate over-generalization but it w i l l limit the degree to which it occurs. It is important to note, however, that while increased richness o f our data may help avoid over-simplifying the causal relationships at work, there are issues with using three measures o f social heterogeneity. Wi th multiple measures, multicollinearity may be inevitable. With multiple variables it is more difficult to identify the magnitude of each cleavage. A n index cannot measure variations in the level o f salience. In particular, Ordeshook and Shvetsova argue that "the particular problem is that ethnic, religious, and linguistic heterogeneity can operate differently when groups are geographically separate than when all groups are mixed ." 5 8 Two territorially based variables and multiple measures of social heterogeneity have been included to take into account of the inherent pathologies of index measurements. Indexes also carry an inherent assumption that all groups have similar organizational abilities and characteristics. Put another way, indexes assume that ceteris paribus, two groups o f a similar size w i l l possess similar levels o f political efficacy. A s Cox and Neto correctly point out, not all social groups have the resources to develop to the point where the creation of politically motivated elements is realized. This aside, 5 7 Ordeshook and Shvetsova (1994), pg 103. 5 8 Ibid, pg 109. 41 indexes do allow social scientists to condense large amounts o f data into manageable parcels and so long as the findings presented recognize the limitation o f macro data they remain a valuable tool. A final comment on the difficulty of using indexes is an assumption that the relationships examined in this thesis are linear. To take this into account, logarithmic values for each variable were tested, none produced robust results. Another area of concern is the accuracy o f data. Census data in Canada is gathered with the best intentions for producing accurate and unbiased information. Nevertheless, there are inevitable biases that occur because o f the difficulty in producing neutral questionnaires. Four issues are relevant to the investigation at hand: First, census questionnaires bias results by explicitly including or excluding variable alternatives. A n example that w i l l be examined later on would be the inclusion o f the option "Canadian" as an ethnicity for the first time in 1996. Second, the primary focus for censuses is providing an accurate snapshot of the population at a given time; a secondary motivation is providing detailed information that can be compared over time. Evidence o f this hierarchy of ideals is seen in the periodic changes that occur with Statistics Canada's counting and reporting methods. A n unavoidable source of error in this study has come from subtle changes in reporting figures between census periods. The third area where we find bias is in generalizations. Considerations of sub-groups are a particular area of difficulty. For example, all people of Asian decent prior to 1961 were classified as Asiatic. Even after this point the differentiation between Asians who were not Japanese or Chinese did not occur until 1971. Finally, the census conducted in the first year of each decade is comprehensive, while the census conducted on the sixth year o f each 4 2 decade is an estimation based on a random sampling of anywhere from five to twenty percent o f the population. The latter census is also not as thorough in the categories that it covers. For the purposes o f this study it was at times necessary to extrapolate some figures. Where necessary an exponential estimation was employed. This form o f estimation takes into consideration that populations do not expand or contract at set rates. Instead growth is a function o f the size o f the current population. Overall, these issues do not jeopardize the validity o f the analysis but they do open the door to error. Ethnic diversity Ethnicity is a relatively straightforward place to begin looking at social heterogeneity in Canada. Canada's national identity is a product o f multiple cultural identities. There is also a strong argument that the most salient social cleavages in Canada are predicated on the divisions between the historical occupants o f North America and the two great powers that colonized i t . 5 9 A s such, consideration o f ethnic diversity is essential for understanding social heterogeneity in provinces. Eleven population groups met at least one o f the two criteria previously set out: English, Scottish, Irish, French, German, Italian, Dutch, Ukrainian, Scandinavian, Asiatic, and Aboriginal/First Nations. The latter three are composites o f multiple ethnicities. 6 0 The Scandinavian counts included Swedish, Finish, Norwegian, and Danish 5 9 See for example Ricketts and Waltzer (1971), Rutan (1971), and McRoberts (1977). 6 0 There are two reasons for using aggregate figures for these three population groups. First, census reports prior to 1976 reported Asian and Scandinavian populations as a single group. Second, most of the ethnic groups that fall in these three categories would be too small to be included under the current parameters of this experiment. That said it is the authors opinion that each of these groups are significant and that the 43 peoples. The Asiatic variable includes people o f Chinese, Japanese, Korean, Taiwanese, and Vietnamese decent. The Aboriginal/First Nations population includes both First Nations and Inuit peoples; it does not include the Metis. A twelfth population group, Canadian, was also included although it falls outside o f the inclusion criteria. It is debatable whether Canadian should be considered an ethnic group, yet the size of this population as recorded by the Census necessitated some consideration. 6 1 groupings represent populations that share enough commonalities that aggregating these populations is not overly troublesome 6 1 Prior to 1981 the number of individuals claiming Canadian as their ethnicity was less than one-percent in all provinces. Since then there has been a meteoric rise in the percentage of the population that self-identifies itself as being of Canadian ethnicity. Across the board the number of people claiming to be Canadian jumped to almost ten percent between 1981 and 1996. The most dramatic of these increases was in Quebec where the ethnic Canadian population accounted for less than 0.1% of the population in 1981. By 1996 this figure had jumped to 37.2%. Even more startling in the Quebec case is that in 1991 the self-identified Canadian population stood at less than 1%. The increase across all provinces in the number of individuals who identify themselves as being ethnically Canadians can be partially attributed to a legitimization of the concept of Canadian as an ethnic group. The concept of "Canadian" as an ethnic identity did not gain purchase until late in the twentieth-century. For this the census questionnaires actually provides a good indicator of when the idea that "Canadian" could be an ethnicity earned a minimum level of acceptance. The inclusion of "Canadian" in the listings of ethnic groups in census forms began in 1996. This inclusion both legitimized respondents who identified as Canadian and introduced the concept to a population that previously may not have hitherto for considered it. This helps explain the steady, and somewhat steep, increase in the size of this population. However, this does not explain the variation between provinces. In the case of Quebec some of the explanation for this phenomenon undoubtedly lies with the impact of the 1995 Sovereignty referendum in Quebec. The referendum and the sentiments of supporters of both sides were a primary, i f not the, casual variable. However, it is important to note that a similar rise in Canadian identity did not occur after the 1980 sovereignty referendum. Two additional intervening variables explain why this was the case. First, it is arguable that from the outset the chance of success for the separatist vote in the first referendum was not as high as in the second. The degree of bifurcation between sovereigntists and nationalists was, therefore, less pronounced and less likely to inspire the strong nationalistic sentiments that manifested after the 1995 referendum. Second, prior to the voting date in the first referendum several powerful interveners, including the Prime Minister of Canada, were successful in persuading the Quebec population that changes would be made to recognize the distinctiveness and importance of Quebecois culture. During the second referendum, Quebec voters were given few assurances that Quebecois cultural concerns would be adequately dealt with. This contributed to both sides becoming firmly entrenched in pro-Quebec and pro-Canada camps which in the latter case is arguable evolved into a national identity that superceded traditional allegiances. The effect of the referendum on the emergence of a Canadian identity was not limited to Quebec. In the region that stood the most to lose if Quebec separated, the Maritimes, there were also significant increases in the number of individuals who considered themselves "Canadian" ethnically. By 1996 over 20% of the population in Nova Scotia, New Brunswick, and Newfoundland considered themselves "Canadian." Compared to the provinces west of the Quebec/Ontario border levels hovered around 10-12%. Ultimately the inclusion of "Canadian" as an ethnicity is important because it provides a powerful indicator of decreasing saliency of ethnic considerations. In all provinces ethnic diversity remaining relatively stable over the first 30 years of this study. After that period there is a variation between provinces with diversity 44 Another area where changes to the census questionnaire reflect major social changes and also impact the quality of the data is the change in how the ethnicity variable is recorded for single versus multiple responses. Prior to 1986 the census questionnaire required ticking a box beside the appropriate ethnic group or writing in a response if it was not already indicated: To claim a second ethnicity the respondent would have to write this in below their initial response. This in addition to the substantially lower number of people with multi-racial heritages resulted in relatively low recorded numbers of multi-ethnic Canadians. After 1986, individuals could tick as many boxes as were appropriate for describing their ethnic background. Added to eroding cultural boundaries regarding interracial marriages, there has been a significant increase in the number of people who claim multi-ethnic heritages. The problem with multiple responses has been addressed by omitting these individuals. The justification for this stems from an assumption that these multiple loyalties will not inspire the same political movements as people who have a single ethnic heritage. There are two ways of omitting this population and both are tested in this study. First, none of the individuals which had multiple-ethnic origins were included in the ethnic population counts. The second way of omitting individuals with multiple ethnic origins is to remove their influence on the index entirely. This is done through an adjusted ethnicity index which uses as the divisor the total number of individuals in the province minus all individuals who claim multiple ethnicities. This does mean there is a increasing in Quebec and Newfoundland, decreasing across the Prairies and the Maritimes and remaining constant in Ontario and British Columbia. This trend is misleading though as overall diversity of Quebec and New Brunswick is significantly lower than the other eight provinces. The dramatic increase of the "Canadian" ethnic population served to bump up measures of ethnic diversity. In Ontario and British Columbia ethnic diversity has always been quite high and the emergence of the "Canadian" ethnicity had a minimal effect of aggregate figures. In the Prairies and Maritimes, the "Canadian" ethnic group served to significantly decrease the ethnic diversity of those provinces. 45 structural difference between data collected in the early and latter parts o f this study because people were not prompted to declare multiple ethnic origins. However, the number o f individuals who this would have applied to in the early years o f this study is likely to be small. Table 3.1 reports the ethnic composition of Canadian provinces between 1951 and 1996. This first figure indicates the average size of the ethnic population during the period under consideration. The second figure reports that net change in the ethnic population percentage from 1951 to 1996. It is important to note that this second figure should not be interpreted as an indicator of consistent long-term trends, though in most cases this is an appropriate assumption. Changes in ethnic make up across provinces are marked by a degree o f consistency. Individuals' claming European ancestry declined in all provinces at relatively steady rates. The biggest decline is among those groups originally from the British Isles which is understandable considering the large amount o f non-European and non-British Isles immigration over the past fifty years. The significance in the ethnic shift in Quebec cannot be overstated. In the five year span between the 1991 and 1996 censuses the population identifying itself as ethnically French dropped from over five-mil l ion to just over two-million, while the "Canadian" population rocketed to over 2.5 mi l l i on . 6 2 Outside o f the "Canadian" ethnic group, large increases were seen in the Chinese and Japanese populations particularly in British Columbia and Ontario. Falling outside o f the parameters set by this study, modest increases in population size were Clearly this phenomenon warrants further investigation, however, this is beyond the scope of this paper. For the purposes of this experiment this dramatic shift raises an important point about the mutability of cultural compositions and the transience of ethnic identity. 46 Table 3.1 - Average size of ethnic populations for single respondents by province as a percentage (1951-1996) Provinces Ethnic Group BC AB SK MB ON QC NB NS NF 27.4 34.5 70.6 -19.0 -25.3 -38.7 9.2 18.0 1.7 -10.4 -17.6 -0.3 9.4 8.3 12.8 -10.6 -9.4 -6.4 34.4 9.0 2.6 -22.2 -7.6 -1.4 0.9 4.2 0.3 0.0 -2.5 0.1 0.2 0.4 0.1 0.0 -0.2 0.0 0.8 2.0 0.1 -0.8 -2.5 0.0 0.5 0.5 0.2 -0.5 -0.3 0.0 0.1 0.2 0.0 0.0 -0.1 0:0 0.2 0.4 0.2 0.1 0.1 0.1 0.6 0.6 0.6 0.4 0.5 1.1 6.3 5.1 5.3 23.7 19.0 20.8 Change in percentage of population for each ethnic group from the beginning to the end of the study is in italics. Percentage of populations in 1951 belonging to a specific group - Percentage of population in 1996 belonging to that same group. Census data between 1966 and 1976 aggregated all individuals who identified themselves as British, Welsh, or Scottish as being from the British Isles. Estimation of each of these populations was calculated by using country of birth as a proxy for ethnic identity. The number of individuals who stated they were born in Britain, Wales and Scotland were aggregated and then the percentage of each group was determined. These figures were multiplied by the total number of individuals whose ethnic identity was attributed to the British Isles. This method produced reasonable estimations for each population but they are also a possible source of error. English63 Scottish Irish French German Italian Dutch Scandinavian Ukrainian Asiatic Aboriginal Canadian 28.3 21.2 -27.0 -16.9 10.5 8.6 -14.5 -11.0 6.4 6.6 -9.2 -8.8 3.3 4.9 -2.3 -4.2 6.3 11.0 -1.2 -6.1 1.8 1.1 -0.2 0.2 2.8 3.0 -1.2 -2.1 4.6 5.4 -4.2 -6.3 2.1 6.8 -0.8 -6.1 4.6 2.3 8.1 4.1 2.3 2.5 -0.4 0.1 3.1 4.4 9.6 12.7 18.1 18.1 -14.5 -16.1 7.5 8.7 -9.4 -11.8 6.4 5.9 -9.0 -8.8 5.0 7.2 -4.4 -5.5 15.7 9.5 -7.1 0.0 0.2 0.8 0.0 0.3 2.1 3.7 -2.9 -4.1 5.9 3.8 -5.7 -3.3 7.8 9.9 -4.9 -7.5 0.8 1.5 0.9 3.2 4.0 4.7 4.6 4.5 3.6 2.9 10.1 8.5 28.2 5.5 -27.6 -6.1 9.0 1.4 -11.6 -1.8 8.9 1.8 -13.7 -1.9 8.1 73.9 -7.6 -53.1 4.7 0.6 -2.6 0.0 4.6 2.2 2.6 1.5 2.3 0.1 -0.6 0.0 1.5 0.4 -0.4 -0.1 1.7 0.3 -1.2 -0.2 2.1 0.7 5.0 1.3 0.8 0.6 -0.2 0.4 4.4 9.4 12.0 37.2 47 recorded in the black and indo-Canadian population. However, neither o f these populations was greater than the 1% minimum threshold until very recently. A final area of concern for putting together a measure o f ethnic heterogeneity was balancing an accurate portrayal of demographic trends in Canada with a measurement that makes intuitive sense. There has been an obvious increase in ethnic diversity in Canada over the past five years, however, basic indexes using the ethnic groupings previous discussed suggested that there were fewer ethnic groups at the end o f the century then there were at the beginning. M u c h o f this anomalous result is due to the emergence of the dominant Canadian ethnicity and the decision to ignore individuals with multiple ethnic identities. The measure that was used in the models overcame some o f the initial short-comings through two modifications. First, individuals o f Brit ish Isles ethnicities were aggregated into a single population. While ethnic identities between these groups may still have some lingering cultural significance, in particular during World Cups, there are few political cleavages that are observable based on these divisions. Second, the Canadian ethnic group was redistributed between the measured ethnic groups on the basis of the magnitude o f the decline o f each group from the previous census figures. This was done to maintain continuity with the data and to take into account that not all groups actually experienced drops i n population counts while the Canadian ethnic population soared. 6 4 An interesting observation made from the ethnic data is that while most ethnic groups saw decreases in populations that were proportional to the increase of the Canadian ethnic population, a few groups were stable. For example, during the final period from 1991-1996 the Canadian population increased by about 20% across the country; for most ethnic groups this was matched by a 20% decline. Some groups were not affected by this. For example, across Canada the Italian population remained quite stable, and the Asiatic populations continued to grow. 48 Religion Religion has an arguable relevance to Canadian politics. The muddiness of the relationship between organized religions and party politics has led some to refer to religion in general as "the unwelcome dinner guest" in the canon o f Canadian party poli t ics. 6 5 For those that have wrestled with the topic some general statements can be made about party politics in Canadian provinces. A t a federal level there are discernable linkages between the Liberal Party and the Catholic Church and a similar historical relationship between the Conservative Party and Protestant faiths. The translation of these connections to the provinces varies, with the strongest connections in the Maritimes. One province where the Liberal Party and Catholicism are not synonymous is in the francophone province of Quebec. The election o f Jean Lesage and the Liberal Party in 1960 herald a period o f extensive change in Quebec that saw the power o f the Catholic Church reduced dramatically. During the "Quiet Revolution" social change was widespread and sudden. During this period o f modernization no institution suffered more than the Roman Catholic Church. Significantly, language replaced faith as the focus of Quebec's distinctiveness. Quebec society became profoundly secularized and Church's connection to politics was irrevocably altered. 6 6 The complex relationship between religion and politics may make it the unwelcome guest; nevertheless, this paper w i l l try to make the connection somewhat clearer. For the purposes of this analysis the variable is constructed in manner somewhat 6 5 Johnston (1991), 108. 6 6 See Barnes (1961), Clarke (1983), and Rawlyk (1995). 49 different than the ethnicity variable. Initial attempts at constructing a religious heterogeneity variable included Catholicism, the five largest denominations within the Protestant domain and, where applicable, individuals with no religious affiliation and Eastern non-Christian religions. 6 7 This index indicated there was significant variation between provinces. The most heterogeneous provinces are located west of the Ontario/Quebec border; Quebec is unsurprisingly the least diverse and the Maritime Provinces fall somewhere in between. This index also suggested that religious heterogeneity was stable over time with only a slight increase in religious diversity over the period o f this study. The average standard deviation o f the religious index for all provinces was 0.013. Figure 3.1 Percentage of Catholic population by Region 1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 Census year -•-West Catholic A Quebec Catholic Ontario Catholic IVfaritirre Catholic 6 7 Figures for Catholic populations are not consistent throughout the examination because census Canada includes Ukrainian Catholics in total counts for the first three censuses used in this study. To overcome this, the sum total for all Catholics is used throughout the study. Individuals who claimed no religious were included after 1966 onward and individuals claiming eastern non-Christian religious after 1976. 50 These results are problematic because there seems to be little evidence that religious membership has been stable over the period o f this study. Figures 3.1 and 3.2 indicate the percentage of Catholic and Protestant populations by region. The four regions considered are the West (British Columbia, Alberta, Saskatchewan, and Manitoba), Ontario, Quebec, and the Maritimes (New Brunswick, Nova Scotia, and Newfoundland). Looking at figure 3.1, the high concentration o f Catholics in Quebec is obvious. The significance of this is debatable, as previously mentioned; in the political arena the politics of language seems more salient than the politics of religion. Additionally, given the hegemony of the Catholic population it is arguable that religion is a non-factor in provincial politics. Outside of Quebec, the general trend for Catholic population is stable and slow growth over the period covered. Figure 3.2 Percertagp ofProtestart Population by Regjon 51 In Figure 3.2 as in the previous graph, Quebec is an outlier with a Protestant population that never exceeds ten percent o f the province's population. Significant decline is the story across all other regions with the largest drops seen in the West and Ontario. Moreover, in all regions outside o f the Maritimes the Protestant population decreased by over fifty percent. These two tables demonstrate a curious aspect o f the Canadian population and prove that the original index was failing to capture a significant trend. O n one hand, the decline o f the Protestant share of the population is not surprising given the influx of non-Western immigrants and the general decline in the popularity o f organized religion. On the other hand, what is sauce for the goose is clearly not sauce for the gander, as the Catholic population appears to have been impervious to demographic change and does in fact seem to have grown over the past fifty years. Explaining this aspect o f Canadian politics is beyond the scope o f this study, however; it does suggest that there may be significant variation in the political unity and strength between religious groups. There have also been sizable increases in the number o f individuals claiming no religious affiliation and affiliation with Eastern non-Christian faiths. This rapid increase might suggest that a new cleavage that has emerged might simply be described as secular-religious. In order to provide a more textured account o f religious heterogeneity a simpler index was developed which included Catholics, all Protestants aggregated together, non-religious individuals, and individuals belonging to non-Christian faiths. The Aggregated Religious Index (ART) successfully captures the significant trends in religious membership over the latter half o f the Twentieth Century. Figure 3.3 demonstrates the intuitive appeal o f this index. The difference in religious heterogeneity 52 between regions remains clear while at the same time reflecting the decreasing religious membership in traditional areas and the emergence of new religious groups. F%ie3.3 - Estirrated rirrber ofReHgous Groups by Regjon 1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 Census Year —•— West —•— Ortario —*— Quebec —•— IVferiurnes Language In a country with as many cross-cutting cleavages as Canada, language is arguably the most salient political division. The two-founding-nations cleavage is addressed partially by considerations of ethnicity and religion; however; it is in language that the bifurcation is most clearly represented. The distinctiveness of language as a source of cleavage is promoted by the fact that at a national level there are no third significant linguistic groups. The two founding nations argument has in many respects 53 evolved into a two official languages argument. However, this debate is not present in all provinces. In 1996 there were two provinces in which three distinct linguistic groups were larger than the French population. In British Columbia the number o f individuals who have German, Cantonese, and Punjabi as a mother tongue was larger than the Francophone population. In Saskatchewan the Cree, German, and Ukrainian populations were larger. A n d in both Alberta and Manitoba the German speaking population was larger than the French speaking population. Having said that, only in Manitoba, where six-percent o f the population has German as their mother tongue, does any non-Official language group exceed five-percent of a province's total population. The absence of a significant non-Official language in any province allows this index to measure almost exclusively the effect that the French speaking populations have on provincial politics. The division between anglophones and francophones is only notable in Quebec and New Brunswick. For the sake o f simplicity the index was created by determining the aggregate number of people who spoke a single official language. Two treatments were applied to these figures. The first method was to create an index using the previously established formula. This index indicated the degree o f linguistic heterogeneity between provinces. Figure 3.3 demonstrates the level of linguistic heterogeneity o f Quebec and New Brunswick. 6 8 With the index 0 represents a completely heterogeneous population, where everyone essentially speaks there own language, and 1 indicates the province is entirely unilingual. A s figure 3.3 indicates, New Brunswick is slightly more heterogeneous than Quebec. Wi th very little variation between the seven provinces outside of Quebec and New Brunswick this variable is In all other provinces the index averaged over 0.950 indicating the provinces were in effect unilingual. 54 limited to describing the unique effects that language has in the latter two provinces and as well as the differences between the two provinces. Figjre 3.4 - Adjusted IJngistic Index for Quebec arriNewBnjnsvvick 0.400 0.300 i 1 1 1 1 r 1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 Census Year —*— Quebec —•— NewRxrewick The second way of treating this data is to create a linguistic ratio between English and French speaking individuals. With this ratio 0 would indicate that population was entirely French speaking and 1 would indicate that the population was entirely English speaking. Employing this method has little effect on the dominantly English speaking Provinces, in all cases across time the ratio was higher than 0.975. The significant change lies in the results for Quebec and New Brunswick. For New Brunswick the ratio ranged from 0.760 in 1951 to 0.851 in 1996, indicating an increase in the English-55 speaking population. In Quebec, the ratio ranged from 0.154 to 0.083, indicating the opposite trend. Although the two measures are derived from the same data they tell two different stories. When regressed on the election results the index indicates how much effect having a heterogeneous society has on party politics. The ratio indicates the effect that the relative size of the French speaking population has on party systems. The ratio is in some respects a dummy variable for Quebec although the distinctive Francophone population in New Brunswick cannot be discounted. This is not necessarily an issue but it does assume that language is an appropriate proxy for cultural differences between Quebec and the R O C . It also assumes that the figures determined for N e w Brunswick correctly reflect the impact that language/culture has on provincial politics. Both measures of linguistic heterogeneity w i l l be employed in the analysis. A final area o f concern for both formats would be they assume that, ceteris paribus, individuals who speak both languages are not as likely to be politically motivated as people who speak only one official language. The Proximity cleavage in Canada The next variable considered in this chapter addresses the centre-periphery cleavage that exists between central Canada and the outlying provinces. Regional alienation has received a significant amount of attention by academics and media alike. It has been used as a rallying cry at times by political parties. A n d it has brought Canada to the brink o f separation on more than one occasion. For these reasons it deserves 56 consideration. However, it w i l l be interesting to see i f alienation has a consistent effect on party systems. Does the magnitude of alienation have a significant impact on party system size? In Canada the centre-periphery cleavage has less to do with geography and more to do with the ties that bind provinces to the federal government. Some studies have explored this relationship but they tend to focus on the effects o f these connections on federal elections. 6 9 Quebec provides ample evidence of this characterization: the province is situated in central Canada. However, it is, by most measures, the province most alienated from the federal government. Similarly, N o v a Scotia and N e w Brunswick have relatively harmonious relations with Ottawa even though they are on the eastern end o f the country. The relationship between alienation and party systems is different from the previously considered variables. To begin with, alienation is not a measure o f social heterogeneity but rather a measure o f a social division. That is to say, alienation does not measure social diversity but instead measures the strength o f the division between the local region and Ottawa. With religion, ethnicity and language it is expected that greater salience w i l l lead to more parties. However, with alienation the opposite effect would be expected. The stronger the degree of alienation between the federal government and the province the more likely provincial parties are to rally around that cleavage. Spatial logic would dictate that under these conditions a party system would be reduced to two viable parties, one on either side o f the cleavage. 6 9 See Cairns (1968) pg. 55-80; Blake (1972), pg. 55-81; Johnston and Ballantyne (1977) pg. 857-66; and Blake (1978), pg. 279-305. 57 Although the idea o f a centre-periphery cleavage based on alienation seems intuitively comfortable there is a significant issue with determining how to quantify this variable. A t the risk o f oversimplifying, this variable has been approximated using a 0 to 1 scale, where 0 indicates no alienation from the federal government and 1 indicates a high level of alienation from Central Canada. A n underlying assumption with this variable is that similar levels of alienation w i l l have a similar effect on party system development. This brings about an important question of whether these effects are similar or would it be better to use a dummy variable for each province to account for the unexplained unique effects o f each province? Regression analysis in Chapter 5 w i l l employ both methods. Figure 3.4 shows the approximated positions o f the nine provinces on an alienation spectrum. Figure 3.5 - Approximated Magnitude of Centre-Periphery Cleavage in Provinces New Brunswick Saskatchewan Nova Scotia British Columbia Alberta Ontario Manitoba Newfoundland Quebec 0.1 0.2 0.3 0.4 0.5 0.6 Alienation Index 0.8 The placement of each province attempts to encapsulate the characteristics of the relationship between provincial and federal levels of government for the duration of this study. Federal transfer payments, jurisdictional debates, constitutional matters, and federal representation are all factors are considered in this ranking: Dependence on 58 Ottawa for funding is a double-edged sword; on the one hand it encourages provinces to be cordial to ensure fiscal support but, on the other hand, it can also be a source o f tension i f there is debate over revenue shares. Jurisdictional conflicts have involved debates over control of natural resources and the provision o f social services. The constitutional battles of the Eighties and Nineties had a profound impact in several provinces on the relationship with the federal government. A n d finally, cabinet representation has a pivotal role in federal-provincial relations, as regions tend to feel ignored i f they perceive a lack of voice in Ottawa. The province with the lowest level of alienation is Ontario. The proximity to Ottawa and common interests has meant the provincial government has rarely felt distanced from its federal counterpart. Nova Scotia and N e w Brunswick both experience relatively low levels o f alienation for two general reasons. Both provinces are heavily dependent on federal transfer payments and both provinces have had success at having local Members of Parliament placed in prominent Cabinet postings. A lso on the less alienated end o f the spectrum is Manitoba where the debate over the constitution has been the lightening rod for provincial-federal tension. Outside o f constitutional matters conflict between Manitoba and Ottawa has been l o w . 7 0 M o v i n g toward the middle of the scale, Saskatchewan is a province which has often felt shut out of federal politics and has been at odds with Ottawa's stance on tariffs and trade regarding natural resource and agricultural commodities. British Columbia's attitude towards Ottawa has been marked more by indifference than tension. Geographic isolation and a healthy provincial economy have left few areas o f major conflict yet several lesser issues dealing with First Nations treaties, fisheries, forestry, and the burden 7 0 A notable exception to this would be the awarding of the CF-18 maintenance contract to Quebec. 59 of being a "have" province have maintained a moderate level o f alienation. Tension over control of offshore fishing, the Hibernia o i l project, and the Churchil l Falls hydro project have all soured relations between Newfoundland and Ottawa. The fiscal nature o f these debates has been exacerbated by Newfoundland's heavy dependence on federal funding. Similar to the economic complaints o f Saskatchewan and Newfoundland, Alberta's frustration with Ottawa is magnified by fundamental differences on social policy regarding healthcare, same-sex marriages, and bilingualism as wel l as being a net donor o f equalization. A t the most alienated end of the spectrum is Quebec. Not much needs to be said about the reasons for this, it should be noted that the reason Quebec is not ranked higher is that the Quebecois population is not a monolithic entity. While it is clear there is a sense o f alienation between Quebec and Ottawa it is important to remember that two separation referendums were defeated in the province and that historically, until recently, the Province tended to vote en bloc for the party that formed the government in Ottawa. Rural/Urban The decline o f rural Canada is a part of a global shift from agrarian to industrial economies. The five decades covered in this study have seen a steady shift in population from rural areas to urban centres. The division has had a significant impact on Canadian politics; Johnston et al surmise that "for the Canadian case, the central cleavage is between export-oriented agriculture and the rest." The recent B C Citizens' Assembly for 7 1 Blake (1996). 60 Electoral Reform final report concluded that the dominance o f the urban centres o f Vancouver and Victoria was a crucial concern to rural voters. Figure 3.4 demonstrates the steady shift in urban and rural populations. The largest shifts are seen in the Prairies where the rural population was larger than the urban population in 1951 but significantly lower by 1996. British Columbia, Ontario, and Quebec and the Maritimes all saw modest shifts of at least 10% over the five decades covered. In the Maritimes, however, the starting point was much higher than the other three regions. Across Atlantic Canada the shift shows a reversal o f the dominant population. Figure 3.6 - Rural Population as a Percentage 1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 Census year —•—British C b l L i r b k —•— Prairies • Ortario —A—Quebec —•— IVferihrres 61 In this study, the urban/rural cleavage w i l l be looked at using three methods which provide a different lens to view the urban/rural divide. The first approach is determining the ratio between urban and rural populations in each province. This provides an elegantly simple measure o f the magnitude of difference between the two populations. The provinces with the higher urban/rural ratios are Alberta, British Columbia, Ontario, and Quebec. For the latter three, urban populations were at least twice that o f rural populations in 1951; by 1996 these ratios had grown to over four to one in Ontario and British Columbia and three-and-a-half to one in Quebec. Alberta's rural population was actually larger than the urban population in 1951, but by 1996 the urban population had increased proportionally to nearly four times that o f the rural population. In the remaining provinces, with the exception o f Manitoba, the rural population was larger in 1951 than the urban population. This remained true in New Brunswick all the way until the end o f this study. In Newfoundland, Nova Scotia, and Saskatchewan the urban population grew about one-and-a-half times larger than the rural population by 1996. In between the mostly urbanized and mostly rural provinces, Manitoba has become somewhat more urban with the urban population growing from 57% to 71%. The second approach for examining the rural-urban cleavage is suggested by Powell where the influence of the agricultural sector over the polity as a whole is coded as 1, 2, or 3; i f the agricultural population comprises 0-19%, 20-49%, or 50-80% of the total populations. For this study these divisions are not as effective as none o f the provinces in this time period have a rural farming population that exceeds 50% of the total population. However, by adjusting the population ranges to 0-10%, 11-30%, and 62 31-100% the evolution o f provinces that are predominantly farm-orientated is quite clear. The results o f this measurement system do a generally good job o f tracking this shift from agricultural to industrial dominance within provinces. The two exceptions are Newfoundland and British Columbia where the farming population never exceeds 10% of the total population during any period of this study. The third approach, employed by Ordeshook and Shvetsova and Eagles, uses a modified fractionalization index. 7 2 Two variations of this index were developed for this study. The simpler index is created by adding the inverse percentage o f the urban and rural populations. The more complex index further divides the urban and rural populations, the urban group is divided into individuals who live in communities below 30,000 and above; 7 3 while the rural group is divided into individuals who live in farm and non-farm residences. A s with the linguistic index described in the previous section, indexes are limited in their explanatory power, for example with the index formula a 40:60 split in urban to rural populations w i l l look the same as a 60:40 split. Wi th these limitations in mind these indexes do offer some valuable insight as to whether the magnitude o f the difference between urban and rural populations is a causal factor in party system development. Methodology Given that this set o f data has both cross-sectional and time-series dimensions, it is unlikely that the E N P of provinces is independently distributed. The reactions of 7 2 Ordeshook and Shvetsova (1994) and Eagles (1995). 7 3 There is some unavoidable error with this classification as census figures prior to 1986 reported the aggregate number of people living in communities below 29,999, in 1986 and for each census thereafter this figure was reduced to 24,999. 63 political actors—political parties, politicians, or voters—to changes in their environment resulting from changes in social heterogeneity are never instantaneous nor are they independent from reactions in previous periods. The changes are l ikely to be distributed over time; and positions of equilibrium in E N P or C N P , i f they are ever attained, are likely to be approached gradually. The slowness to respond may be due to two factors. In the first place, there w i l l be time delays in the transmission and the reception o f the information upon which the political actors base their actions. In the second place, costs w i l l be entailed in the process o f adapting to the new circumstances, and these costs are likely to be positively related to the speed and to the extent of the adjustments. For these reasons, it is appropriate to make some provision for correlations across time and provinces. Given the relatively lengthy period (1951-1996) under analysis, the customary approach to this kind of time-series cross-sectional data used in political science—OLS with panel corrected standard errors, including fixed effects—should be satisfactory. 7 4 A possible area where bias might arise would be that five decades o f elections covered in this project are encapsulated in ten observations for each province which correspond to the years in which census data was collected. The estimation model assumes that the various measures of social heterogeneity examined in this chapter have independent effects modeled by a simple linear relationship with E N P and C N P . E N P = /30 + (ethnicity) ft + (religion)/32 + (language)/33 + (rural/urban)/34 + (agriculture)j85 + (alienation)/36 + (DM)j37 + c 7 4 Beck and Katz (1995). 64 Within this linear model the two measures of linguistic heterogeneity and all three of the measures of the urban/rural divide are employed. For both variables, the different measures make intuitive sense and they both produced robust results. Each variation provides a different lens for examining party systems. In all of the models the agriculture index is employed. The index and ratio measures are each used separately. A basic assumption with this variable is that its effect should be binary. That is to say, if a population is equally split into an urban and rural population we would expect this to foster a political cleavage that was also equally split. The rural side of this cleavage would champion agrarian and family centric concerns and the urban side would champion union and post-material issues. The rural-urban index (RUI) is successful in describing the balance between urban and rural populations but it is unable to determine the direction of the relationship between the urban/rural cleavage and party system size. The rural-urban ratio (RUR) does provide directionality but it less sensitive to the balance between urban and rural populations. As with the religious variables the strength and weakness of each measure warrants separate investigation of each. 65 Chapter IV - Social Heterogeneity and Party Systems The previous chapters have described how party systems have evolved over the latter half of the Twentieth Century, examined social structural models, and explained how this approach could be applied to Canadian provincial politics. The first half o f the previous chapter outlined the social cleavage structures o f the Canadian population. This chapter tests the validity of social structural approaches in a Canadian setting. The chapter is divided into two parts, the first reports the results and the second part summarizes the findings and indicates where future research is needed. The first two tables in this chapter indicate that even when party systems are constrained by institutional factors, social heterogeneity has a significant influence on the size of party systems. A s anticipated, where the estimated coefficients for social heterogeneity are significant, they are also positive. This supports part o f the social structural hypotheses: populations divided by multiple social cleavages w i l l have more parties participating at an electoral and legislative level than polities with fewer social divisions. Briefly looking at each variable, the effects of ethnic and religious heterogeneity are significant in a few o f the equations. They generally have more power in the C N P models. The significance of the linguistic variables is dependent on the format o f the linguistic heterogeneity measurement. The same applies for the urban/rural variable which is only significant in the ratio format. For the alienation variable the regressions indicate that alienation is negatively correlated with party systems. The following sections w i l l examine the E N P and C N P models separately and w i l l unpack the variables in further detail. 66 Social Heterogeneity and ENP E N P is the dependent variable in the regressions in Table 4.1. The four models described in the previous chapter are all tested with varying degrees o f significance. A l l models indicate that there is a relationship between social heterogeneity and the number of parties that are elected. Three o f the four models are statistically significant at the 0.05 level and several o f the independent variables are significant at the 0.10 level or better. The reasonable degree of consistency in the direction o f the coefficients encourages further confidence in the results. Negative coefficients for variables that use index measurements indicate that there is a positive relationship between social heterogeneity and party system size. This hypothesis holds true for several of the regression coefficients. This does not apply for the ethnicity and religion coefficients which, although all negative, are with one exception, negligible in size. The steady increase in ethnic heterogeneity, in particular west of the Ontario- Quebec border, has not had a significant impact on the number o f parties elected. The same can largely be said for religion. Using the index as a comparative tool, religious diversity in provinces west of Ontario has increased three-fold while a proportionally much smaller increase has occurred in Quebec and the Maritimes. The regression results indicate that this has had a negligible effect on the number of parties. 67 Table 4.1 Estimates of the Effective Number of Parties (ENP) in Canadian Provinces (1951-1996) Model 1 Model 2 Model 3 Model 4 Ethnicity Language Index Linguistic Ratio Basic Religious Index Rural-Urban Index Rural-Urban Ratio Agricultural Alienation District Magnitude 7 5 Constant R 2 Wald chi 2 Prob > chi 2 -.1230 -.0775 (.2340) (.2156) .0737 (.5525) -1.1417*" (.3446) -.0198 -1.4323" (.4650) (.5830) -.5508 .0483 (.5862) (.5855) -.0217 -.0432 (.0652) (.0642) -.2297 -1.0333*" (.2709) (.3194) .2005 .8668 (.5284) (.5703) 2.0583" 3.0029*" (.8419) (.7117) 0.7613 0.7692 4.93 16.63 0.6689 0.0199 -.0381 .0026 (.2005) (.1977) .1245 (.4904) -.7414* (.3889) .0737 -.9678 (.3852) (.6137) .1732 .0989 (.0564) (.0685) -.0375 -.0378 (.0606) (.0615) -.3944* -.8286*" (.2410) (.3081) .2299 .6857 (.4483) (.5554) 1.5101" 2.450"* (.6198) (.6798) 0.8412 0.7924 14.49 18.21 0.0431 0.0111 OLS with panel corrected standard errors, standard errors in parentheses, ***p<.01, **p<.05, *p<.10 Neither coefficient for the linguistic index is significant but each indicates that there is a negative relationship between linguistic heterogeneity and party system size. In this case, more diversity leads to fewer parties. A s previously established, the linguistic variable is essentially identical for all provinces except New Brunswick and Quebec. O f 7 5 District magnitude was included to limit error attributed to institutional bias. No control was necessary for the electoral system as this can be held constant for the entire period of study. 68 the two New Brunswick has a more balanced and thus more diverse linguistic population. New Brunswick and Quebec's linguistic uniqueness means that the models predict how much linguistic diversity affects party systems in those provinces. The first model indicates there w i l l be on average 0.023 fewer parties in New Brunswick and 0.015 fewer parties in Quebec than in the rest of Canada. The third model predicts that there w i l l be 0.038 fewer parties in N e w Brunswick and 0.026 fewer in Quebec. Clearly the effect that this variable has on party system size is limited. The coefficients for the linguistic ratio are both negative and significant at least at the p=0.10 level. The coefficients indicate that there is a negative relationship between the proportions of English speakers, relative to the French speaking population. These results indicate that the existence o f a significant French speaking population appears to increase E N P . A simple explanation for this could be that French-English linguistic cleavages tend to be a dominant political issue. Even in provinces without large French speaking populations' language has been a divisive issue. The addition o f a significant cleavage provides an additional spectrum which opens the arena for more parties. Similar to the findings for the linguistic index the unique linguistic nature o f Quebec and New Brunswick allows predictions to be made based on the ratio o f French to English speakers in those provinces. The second model predicts there w i l l be on average 0.994 more parties in Quebec and 0.223 parties in New Brunswick. The fourth model predicts there w i l l be 0.646 more parties in Quebec and 0.146 more parties in New Brunswick. The coefficients for the rural-urban index (RUI) are also either negative or negligible in size, however, the story this tells is slightly different from that o f the ethnicity and religious index variables. The negative coefficient indicates that provinces 69 with balanced urban and rural populations have more parties than provinces with unbalanced populations. This suggests that provinces with unbalanced sized urban and rural populations develop stronger political cleavage structures around the urban/rural bifurcation than provinces with more equal urban and rural populations. This conclusion verifies the hypothesis that the urban-rural cleavage is negatively correlated with party system size. Provinces with equally sized urban and rural populations have more parties than provinces with unequal distributions. These results are reinforced by the rural-urban ratio (RUR) and agricultural variables. These variables also indicate the direction o f the relationship between urban and rural populations and party systems. The coefficients for the R U R variables in the third and fourth models are positive and for the latter model significant at the p=0.0T level. These results suggest there is a positive relationship between the size o f party systems and the proportion of a population that resides in urban centres larger than 30,000 people. A s a province becomes more urbanized the number o f elected parties increases. Given the steady shift in population counts from rural to urban centres, these coefficients indicate that population movements have a powerful effect on E N P . The coefficients for the agricultural index are consistently negative but not significant. They support the conclusion that there is a positive relationship between urbanization and E N P . Provinces with larger agricultural populations w i l l have fewer parties. A s the most agrarian province, the four models predict that over the course o f this study Saskatchewan w i l l have between 0.043 and 0.085 fewer parties that the least agrarian provinces o f British Columbia, Newfoundland and Ontario. 70 Union membership figures are not available over the course o f this study so urban growth serves as a proxy for measuring societal class structures. B y using urban/rural dispersion there is an implicit assumption that urban growth is positively correlated with union membership as wel l as post-materialist and environmental movements. A s a proxy for the class cleavage, the coefficients of the urban/rural variables tell an interesting story. The variables suggest that for E N P , the balance between urban and rural population is a less powerful indicator of party system size than urban growth. Party systems are less affected by the cleavage between urban and rural populations and more so by the concentration o f populations in urban centres. The coefficients for the Alienation scale are consistently negative and significant at the 0.10 level for all but the first model. These coefficients indicate that there is a negative relationship between levels of alienation and E N P . Higher levels of alienation reduce the number o f parties that are elected to legislatures. Comparing Ontario and Quebec, the least and most alienated provinces in this study the first model predicts that Quebec w i l l have 0.159 fewer parties than Ontario, 0.723 fewer in model 2, 0.276 fewer in model 3, and 0.580 fewer in model 4. Social Heterogeneity and CNP In Table 4.2 C N P is the dependent variable. The results are encouraging as all o f the models are significant at the 0.001 level and several o f the independent variables are significant at the 0.10 level. There some consistency in the direction and magnitude o f the coefficients. Because voter preferences are more pronounced in measurements of 71 C N P it was expected that the greater variation in C N P over the period o f this survey would produce more robust results and these models do not disappoint. The coefficients for index measurements are less consistent than the E N P models. Depending on the variable, social heterogeneity can have both positive and negative relationships with the number of parties that garner vote support. Coefficients for the ethnicity and religious indexes are significant at least at the 0.10 level in three o f the four models. A s with the E N P , models the positive relationship between ethnicity and C N P is not surprising. However, the negative relationship between religious heterogeneity and the number o f parties is somewhat unexpected. The coefficients for the religious variable are significant at the 0.05 level for all but the second model. Why is religion not positively correlated with C N P ? Part o f the answer is that the number of religious cleavages does not impact vote support for parties but rather the magnitude o f those cleavages. In North America, the key political-religious cleavage is centred on social policies that are either founded in some way on Christian Scriptures or are nonsectarian. It would be expected that the C N P in a province with fewer major religions and more clearly defined religious boundaries would be lower than a province with multiple significant religious groups with less definable religious divisions. The existence o f fewer religious cleavages is coupled with an assumption that cleavages between these religious groups are stronger and w i l l act as a constraining force on the number of parties that receive votes. 7 2 Table 4.2 Estimates of the Competitive Number of Parties (CNP) in Canadian Provinces (1951-1996) Model 1 Model 2 Model 3 Model 4 Ethnicity Language Index Linguistic Ratio Basic Religious Index Rural-Urban Index Rural-Urban Ratio Agricultural Alienation District Magnitude Constant R 2 Wald chi2 Prob > chi2 -.6537 (.3401) 1.3577 (.5013) 1.0366 (.4171) -.6896 (.6078) -.6433 (.3347) .4497 (.3287) .8666 (.6465) -.9148 (.6225) -.6898 (.3764) 1.0660*** (.4030) 1.0085 (.4876) -.5886 (.3818) .8912*** (.3122) 1.7880*** (.5871) .1387** .2562*** .0154 .0011 (.0576) -.0085 (.0528) .0108 (.0539) -.4269* (.0580) .1474 (.0522) -.4406** (.0519) -.0027 (.2627) .5906 (.2787) .9573* (.1955) .8165 (.2341) .7105 (.5422) 1.1392 (.5682) 1.5635* (.5155) .6867 (.5060) .3134 (.7215) (.7715) (.5956) (.6194) 0.8954 0.8669 0.9297 0.9436 34.32 20.88 39.81 45.73 0.0000 0.0040 0.0000 0.0000 OLS with panel corrected standard errors, standard errors in parentheses, ***p<.01, **p<.05, *p<.10 Both of the coefficients for the linguistic index are positive and significant at the 0.01 level indicating that there is a negative relationship between linguistic heterogeneity and party system size; more diversity leads to fewer parties. A s with the analysis of the E N P models the lack of variation in the linguistic index outside o f New Brunswick and Quebec's allows prediction o f how much linguistic diversity affects C N P . The first model predicts there w i l l be on average 0.419 fewer parties in New Brunswick and 0.282 fewer 73 parties in Quebec than in the rest o f Canada. The third model predicts that there w i l l be 0.329 fewer parties in New Brunswick and 0.222 fewer in Quebec. The coefficients for the linguistic ratio are both positive though only one is significant at the p=0.01 level. The regression results indicate that there is a positive relationship between the proportions of English speakers, relative to the French speaking population, and C N P . Unlike the E N P models these results indicate that a significant French speaking population has a limiting effect on party systems. A possible explanation for this could be that at a voting level the party systems becomes constrained through the splitting of vote support into two camps; camps that are either explicitly or implicitly pro-French party or pro-English. In these provinces the mechanical and psychological effects of the electoral system have a powerful effect on the size of the party system. The third model predicts there w i l l be on average 0.391 fewer parties in Quebec and 0.088 parties in New Brunswick. The fourth model predicts there w i l l be 0.776 fewer parties in Quebec and 0.174 fewer parties in New Brunswick. The coefficients for the R U I are negative in the first and second model. A s with the E N P models this is an unexpected result. These results indicate that provinces with balanced urban and rural populations w i l l have more parties receiving vote support than provinces with unbalanced populations. Further evidence of the nature o f this relationship is provided by the R U R variables. The coefficients for the ratio in the third and fourth models are both positive and for the third model significant at the p=0.01 level. The results indicate that there is a positive relationship between the size of party systems and the proportion of a population that resides in urban centres larger than 30,000 people. A s a province becomes increasingly urbanized a greater number o f 74 parties receive votes in elections. The coefficients for the agricultural index are not significant or consistent in the four models. Unlike the E N P models the agricultural variable is not useful for either confirming or denying the findings o f the other urban/rural variables. The coefficients for the alienation scale are also inconsistent. Three of the models are negative and the first and third models are significant at the 0.10 level. The inconsistency for the second and third models can be attributed to the use o f the linguistic ratio in these regressions. Both the linguistic ratio and the alienation variable have Quebec as an outlier which results in a high degree o f correlation between the two variables thereby limiting the explanatory power each variable has individually. For the first and third models negative coefficients indicate that there is a negative relationship between levels o f alienation and E N P . Higher levels o f alienation reduce the number of parties that are elected to legislatures. Comparing Ontario and Quebec, the least and most alienated provinces in this study the first model predicts that Quebec w i l l have 0.299 fewer parties, with 0.308 fewer in model 3. Comparing ENP and CNP models This analysis, unsurprisingly, found a stronger link from heterogeneity to C N P than to E N P . It is assumed that most voters understand the restrictive elements o f the electoral system, at least in as much as they recognize that only a limited number o f candidates have a chance o f winning. Under this assumption it was expected that the distribution o f votes offers a 'truer' representation of the voters' desires than the 75 distribution of seats. Estimations of C N P should be more robust because they more closely reflect the political affiliations of the population. These results also indicate that not all voters are fully aware o f the restrictive effects, or it verifies that there are voters who chose to support the party of their choice regardless o f its chances. The two sets of regressions indicate that there is a previously unexplored connection between social heterogeneity and party system size. O f the two sets o f models the C N P models are the more robust with all regressions significant at the 0.05 level or better. The E N P models are still useful given the small amount o f variation in the dependent variable. This section w i l l compare the results o f the two regression results and w i l l also attempt to explain some of the inconsistency issues in the regression results. The ethnicity coefficients are weakly significant in the C N P models but are not significant in E N P models. This is a reoccurring theme in this analysis, several o f the heterogeneity variables are correlated with C N P but not with E N P . Wi th these variables it appears that some cleavages are salient enough to alter voting behaviour among a section o f the populace. These divisions convince some voters to move their support from traditional parties to minor and emerging parties. However, the shift o f votes does not seem to be significantly high enough to push these smaller parties over the restrictive electoral-system thresholds. There is consistency in both E N P and C N P models for both o f the linguistic variables but there is a surprising variance between the two sets o f regressions. The linguistic index variable is positive in both sets o f models but it is only significant in the C N P models. Similar to the ethnic variable, linguistic diversity is positively correlated 76 with the number of parties that get votes but the strength o f this correlation is not high enough for it to significantly affect the number parties that win seats. A t least one coefficient for the linguistic ratio variable in each of the sets o f models is significant. However, the coefficients are negative in E N P models and positive in C N P models. This suggests that on the one hand provinces with a significant exclusively francophone population elect more parties than exclusively anglophone provinces, but on the other hand, few parties receive votes in French speaking provinces. This is an interesting result. W h y does a cleavage that reduces the number of parties that receive votes in an election also seem to increase the number o f parties that are ultimately elected? This result is not as paradoxical as it seems. The linguistic ratio compares English to French speakers in each province and the outlier in this variable is Quebec. A s intended the linguistic ratio acts as a proxy for measuring the social cleavage between the sovereigntist and nationalist populations. The regression results indicate how significant this division is. The oddly inverse relationship suggests that language is a powerful variable that is capable o f narrowing the playing field. Its dominance has reduced the number o f dimensions that parties compete on. Party politics in Quebec has been played out on two dimensions, one ideological and the other nationalistic. In this two dimensional space three parties successfully staked out positions with relatively wide voter appeal. From the 1950s to the 1970s the three parties were Union Nationale, Social Credit and the Parti Quebecois. In more recent years the three parties have been the Parti Quebecois, Liberal, and Act ion Democratique. Over the past fifty years these parties have been able to build support bases that win themselves seats and effectively shutout smaller parties. The consolidation o f vote support has led to more proportional results. 77 In an electoral arena where institutional effects tend to constrain party system size this effect is an interesting inconsistency. This existence of this curious relationship is reinforced by the religious variables. In CNP models the religious index is positive and significant indicating that religious diversity is negatively correlated with the number of parties receiving votes. As the least religiously diverse province, these regressions indicate that, ceteris paribus Quebec will have more parties receiving votes than any other province. For ENP models the two that include the linguistic index are not significant. However, the two models that use the linguistic ratio are significant and the coefficients are negative. This indicates that there is a positive relationship between ENP and religious heterogeneity. Quebec will, all else being equal, have fewer parties getting elected to legislatures than other provinces. In the four regression equations that include the linguistic ratio the effects of this variable and the religious index seem to cancel each other to a certain degree. There is a high degree of correlation between the two variables but they are both significant in two of the regression models. Coefficients for the alienation scale are generally negative and most are significant at the 0.10 level. This indicates that there is a negative relationship between levels of alienation and number of parties that win votes and seats. Provinces with high levels of alienation will have fewer parties competing for and winning seats. The coefficients for the battery of urban/rural variables indicate that residence is positively correlated with party system size. As provincial populations become more urbanized there is an increase in the number of parties that receive votes and win seats. While this result is not wholly unexpected it does seem to refute a traditional connection 78 made in Canadian politics between rural alienation and populist parties. The success o f the Social Credit and Reform/Alliance parties at a federal level are past examples o f political movements that have at their core rural grassroots followings. A t different times in history, both parties successfully translated rural discontentment with urban centres into political representation for with considerable success. The results o f this thesis do not disprove the effects o f rural alienation; but they do suggest that the magnitude of the difference between rural and urban populations is not a significant catalyst in populist movements. Explaining the dynamics o f rural politics is beyond the scope o f this thesis but this is a clear area of study that deserves further attention with comparative provincial data. Testing the Models Tables 4.3 and 4.4 report the estimations o f E N P and C N P for all provinces for two sets of election results. The estimates are derived from election data from around 1960 and 1990. The thirty year gap tests the validity o f the models over a set time period and also gives a basic idea o f shifts in party systems over time. The results in both tables are encouraging. The models do a decent job of estimating the E N P and C N P with the results reasonably matching the actual figures. On the whole, the E N P models have a slight tendency to over-estimate while the C N P models have an equally slight tendency to underestimate. 79 Table 4.3 Estimated E N P for Selected Provincial Cases - with comparisons between early period and late period in study Province (election year) Predictions Actual E N P Model 1 M o d e l 2 M o d e l 3 Model 4 British Columbia (1991) 1.914 1.868 1.964 2.039 2.023 British Columbia (1963) 2.064 1.917 1.759 1.864 1.780 Alberta (1993) 1.900 1.843 1.766 1.919 1.835 Alberta (1959) 1.376 1.831 1.486 1.750 1.563 Saskatchewan (1986) 1.979 1.880 1.830 1.762 1.779 Saskatchewan (1956) 1.871 1.749 1.645 1.651 1.599 Manitoba (1986) 2.060 1.928 2.021 1.969 2.005 Manitoba (1962) 2.145 1.902 1.835 1.861 1.834 Ontario (1991) 2.135 1.931 2.232 2.168 2.250 Ontario (1959) 1.730 1.911 1.882 1.992 1.935 Quebec (1989) 1.700 1.756 1.853 1.813 1.700 Quebec (1962) 1.830 1.692 1.750 1.570 1.669 New Brunswick (1987) 1.536 1.866 2.115 1.803 1.993 New Brunswick (1956) 1.899 1.793 2.085 1.731 1.943 Nova Scotia (1988) 2.198 1.892 1.956 1.870 1.928 Nova Scotia (1956) 2.052 1.885 1.781 1.851 1.798 Newfoundland (1993) 1.825 1.691 1.663 1.658 1.651 Newfoundland (1962) 1.463 1.699 1.607 1.670 1.615 In terms o f long-term trends there is a divergence between what the two sets of models suggest is occurring over time. The E N P models do not corroborate initial observations that E N P is changing over time. However, they do support the hypothesis that there are stable differences between provinces. There is evidence o f variation over time and space for the C N P models. The inclusion o f social heterogeneity in explanatory models of party systems indicates that provinces in polities with identical electoral 80 systems and only slight variations in district magnitude there are variables that heretofore have not been accounted for that affect the size o f party systems. In British Columbia the estimations for E N P are consistently lower than the actual figures while the estimations for C N P are consistently higher. This suggests that more parties then would be expected are winning seats in the province. The first and third models are better estimators o f E N P for Alberta which indicates that the difference in treatment o f linguistic cleavages has an impact on approximations o f E N P . A t first this is a perplexing finding as the linguistic ratio and linguistic index are both very close to one for all census periods in this study. This result indicates that overall it is not simply the existence o f a linguistic cleavage but also its direction that is critical. The index which simply measures the magnitude o f the cleavage between English and French speakers is not as significant as the ratio variable which measures the difference between the two linguistic camps. It is not just the lack o f a linguistic cleavage but the lack o f a significant population of French speakers that has a constraining effect on the number of parties elected. This effect is not limited to Alberta; however, it is with this province's elections that the effect is quite noticeable. In models that do not use the ratio variable the estimations for E N P are consistently low. There is no discernable difference between models that estimate C N P for Alberta which indicates that the absence o f a significant French speaking population does not affect the number o f parties receiving votes. For Saskatchewan and Manitoba the models tend to under-estimate E N P and C N P , the notable exception to this trend are the approximations for C N P in Saskatchewan for 1991 elections. In multivariate models it is hard to determine the specific cause for this but it does indicate that the effect of one or more of the variables is consistently 81 different from the other provinces. In both provinces the estimations are higher for the more recent period which mirrors the actual changes in party system size in these provinces. Estimation of C N P for Ontario generally hit the mark while approximations of E N P are over-estimated. The cause of this effect is identical to that identified in estimations for Saskatchewan and Manitoba except it is an opposite direction. Table 4.4 Estimated CNP for Selected Provincial Cases - with comparisons between early period and late period in study Province Actual Predictions (election year) C N P Model 1 Mode l 2 Mode l 3 Mode l 4 British Columbia (1991) 2.985 2.578 2.516 2.694 2.682 British Columbia (1956) 2.946 2.923 2.916 2.929 3.043 Alberta (1986) 2.762 2.584 2.638 2.607 2.661 Alberta (1959) 2.546 2.710 2.738 2.567 2.648 Saskatchewan (1991) 2.627 2.778 2.703 2.651 2.537 Saskatchewan (1964) 2.808 2.675 2.523 2.598 2.537 Manitoba (1986) 2.856 2.752 2.641 2.749 2.637 Manitoba (1962) 2.807 2.805 2.683 2.707 2.661 Ontario (1991) 2.910 2.736 2.501 2.948 2.874 Ontario (1963) 2.642 2.827 2.603 2.716 2.626 Quebec (1994) 2.485 2.420 2.316 2.469 2.383 Quebec (1956) 2.111 2.326 2.378 2.230 2.202 New Brunswick (1987) 2.228 2.348 2.448 2.314 2.333 New Brunswick (1956) 2.073 2.256 2.397 2.242 2.322 Nova Scotia (1988) 2.698 2.578 2.400 2.454 2.332 Nova Scotia (1956) 2.121 2.808 2.632 2.677 2.677 Newfoundland (1989) 2.208 2.169 2.127 2.077 2.046 Newfoundland (1962) 2.086 2.234 2.190 2.161 2.162 82 In Quebec, there is a large variation between elections in E N P but only a small variation in C N P . There is some variance in the measures o f social heterogeneity although not very large. The difference in magnitude between the two estimations is largely due the deviation between the estimated coefficients for the linguistic heterogeneity variable. For the E N P model the estimated coefficient is 2.595, for the C N P model it is 1.094. Overall, Quebec has the highest variation between time periods for the language variables and this is demonstrated by the estimations for the province. The difference in linguistic heterogeneity between 1960 and 1985 (0.255) translates into approximately 0.66 parties on the E N P measure but only about 0.28 parties on the C N P measure. This partially describes why there is a sizable deviation in estimations between the two elections for E N P but not for C N P . The models struggle to accurately explain the shape o f party systems in the Maritimes. The models consistently over-estimate E N P and C N P for N e w Brunswick indicating that there is some unexplained aspect o f New Brunswick which is not captured by any o f the models. The same can be said for Nova Scotia except the models tend to under-estimate E N P and C N P . Some o f the unexplained variation may be attributable to the longstanding party loyalties identified by Stewart. The province with the poorest overall fit is Newfoundland. It is l ikely that low levels of ethnic and linguistic heterogeneity are contributing to under-estimations o f C N P and over-estimations of E N P . The difficulties with Maritime Provinces make a strong case for using dummy variables to capture the unique unexplained effects in each province. However, using dummy variables runs the risk o f obscuring or masking the effects o f the heterogeneity variables which is why they have not been used in this analysis. 83 Summary It is clear that there is a relationship between social heterogeneity and party systems. Even when party systems are constrained by low district magnitude and plurality electoral systems, social cleavages are significantly related to the number of parties that win votes and seats. The two measurements of party system size used in this thesis have looked at the number of votes parties receive (CNP) and the number of seats that parties win (ENP). There is a correlation between both measurements and social heterogeneity. There is a reasonably strong relationship between CNP and social heterogeneity, with three of the four models significant at the 0.001 level or better. The strength of the relationship between ENP and social cleavage structures is moderately strong. Three of the four models are significant at the 0.05 level or better. Social heterogeneity has been measured on several dimensions: ethnicity, religion, language, and the proportion of rural and agrarian populations. From this some generalizations are possible. Ethnicity is positively correlated with party system size for both ENP and CNP. More ethnic groups lead to slightly more parties wining seats and votes. There is a complex relationship between religion and party system size. Religion is positively correlated with ENP but negatively correlated with CNP. More religious groups tend to increase the number of parties that win seats but it also constrains the number of parties that receive votes. A similar relationship occurs with linguistic diversity. Two variables were used and they both indicate that the anglophone/ francophone cleavage does have an effect on party system size. The size of the French speaking population has a positive relationship with the number of parties that win seats but a negative relationship with the vote distribution among parties. The size of a 84 province's rural population is negatively correlated with party system size. Interestingly though, the proportion of a population that lives in agrarian communities has no effect on party system size. These findings build upon the work of Taagepera and Shugart; Lijphart; Ordeshook and Shvetsova; and Cox and Neto. Their studies successfully argued that social heterogeneity is related to the size of party systems. However, few have been able to demonstrate this quantitatively. This paper has advanced their work by identifying specific areas of social heterogeneity that can be linked to party system size. In addition, this study has been able to test a model that shows promise as a framework for studying other states. An interesting application could be to the United States which for a variety of reasons is rigidly two-party. Applying this model would not be looking for cases why and where a third party might emerge but rather why one party tends to dominate another. Another set of countries that would be fruitful to explore are states that do not use plurality or majoritarian electoral systems and/or employ multi-member districts. Two important conclusions can be drawn from this thesis. First, there is substantial evidence that social heterogeneity influences ENP and CNP. In the provinces studied, district magnitude is equal to or approaches one and all of the elections have been decided using the plurality electoral formula. Under these conditions, institutional approaches to party systems would suggest that the effects of societal fractionalization would have little or no effect on the ENP and the CNP. This thesis rejects the exclusively institutional explanation. The regression results suggest that social cleavages have a significant affect on the size of party systems. Ordeshook and Shvetsova (1994). 85 The second conclusion is that this thesis has identified a largely untouched area o f research on provincial party systems. For the last fifty year European theorists have used social structural approaches to build a richer understanding o f the way in which societies and political parties co-evolve. This thesis has proved that these approaches have an important role to play on this side o f the Atlantic. With respect to provinces, social structuralism provides an effective framework for understanding how social cleavages affect who wins votes and seats. B y finding fruit this thesis proves there is an orchard to be explored. 86 Works Cited Barnes, Samuel H . "Quebec Catholicism and Social Change." The Review of Politics, V o l . 23, N o . 1. (Jan, 1961), 52-76. Bartolirii, Stephano, and Peter Maif . Identity, Competition, and Electoral Availability: the Stabilization of European Electorates 1885-1985. Cambridge: Cambridge University, 1990. Beck , Nathaniel; and Jonathon N . Katz. "What to do (and not to do) with Time-Series Cross-Section Data." American Political Science Review. Vol. 89, No . 3. (Sep, 1995), 634-647. Blais, Andre, Elisabeth Gidengil, Richard Nadeau, and N e i l Nevitte. "Measuring Party Identification: Britain, Canada, and the United States." Political Behaviour. V o l . 23, Issue 1,(2001), 5-22. Blais, Andre. "Accounting for the Electoral Success of the Liberal Party in Canada" Presidential Address: Canadian Political Science Association. June 2005. Blake, Donald. "Constituency Contexts and Canadian Elections: A n Exploratory Study." Canadian Journal of Political Science. V o l . 11, N o . 2 (Jun., 1978), pp. 279-305 —. "The Consistency of Inconsistency: Party Identification in Federal and Provincial Politics." Canadian Journal of Political Science. V o l . 15, N o . 4. (Dec, 1982), 691-710. —. "The Measurement o f Regionalism in Canadian Voting Patterns." Canadian Journal of Political Science. V o l . 5, No . 1. (Mar, 1972), 55-81. —. Two Political Worlds. Vancouver: U B C , 1985. —. "Value Conflicts in Lotusland: B C Political Culture." In R. K . Carty Ed . Politics, Policy, and Government in British Columbia. Vancouver: University o f British Columbia, 1996. Budge, Ian; Ivor Crewe and Dennis Farlie. Party Identification and Beyond: Representations of Voting and Party Competition. London: Wiley, 1976. Butler, David and Donald Stokes. Political Change in Britain; Forces Shaping Electoral Choice. New York , St. Martin's, 1969. Cairns, Ala in . "The Electoral System and the Party system in Canada, 1921-1965." Canadian Journal of Political Science. V o l . 1, No . 1. (Mar, 1968), 55-80. Clarke, Harold D . ; and Marianne C. Stewart. "Partisan Inconsistency and Partisan Change in Federal States: The Case o f Canada." American Journal of Political Science, V o l . 31, N o . 2. (May, 1987), 383-407. Cox, Gary. Patterns of Democracy, New Haven: Yale , 1999. 87 Cox, Gary; and Octavio Neto. "Electoral Institutions, Cleavage Structures, and the Number o f Parties," American Journal of Political Science, V o l . 41, Issue 1 (January 1997), 149-174. Campbell, Angus; et al. The American Voter. Chicago: University o f Chicago, 1960. Campbell, Angus; et al. The Voter Decides. Evanston: Row, Peterson & Co. , 1954. Carty, R. K . " O n the Road Again: 'The Stalled Omnibus' Revisited," Canada's Century. C. E . S. Franks et al Ed . Montreal: McGill-Queen's University, 1995. Chhibber, Pradeep; and Kenneth Kollman. The Formation of National Party Systems: Federalism and Party Competition in Canada, Great Britain, India, and the United States. Princeton: Princeton University, 2004. Clarke, Harold D . "The Parti Quebecois and Sources o f Partisan Realignment in Contemporary Quebec." The Journal of Politics, V o l . 45, N o . 1. (Feb, 1983), 64-85. Crewe, Ivor, and David Denver. Electoral Change in Western Democracies: Patterns and Sources of Electoral Volatility. Beckenham: Croom Helm, 1985. Converse, Phil ip E . , and Pierce, R o y E . (1987). "Measuring partisanship." Political Methodology 11: 143-166. Dalton, Russell J. Citizen Politics in Western Democracies : Public Opinion and Political Parties in the United States, Great Britain, West Germany, and France. Chatham: Chatham House: 1988. Dalton, Russell J., Scott C . Flanagan, and Paul A l l en Beck. Electoral Change in Advanced Industrial Democracies: Realignment or Dealignment? Guildford: Princeton, 1984. Downs, Anthony. An Economic Theory of Democracy. New York : Harper & Brothers, 1957 Duverger, Maurice. Political Parties, Trans. Barbara and Robert North, London: Methuen, 1964. Enelow, James M . ; and J. Me lv in Hinich. The Spatial Theory of Voting: An Introduction. Cambridge: Cambridge University, 1984. Engelmann, Frederick C ; and Mildred A . Schwartz. Canadian Political Parties: Origins, Character, Impact, Scarborough: Prentice-Hall, 1975. Ersson, Svante, and Jan-Erik Lane. "Democratic Party Systems in Europe: Dimensions, Change and Stability." Scandinavian Political Studies. V o l . 5 (1982), 67-96. Franklin, Charles, and John Jackson. "The Dynamic o f Party Identification." American Political Science Review. V o l . 77 (1983), 957-73. 88 Gidengil, Elisabeth. "Canada Votes: A Quarter Century o f Canadian National Election Studies." Canadian Journal of Political Science. V o l . 25, N o . 2. (Jun, 1992), 219-248. Green, Donald, and Bradley Palmquist. " O f Artifacts and Partisan Instability." American Journal of Political Science. V o l . 34 (1990), 872-902. Green, Donald Phil ip, Bradley Palmquist and Eric Schickler. Partisan Hearts and Minds: Political Parties and the Social Identities of Voters. New Haven: Yale . 2002 Inglehart, Ronald. Culture Shift in Advanced Industrial Society. Princeton: Princeton, 1990. —. The Silent Revolution: Changing Values and Political Styles among Western Publics. Princeton: Princeton, 1977. Irvine, Wi l l i am P. "Explaining the Religious Basis o f the Canadian Partisan Identity: Success on the Third Try." Canadian Journal of Political Science. V o l . 7, No . 3. (Sep. 1974), 560-63. —. "Comment on 'The Reproduction of the Religious Cleavage in Canadian Elections." Canadian Journal of Political Science. V o l . 18, N o . 1. (Mar. 1985), 115-117. Jenson, Jane. " Party Loyalty in Canada: The Question of Party Identification." Canadian Journal of Political Science V o l . 8, N o . 4 (Dec, 1975), 543-53. Johnston, Richard and Janet Ballantyne, "Geography and the Electoral System." Canadian Journal of Political Science. V o l . 10, N o . 4. (Aug, 1977), 857-66. Johnston, Richard et al. Letting the People Decide, Stanford: Stanford University, 1992. Johnston, Richard. "The Geography o f Class and Religion in Canadian Elections," in Joseph Wearing, Ed. , The Ballot and its Message: Voting in Canada. Toronto: Copp-Clark Pitman, 1991. —. "The Reproduction of the Religious Cleavage in Canadian Elections." Canadian Journal of Political Science. V o l . 18, No . 1. (Mar, 1985), 99-113. Katz, Richard S., A Theory of Parties and Electoral Systems. Baltimore: John Hopkins University, 1980. Koop, Royce; and Kashi Tanaka. " A V on Trial : The Impact o f Alternative Vote Electoral Systems in Canadian Politics." Unpublished paper (2003). Romberg, Al len ; Harold D Clarke; and Marianne C . Stewart "Federalism and Fragmentation: Political Support in Canada." The Journal of Politics, V o l . 41, No . 3. (Aug, 1979), 889-906. Kriesi , Hanspeter. "The Transformation of Cleavage Politics: The 1997 Stein Rokkan lecture." European Journal of Political Research. V o l . 33 (1998), 165-85. 89 Layman, Geoffrey C . and Edward G . Carmines. "Cultural Conflict in American Politics: Religious Traditionalism, Postmaterialism, and U . S . Political Behaviour," The Journal of Politics, V o l . 59, Issue 3 (August 1997), 751-77. LeDuc, Lawrence; Harold D . Clarke; Jane Jenson; and Jon H . Pammett. "Partisan Instability in Canada: Evidence from a New Panel Study." The American Political Science Review > V o l . 78, No . 2 (Jun, 1984), 470-484. Lijphart, Arend. "Religious vs. Linguistic vs. Class Voting: The "Crucial Experiment" of Comparing Belgium, Canada, South Africa, and Switzerland." The American Political Science Review, V o l . 73, No . 2. (Jun, 1979), 442-58. —. "The Political Consequences o f Electoral Laws, 1945-85," The American Political Science Review, V o l . 84, Issue 2 (June 1993), 481-496. Lipset, Seymour Martin. "Political Cleavages in 'Developed' and 'Emerging' Polities," Reprint No. 244, Berkley: Institute of International Relations, 1964. Lipset, Seymour Martin; and Stein Rokkan. Party systems and voter alignments : cross-national perspectives. New York : Free Press, 1967. McRoberts, Kenneth. "Quebec and the Canadian Political Crisis ." Annals of the American Academy of Political and Social Science, V o l . 433, Ethnic Conflict in the World Today. (Sep, 1977), 19-31. Meisel , John. "Political Culture and the Politics of Culture." Canadian Journal of Political Science /Revue canadienne de science politique, V o l . 7, N o . 4. (Dec , 1974), pp. 601-615. —. "The Stalled Omnibus and the Decline o f Party" in R. K . Carty Ed . Canadian Political Party Systems. Peterborough: Broadview, 1992. Ordeshook, Peter C ; and Olga V . Shvetsova. "Ethnic Heterogeneity, District Magnitude, and the Number o f Parties," American Journal of Political Science, V o l . 38, Issue 1 (February 1994), 100-123. Powell , G . Bingham. Contemporary Democracies: Participation, Stability, and Violence, Cambridge: Harvard, 1982. Rae, Douglas. The Political Consequences of Electoral Laws, New Haven: Yale , 1971. Rabinowitz, George; and Stuart Elaine MacDonald. " A Directional Theory of Issue Voting." The American Political Science Review, V o l . 83, Issue 1 (March 1989), 93-121. Rawlyk, George. "Religion in Canada: A Historical Overview." Annals of the American Academy of Political and Social Science, V o l . 538, Being and Becoming Canada. (Mar, 1995), 131-142. 90 Ricketts, Edmond F; and Herbert Waltzer "Electoral Arrangements and Party System: The Case of Canada." The Western Political Quarterly, V o l . 23, N o . 4. (Dec, 1970), 695-714. Rose, Richard, and Derek Urwin . "Social Cohesion, Political Parties and Strains in Regimes." Comparative Political Studies V o l . 2 (1969), 7-67. Rutan, Gerard F. "Two Views o f the Concept of Sovereignty: Canadian, Canadien." The Western Political Quarterly. V o l . 24, No . 3 (Sept, 1971), 456-466. Sartori, Giovanni. "The Influence o f Electoral Systems: Faulty Laws or Faulty Method?' Ed . Bernard Grofrnan and Arend Lijphart. Electoral Laws and their Political Consequences. New York: Agathon, 1986 Siaroff, Alan . "Two-and-a-Half-Party Systems and the Comparative Role o f the Half," Party Politics, V o l . 9, Issue 3 (May 2003), 267-290. Shamir, Micha l . "Are Western Party Systems 'Frozen'? A Comparative Dynamic Analysis." Comparative Political Studies. V o l . 17 (1984), 35-79. Stasiulis, Daiva; and Yasmeen Abu-Laban. "Unequal Relations and the Struggle for Equality: Race and Ethnicity in Canadian Politics," in Michael Whittington and Glen Wil l iams (eds.), Canadian Politics in the 21st Century. Scarborough: Nelson, 2000 Stevenson, H . Michael . "Ideology and Unstable Party Identification in Canada: Limited Rationality in a Brokerage Party System" Canadian Journal of Political. V o l . 20, N o . 4. (Dec. 1987), 813-850. Stewart, David. "Political Realignment in Atlantic Canada?" Regionalism and Party Politics in Canada, Ed . L isa Young and Kei th Archer. Oxford: Oxford, 2002. Taagepera, Rein and Matthew Soberg Shugart, Seats and Votes: The Effects and Determinants of Electoral Systems. New Haven: Yale University, 1989. —. "Predicting the Number of Parties: A Quantitative Model o f Duverger's Mechanical Effect," The American Political Science Review, V o l . 87, Issue 2 (June 1993), 455-464. Uslaner, Eric M . "Splitting Image: Partisan Affiliations in Canada's "Two Political Worlds." American Journal of Political Science, V o l . 34, N o . 4. (Nov, 1990), 961-981. Government Publications Dominion Bureau o f Statistics. 1951 Ninth Census of Canada: Population (Origin, Bulletin: 1-14; Religious Denomination, Bulletin: 1-15; Official Language and 91 Mother Tongue, Bulletin: 1-17; and Population characteristics for census metropolitan areas, Bulletin: 1-19). Ottawa: Queen's Printers, 1953. Dominion Bureau o f Statistics. 1956 Census of Canada: Population (Rural and urban distribution, Bulletin: 1-7). Ottawa: Queen's Printers, 1957. Catalogue number 92- 507. Dominion Bureau o f Statistics. 1961 Census of Canada: General Review (Growth of Population in Canada, Bulletin: 7.1-1; Rural and Urban Population, Bulletin: 7.1-2; Origins of the Canadian Population, Bulletin: 7.1-6; Mother Tongue and Official Language, Bulletin: 7.1-9; and Religious Denominations on Canada, Bulletin: 7.1-11). Ottawa: Trade and Commerce, 1963. Catalogue numbers 99-511, 99-512, 99-516, 99-519, & 99-521. Dominion Bureau o f Statistics. 1966 Census of Canada: Population (Rural and Urban Distribution, Vol. 1 [1-8]). Ottawa: Trade and Commerce, 1968. Catalogue number 92-608. Statistics Canada. 1971 Census of Canada: Population (Ethnic Groups, Bulletin: 1.4-3; Religious Denominations, Bulletin: 1.4-4; and Language, Bulletin: 1.4-8). Ottawa: Trade and Commerce, 1974. Catalogue numbers 92-731, 92-732, & 92-736. Statistics Canada. 1976 Census of Canada: Populations: Geographic Distributions (Urban and Rural Distribution, Bulletin: 1.8). Ottawa: Industry, 1978. Catalogue number 92-807. Statistics Canada. 1981 Census of Canada: Population (Ethnic Origin, Vol. I; Highlight Information on: Ethnicity, Place of Birth, Citizenship, Income, Language, Immigration, Religion, and Shelter Costs). Catalogue number 92-911. Statistics Canada. 1986 Census of Canada: Population (British Columbia, Alberta, Saskatchewan, Manitoba, Ontario, Quebec, New Brunswick, Nova Scotia, PEI, and Newfoundland). Ottawa: Supply and Services Canada, 1987. Catalogues 92-109, 92-110, 92-111, 92-112, 92-113, 92-114, 92-115, 92-116, 92-117, & 92-118 Statistics Canada. 1986 Census of Canada: The Nation (Language: Part II; Ethnicity, Immigration and Citizenship. Ottawa: Industry Canada, 1989. Catalogue numbers 93- 103, 93-109 Statistics Canada. 1996 Census of Canada: Profile of Census Divisions and Subdivisions. Ottawa: Industry Canada, 1999. Catalogue numbers: 95-182-XPB, 95-183-XPB, 95-184-XPB, 95-185-XPB, 95-186-XPB, 95-187-XPB, 95-188-X P B , 95-189-XPB, 95-190-XPB, & 95-191-XPB. Statistics Canada. 2001 Census of Canada: Profile of Census Divisions and Subdivisions. Ottawa: Industry Canada, 2004. Catalogue number 95-2222-XPB. 92 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items