Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Voters’ evaluations of prime ministerial candidates : the impact of leader traits in the 2000 Canadian… Nakai, Emily 2003

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

Item Metadata


831-ubc_2003-0416.pdf [ 4.54MB ]
JSON: 831-1.0091009.json
JSON-LD: 831-1.0091009-ld.json
RDF/XML (Pretty): 831-1.0091009-rdf.xml
RDF/JSON: 831-1.0091009-rdf.json
Turtle: 831-1.0091009-turtle.txt
N-Triples: 831-1.0091009-rdf-ntriples.txt
Original Record: 831-1.0091009-source.json
Full Text

Full Text

VOTERS'EVALUATIONS OF PRIME MINISTERIAL CANDIDATES: THE IMPACT OF LEADER TRAITS IN THE 2000 CANADIAN FEDERAL ELECTION by E M I L Y N A K A I Bachelor of Arts, The University of British Columbia, 1996 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS in THE F A C U L T Y OF G R A D U A T E STUDIES (Department of Political Science) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A July 2003 © Emily Nakai, 2003 in presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department The University of British Columbia Vancouver, Canada DE-6 (2/88) ii ABSTRACT This study examines the impact of perceived personality traits of the political party leaders on voting decisions in the 2000 Canadian federal election, replicating Richard Johnston's research that is based on the 1997 election. Employing data from the 2000 Canadian Election Study (CES), the research uses Ordinary Least Squares regression analysis to estimate how evaluations of leader personality traits over two aggregated dimensions -competence and character - moved votes. The changes in the design of the 2000 CES from prior years created many difficulties in assessing voters' evaluations of the party leaders and limited the comparability of the results from the study. The key methodological differences are: (1) leaders were not evaluated individually; (2) it did not measure the degree of applicability of the trait labels; (3) it included significantly fewer leader personality questions, and (4) the "new ideas" variable does not fall squarely into either the competence and character domains and seems to favour the new Alliance Party leader. This study finds that leader effects are more critical to the parties struggling for their political survival. A counterfactual party leader-switching exercise suggests that the distance between the frontrunner parties and the others was too great for leader-switching effects to make a difference in determining which parties would form the government and the Official Opposition and whether the winning party would form a majority or minority government. Joe Clark improved his party's standing during the campaign and helped it to retain its official party status while evaluations of Stockwell Day declined. The relevance of judgements of Day and Clark on pre-election vote intentions moved in the same direction as I l l voters' respective evaluations of the leaders over the campaign. This study confirms that campaigns can have an effect on voters. The study supports earlier research findings that suggests that Canadian elections are vulnerable to leader effects. Conventional wisdom that is driven by the media's focus on the personalities suggests that leaders are significant factors in Canadian federal elections, but the empirical research reported in this study and others before it suggest otherwise. CONTENTS A B S T R A C T ii LIST OF FIGURES v LIST OF T A B L E S ' INTRODUCTION 1 CHAPTER 1. THEORY A N D LITERATURE REVIEW 4 The Canadian Political Environment 5 Which Box Should I Tick? The Voters' Decision-making Process 9 The Evolution of Research on Leader Personality Effects on Voting Behaviour 12 2. M E T H O D O L O G Y 29 Selection of Variables and Statistical Methods Employed 30 3. D A T A A N D FINDINGS: ELECTION D A Y 41 How the Leaders' Personalities Measured Up in 2000 41 Leader Evaluations and their Impact on Voter Choice 43 Leader Evaluations and their Implications for the Outcome of the 2000 Election 50 Leader Switching: What-If Scenarios in the 2000 Election 54 4. D A T A A N D FINDINGS: ON THE W A Y TO THE B A L L O T B O X 58 Defining Momemts? The (non-)Dynamics of the Campaign of 2000 58 The Crossing Paths of Joe Clark and Stockwell Day 62 5. DISCUSSION 67 Research Findings and Their Significance 67 Concluding Remarks 82 APPENDIX 84 BIBLIOGRAPHY 86 LIST OF FIGURES FIGURE 1. Survey Questions ("Q") Employed 33 2. Personality Data Recoding Scheme 34 3. Statistical Control Variables Recoding Scheme 36 4. The Relative Merits of the 2000 Canadian Federal Leaders 42 5. The Net Impact of Canadian Leaders' Perceived Traits 51 6. Impact of the Campaign 59 vi LIST OF TABLES T A B L E 1. The Impact of Canadian Leaders' Perceived Traits 44 2. Average Impact of Canadian Leaders' Perceived Traits, 1988-2000 48 3. Electoral Districts =/< 4.2 Percent Margin of Victory in 2000 (+ won / - lost) 55 4. Stockwell Day's and Joe Clark's Journeys to November 27 62 1 INTRODUCTION Are Canadian elections particularly vulnerable to leader effects? That they are is a contention dating back at least to Andre Siegfried's remarkable early study and seems to be a stylized fact of recent research. It is not hard to outline a case as to why this might be so. But the case is not clearly about leaders' personalities. And the evidence for it is often weak, open to rival interpretations, or not the main issue. Richard Johnston, Prime Ministerial Contenders in Canada^ The President of the United States serves as a focal point for American voters, a political leader whose personality understandably would be expected to significantly affect election outcomes. American executive powers are wide-ranging and clearly packaged in this commanding, highly visible personification of world power, and this reality is underscored by an electoral contest that is frequently portrayed as a horse race of personalities battling for the coveted position. The visual nature of television makes the projection of personalities and concrete leader images easier and more desirable than the mere discussion of issues and political parties, and the voting public uses these images as a key source of information during election campaigns. In Canada, television is the medium of choice for the majority of voters - the 2000 Canadian Election Study ("CES") reveals that 52 percent of voters obtained election information from television sources (as compared to 23 percent and 11 percent from newspapers and radio sources, respectively) - with political party leaders seemingly at the centre of attention during election campaigns. 1 Richard Johnston, "Prime Ministerial Contenders in Canada," in Leaders' Personalities and the Outcomes of Democratic Elections, ed. Anthony King (New York: Oxford University Press, 2002), 158. 2 Institutional context makes for different leadership scenarios on either side of the forty-ninth parallel. Leaders and their personalities operating in parliamentary systems like that of Canada are popularly assumed to matter. Instead of having a (near) direct connection with the electorate as in presidential systems like that of the United States, the Prime Minister is embedded within a structure where s/he is not directly elected by the Canadian people -rather s/he is voted in as one of the 301 Members of Parliament and is only selected as leader of the party by its members. Notwithstanding the institutional differences, the American examples of leadership effects on vote choice are relevant and are included in the literature review supporting this study. This is not to, say that political party leaders do not matter in Canada. The media leads Canadians to believe that leaders matter by commonly talking about the government in ways that label its actions as those of the current Prime Minister - e.g. "The Chretien government today announced..." - and hence, media and public blame or responsibility for its actions is often attributed to the executive leadership. The question relevant to this discussion, however, is whether Canadian elections are actually vulnerable to leader effects. Put another way, are leaders relevant when one must decide how to vote in Canada, and if so, what level of effect do they have in the decision making process? This analysis attempts to extend the research done by Richard Johnston in his chapter entitled "Prime Ministerial Contenders in Canada" in editor Anthony King's book, "Leaders' Personalities and the Outcomes of Democratic Elections." Johnston's study analyzes the effects of leader personalities in the 1988, 1993, and 1997 Canadian federal elections; this 2 Richard Johnston, "Prime Ministerial Contenders in Canada," 158-83. 3 research, in turn, attempts to follow the same methodology in examining prime ministerial contender effects in the successive election in 2000. The study is divided into several parts. The first chapter examines the existing literature and theories relevant to the area of voting evaluations and behaviour that this research in the Canadian realm explores and expounds. The second chapter outlines the methodology of the study and explains its parameters. The results of the quantitative analyses that are based on the 2000 CES data are presented and explained in the third and fourth chapters, and lastly, in Chapter Five, I discuss the results of the analyses and the relevance and significance of the findings. It should be noted that since commencing this study, research has been published that analyzes the 2000 Canadian federal election. This study, however, hopes to contribute to the existing body of research by dissecting and analyzing the data at different points in the election campaign period and considers the contributions, or lack thereof, of factors which are central to the research carried out by Johnston for the 1988, 1993, and 1997 elections, many of which were not included in the research published thus far. In doing so, I hope that some richness will be added to the existing body of knowledge about the impact of leaders in Canadian federal elections in general, and specifically in 2000. 4 CHAPTER 1 THEORY AND LITERATURE REVIEW It is easy to understand the widespread assumption that leaders' personalities must influence results in presidential systems such as those of the United States, France, and Russia. In those countries, voters select individuals to exercise wide-ranging executive powers, whether it be to wage war, appoint the senior judiciary or dissolve the legislature. Those sorts of decisions are very likely to be heavily influenced by the personal qualities and quirks of the decision maker. It might therefore be supposed that it is only natural for voters to take into account the personal traits of presidential candidates -their intelligence, honesty, political acumen, and sheer likeability - when deciding for whom to vote. John Bartle and Ivor Crewe, The Impact of Party Leaders in Britain: Strong Assumptions, Weak Evidence* There is a prima facie case for the argument that political leaders are an important component in the minds of voters on election day in presidential systems. However, in the Canadian context, leader effects are clouded by factors inherent in the country's parliamentary system of government. Various theories and research concerning the institutional differences and other relevant considerations have helped to support this study of the impact of leader personalities on voting decisions, the foundation of which is outlined in the text that follows. The review of the research and theories is grouped into three topic areas: 1. The Canadian Political Environment 2. The Voters' Decision-making Processes 3. The Evolution of Research on Leader Personality Effects on Voting Behaviour 3John Bartle and Ivor Crewe. "The Impact of Party Leaders in Britain: Strong Assumptions, Weak Evidence," in Leaders' Personalities and the Outcome of Democratic Elections, ed. Anthony King (New York: Oxford University Press, 2002), 70. 5 The Canadian Political Environment Few would argue that leaders are unimportant actors in Canadian elections - at the very least they serve as visual representations of the more abstract concept of political parties - but what seems to be questioned in Canada is the substance of how leader traits have an impact on voting decisions. The Canadian institutional setting, the Westminster-style parliamentary system inherited from its colonial mother country, differs significantly from the American system, which poses a problem when attempting to apply the abundance of American and other non-parliamentary system leader trait research findings to the Canadian context. Case in point, the American presidential system tends to focus voters' attention on presidential candidates, whereas the Canadian parliamentary system tends to view a prime ministerial candidate as a leader-member of a party collective. The Canadian Prime Minister is a fellow Member of Parliament who was selected to lead the party that won the most seats in the House of Commons; Canadian voters select a candidate to represent their constituency as a Member of Parliament and this effectively translates into a vote for the leader of the local candidate's political party. Canadian prime ministers are fused into and are both accountable to the legislature and dependent on the confidence of the legislature for their political survival as party leaders, whereas American presidents are not tied institutionally to either of the two federal legislative chambers. Despite the institutional structure that fuses the Prime Minister, leaders are central, visible actors in Canadian electoral politics. Additionally, Americans have three federal voting opportunities ([1] US Senate members are elected, not appointed as in the Canadian case, [2] the people essentially elect the President by way of the Electoral College vote, and [3] voters select a local candidate to 6 represent their constituency in the House of Representatives) during each election that give them the ability to "split their ticket" which means that the three candidates selected may be from different political parties. Canadians only have one vote by which they choose whom will govern their country. Since Americans have multiple federal voting decision points, they have more outlets for compromise and, for example, may "split their ticket" and vote for a president based on personality traits and their Senate and/or House representatives based on political party affiliation and/or other relevant factors. The ability to vote in this manner enables Americans to view presidential candidates as individuals, not bundled in packages with their parties as in the parliamentary system case. Americans inherently have leeway to consider factors including the presidential candidates' personal traits and other non-partisan factors. Furthermore, unlike in parliamentary systems where party discipline is critical to the governing party's political survival, knowing that one's elected Congress member may vote based on conscience or otherwise not strictly along party lines, American voters need not be concerned to the same degree as Canadians about political party considerations. A l l told, there are many substantive reasons for a candidate-centred approach to the electors' decision-making processes in the United States. Some scholars suggest a "presidentialization thesis"4 which contends that parliamentary elections are similar to presidential elections as increasing attention is paid to leaders, instead of parties. Supporting this theory is the idea that the Canadian prime minister is seen as accountable in similar ways as a president is to his/her national constituency. The primacy of the Canadian prime minister is exemplified by the media focus on the leader - for 4 Lawrence LeDuc, Richard G. Niemi, and Pippa Norris, eds., Comparing Democracies: Elections and Voting in Global Perspective (Thousand Oaks: Sage, 1996), 286. 7 example, the media encourages the personification of the Canadian government by labelling its actions as those of the current prime minister's government (i.e. the Chretien government). It can perhaps be said that the linkage is largely a media construct that often results in the attribution of blame or responsibility for the policy outcomes to the executive leadership. On the other hand, Harold Clarke et al. argue against the presidentialization thesis in the Canadian case.5 They contend that although party leaders are important to campaigns today, leader assessments do not transform voters' perceptions of political parties in the short term. Despite the differences in opinion about the presidentialization of Canadian elections, the abundance of scholarly research in the area of leader effects on voting decisions at the very least supports the basic notion that party leaders are meaningful actors in parliamentary elections. Partisanship and Strategic Voting Clarke and his colleagues report that partisanship in Canada is relatively weak, this finding being based on data from the 1984 Canadian National Election Study. They observe that about two-thirds of those surveyed were categorized as "flexible partisans" - defined as those who identify with different parties in federal and provincial politics, identify only weakly with a federal party, or have changed their partisan identification at some time -differing from "durable partisans" who have deep-seated attachments to a political party. Lacking strong attachments to political parties, other electoral factors must fill the void, 5 Harold D. Clarke, Jane Jenson, Lawrence LeDuc, and Jon H . Pammett, Absent Mandate: Canadian Electoral Politics In an Era of Restructuring (Vancouver: Gage, 1996), 75. 8 effectively increasing the contribution of political party leaders' personalities in the vote decision-making process. Andre Blais, Elisabeth Gidengil, Richard Nadeau, and Neil Nevitte report that in 2000, 56 percent of Canadian Election Survey respondents identified themselves very or fairly strongly with a political party.6 Of this group of party identifiers, 85 percent voted for the party for which they indicated an attachment, implicitly leaving 15 percent of this group (8.4 percent of all eligible voters), or a total of approximately 52 percent of all eligible voters (including the 44 percent of people reporting no party identification), who were not driven to vote based on partisan attachments. Without political allegiances dictating the majority of voters' decisions, Blais et al.'s findings suggest that a plurality of voters may have given greater consideration to factors such as leaders and issues in their voting equations in the 2000 federal election, thus adding more backing to the argument that leaders are meaningful considerations in Canadian elections. Additionally, Lawrence LeDuc notes in his book chapter stemming from the "Absent Mandate" series which he co-authored, that voters with relatively low levels of interest in politics place greater emphasis on their evaluations of leaders, in conjunction with issues, into their voting equations.7 He observes that almost half of the "flexible partisan" group reported having low levels of political interest, and comprised the largest single group of 6 Andre Blais, Elisabeth Gidengil, Richard Nadeau, and Neil Nevitte, Anatomy of a Liberal Victory: Making Sense of the Vote in the 2000 Canadian Election (Peterborough: Broadview Press, 2002), 117. 7 Lawrence LeDuc, "Leaders and Voters: The Public Images of Canadian Political Leaders," in Leaders and Leadership in Canada, eds. Maureen Mancuso, Richard G. Price, and Ronald Wagenberg (Don Mills: Oxford University Press, 1994), 71. 9 voters in the 1984 election study's sample. Declining voter turnout in recent years suggests political interest is waning in Canada, combined with the high number of low interest-flexible partisan voters, points to the strengthening of the role of party leaders in Canadian electoral politics. With generally weak partisanship coupled with a first-past-the-post system of voting in a parliamentary system of government, Canadian voters may find reasons in particular elections to use the tool of strategic voting. Strategic voting occurs when electors do not vote for their first choice, but rather they make their decisions with the intent to block a party from winning by voting for the next strongest party. If used by a significant number of electors, there may be a negative impact on this type of study. When strategic voting is evident, the connection between voters and their use of leader personality trait considerations in their voting equations will be veiled. Blais et al. report that a limited amount of strategic voting -only three percent of voters outside of Quebec voted strategically8 - occurred in the election of 2000 and the impact on the election outcome was insignificant. Given the lack of significance to the election outcome, I did not find it necessary to make any adjustments to the analyses reported in this study. Which Box Should I Tick? The Voters' Decision-making Process There are numerous factors worthy of consideration when thinking about how people go about deciding how they will vote. Intuitively, working within the framework of the given 8 Blais et al., Anatomy of a Liberal Victory, 187. The losers were the Progressive Conservatives and the New Democrat Party which each suffered a loss of about one percent of the popular vote due to strategic voting - and the winners were the Canadian Alliance and the Liberals. 10 institutional structure, it would seem fair to deduce that decisions are based on factors that are arguably more substantive, including the state of the economy, the performance of the government currently in office, ideology and political party loyalties, the parties' stands on issues, and leadership. Media and election campaign influences, gender, regional links, and social factors such as ethnicity, religion, class, an individual's own personality, and a candidate's likeness to the voter, amongst other influences, are assumed to play roles of varying degrees in the vote decision-making process. Voters' education levels, political awareness, interest, and knowledge, to name a few, are personal elements that also appear to contribute to differences in voting behaviour. This study isolates leader personality traits and examines their effects on voting decisions in an attempt to better understand what is popularly assumed to be a key element of Canadian electoral behaviour. Leader personalities have been found to be more important in overall candidate evaluations than most other factors mentioned previously, including issue positions and political party affiliations in American presidential elections.9 Similarly, the "leader factor" is predominant in Canadian elections. One study that supports this theory is that of Brown, Lambert, Kay, and Curtis. This group reports that over 75 percent of responses to open-ended questions posed to voters about the positive and negative attributes of Canadian political figures referenced personality characteristics (for further discussion about this study, please refer to page 14).10 Therefore, there is research supporting the idea 9 Arthur H . Miller, Martin P. Wattenberg, and Oksana Malanchuk, "Schematic Assessments o f Presidential Candidates," American Political Science Review 80 (June 1986): 525. 1 0 Steven D. Brown, Ronald D. Lambert, Barry J. Kay, and James E. Curtis, "In the Eye o f the Beholder: Leader Images in Canada," Canadian Journal of Political Science 21 (December 1988): 739-40. 1 1 that political party leaders' personalities stand out when voters decide how they will mark their ballots on election day in both Canada and the United States. The visual media age brings political leaders and their personalities to the forefront, in particular during the run-up to elections when voters are subjected to a plethora of carefully crafted images of political leaders. Graphic leader images tell a more attention-grabbing story through the means of television than political parties and other abstract electoral elements that do not project nearly as effectively through this medium. Voters are subjected to largely polished imagery that projects very effectively and efficiently into the limited attention space available in the citizenry's busy daily lives. Political leaders serve as focal points, particularly during election periods but also more broadly as government icons, which leads to a natural assumption that leader assessments are a central factor in the electorate's voting equations. Ellen Riggle's political cognition-based study underscores the importance of leaders in voting equations. This study focuses on the number of candidates voters have to assess and the associated impact on candidate evaluations." Riggle observes that i f an individual comparatively evaluates two or more candidates, s/he uses heuristics, or mental shortcuts, such as party identification and ideology to do so, thereby reducing the volume of mental processing required when issue stands are judged on a comparative basis. When evaluating single candidates, individuals compare the candidate's ideological and issue stands to their own. She finds that voting based on leader images occurs in campaigns where significant image-related information is available to the public. As mentioned earlier, the visual media " Ellen D. Riggle, "Cognitive Strategies and Candidate Evaluations," American Politics Quarterly 20 (April 1992): 240. 12 tends to focus on leaders and offers the public a substantial volume of leader information, therefore, based on Riggle's supposition, it may be inferred that leader image voting prevails in Canada. The Evolution of Research on Leader Personality Effects on Voting Behaviour With the continued strength of the role of television in people's lives and the personification of government and politics, popular wisdom suggests that leaders play a vital role in voting decisions in Canada. Research interest in the area of voting behaviour has increased with the growth of the role of the media, particularly television, leading to the production of a substantial volume of studies that are important to our understanding of how political party leaders fit into the voting decisions of Canadians. Although the institutional settings, players, and dynamics are different in the various studies cited below, as a whole, they provide the foundation for which this analysis of leader personality effects is ultimately based. Kinder, Peters, Abelson, and Fiske, 'Presidential Prototypes' (1980)12 Pioneering the wave of schema-based leader trait studies, Kinder et al. coined the term "presidential prototypes" to represent voters' conceptions of an ideal president. In this study, Kinder and his associates speculate how the public uses these prototypes to appraise American presidential candidates. They observe that Americans use prototypic information in defining an ideal president, but they found that prototypes did not provide standards that 1 2 Donald R. Kinder, Mark D. Peters, Robert P. Abelson, and Susan T. Fiske, "Presidential Prototypes," Political Behavior 2 (1980): 315-37. 13 were used to evaluate real presidential candidates. Conceptions of an "ideal president" only affected citizens' appraisals of the incumbent president. The group suggests that some standards of appraisal are widely shared, whereas others are idiosyncratic in that they are tied to individual candidates and that the most salient qualities, based on evaluations of candidates' personalities and performance, of the public's ideal president are competence and political trustworthiness. Interestingly, personal moral transgressions, reflecting personal trustworthiness, were of minor consequence to the public's perception of the ideal American president. They found that some voters single out distinctive qualities tied to particular candidates (using Kinder et al.'s example, Senator Edward Kennedy's widely known personal transgressions) which may accordingly have implications for their evaluations of the candidates. The study found that less substantive personal traits such as the projection of a "winning personality", warmth, humility, and selflessness, won few points with the public. Their findings were widely held across genders, regions, age, and party identifier and ideological groups; however, they found that the implications might, as noted earlier, be confined principally to the appraisal of the incumbent president. Lastly, Kinder et al. report that well-educated individuals weigh competence more heavily than did the less-educated group of voters, and comparably, the less-educated group placed more weight on candidate likeability and personal morality. 14 Graetz and McAllister, 'Popular Evaluations of Party Leaders in the Anglo-American Democracies'(1987)13 A study of leader evaluations in Anglo-American democracies, specifically Australia, Britain, Canada, and the United States in their respective election years in 1979 and 1980, further underscores the importance of the leader factor. Brian Graetz and Ian McAllister observe the significance of partisanship in leader evaluations, particularly in Australia and Britain, with Canadian voters found to be especially influenced by party association in their evaluations of incumbent leaders. They identify a mutual interaction between partisanship and leader evaluations with a cross-national commonality that leader judgements have a greater influence on party identification, rather than vice versa. Graetz and McAllister do, however, note an exception in the Canadian case in that party identification has notably less significant effects on leader judgements. An interesting observation by the researchers is that voters' evaluations of leaders weigh equally with their feelings of political efficacy (the belief that their views are considered by government) and are twice as important than their trust in government. Brown, Lambert, Kay, and Curtis, 'In the Eye of the Beholder: Leader Images in Canada' (1988)14 In their study of Canadian leader images using election studies data from 1974 through 1984, Brown, Lambert, Kay, and Curtis observe that Canadian voters do not evaluate leaders idiosyncratically but instead employ stable leader schemas when processing 1 3 Brian Graetz and Ian McAllister, "Popular Evaluations of Party Leaders in the Anglo-American Democracies," in Political Elites in Anglo-American Democracies: Changes in Stable Regimes, eds. Harold D. Clarke and Moshe M . Czudnowski, (Dekalb, IL: Northern Illinois University Press, 1987), 44-64. 1 4 Brown, Lambert, Kay, and Curtis, "In the Eye of the Beholder." 15 information about prime ministerial candidates. The authors state that more sophisticated15 voters tend to focus on task-relevant and dispositional personality traits and that level of political interest dictates the degree to which voters make inferences based on issue position.1 6 Brown et al. found education and political interest levels are positively correlated to the frequency that respondents made task-relevant inferences when describing candidates, as opposed to relatively superficial and/or readily observable, personality traits (e.g. personal style traits such as general likeability, warmth, arrogance). Although this study does not analyze the link between leader evaluations and voting behaviour, its relevance is that it supports the theory that voters in Canada largely focus on the substantive, task-relevant attributes of political leaders - integrity (the authors report that an average of 14 percent of survey respondents in each of the election-year surveys cited this attribute), political style (14 percent), personal style (13 percent), dynamism (13 percent), and competence (9 percent) -and place significantly less emphasis on leaders' socio-political attributes - social background (4 percent) and party affiliation (2 percent). 1 5 Ibid., 735. Based on Philip E. Converse's work in the area, differences in "sophistication" are due to differences in voters' capacities to organize political evaluations around a small number of abstract programmatic or ideological dimensions. Highly sophisticated voters had a broad contextual grasp of political events understood in abstract ideological terms whereas less sophisticated voters had a narrower grasp and used more concrete policy tools. 1 6 Ibid., 747. 16 Bean and Mughan, 'Leadership Effects in Parliamentary Elections in Australia and Britain' (1989)17 Clive Bean and Anthony Mughan posit that, based on election survey data from the 1987 Australian and 1983 British parliamentary elections, leaders' perceived effectiveness strongly influenced voters. The electorate uses schematic evaluations as its guide when deciding for which party they will vote. The research findings reinforce the suggestion that electors use schemas as tools to aid in simplifying the task of voting, helping them make their way through the myriad of information about leaders, issues, parties, and the like, linked to the process of voting. From their research, Bean and Mughan infer that leadership appeal in parliamentary elections is not idiosyncratic as they observe that voters uniformly focus on particular leader characteristics across parties and leaders. In addition to leader effectiveness, the researchers state that listening to reason, caring, and sticking to principles are leader traits ' that are statistically significant across the countries and leaders studied and can influence the distribution of votes accordingly. Comparably, trivial leader characteristics such as likeability take a backseat to the more substantive traits to which voters are inclined to consider in their prime ministerial candidate evaluations. In the context of leader evaluations and electoral impact, Bean and Mughan introduce a unique concept in which they switch leaders between political parties and analyze the comparative electoral effects. This counterfactual exercise created a new way to look at leader personality effects, a derivative of which is employed in this study of the 2000 Canadian election. 1 7 Clive Bean and Anthony Mughan, "Leadership Effects in Parliamentary Elections in Australia and Britain," American Political Science Review 83, no. 4 (December 1989): 1165-1179. 17 Johnston, Blais, Brady and Crete, 'Letting the People Decide: Dynamics of a Canadian Election' (1992) 1 8 In this text, Johnston and his colleagues carry out a comprehensive analysis of the 1988 Canadian federal election. The group observes that leader evaluations can change over time, even over the short period of an election campaign, therefore, campaign strategists may be rewarded for their efforts on election day by skillfully manipulating leader images. In addition to the credibility issue of Ed Broadbent's New Democratic Party, many seriously questioned Prime Minister Brian Mulroney's character while Liberal leader John Turner's competence lay very much in doubt - personality trait-driven situations that underscore the salience of party leaders. Johnston et al. identify a significant relationship between the media and Canadian leader evaluations; they find that advertising is persuasive to voters, and that the electorate is also sensitive to the news that is in part affected by political agenda setting but is largely considered episodic in nature. News and advertising, in the short run, strongly affected debate-watching voters' opinions on the controversial issue of the Free Trade Agreement and their ratings of the Prime Minister, to a lesser degree, advertising alone affected "inattentive" voters. In addition to a voter's own social and political considerations, Johnston et al. observe that overall leader evaluations are driven by somewhat independent ratings of competence. Again referencing Ed Broadbent's leadership of the NDP, the group speculates that apostate Liberals needed to find a way to relieve their cognitive dissonance about the leader's competence and fortunately, John Turner's strong showing at the leader debates and in debate 1 8 Richard Johnston, Andre Blais, Henry E. Brady, and Jean Crete, Letting the People Decide (Montreal: McGill-Queen's University Press, 1992). 18 commentaries and analyses thereafter, provided them with sufficient reasons to return to the Liberal camp. Additionally, it is noted that voters' levels of doubt about a leader may make perceptions of the leader in question more malleable. The implication is that challengers, as opposed to incumbents for which images and opinions are for the most part already established, may be more in doubt and hence perceptions of challengers are more susceptible to change. Based on this study of the 1988 Canadian federal election, voters' cognitive dissonance resolution and the public's doubt about challengers - Liberal Party leader John Turner in this particular case - seem to be two plausible cognitive responses by voters that can lead to changes in leader evaluations. A key feature of this research is the introduction of leader trait pooling. In studies to date, researchers analyzed leader traits and their effects individually with the understanding that the measures fell into broader trait categories. Johnston and his colleagues bundle the personality measures into the character and competence dimensions and study the aggregated effects accordingly. LeDuc, 'Leaders and Voters: The Public Image of Canadian Political Leaders' (1994)19 Along the same vein as the study by Clarke et al. (see next subsection), Lawrence LeDuc's research reveals that political party leaders in Canada were of secondary importance to voters when asked to rank order the contributions of political party, leader, and local candidate factors in the electorate's voting decisions from 1974 to 1984. Respondents who ranked leaders as the most important voting consideration were asked to indicate the relative importance of leaders' personal qualities and their stands on issues. The study states that 1 9 LeDuc, "Leaders and Voters." 19 personal qualities figured prominently as a higher-ranking factor contributing to vote decisions in three of the four elections in the period analyzed, with leaders' issue stands barely edging out personal qualities in the 1974 election period. LeDuc and other observers discussed in this chapter suggest that because personal traits do not tell the whole story about voting decisions, leaders who are held in high public esteem do not always see favourable voting results. Since party leaders are secondary to political parties in terms of their contributions to voting decisions, voters may support a party on issue or policy grounds while rating the party leader behind all others. LeDuc observes the steady waning in support of Canadian political party leaders evidenced by the overall decline in CES "feeling thermometer" ratings (see the next chapter for more information). He reports on the declining trend in the public esteem of party leaders and states that the voting public generally holds leaders in lower regard than it did in past years. Clarke, Jenson, LeDuc and Pammett, 'Absent Mandate' (1996)20 Harold D. Clarke, Jane Jenson, Lawrence LeDuc, and John H. Pammett, in their third analysis of Canadian federal elections spanning three decades up to and including the 1993 election, argue that leader traits stand out over issue characteristics in voting decisions and evaluations of incumbent leaders are of a relatively static nature. They find that, in line with the findings of Johnston and his colleagues, short-term forces including debates and the media have a noteworthy impact on leader evaluations, with the exception of the 1993 debates that were relatively inconsequential. Their analyses of short-term forces on voting in 2 0 Clarke et al., Absent Mandate. 20 1988 and 1993 reveals the relatively strong influence of unique leader effects among "flexible partisans" (definition on page 7) with low levels of political interest, whereas issue-driven evaluations appear to more strongly influence flexible partisans with high levels of interest. However, they note that the relative strength of these short-term forces differ from election to election and state that these forces more strongly affect the voting behaviour of flexible partisans. Clarke et al. contend that the effect of party leaders on voting decisions is on the decline; the rank order importance of party leaders has declined from a high of 37 percent (voters who ranked leaders as the most important decision factor) in 1979 and steadily decreased to 20 percent in 1988. By 1993, party leaders were the top ranking factor in vote decisions for 21 percent of electors; local candidates and parties were the most important factors for 22 and 57 percent of voters, respectively. In this cross-election study, the group finds that political parties consistently top the voters' rankings in Canada. Miller, Wattenberg, and Malanchuk, 'Schematic Assessments of Presidential Candidates' (1996)21 Cognition-based research conducted in the United States suggests that individuals use political information selectively based on schemas. Schematic assessments of political candidates based on a limited set of general and enduring criteria grounded in voters' pre-existing beliefs and knowledge structures - e.g. how information fits into voters' mental views of the world, and/or knowledge of which personal characteristics a leader should possess - are a cognitive tool employed by some voters when evaluating political candidates. Miller, Wattenberg, and Malanchuk observe based on American national election studies data 2 1 Miller, Wattenberg, and Malanchuk, "Schematic Assessments of Presidential Candidates." 21 from 1952 to 1984, that voters use schemas focused on leader personality characteristics -most notably the competence dimension, rather than issue and partisan group cues - as a key evaluative tool in American presidential elections. The group also notes that college-educated voters are more likely to make inferences about candidates' prospective performance in office and their internal dispositions; less educated voters tend to make their judgements based on candidates' more readily observable attributes. Miller et al. report that voters with higher levels of political interest favour making inferences based on performance. More specifically, the authors suggest that this subgroup has a greater tendency to make competence judgements when assessing political candidates. Overall, the group found linkages between education and political interest level and voters' evaluations of presidential candidates. King, 'Do Leaders' Personalities Really Matter?'22 and 'Conclusions and Implications' (2002) 2 3 Anthony King's chapter contained within the text that he edited, Leaders' Personalities and the Outcomes of Democratic Elections, addresses the central question to this thesis - "do leaders' personalities really matter?" He identifies two types of leader effects, "indirect" and "direct", which refer to the type of influence that leaders exert on voters. "Indirect" effects are those that result from a leader's actions that influence the political party and/or government that s/he leads, whereas "direct" effects are those exerted 2 2 Anthony King, "Do Leaders' Personalities Really Matter," in Leaders' Personalities and the Outcomes of Democratic Elections, ed. Anthony King, (New York: Oxford University Press, 2002), 1-43. 2 3 Anthony King, "Conclusions and Implications," in Leaders' Personalities and the Outcomes of Democratic Elections, ed. Anthony King, (New York: Oxford University Press, 2002), 210-221. 22 on voters by a leader's being, i.e. the leader's appearance and public persona. King offers three reasons supporting his suggestion that leader preferences largely offset one another in the vote decision making process: (1) leader effects may be neutralized by the fact that voters are not offered a wide-ranging variety of candidates - candidates chosen to lead political parties generally fall within a relatively narrow range of personalities and abilities that are considered acceptable by the public for the powerful position; (2) voters' internal mechanisms do not seem to permit them to heavily skew their views of the leaders which tempers the probability that voters will base their decisions on leader personalities; and (3) the electorate, in general, is not extreme, people's personalities are roughly in balance and do not elicit strong opinions in one direction or another. He notes that voters deal with the complexities of elections by using shorthand, specifically leader personalities, and rationalization in their discussions and decision-making processes. The electoral significance of leader characteristics is said to not be an all-or-nothing situation, rather, leaders are important in particular circumstances. When party attachments are weak, King states that leader personalities have a greater impact on voters. Also, he suggests that when differences in policy stands and performance are difficult to discern, leaders play a larger role. King presents the distinction between "gross" and "net" effects of leaders: gross effects appear at the level of the electorate's vote decisions; the net level is illustrated by the overall election outcomes. When voters are presented with leaders with great differences in their perceived levels of capabilities, King found gross effects are significant. When the differences between leaders are obvious and when a large proportion of the electorate is influenced by the leaders' characteristics, election outcomes, or net effects, are significant. The degree of gross and net effects of leader characteristics varies 23 from election to election, but in the more general sense, they play a part in electoral contests. Bartels, 'The Impact of Candidate Traits in American Presidential Elections' (2002) 2 4 Analyzing the impact of candidates' traits in the United States over six presidential elections from 1980 through 2000, Bartels found that the public's evaluations of leaders generally fall within a narrow range; from this, he states that the lack of significant differences in perceptions of leaders limits the net impact of personalities on election outcomes. Statistical evidence shows that political biases, including partisan and ideological, colour voters' perceptions of leaders' traits, as marked by their slanted ratings based on a 200-point scale - Bartels observes a largely unidirectional, causal relationship. He notes that voters' assessments of presidential candidates vary more within the electorate in each election than they do across elections. The impact of leader assessments fell within the range of 1.6 percent to 3.5 percent of the popular vote in the election years analyzed. In these contests, leader effects were not of a sufficient magnitude to make a difference to the electoral outcome; in fact, in half of the elections, the winners would have won a greater share of the vote with the removal of leader trait effects from the voting mix. Interestingly, the presidential candidates popularly known for their strong personal images, specifically Ronald Reagan and Bi l l Clinton, did not see electoral benefits from being outstanding election campaign performers. Reagan's net advantage over his Democratic rivals in two elections was in the range of one percent whereas Clinton's net disadvantage in 1992 was 3.5 percent. In 1996, Clinton's net 2 4 Larry M . Bartels, "The Impact of Candidate Traits in American Presidential Elections," in Leaders' Personalities and the Outcomes of Democratic Elections, ed. Anthony King, (New York: Oxford University Press, 2002), 44-69. 24 disadvantage was 2.5 percent to his relatively sedate rivals, George H. Bush and Bob Dole respectively. A l l told, Battels states that candidate evaluations overall have a weak statistical relationship with vote choice, particularly that trait evaluations largely offset one another resulting in a weak effect on election outcomes. Combined with voters' political predispositions, he concludes that the effect of leader traits in the United States is less than is popularly assumed. Johnston, 'Prime Ministerial Contenders in Canada' (2002) 2 5 As stated previously, this thesis is an extension of the research conducted by Richard Johnston that analyzes the impact of prime ministerial contenders in the 1988, 1993, and 1997 Canadian federal elections. At the first level of analysis, in viewing assessments of the 13 leaders' perceived "character" and "competence"26 in this period, few were distinct - less than half of the leaders deviated significantly from the mean values for these traits measures which is indicative of the weakness of leader personality effects. Looking at the statistical impact of the. individual leaders' perceived competence and character over the three elections, Johnston found character matters more than competence to voters. The "net effects" of the leaders' perceived traits - a comparative look at voters' leader assessments and their impact on voting decisions - reveals strong leader effects on vote choice in 1988. In this election, the three party leaders had positive or negative effects on their respective parties' vote shares of about two percent. Translating Bean and Mughan's leader substitution exercise to the 1988 context, Johnston notes the potentially substantive 2 5 Johnston, "Prime Ministerial Contenders in Canada," 158-183. 25 implications on the election's outcome and the subsequent impact on the implementation of the Canada/U.S. Free Trade Agreement had the popular leader of the New Democratic Party, Ed Broadbent, instead led the second place Liberal Party. The 1993 and 1997 elections offered less intriguing counterfactual opportunities to explore as perceptions of the leaders indicate that particularly in 1993, voters assessed the leaders very similarly. In 1997, the leader of the Progressive Conservative Party, Jean Charest, stood out from the group but similar to the situation in 1988, his impact on his party's vote share was about two percent. The more successful political parties, those that are closer to winning elections, as well as newer parties, appear to experience greater leader effects. Johnston indicates that personality effects on vote choice are marginal, but he states that they potentially could matter when vote margins are close which has generally not been the case in the Canadian electoral system of late. Using the "feeling thermometer" 100-point scale to measure voters' assessments of the leaders, Johnston graphically presents campaign effects that indicate significant changes in assessments occurred over the course of the campaigns for one leader in each election. Considering the changes of evaluations from the beginning to the end of each campaign period, the outcomes of the elections could have been different, albeit not drastically given the magnitude of effect of leader evaluations on voting decisions. Changes in the impact of leader evaluations on vote decisions in the 1993 election show, in addition to the feeling thermometer changes just mentioned, that campaigns do matter to voters in their assessments Johnston, "Prime Ministerial Contenders in Canada," 167. These broad, bundled measures are widely employed in the study of leader trait effects on voting decisions. 26 and with regard to the impact of party leaders on vote choice in Canadian elections. Further details of this Johnston's research findings are included in later sections as it pertains to the discussion at hand. Blais, Gidengil, Nadeau, and Nevitte, 'Anatomy of a Liberal Victory: Making Sense of the Vote in the 2000 Canadian Election' (2002)27 and 'Campaign Dynamics in the 2000 Canadian Election: How the Leader Debates Salvaged the Conservative Party' (2003)28 Andre Blais, Elisabeth Gidengil, Richard Nadeau, and Neil Nevitte, co-investigators of the 2000 Canadian Election Study, report on the effects of campaigns and campaign-period debates in their comprehensive analysis of the 2000 Canadian general election. The authors contend that Joe Clark's superior performance in the English debate was the critical element that created a surge in the leader's profile and kept the Progressive Conservative Party afloat. Evaluations of Clark, however, slipped mildly as the campaign progressed but the researchers note that Clark's debate performance earned him a permanent four point rating increase and simultaneously accrued his party a boost in vote intentions of the same value. The investigators found that socio-economic variables, such as perceptions of the state of the economy and government performance, weakly affected voting behaviour while they indicate that voters' gender and socio-demographic factors, including region of residence, religious affiliations, and voters' ethnic origins, have at least as strong effects on vote choice in Canada now as they did 30 years ago. From their findings, Blais' group contends that social cleavages are not eroding in Canada, unlike the trend reported in other 2 7 Blais, Gidengil, Nadeau, and Nevitte, Anatomy of a Liberal Victory. { 2 8 Andre Blais, Elisabeth Gidengil, Richard Nadeau, and Neil Nevitte, "Campaign Dynamics in the 2000 Canadian Election: How the Leader Debates Salvaged the Conservative Party," PS: Political Science and Politics 36 (2003): 45-50. 27 western democratic societies. They found that the enduring Canadian cleavage, religion, still made a mark - the survey data shows that Catholics are inclined to vote Liberal, and voters with no religious ties tend to support the New Democratic Party more than any other federal political party. The strength of the electorate's beliefs of economic market forces correlated significantly with voting - those with strong beliefs that markets have a proper role in Canadian society tended to vote for parties of the ideological-right, and voters who were more sceptical about the value of market forces tended to support left wing and centrist parties. Leader evaluations reflect partisan predispositions and are especially noticeable when compared to leader assessments by voters with no party identification. Although overall leader evaluations fall within a very narrow range of values based on a scale from zero to 100 - Canada-wide, the electorate rated the five leaders were rated in the 45 to 48 point range -the variability increases to a five point range from 45 to 50 for voters who indicated no party identification. Most telling is the Liberal Party leader's move from being rated the most popular leader for all Canadian voters to being the least popular leader amongst voters with no party identification. Regionally, the authors found similar differences in leader ratings between the partisan and non-partisan groups. Blais and his colleagues remind readers that a great deal of political power is vested in the Canadian Prime Minister, by virtue of the fact that majority governments largely prevail and that voting patterns appear to reflect this understanding as evidenced by the significant degree of consideration given to party leader evaluations in voting decisions. The group states that leader assessments had a substantial effect on vote choice - when voters rated a leader at 100 on the 100-point scale, this increased the probability of voting for the leader's 28 party by 25 to 30 percent. They also observe that for a substantial proportion of voters - 21 percent outside the province of Quebec - leaders were a decisive factor to their vote choices, however, this had only a small impact on the election's outcome because voters found very little difference between the party leaders in 2000. Lastly, the group reports that economic considerations had a minimal impact and strategic voting played only a small part in the 2000 election. The examination of the relevant theory and literature sets the stage for the next chapter in which I outline the methodology followed to analyze the impact and relevance of the party leaders in the 2000 Canadian federal election. 29 CHAPTER 2 METHODOLOGY There is no democracy without free and fair elections. Elections provide a crucial opportunity to assess the actual workings of a democracy. Canadian elections are the primary focus of the Canadian Election Study (CES). The main objective is to explain what makes people decide to vote (or not to vote), and, if they do, what makes them decide to support a given party or candidate, and why parties gain or lose ground from one election to another. Excerpt from the introduction to the 2000 Canadian Election Study29 To achieve the goal of this study and to maintain consistency with the research being replicated, I employ quantitative methods of analyses. In using a quantitative approach in the area of the social sciences, according to Gary King, Robert E. Keohane, and Sidney Verba, the research should have the following characteristics:30 1 . Descriptive or explanatory/causal inference is the goal of scientific research. 2. The procedures used in carrying out scientific research are public, explicit, and codified. 3. Uncertainty of conclusions is central to all research; certainty is impossible from uncertain data. 4. Content, not subject matter, is what drives the scientific methods and rules of research. This research works to reach the desired end - inference - about the impact of political leaders' traits on voting behaviour in the 2000 Canadian federal election. Next, I meet the second requirement of scientific research in the social sciences - the explicit description of the methods and procedures used to reach the desired research end. 2 9 Andre Blais, Elisabeth Gidengil, Richard Nadeau, and Neil Nevitte, 2000 Canadian Election Study [study on-line]; accessed March 2003; available from; Internet. 3 0 Gary King, Robert O. Keohane, and Sidney Verba. Designing Social Inquiry: Scientific Inference in Qualitative Research. (Princeton: Princeton University Press, 1994), 7-9. 30 Selection of Variables and Statistical Methods Employed Dating as far back as 1965, empirical election studies in Canada have included leadership components in their batteries of survey questions.31 Only over the last two decades, however, have the questions been of a nature employable in the type of research reported here. This study uses the data from the Canadian Election Study as its primary tool to analyze the voting behaviour of Canadians relating to the 2000 federal election. The CES is conducted in three waves consisting of a pre-election, post-election, and a mail-back questionnaire wave. The 2000 survey results are the outcome of comprehensive interviews with 3,651 respondents in the first wave over the 36 day election campaign period leading up to the November 27 election, 2,862 respondents in the post-election wave, and the responses from 1,535 post-election, self-administered, mail-back questionnaires. The survey purports to represent the opinions of the 12,997,185 Canadian voters (of the 23,786,167 eligible voters / 21,243,473 registered voters) who turned out at the polling stations to vote for candidates Philip Converse, John Meisel, Maurice Pinard, Peter Regenstreif, and Mildred Schwartz, 1965 Canadian National Election Study [study online] (Ann Arbor: Inter-University Consortium for Political Research, 1972, accessed March 2003); available from; Internet. Following is an example of the type of leadership-related questions included in the 1965 election study: Question 7H: Did the issue of strong leadership in Ottawa strike you as being very important, fairly important or not too important in deciding how to vote? 3 2 Both the post-election and mail-back survey respondents participated in the pre-election/first wave of the survey. 31 representing the 11 registered political parties running in the Thirty-seventh General Election of Canada.33 Survey Indicators The 1997 Canadian Election Survey, on which the research by Richard Johnston is based, included several personality trait measures for the individual prime ministerial contenders in its pre-election survey component. The survey asked respondents to indicate their opinion, from a list of possible responses offered, of each leader's trustworthiness, arrogance, compassion, and the strength of his/her leadership, and whether he/she was in touch with the times.34 In 2000, in a marked departure from the 1997 and other recent Canadian Election Surveys, respondents did not evaluate the individual leaders' personality traits, instead, respondents could only select one, all, or none of the leaders that they felt exhibited the said traits. The result, in my opinion, is a substantial reduction in the value and utility of the leader trait measures, and this change has significant implications for the continuity of the study of the effects of leader personality traits on voting decisions in Canada. Specifically, the 2000 pre-election survey wave measured three personality traits, compared to the five traits included in the 1997 pre-election survey, in this manner: arrogance, trustworthiness, 33 Past Elections: 37th General Election, November 27, 2000 [report online] (Ottawa: Elections Canada, 2001, accessed March 2003); available from^as&document=index&dir=37g&lang=e&textonly=false. 3 4 Questions in the 1997 survey were: <cpsh#a> Does "strong leader" describe {leader name}...; <cpsh#b> Does "trustworthy" describe {leader name}...; <cpsh#c> Does "arrogant" describe {leader name}?; <cpsh#d> Does "compassionate" describe {leader name}...; <cpsh#e> Does "in touch with the times" describe {leader name} Respondents were asked to indicate whether the trait described the leader: very well, fairly well, not very well, not at all, and were able to respond with: don't know or don't know leader very well, or refuse to answer the question. 32 and having new ideas.35 The 2000 post-election survey leader trait questions fell more closely in line with the 1997 survey variables employed by Johnston in his research and thus I selected them for the purposes of this analysis. Although the questions associated with the variables offered in the post-election survey are more consistent with the 1997 trait set than are the pre-election survey's offerings, there is an evident discrepancy in that none of the survey items speak clearly to the question of the leaders' measures of competence. Similar to the situation noted above, the change in survey items is clearly problematic from both a comparative and a continuity perspective for the study of leader personality effects on voting decisions in Canadian elections. Figure 1 lists the survey's leader trait questions and other variables employed as statistical controls for the purposes of the regression analyses that serve as the starting points for the empirical portion of this study. Of special note is the exclusion of the trait variable associated with the survey question that asked respondents to indicate which leader and/or leaders were "extreme". As this measure reconciles with neither the character nor the competence domains, I omitted this trait variable from the analyses. Questions in the 2000 pre-election survey were: <cpsel> Which leader would you describe as arrogant?... <cpse2> Which leader would you describe as trustworthy?... <cpse3> Which leader would you describe as having new ideas?.... Respondents were offered the following list of options to choose from: Chretien, Clark, Day, Duceppe, McDonough, none of the leaders, all of the leaders, don't know, refused to answer the question. 33 Figure 1. Survey Questions ( " Q " ) Employed ! Competence Var iables ( harucliT \ . ir i. iblcs ( o u l i o l \ a m b l e s Q pesi3: "new ideas" Which leader would you describe as having new ideas? Q pesi l : " a r r o g a n t " Which party leader would you describe as arrogant? Q cpa2g: "cu t taxes" Cutting taxes. To you personally, in this Federal election, is it very important, somewhat important, or not very important? Q pesi6 : " in te l l igen t " Which leader would you describe as intelligent? Q pesi2: " t r u s t w o r t h y " Which party leader would you describe as trustworthy? Q cpscl2 : "Quebec" How much do you think should be done for Quebec: much more, somewhat more, about the same as now, somewhat less, or much less? Q pesi8: " w e a k leader" Which leader would you describe as a weak leader? Q pesi4: "compassionate" Which leader would you describe as compassionate? Q cpsa2i: "defend interests of Quebec" (Quebec respondents only) To you personally, in this federal election, is it very important, somewhat important, or not important at all? Q pesi5 : "d ishonest" Which leader would you describe as dishonest? Q cpsc l : "personal f inance" Financially, are you better off, worse off, or about the same as a year ago? Q cpsc5: "unemp loymen t " Do you think unemployment in Canada has gone up, gone down, or stayed about the same since the last Federal election? Q cpsk3a: " p a r t y I D " Do you generally think of yourself as being a little closer to one of the federal parties than to the others? Q cpsk4a: Which party is that? Q province: " p r o v i n c e " Province of interview. Q cpsage: "age" In what year were you born? Q cpsm9: " u n i o n " Do you or anyone in your household belong to a union? Q cpsmlO: " r e l i g i o n " What is your religious affiliation? Is it Protestant, Catholic, Jewish, Islam, another religion, or none? Q cpsn6a: "e thn i c i t y " To what ethnic or cultural group do you belong? Vote Choice Var iables Note In tent ion Yar iabk-Q pesa3a (Quebec respondents on ly) : Which party did you vote for? Q pesa3b: Which party did you vote for? Q cpsa4: Which party do you think you will vote for? 34 The variables used for statistical control purposes - demographic, socio-economic, party identification, and issue variables - from the 2000 CES were carefully selected to match as closely as possible the control variables included in Johnston's analyses of the 1997 election.36 Fortunately, unlike the case with the personality trait measures, the survey questions and response options relating to the control variables (see Figure 1) did not change materially from 1997 to 2000. Figure 2 shows the scheme used to replicate the recoding of the leader trait variables employed, and following in Figure 3 is the recoding scheme used for the control variables. Figure 2. Personality Data Recoding Scheme ( oinpi-U-iui- \ ai iahli-K Specification Associated Duniniv Variable Recoding Positive competence trait Chretien Clark Day Duceppe McDonough A l l of the leaders Refused None of the leaders Don't know A new variable were created for each leader. Selection of an individual leader = +1.00 A l l of the leaders = +1.00 for each leader variable Response eliminated from dataset A l l other responses = 0.00 measures Question pesi3: "new ideas" Question pesi6 : "intelligent" Negative competence trait measures Question pesi8: "weak leader" Chretien Clark Day Duceppe McDonough A l l of the leaders Refused None of the leaders Don't know A new variable was created for each leader. Selection of an individual leader = -1.00 A l l of the leaders = -1.00 for each leader variable Response eliminated from dataset A l l other responses = 0.00 i b 1997 survey questions were: <cpselb/c> Should taxes be cut a lot, somewhat, or a little? Should taxes be increased a lot, somewhat, or a little?; <cpse3a/b/c> How much do you think should be done for Quebec: more, less, or about the same as now? Should a lot more be done for Quebec, somewhat more, or a little more? Should a lot less be done for Quebec, somewhat less, or a little less?; <cpscla/b> Financially, are you better off, worse off, or about the same? Is that much better or somewhat better? Is that much worse or somewhat worse?; <cpsc5> Do you think unemployment in Canada has gone up, gone down, or stayed about the same since the Liberals came to power?; <cpskl> In federal politics, do you usually think of yourself as a Liberal, Conservative, NDP, or none of these? 35 Figure 2. - Continued Character Variables S])iiitu:ilii)ii Associated Dummy Variable Recoding Positive character trait measures Question pesi2: "trustworthy" Question pesi4: "compassionate" Chretien Clark Day Duceppe McDonough A l l of the leaders Refused None of the leaders Don't know A new variable was created for each leader. Selection of an individual leader = +1.00 A l l of the leaders = +1.00 for each leader variable Response eliminated from dataset A l l other responses = 0.00 Negative character trait measures Question pesil: "arrogant" Question pesi5: "dishonest" Chretien Clark Day Duceppe McDonough A l l of the leaders Refused None of the leaders Don't know A new variable was created for each leader. Selection of an individual leader = -1.00 A l l of the leaders = -1.00 for each leader variable Response eliminated from dataset A l l other responses = 0.00 36 Figure 3. Statistical Control Variables Recoding Scheme Variables Specification Associated Dunum Variable Recoiling Question cpa2g: Very important + 1.00 "cut taxes" Somewhat important + 0.66 Not very important 0.00 Do not plan to vote Response eliminated from dataset Don't know Response eliminated from dataset Refused Response eliminated from dataset Question cpscl2: Much more + 1.00 "Quebec" Somewhat more + 0.50 About the same as now 0.00 Somewhat less - 0.50 Much less - 1.00 Don't know Response eliminated from dataset Refused Response eliminated from dataset Question cpsa2i (Quebec Very important + 1.00 respondents only): Somewhat important + 0.50 "defend interests of Not very important 0.00 Quebec" Do not plan to vote Response eliminated from dataset Don't know Response eliminated from dataset Refused Response eliminated from dataset Question cpscl: Better off + 1.00 "personal finance" Worse off - 1.00 About the same 0.00 Don't know Response eliminated from dataset Refused Response eliminated from dataset Question cpsc5: Gone up - 1.00 "unemployment" Gone down + 1.00 Stayed about the same 0.00 Don't know Response eliminated from dataset Refused Response eliminated from dataset Question cpsk3a: No action: i f "yes", a response will be found to Question "party ID" - closer to a cpsk4a below particular party? Question cpsk4a: Liberal A new variable was created for each of the five political if "Yes" to question cpsk3a Bloc Quebecois parties. Alliance Conservative For each party variable, if selected = +1.00 NDP A l l other party responses = 0.00 None Other Don't know Response eliminated from dataset Refused Response eliminated from dataset 37 Figure 3. - Continued Yariablcs Specification Associated Dummv Variable Recodinu Question province: "province" (for Bloc Quebecois leader analyses only) Quebec A l l others Three new variables were created: "Quebec", "Atlantic provinces", and "Western provinces". + 1.00 ("Quebec") 0.00 Newfoundland Prince Edward Island Nova Scotia New Brunswick A l l others + 1.00 ("Atlantic provinces") + 1.00 + 1.00 + 1.00 0.00 Manitoba Saskatchewan Alberta British Columbia A l l others + 1.00 ("Western provinces") + 1.00 + 1.00 + 1.00 0.00 Question cpsage: "age-Year of birth Refused (2000-Year of birth) Response eliminated from dataset Question cpsm9: "union" Yes No Don't know Refused + 1.00 0.00 Response eliminated from dataset Response eliminated from dataset Question cpsmlO: "religion" Protestant Catholic Jewish Islam Other religion None Don't know Refused Two new variables were created: "Catholics" and "none". For each of the two religions, i f selected = +1.00 A l l other religion responses = 0.00 Response eliminated from dataset Response eliminated from dataset Question cpsn6a: "ethnicity" 60 ethnic/cultural group options (see Appendix) Five new variables were created: "Visible minority", "northern European", "eastern European", "southern European", and "French". For each of the five 1.00 i f relevant ethnic/cultural groups, i f selection = +1.00 A l l other responses = 0.00 Don't know Refused Response eliminated from dataset Response eliminated from dataset 38 Survey Fieldwork and Design The pre-election survey's fieldwork was conducted over the entire span of the campaign period. The daily samplings enabled the breakdown of the dataset into phases with a sufficient number of respondents falling into each time period for comparison purposes. The daily samplings ranged from a low of 59 respondents to a high of 220 respondents. An explanation of the division of the campaign period for the purposes of analyzing the pre-election changes in leader effects follows in the next chapter. As stated earlier, Johnston's study used the 1997 pre-election survey variables and this analysis uses the post-election variables. There is a distinct possibility that respondents will answer survey questions about party leaders differently before, as opposed to after, an election. For example, knowledge of the outcome of the election and exposure to a full election campaign may affect a respondent's judgements of the party leaders, and these assessments could differ greatly from those s/he would give i f asked the same questions early in the campaign period. I found that despite the timing difference issue, after considering both the 2000 survey's pre- and post-election questions, the spirit of this analysis is as near as possible to that of Johnston, and overall there was more to be gained by using the broader and more relevant range of personality trait measures from the post-election survey. A discussion about the analytical problems arising from the timing difference follows in the final chapter of this paper. Analytical Methodology This study is centred on six analytical components that I present later in the following order: 39 Static analyses considering the dataset as a whole 1. An analysis comparing the average personality ratings of the individual leaders against the mean rating for the group. 2. Regression analyses producing basic personality impact coefficients for each leader. 3. Analyses of the net impact of the leaders' personality measures on their respective parties' vote share. 4. Counterfactual leader-swapping analyses that present the impact on the parties' vote shares i f one hypothetically moves the prime ministerial candidates to lead parties other than their own. Dynamic analyses breaking down the dataset into phases or subsets within the campaign period 5. An analysis of the movement in feeling thermometer ratings of the individual leaders during the campaign period. 6. Two-leader, three-phase regression analyses looking at the changes in ' the personality impact coefficient values over the campaign period. The competence and character measures for each leader serve as the independent variables, and the post-election survey data representing the survey respondents who indicated that they voted for the party that the individual leader represents serves as the dependent variable for the Ordinary Least Squares ("OLS") regression analyses. The two-leader analysis focuses on Stockwell Day and Joe Clark, and the dependent variable is derived from the pre-election survey data representing the vote intentions rather than the party for which the respondents actually voted. Additional pertinent information accompanies the analytical output in the next chapter. The Bloc Quebecois only ran candidates in Quebec and personality questions were only presented to Quebec survey respondents for that party's leader, so taking this consideration, the dataset was restricted to only include this subsection of survey respondents in the analyses of Bloc leader, Gilles Duceppe. Unlike in 1997, Quebec respondents were asked their opinion of the Canadian Alliance Party leader as the Alliance moved toward brokerage party status with its near full-slate of candidates across Canada's provinces and 40 territories, therefore enabling the use of the full dataset in the 2000 election analyses of Stockwell Day. While both the Bloc and the Alliance came into being in Canadian electoral politics in 1993, the Bloc remains a ethno-regional party whereas the Alliance continued its growth in 2000 and ran 298 candidates in the 301 Canadian electoral districts,37 a marked increase from the 227 hopefuls who ran in 1997 under the Reform Party banner. Lastly, the analysis of the campaign dynamics of leader perceptions uses the 2000 CES pre-election feeling thermometer measures, individual evaluations based on the respondents' ratings of the leaders on a zero-to-100 scale. To smooth the daily mean leader ratings, I calculated the five-day moving average in a manner yielding acceptably similar •JO results as that presented by Johnston. The Canadian Alliance Party did not run candidates in one riding in each of Quebec and Manitoba, and in the territory of Nunavut comprised of one electoral district. 3 8 Johnston smoothed (to control for respondents' accessibility and sampling variance) the 1997 feeling thermometer data using the "loess" (locally-weighted least squares) method. For a discussion about the ioess" method, please refer to Johnston, "Prime Ministerial Contenders in Canada," 174. CHAPTER 3 DA TA AND FINDINGS: ELECTION DA Y 41 This chapter presents analyses based on reported post-election vote data. It shows that significant variations in evaluations and estimated impact exist between the Canadian federal political party leaders in the 2000 election. In counterfactual scenarios where leaders are swapped between parties, I analyze and discuss the implications and effects of the hypothetical situations. How the Leaders' Personalities Measured Up in 2000 Figure 4 shows the variances in voters' perceptions of the individual federal party leaders from the group mean on each of the personality trait clusters. To calculate the individuals leaders' competence and character ratings that are reported graphically in Figure 4, I recoded the data for each trait variable as indicated in the previous chapter, pooled the variables in the appropriate trait classifications, and the sum was divided by the number of traits in the particular trait group to yield a value representing the leader's mean rating for that trait category. Each bar on the graph represents an individual leader's rating relative to the mean rating of all five leaders on the indicated trait cluster. What stand out most conspicuously on the graph are the sharp variances of Jean Chretien's character assessments and Stockwell Day's competence assessments. Judgements of Chretien suggest that he had the ability to hurt his party while Day oppositely had the prime potential to help the party that he led. Chretien was ripe for character attacks; the next weakest link was McDonough's competence, but since her party was not a direct threat to the 42 forerunners, campaign strategies undoubtedly passed on exploiting this situation when the assumed payback was greater from focussing their attacks on the true competition, Stockwell Day. Day's high competence rating is somewhat puzzling and is discussed in greater detail in subsequent sections of this text. Given the information at face value, the Alliance Party's leader was clearly perceived by voters as being the most competent of the group and in turn, his rivals would have to find ways to offset and/or reduce the public's favourable assessments to temper the leader's electoral threat. Overall, with the exception of the Bloc Quebecois' Gilles Duceppe - who in this analysis of Canadian federal political leaders is largely overlooked due to the party's singular, ethno-regional nature and its fading presence in the House of Commons - the leaders had manifest variances from the group's mean character and competence values. Figure 4. The relative merits of the 2000 Canadian federal leaders mmmmmmmmmmmmmmmmmmmmm Liberal (Chretien) Conservative Alliance (Day) NDP Bloc (Duceppe) (Clark) (McDonough) • Competence • Character 43 Of the four leaders who show distinct personality patterns, two - Chretien and Day -led the parties with the greatest potential for political success, that is the parties with best chances of forming the government, and these leaders display the strongest personality impressions at this level of analysis. However, after pooling the 2000 election information with Johnston's findings from the 1988, 1993, and 1997 elections, seven of the 18 leaders do not stand out, and of the 11 distinct leaders, seven led weak parties with no governing prospects. Specifically, the three party leaders (Liberal, PC, NDP) in 1988 each stood out against the group's mean ratings on the competence and character trait measures; in 1993, two (NDP and Bloc) of the five party leaders stood out; and the Conservative and Bloc Quebecois leaders were two distinct personalities in the 1997 election. 3 9 Leader Evaluations and Their Impact on Voter Choice The next question to be asked is about how voters' perceptions of leaders affect their choices at the ballot box. Table 1 presents the coefficients representing the impact of voter evaluations. As outlined in Chapter 2, the estimations are the outcomes of the OLS regression analyses using the leader trait clusters and statistical control variables (the control variables are included to eliminate partisanship effects) listed in Figure 1; I converted all the variables to values within the -1.00 to +1.00 range as detailed in Figures 2 and 3 in the previous chapter. The five dependent variables used in this analysis (see Figure 1 - one variable created for each political party) represent the survey respondents who stated they Johnston, "Prime Ministerial Contenders in Canada," 169. 44 voted for the particular leader's party - 1.00 indicating that the respondent voted for the party, 0.00 indicating s/he voted for another party. 4 0 Table 1. The impact of Canadian leaders' perceived traits ( omniume Character I 'ar iv leader s.e. (B) Liberal 0.354 0.031** 0.182 0.022** (2,664) Conservative 0.141 0.020** 0.177 0.020** (2,664) NDP 0.087 0.018** 0.173 0.018** (2,664) Alliance 0.389 0.021** 0.318 0.021** (2,664) Bloc 0.444 0.066** 0.457 0.062** ( 891) Average 0.283 0.261 Avg. excl. BQ 0.243 0.213 *p= O . 0 1 ; **p= < 0 . 0 0 1 . Note: Estimated by OLS. Details in this chapter. The coefficients reported in Table 1 describe the "hypothetical effects"4 1 of the individual leaders, that is the difference in the probability of choosing a political party when a shift of a whole unit occurs on the personality trait judgement scale. Functionally, a whole unit shift means, when using the example of the competence measure, a shift from saying that the leader is judged as being incompetent or competent, and moving to a neutral rating of zero. A move from one extreme to another, that is, where "competence" is perceived to 4 0 As in the case with the independent variables detailed in Figures 2 and 3, irrelevant responses were deleted from the dataset. In the case of the vote proper and vote intention variables, the eliminated responses were: "will not vote", "none", "don't know/undecided", and "refused". 4 1 Johnston, "Prime Ministerial Contenders in Canada," 171. 45 accurately describe a leader to where the trait is perceived to not fit the leader at all, therefore, represents a two-unit shift from +1 to -1. Using Stockwell Day's competence coefficient from Table 1 to illustrate, i f voters' judgements of the leader shifted from describing him as "incompetent" to the other end where he is described as being "very competent", then the probability of voting for the Canadian Alliance would increase by 77.8 percent (2 units x 38.9 percent per unit). Therefore, it can be said that the larger the coefficient, the stronger the impact of the related personality trait for the particular party leader, and in turn, the stronger the impact, the more relevant that trait is to voters for the particular leader. A trait may be more relevant to voters with regard to a particular candidate when, for example, a controversial situation surrounds the leader or when the leader's perceived competence or character otherwise draws the public's attention. To exemplify this effect, recall the general public's perceptions of Brian Mulroney's level of trustworthiness that drew questions in 1988 as voters linked him to the highly contentious issue of the Canada-United States Free Trade Agreement.42 Mulroney's distinct character coefficient (0.307) in that election year suggests that voters held him accountable for the Agreement and made judgements of his character a relevant consideration on election day. More specifically, i f voters who perceived the leader as having a poor character instead judged him as being of high character ("character" includes the variables "trustworthy", "compassionate", "moral", "really cares"), they would have been 61.4 percent more likely to have voted for him in 1988. Therefore, it can be said that based on Mulroney's distinct character coefficient, judgements of the leader on this measure were very relevant to the election since extreme changes in Ibid., 169-170. 46 voters' assessments could have improved the likelihood of voting for the Conservatives by a substantial amount. The averages of the five party leaders' competence and character coefficients weighted in at 0.283 and 0.261, respectively, which indicates that voters gave quite similar consideration overall to their judgements of the leaders on the two trait dimensions in 2000. This finding runs contrary to the qualified suggestion made by Johnston that competence counts less than character - a suggestion that was qualified in light of the special circumstances surrounding the perceptions of the leaders of the Liberals and the Progressive Conservatives in 1988.43 Viewed individually, however, the measures were weighted quite differently for two of the five leaders. Perceptions of Chretien's competence (0.354) had a significantly stronger impact on voters than those of the leader's character (0.182). Conversely, the impact of judgements of the NDP leader Alexa McDonough's competence (0.087) was approximately half that of judgements of her character (0.173). On these two aggregated trait measures, Clark, Day, and Duceppe were roughly balanced in comparison. The peculiar salience of Gilles Duceppe, as shown by his competence coefficient of 0.444 and character coefficient being similarly pronounced at 0.457, calls into question the campaign dynamics and/or the political environment in the period leading up to the election in Quebec. More importantly perhaps is the situation surrounding the translation of the 2000 CES to French.4 4 Stockwell Day's leadership impact takes a strong second place with coefficients of 0.389 and 0.318 for competence and character, respectively. The significance 4 3 Ibid., 169. 4 4 There is a strong possibility that language and cultural translation-related inconsistencies contributed to the notable variances in Gilles Duceppe's trait coefficients. Given the inconsistency of voters' evaluations of 47 of these two leaders in 2000 appears to mirror the comparative impacts of the parties' leaders in 1993 - the Bloc Quebecois' Lucien Bouchard and the Alliance Party's Preston Manning -the breakthrough year for both parties. At first glance, the election of 2000 was one in which leaders were significant considerations in voting decisions, based on the information provided in Table 2. The combined average leader trait coefficient (0.544) for the five federal parties clearly exceeds that of the prior three elections, however, it seems that the extraordinary effect is largely attributable to Gilles Duceppe. The removal of the BQ leader's coefficients from the mix effectively normalizes the group's average and brings this measure (0.456) into closer alignment with the values for the previous three elections. The consistent effects of perceived competence and character traits on voting decisions over the last four Canadian federal elections supports the theory that leaders do matter and suggests that campaign strategy consultants should continue priming their candidates and hammering the rivals given the evidence that party leadership is a valid consideration for voters. Duceppe due to the survey design issues, as well as the implicit limitations due to the Bloc Quebecois' presence only in Quebec, further discussion about Duceppe in this study is limited. 48 Table 2: Average impact of Canadian leaders' perceived traits, 1988-2000 1 lection \ ( omncu-iu'o ( h:i rac ier A v g . p Combined A v g . B ( oml) t \ i l UO A v g . p 1988 0.109 0.238 0.347 0.347 1993 0.155 0.272 0.427 0.432 1997 0.158 0.185 0.343 0.400 1988-1997 average 0.146 0.231 0.377 0.397 2000 0.283 0.261 0.544 0.456 Notwithstanding the language translation issue discussed earlier, the coefficients attributable to Duceppe beg the question - was there any issues within this particular election that brought Quebec voters' foci to the Bloc's leader? I conducted a brief search for an answer in order to disqualify the possibility that a particular situation or issue (for example, the Canada-U.S. Free Trade Agreement in 1988) affecting the leader had a notable effect in the broader election landscape in 2000. One potential answer may be related to the passage of a highly unpopular piece of legislation during the campaign period by Quebec's governing party, the Parti Quebecois. The legislation forced municipal amalgamations in Quebec City, Montreal, and Hull, and it is speculated that there was a public backlash against the party's federal counterpart, Gilles Duceppe's Bloc Quebecois. From the accounts of the events surrounding the passage of the legislation, it is possible that Duceppe's leadership became a more questionable element for Quebeckers. A scan of the campaign period media stories did not reveal any other significant events to warrant the strong focus on the BQ leader. With only regionalized potential leader effects from the municipal amalgamation legislation, it appears that this helps to partially 49 explain Duceppe's salience and that broader leader implications are not an issue that needs any further consideration in this study. The relationship between party size and leader personality impact that Johnston questioned in the 1997 election analysis is further called into question in 2000.4 5 In the three years since the last election, the Canadian Alliance established itself as a nationwide brokerage party and Stockwell Day's strong coefficients support the party size-leader claim; however, Jean Chretien's personality impact was mixed as his competence coefficient was high but his character effects were unremarkable. Given these mixed findings, i f any strong argument can be made it seems that it is from the bottom up; the lack of impact of leaders of weak political parties seems consistent from election to election - the lowest coefficients relate to the leaders of the NDP and the Progressive Conservatives which were arguably the two weakest political parties and most distant from power. Assuming that campaigns make an impression on voters, it seems logical that leaders of larger parties have a stronger impact given that the more competitive parties are at the front and centre during campaign periods, but in 2000, the argument is supported rather weakly. And finally, the temptation suggested by Johnston to read the decline in the relevance of leaders in 1997 as paradigmatic following the year of electoral flux in 19934 6 is unsupported by the findings reported in Table 2 - leader considerations in 2000 actually made a reversal and drew more attention than they did in 1993, rather than slip downward from the 1997 levels. The rise in leader personality impact in 2000, however, may be an artefact of the survey design issues that I discuss in greater detail in the final chapter. 4 5 Ibid., 170. 50 Leader Evaluations and Their Implications for the Outcome of the 2000 Election In the last section, I explained the coefficients representing the discrete impact of each leader on his/her respective party. This section moves a step forward and views the effects of the leaders comparatively by looking at the net effect of leader traits on voting. To reach this end, the character and competence coefficients reported previously in Table 1 are, in essence, disentangled of their overlapping values, producing values representing the impact on the popular vote of each leader to his/her party. To carry out this process, the trait coefficient for each leader (Table 1) is multiplied by the leader's respective trait rating variance from the group's mean, that is the value that underlies the graph in Figure 4. This equation takes the popular vote impact of the leader on a trait dimension and considers the effect o f the leader being judged at the mean for the trait measure, instead of his difference from the mean. The result shown in Figure 5 is the measure of the net effect of judgements of the leader on the particular personality trait cluster on his party's share of the popular vote, compared to a hypothetical leader standing at the mean. To complete the figure, the net impact values of the competence and character trait dimensions for each leader are added together, this representing the total net impact on the respective parties' vote shares of the two bundled trait measures. Ibid., 171. 5 1 Figure 5. The net impact of Canadian leaders' perceived traits (Note: Impact in percentage points of vote) Lib-Chretien PC-Clark Alliance-Day NDP- Bloc-Duceppe McDonough • Competence • Character • Total What seems certain after considering the information contained in both Figures 4 and 5 is that the Liberal Party leader and the Canadian Alliance Party leader made strong impressions on voters in 2000. Figure 5 tells us that, given the CES trait measures offered, Stockwell Day effectively projected himself as the most able, competent leader of the group. Meanwhile, the competition made a concerted effort to intertwine his image, particularly his character, with religious fundamentalist values linked to the former lay pastor and it seems that the media took full advantage of the opportunity to exploit the Alliance leader. It might be said that Day's highlighted religious beliefs did not resonate as strongly with voters at the polls as did the positive image of his perceived competence. The information presented thus far seems to indicate that Stockwell Day was a well-chosen leader for the Alliance Party as he 52 stood out well from the crowd in this particular election and where it mattered most, at the voting box. Comparing Day's leadership effects to that of his predecessor, Preston Manning, it is reasonable to suggest that the change was a step in the right direction for the relative upstart party. The direct result in 2000 was a dividend of over two percentage points paid to the Canadian Alliance - 3.5 percent competence impact offset by 1.2 percent of negative character effects - a reversal from 1993 and 1997 when the Reform Party suffered under the leadership of Preston Manning. The increase in the party's share of the popular vote in 2000 over 1997 was 6.1 percent, therefore, other factors besides leadership account for about 3.8 percent of the inter-election gain. The party's steady popularity growth suggests that it is successfully establishing itself in the Canadian electoral arena, so now more than ever, leadership is a critical factor as the Canadian Alliance is moving in a direction in which it has an opportunity to seize the position formerly held by the Progressive Conservatives as the first runner-up to the Grits. The changeover to Stockwell Day made a difference in maximizing the party's oppositional strength in the House of Commons - given Manning's dampening effect on his party's vote share of 0.3 percent, Day can be credited for increasing the party's popularity by a total of 2.6 percent over 1997 which accounts for a large portion of the improvement in the Alliance's vote share in 2000. Prime Minister Chretien was a liability to his party to nearly the same degree as Stockwell Day was an asset to the Alliance. In the Liberal Party's leader's case, the impact of each trait dimension was reversed, with his perceived character being the detrimental element to his party's fortunes. Chretien had a combined negative impact on the vote for the Liberals of about 1.9 percent - positive competence effects of 1.1 percent were offset by his 53 negative character impact of 3.0 percent - that was not enough to make a difference to the overall election outcome given that voters chose the Liberals to form the government by a vote margin of over 15 percent. The Liberal Party's current dominance in Canadian electoral politics appeared insurmountable under the snap election circumstances, and the selection of a leader with a negative effect on the party he led did not diminish the ability of the Liberals to win the 2000 election. The rehabilitating Progressive Conservative Party is the third and last party that had any realistic chance of challenging the Liberals in 2000. The Conservatives did not benefit to the same degree that the Canadian Alliance did from its leadership change. Although it can be said that Joe Clark was a positive force that boosted his party's vote share by 1.3 percent, the Conservatives lost ground after the loss of the party's popular leader in 1997, Jean Charest. The Charest-led Conservatives won 18.8 percent of the vote (20 seats), but behind the leadership of Clark, the party suffered a drop of 6.6 percent resulting in a loss of eight parliamentary seats; the 12 seats that the Conservatives held onto were just enough to salvage its official party status. The 1997 CES data suggests that voters judged Charest the most competent federal party leader in 1997 and his positive leadership effects landed his party two percent more of the vote than it would have earned without him. The Conservatives suffered a notable loss in Quebec in 2000, the province from which Charest hails, which hints more at what could be called a leader-"hometown" effect in 1997 than a personality effect.47 As for personality effects, the Conservatives only lost one seat by a margin of less than 0.7 percent (see Table 3) - 0.7 percent being the difference in vote impact between Charest and Johnston, "Prime Ministerial Contenders in Canada," 165. Refer to the discussion regarding prospective leaders' province of residence. 54 Clark - which suggests that leader personality effects are responsible for only a small portion of the party's vote share loss from 1997 to 2000. Leader Switching: What-if Scenarios in the 2000 Election The vote impact values reported above provide the basis for a compelling counterfactual analysis of the leaders. A hypothetical leader-substitution exercise can be conducted by calculating the difference between pairs of leaders' impact on vote choice. Some combinations are less plausible than others considering ethno-regional and ideological concerns but I present them here nonetheless as they illustrate the extent to which leaders can affect the fate of their parties come election day. Starting with the example of the most positively charged leader, Stockwell Day, i f he were to replace the most negatively assessed leader, Jean Chretien, I estimate a net gain of 4.2 percent (values taken from Figure 5: Chretien -1.9; Day +2.3) in the Liberal Party's share of the popular vote. The Liberals lost by a margin of 4.2 percent or less in 11 ridings in 2000 (see Table 3 for details about the 29 ridings that were won by a margin of 4.2 percent or less) translating to a gain, ceteris paribus, of 3.7 percent of the 301 House of Commons seats. The Liberals won 57.1 percent of the seats, therefore, a gain of 3.7 would only strengthen the majority it had already earned, but would have done so at the expense of three NDP (reducing the NDP to 10 seats, resulting in a loss of its official party status), two Alliance, and six Bloc seats. 55 Table 3: Electoral Districts =/< 4.2 Percent Margin of Victory in 2000 (+ won / - lost) Kleiloral District Vote Liberal \ l l l .HIU' PC M)P HQ Margin % 1) PEL Cardigan 1.6 + -2) NS: Dartmouth 3.1 + 3) NS: Sackville-Musquodoboit 1.9 - + 4) NS: West Nova 2.0 + _ 5) N B : Tobique-Mactaquac 0.5 + 6) PQ: Argenteuil-Papineau-Mirabel 1.1 - + 7) PQ: Champlain 0.0 - + 8) PQ: Charlesbourg-Jacques-Cartier 1.4 - + 9) PQ: Fronenac-Megantic 3.7 + 10) PQ: Laval Centre 0.1 - + 11) PQ: Laval East 2.2 + _ 12) PQ: Matapedi-Matane 0.9 - + 13) PQ: Quebec East 1.1 + 14) PQ: Richmond-Athabaska 0.7 + _ 15) PQ: Shefford 2.0 + _ 16) PQ: Trois-Rivieres 3.7 - + 17) ON: Elgin-Middlesex-London 4.1 + -18) ON: Haliburton-Victoria-Brock 2.3 + 19) ON: Lanark-Carleton 2.9 + 20) ON: Leeds-Grenville 0.1 + 21) ON: Nepean-Carleton 3.8 + -22) ON: Windsor-St. Clair 1.0 - + 23) SK: Saskatoon-Rosetown-Biggar 0.3 + 24) SK: Palliser 0.7 - + 25) SK: Regina-Lumsden-Lake Centre 0.5 + 26) SK: Regina-Qu-Appelle 0.6 - + 27) A B : Edmonton-West 1.5 + 28) B C : Richmond 2.4 - + 29) Y T : Yukon 0.5 + -Totals: +13 +4 +1 +5 +6 -11 -7 -3 -3 -5 Source: Past Elections: 37' General Election, November 27, 2000. In the reverse of the swap, the Canadian Alliance would have lost four seats, or one percent of the seats in the House, that it won by a margin of 4.2 percent or less - two to each of the Liberals and the NDP - had Chretien been the leader of Day's party in 2000. Going into the election, the Alliance Party was poised to become the Official Opposition and again, the leadership switch does not change matters in 2000. However, at this early stage in the life of the party, the loss of four seats would result in the Alliance earning 62 seats, only two 56 more than the number it earned in 1997. This would surely take some momentum away from the burgeoning party that is arguably on the cusp of an electoral breakthrough. A l l told, the Alliance had the most potential upside in this election and it had the strongest leader of the pack at that, any other leader-switching analysis could only result in a smaller loss of seats and vote share for the party which would be insufficient to bump the party from its strong, second place finish. A particularly intriguing leader exchange is the prospect of a moving the most negatively charged leader, Jean Chretien, to the two parties trying to keep afloat - the New Democrats and the Progressive Conservative Party. In the case of the NDP, it would lose 3.5 percent (Chretien -1.9; McDonough +1.6) of the popular vote resulting in the loss of the five seats that it won by that margin or less (Table 3). Such a loss translates to the NDP winning a mere eight seats in 2000; therefore, it would have again lost its official party status having fallen below the required threshold of 12 seats, possibly decimating the party once and for all. The severity of the loss for the Conservatives is not as great as that of the NDP, but the 3.2 percent (Chretien -1.9; Clark +1.3) reduction in Canadian voter support results in a loss of one seat leaving it with 11 seats. The party would have also lost its official party status with future implications being possibly worse than that of the NDP since the PCs have a ready replacement of sorts in the Alliance Party. The NDP's strength can be said to rest on the fact that it stands virtually alone as the most viable and rooted of the left-wing parties and does not have a realistic rival that can swiftly fill in the gap i f it were to drop back once again. Lastly, switches of Day for McDonough and Clark would have hypothetically given the NDP an additional two seats (seats lost by 0.6 percent or less) that the Canadian Alliance won, and the PCs one more seat (the lone seat won by 1.0 percent or less) at the expense of the Liberal 57 Party. In the bigger picture, the increases resulting from the substitution by Day do little to improve the lot for these struggling parties at this point in time. Canada is three years into a relatively sedate term since the Liberals took office once again after the 2000 election, and the impact of the hypothetical leadership exchanges just presented does not change matters with respect to the formation of a majority or minority government and the determination of the governing party and the Official Opposition. The lack of political turbulence offers us virtually no opportunities to present compelling counterfactual scenarios as those presented in the analysis of the 1988, 1993 and 1997 elections. The outcome of the 2000 election came as no surprise, which, no doubt, was the reason Chretien called the election at the opportune time that he did. And for all the Liberal Party's efforts to discredit the leadership of the Canadian Alliance Party, what was to be gained in the popular vote was too modest to matter materially this time around. Johnston, "Prime Ministerial Contenders in Canada," 171-172. One such counterfactual that was presented relates to the 1988 election and the Progressive Conservative Party's leader, Brian Mulroney. The Canada/U.S. Free Trade Agreement that was negotiated while Mulroney was Prime Minister became associated with the leader during his re-election campaign. The situation presented an opportunity for a compelling counterfactual analysis concerning the leader's effects on the election's outcome and the fate of the Agreement. 58 CHAPTER 4 DA TA AND FINDINGS: ON THE WA Y TO THE BALLOT BOX In Chapter 3, actual vote choice, or vote proper, data was the basis of the analyses. In this chapter, the data stems from the pre-election/campaign period dataset that includes vote intention data. In the first section of the chapter, I analyze campaign dynamics by way of the feeling thermometer leader ratings over the course of the pre-election period. In the second half of the chapter, the discussion moves to one centred on the relevance to voters of the personalities leading the two conservative parties, Stockwell Day and Joe Clark, over the course of the campaign period. Defining Moments? - the (non-) Dynamics of the Campaign of2000 The "feeling thermometer" data (explained in Chapter 2) provides the basis of the campaign dynamics visual that follows in Figure 6. The movements in Stockwell Day's and Joe Clark's feeling thermometer ratings during the campaign stand out most distinctly. First and foremost, we see Day's precipitous 11 percent decline from a mean rating at the start of the campaign of 52, on a scale of 100, to 41 at the campaign's end - equal to a 21 percent drop from the starting point. Secondary to Day's movement is that of Clark. Feeling thermometer judgments of Clark cover a seven point range over the 36 day period: an early campaign period dip from his starting point of 48, to a low of 42 about one-third of the way through the campaign, and the steady climb to a peak of 49 and a finish back where he began at 48. Both these leaders' feeling thermometer changes suggest that there are some campaign period stories to be told. 59 Figure 6. Impact of the campaign Conservative (Clark) Liberal (Chretien) NDP (McDonough) Alliance (Day) Weeks to election In an attempt to look beneath the fluctuations in the leader ratings, I unfold the campaign and divide it into three phases with the hope of better understanding what brought about the shifts. Fieldwork for the 2000 CES took place over 35 days of the pre-election period. The survey dataset was broken down into three periods of roughly the same length -12, 11, and 11 days, respectively. Rather than analyze all five leaders over the campaign period, I selected the two who stood out most prominently because voters' evaluations of Day and Clark changed most dramatically leading up to election day. Since the race to form the government was never really called into question, it was the competition between the two leaders and their respective parties that had the greatest chance of winning the position of the Official Opposition - the Conservatives and the Alliance having the strongest current or historical electoral support - that was more telling in this election. 60 The data presented in Figure 6 suggests that the campaign mattered to voters as their perceptions of the party leaders were not static - movement is apparent during the campaign period for all the party leaders and as a group, their ratings declined by about five points from the beginning of the campaign to the end. As discussed earlier, manipulation of leader images is a tool widely used by the rival political parties, as personalities are easy to project and vilify in the media and party strategists respond accordingly. Policy viewpoints, the country's economic performance during the previous term, amongst other abstract factors, are personified in the projections of the leaders in the media and the personalities become the focal point of the elections. This leadership focus occurs despite the relevance of the package of more substantive components that should matter to voters, but the media assumingly looks for stories that maximize their audiences and often bypasses the substance in favour of the personalities. The feeling thermometer data suggest that Stockwell Day suffered from the rival parties' attacks on his religious beliefs and healthcare proposals. Day's seven-point slide in the first 12 days of the campaign was the greatest popularity decline Day experienced during the campaign. Clark suffered a similar decline losing six points in the same period. The movement in the second phase of the campaign period suggests that the Clark's strong performances in the November 9 and 10 debates positively charged the leader as we see a notable change in trajectory in assessments of Clark who is in the midst of a notable four-point ratings ascent. Stockwell Day, on the other hand, did not have outstanding performances at the two debates and the feeling thermometer assessments suggest the same as they are stable as he begins and ends the period at 45. It is at this point in the campaign that Clark overtook Day, when the public's assessments of the two leaders crossed one I 61 another in their respective trajectories through to the end of the horserace. In the last 11 days, Day dropped an additional four points to 41 while Clark ended at 48 after climbing two points in the final days of the campaign. Overall, the magnitude of difference of the group of party leaders from the top to the bottom was stable at about seven points at the beginning (Chretien at the top at 53; McDonough at the bottom at 46) and at the end (Clark top at 48; Day bottom at 41) of the election campaign period. This measurement suggests that the campaign produced no change in net leader effect by election day, however, the campaign had an impact on evaluations of all the leaders given that movement of varying degrees is evident throughout the pre-election period. Put into the context of the earlier analysis of leader trait impact on the vote outcomes, if I assume that there is a direct relationship between feeling thermometer and personality trait ratings, Day's declining thermometer ratings can be translated to a campaign period change in vote share earned by the Alliance Party. Stockwell Day's feeling thermometer rating dropped from a high of 52 at the beginning of the campaign to 41 at the end - the 11-point decline translates to a 21 percent drop (11/52) from his original standing. Using a rough linear calculation, the 21 percent change in ratings of Day suggests a cost to the Alliance Party in popular vote share of 0.6 percent (Day's 2.3 percent impact on the Alliance's vote share per Figure 5 / [100-21 percent]) in the 34-day survey period. The vote impact is inconsequential to the outcome of the 2000 election but given the evidence that campaigns do matter, this exercise serves to illustrate their potential effect which in a very close election context, could make a difference in determining which parties form the government and become the Official Opposition, and in the current environment, may also be a factor for parties in retaining, losing, or regaining official party status. 62 The Crossing Paths of Joe Clark and Stockwell Day As explained above, i f there is any interesting story to be told about the impact of leaders in election of 2000, it seems the focus should be directed to the evaluations of Stockwell Day and Joe Clark, the leaders of the two parties battling for support from the voters seeking a conservative solution to Canadian government. Table 4 presents the impact estimations of each leader's competence and character evaluations, replicating the methodology used to produce Table 1. In this analysis, the data is split over the three phases of the campaign defined previously. In place of the post-election survey vote proper, the labile, campaign period vote intentions serves as the dependent variable in this analysis that measures the changes in impact of the leaders' perceived personality traits - in essence, this is a measurement of the campaign dynamics within the leader impact framework presented earlier. Table 4: Stockwell Day's and Joe Clark's Journeys to November 27 Competence Character 1 eailt-r Phase s.e. (B) s.e.(B) HPHI Day / Oct 24-Nov 4 0.301 0.046**** 0.393 0.046**** ( 717) Day/Nov 5-15 0.242 0 039**** 0.354 0.038**** ( 975) Day /Nov 16-26 0.237 0.036**** 0.311 0.036**** (1,101) Clark / Oct 24-Nov 4 0.039 0.033* 0.071 0.032** ( 717) Clark /Nov 5-15 0.081 0.029** 0.105 0 029**** ( 975) Clark /Nov 16-26 0.054 0.030** 0.157 0.029**** (1,101) *p=<0.00;**p=<0.1;"'*p= <0.01;****p= <0.001. Note: Estimated by OLS 63 The coefficients for Day in the initial phase reflect the introduction of the new leader of the Canadian Alliance, chosen by party members to replace prosaic Reform Party leader Preston Manning. Efforts by Day's strategists to project a dynamic image of the leader of the re-fabricated Reform Party, including using tactics like the memorable media event showing a fit and healthy-looking Day approaching the shore of British Columbia's Lake Okanagan on a personal watercraft arriving at his first news conference after his by-election victory. Adding to his "man of action" profile, Stockwell Day used the opportunity to talk to voters' pocketbooks by announcing that he wanted to shut down the House of Commons on Fridays in an effort to shave government expenditures and save taxpayers some money. The leader's pronounced character and competence coefficients (0.301 and 0.393) are evidence of the public's awareness of Stockwell Day in the first phase of the campaign. Joe Clark offered a vastly different picture to Canadians. A n average voter, at first glance, would likely think that Clark's stodgy public image did not help his wounded party that was in a desperate rebuilding mode and in a time of need. He arrived with a whimper at the beginning of the campaign - the coefficients representing Clark's first phase impact of his perceived character and competence, or rather lack thereof, (0.071 and 0.039), suggesting that judgements of leader's did not matter to supporters of the Conservative Party. The leader of the once-mighty Progressive Conservative Party did not seem to register with voters as a factor of import at this early stage in the campaign. The second phase brought about some change in the electorate's focus regarding Stockwell Day. The impact of perceptions of both his character and competence (0.354 and 0.242) decreased somewhat but Day was still relevant to voters. Clark seemed to gain some visibility in this period - his competence coefficient doubled and his character increased by 64 about 50 percent. The November 8 (French-language) and November 9 (English) televised debates took place in this period and Clark gave strong performances that were noted by the media and pundits alike. Nevertheless, Joe Clark's impact on voting decisions remained low. Day's performance was sound but largely unremarkable and the relatively stable measurements of his impact on voters intending to vote for the Alliance Party during this period fall in line with the judgements about his debate performances. Clark was unremarkable overall at this point in time and he was not making an imprint on voters whereas Day did seem to matter to those intending to vote for the Alliance Party - the impact of judgements of Day's personality traits was about three times greater than Clark's at this stage of the campaign. It is clear that the Tory leader, along with both McDonough and Duceppe, were sideshows to the main event centred on the leaders of the frontrunner parties. Apparently there was good reason for the Liberals to take a strong offensive approach with the Alliance leader at this point in the campaign period. Liberal pollster Michael Marzolini revealed only after the election that Stockwell Day hit the 34 percent popular vote mark at around the time of the debates, as measured by the party's private tracking polls. 4 9 Strangely, the feeling thermometer ratings do not reflect such extraordinary evaluations of Day but these ratings, as I suspect is the case with the Liberal poll indicators, are infused with partisan effects that may not be reflected in a parallel manner between these two measures. In any case, according to Marzolini's report, Day was a legitimate threat to the Liberal Party's bid to form the government for a third consecutive term and the party took measures to quell Day. The Liberals had the tools and experience for 4 9 Watson, "Day did better than we knew: A newly released poll shows the Liberals had reason to worry." 65 a pre-emptive strike; since the Liberal Party's polling information was private, no momentum boost was created for the Canadian Alliance. This disclosure helps to explain why Stockwell Day and his party did not develop into the full-blown threat to the Liberals that they looked to have the potential to become in the 2000 election. The last phase of the campaign shows small declines in the impact of Stockwell Day's character (0.237) and competence (0.311) coefficients, but the leader remained very relevant to voters. Day's threat of legal action on November 12, in response to what he thought were defamatory comments put forth by the Liberals about the Alliance's stand on health care, did not appear to go anywhere. As the election drew closer, Day dodged more bullets and continued to take criticism for the hand-scribbled sign that he displayed during the English debate - signs are disallowed by debate rules and he carried out the act out without the approval of his campaign handlers.50 The media picked up on the Liberals' sharp attacks on the Alliance Party leader's religious beliefs and numerous "hidden agendas". Meanwhile, Chretien was trying to draw attention away from himself and redirecting it to Liberal Party label since polling results indicated that the Prime Minister's popularity numbers were falling.51 Chretien also took a step back by accepting some responsibility for the increase in private health-care facilities and declared that additional funding would be provided to i U The sign read: "No Two-Tier Health Care". 5 1 Brian Laghi, "Chretien evokes image of team spirit," [newspaper article on-line] The Globe and Mail, 13 November 2000, accessed April 2003); available from cf=realtime/config-neutral&slug=wlibsteam&date=20001113; Internet. Specifically, Chretien said, "Politics is a team. I'm the coach, but I was a member of the team of Mr. Trudeau for many years." 66 purchase more medical equipment.52 Lastly, in a more trivial look at evaluations of Stockwell Day in this period, but noteworthy because of its possible contribution to voters' judgements of the leader's character, Rick Mercer, a television personality from a popular Canadian Broadcasting Corporation weekly comedy program, spearheaded a well-publicized effort to change the leader's name to that of the actor, "Doris Day". With supposedly over one million people who signed the online petition supporting the name change, the gibe drew public attention to the ridiculing joke aimed at the character, which was already on voters' minds, of the Alliance Party leader. Joe Clark continued to ride his post-debate wave into the third campaign phase appearing positioned to reap the benefits of the fallout from the war between Stockwell Day and Jean Chretien. Clark's relevance to voters remained low overall but judgements of his character seemed to matter more to voters relative to the prior phases as well as to voters' judgements of his competence. Despite strong debate showings and the opportunity that he and his party had to exploit the situation and pick up votes from presumably frustrated voters, as detailed earlier, the Conservatives lost both seats and vote share in 2000 behind Joe Clark. Voters judged the leader favourably but it is clear that other factors offset and perhaps overrode Clark's personality effects to produce the incongruent election results. The findings are largely inconclusive. It seems reasonable to say only that party leaders are but a piece in the Canadian vote decision-making puzzle. Specifically, private health-care facilities have been flourishing providing diagnostic services such as MRIs (magnetic resonance imaging machines), allowing those willing to pay for the service to not have to wait in line within the publicly funded medical system. The promise of additional funding to purchase medical equipment was an obvious shot back to the suggestion that there is a need for private health-care facilities due to the lack of needed resources. 67 CHAPTER 5 DISCUSSION The fundamental problem of causal inference: ... no matter how much data we collect, no matter how perceptive the observers, ho matter how diligent the research assistants, and no matter how much experimental control we have, we will never know a causal inference for certain. - Paul Holland, Statistics and Causal Inference53 Research Findings and Their Significance At the most basic level it can be said that survey responses in 2000 to questions about the leaders' personality traits indicate that the voters differentiated between the individual political party leaders. Moreover, the findings from this research reinforce earlier findings that Canadians consider the party leaders' personality traits when determining how they will vote at the federal level. In the 2000 election, however, the impact of leader assessments on the election's most popularly viewed outcome, that is the determination of which party will form the government, was negligible since the Liberal Party, in tandem with its leader, Jean Chretien, had many comparative advantages going into the 36 day campaign period leading up to the November 27 election. Chretien and his team of campaign strategists clearly knew that the timing was ripe for an early election call; with budget surpluses for which to take credit, a Paul Holland, "Statistics and Causal Inference," Journal of the American Statistical Association 81: 845-60, quoted in Gary King, Robert O. Keohane, and Sidney Verba, Designing Social Inquiry: Scientific Inference in Qualitative Research (Princeton: Princeton University Press, 1994), 79. 68 substantial lead in the public opinion polls, 5 4 unorganized opposition parties, and the two parties that could cause the most grief for the Liberals being without well-entrenched leadership, the conditions could not have been more ideal for a victory. In decrying the snap election call, Joe Clark went so far as to accuse the Prime Minister of "walking out in the middle" of his mandate.55 Jean Chretien's retort supporting his decision to bring electors to the polls one-and-a-half years early, stated that Stockwell Day in particular had been campaigning and otherwise positioning himself and his party for several months leading up to the October 22 announcement. In this light, the election call could be viewed as a defensive response to a feeling of threat as could be read between the lines of the Prime Minister's statement, "Mr. Day dared me to call an election". Nevertheless, the strategy was effective and with relative ease, Jean Chretien and the Liberal Party returned to power with an extended and well-supported mandate granted in the exchange. Considering the fact that voters did take leaders' personal traits into account in the 2000 and other recent Canadian federal elections, some support for the presidentialization thesis appears to be in order. The systemic context of Canadian elections diffuses the attention paid to party leaders so that it includes the leaders' parties and the local candidates, 5 4 Allison Dunfield, "Chretien calls election for Nov. 27," [newspaper article on-line] (The Globe and Mail, 22 October 2000, accessed April 2003); available from hr^i/Avww.theglobeandmail.conVservlet/RTGAMArt cf^ealtime/config-neutral&slug=wmain_election&date=20001022; Internet. According to an Ipsos-Reid poll taken in early October 2000, the Liberal Party were standing with public support of more than 50 percent, the Canadian Alliance near 20 percent, the Bloc Quebecois at 10 percent, and both the Progressive Conservatives and the New Democratic Party each near eight percent. 5 5 Jeff Gray and Allison Dunfield, "The opposition reacts," [newspaper article on-line] (The Globe and Mail, 22 October 2000, accessed April 2003); available from cf=realtime/config-neutral&slug=wreax&date=20001022; Internet. "He's walked off the job with a lot of business yet to be done. He was elected in 1997 for a full mandate; he's walking out in the middle of it," Mr. Clark said. 69 amongst other factors. Nevertheless, leaders serve as focal points in the media and this, in part from a campaigning standpoint, fuels support for the notion of the presidentialization of Canadian elections. Complementing the theory about what voters include in their decision-making processes, next to be considered is what voters do with the information that is available to them. The voting behaviour of Canadian electors evidenced in this study backs up the theory that voters use heuristics and/or schemas (which could arguably be called stereotypes) to help sort through the information in their vote decision-making processes, but the support is limited by survey design issues that I discuss later. It is interesting to note that over 80 years ago, several decades before Kinder and his colleagues broke ground in the leader personality field, Walter Lippman alluded to the concept in his oft-quoted treatise contained within the chapter entitled, "The World Outside and the Pictures in Our Heads": For the real environment is altogether too big, too complex, and too fleeting for direct acquaintance. We are not equipped, to deal with so much subtlety, so much variety, so many permutations and combinations. And although we have to act in that environment, we have to reconstruct it on a simpler model before we can manage with it. 5 6 Lippman seemed to grasp the concept of schemas but Kinder et al. and others took the general concept to another level applying it to the field studying leader assessments and their effects on voting behaviour. Despite employing less direct measures of voters' assessments of the individual leaders than in prior election studies, when evaluating the competence and character trait batteries, with the vote proper as the dependent variable (and vote intention in the case of the campaign period analysis of the changing relevance of Stockwell Day and Joe Clark), the Walter Lippman, Public Opinion (New York: Macmillan, 1922), 16. 70 resultant coefficients reported earlier lend support to the argument for the enduring impact of leader schemas in Canadians' vote decision-making processes. Combined with the data from the three prior federal elections, it is reasonable to generalize that the electorate regularly assesses prime ministerial candidates against a schematic framework, specifically using the competence and character trait measures, in deciding how to vote. In further determining whether Canadians use schematic processing, voters' considerations of Jean Chretien's competence and character from election to election are more telling. The three elections in which Chretien led the Liberal Party give us the opportunity to test the schema hypothesis from a different perspective. The impact of Chretien's combined perceived traits has generally been high (1993: 0.231 competence, 0.238 character; 1997: 0.231, 0.256; 2000: 0.354, 0.182 - again, see the discussion later about the variation in the 2000 survey's design) and seems to support the idea that voters tend to evaluate leaders schematically rather than idiosyncratically. The "character" and "competence" leader trait composites in themselves suggest that voters use schemas when evaluating political party leaders. Scholars developed these particular trait measures in large part as a response to research findings from the last two decades that suggests voters in western liberal democracies judge leaders against schemas, and using heuristics, that are built around the public's mental image of what personality traits leaders should possess, or at least be perceived to possess, in order to earn the title of President, or in the Canadian case, Prime Minister. Canadians vary the weight that they place individually on the competence and character measures from election to election but the variance may be more a by-product of survey design rather than being an indicator of idiosyncratic behaviour. Despite the 71 fluctuations in degree of impact of the leaders, the evidence does show that voters factor personality traits in to their voting decisions with consistency. Further exploration of the theory of schemas and heuristics in Canadian leader evaluations is limited by the changes in design of the Canadian election surveys over the years; the greatest limitation in this area of research is due to the change from questions requesting open-ended responses to questions with supplied-lists of possible responses. Recent surveys offer a finite list of possible responses to each question, which limits the response set. This survey method, no doubt, is more efficient and less expensive from an administrative and analysis perspective. The limitation of this methodology is that it does not permit the free flow of responses that is needed in order to study the nature of schematic information processing being carried out, therefore, the CES design changes make it impossible to conclude that Canadian voters use schemas in their vote decision making processes. The CES pre-election survey (which is the data set used by Johnston in his studies covering the 1988 to 1997 elections) in 2000 employed a substantively different set of leader personality trait questions resulting in a peculiar narrowing of the personality measurements compared to the surveys of past years. This departure led to problems starting from the earliest step in this research when attempting to measure the impact of leader evaluations. The most substantive change was to the overall manner in which voters assessed the leaders. Specifically and as mentioned in brief earlier, the survey did not ask respondents to rate the leaders on an individual basis, instead, the questions asked whether a particular trait described a single leader, all of the leaders, or none of the leaders. This survey approach limited the results considerably as it eliminated the ability to measure the degree of applicability, or inapplicability as the case may be, of a particular trait label. Also, further to 72 the discussion about schemas, in previous election studies, the personality trait questions addressed each leader individually and asked the respondents to state the degree to which a trait described each leader with four gradation options available ranging from "very well" to "not at all". The change significantly decreased the number of leader personality survey questions from 25 (five measures for each leader) in the 1997 pre-election survey to a total of 11 questions in 2000 (three in the 2000 pre-election survey and eight in the 2000 post-election survey); the 1997 CES did not include any leader trait questions in its post-election component. The pre-election questions in the 2000 CES covered only three personality traits -arrogance, trustworthiness, and having new ideas - which did not adequately fill the competence and character trait groups, as well, some uncertainty surrounds the categorization of the last trait which I discuss in greater detail below. These issues, coupled with the restricted ability to measure assessments of the individual leaders, were clearly problematic. If I employed the pre-election rather than the post-election data set, the value of the results would have been severely limited and inadequate for comparative purposes given the very thin volume of data used in the analyses. To deal with this predicament, I used the data from the responses to seven of the eight post-election questions (the question asking whether any of the leaders were "extreme" was excluded as it did not fit into either of the trait categories) because the post-election survey offered a broader range of questions, but with this change there lies an implicit pre- versus post-election timing issue. The concern is that recall-based evaluations of leaders in the weeks after the election may be tainted by the fact that voters have more campaign period information with which to base their judgements, as well, they know the outcome of the election. This obvious concern potentially results in quite different 73 leader evaluations which this study uses at the core of its analyses, and limits its comparability against prior, and possibly future studies of political party leader effects. As stated above, the one personality trait measure - having "new ideas" - of the seven employed in this study was ambiguous for classification purposes. It is plausible that this trait could fall into both the competence and character domains, however, I chose to classify it as a competence measure. I made this decision in part to have a more balanced mix of traits in each group (three competence variables and four character variables), but I did so largely because it seems that the measure more strongly reflects one's leadership ability, although a smattering of character seems infused in its definition. In any case, the ambiguity is problematic as I explain in greater detail below. Voters' judgements of the leaders indicate that the two most prominent party leaders show the sharpest contrast in personalities. It appears that voters questioned the Prime Minister's character - as measured by perceptions of his trustworthiness, compassion, arrogance, and dishonesty - whereas they ranked Stockwell Day the most politically competent - as measured by judgements of his intelligence, strength of leadership, and as having new ideas - of the group of five. Again, survey design concerns pervade these findings. Day's superior competence is particularly questionable. His feeling thermometer ratings declined significantly as election day grew nearer but his post-election competence evaluations are remarkably high, therefore the two datasets are inconsistent with one another. The problem seems to be largely attributable to evaluations of the "new ideas" trait. As the new leader of the young political party, it is not surprising that voters rated Stockwell Day strongly on this particular measure. On the other hand, this may be an artefact of the post-election recall/knowledge problem where respondents had more information that caused them 74 to judge Day differently given their knowledge of the election's outcome. In any case, survey design and timing issues seem to be at the root of the problem causing the marked difference in assessments of the leader. To test for a biased relationship between Day and the "new ideas" measure, I ran a regression analysis with the character and competence variables disaggregated that produced impact coefficients for the individual trait variables. The results confirm the significance of three traits in particular: compassionate (0.203), new ideas (0.173), and intelligent (0.101). The strong relevance of the "new ideas" and "intelligent" measures (two of the three traits included in the competence cluster) related to Day validate his high competence rating, although the rating is arguably artificial given the weak classification link of the "new ideas" trait. Comparatively speaking, the inclusion of the "new ideas" variable reduces the value of the impact of Day's high competence ratings on voting decisions because the measure seems skewed to favour the new leader of the young and arguably progressive party. The change in survey design created great limitations in its utility as a comparative tool, nevertheless, there is still a great deal of value in the data available as an evaluative tool about the 2000 election. Taking into consideration the limitations of the available data for this study, I now return to the discussion about the election at hand. Unfavourable evaluations of Chretien did not hinder the Liberal Party's ability to form a strong majority government although based on the empirical analysis reported here, Chretien's impact on the vote for the Liberals cost the party a larger share of the popular vote and a greater majority in the House of Commons. On the other hand, Stockwell Day's positive effects contributed but were not the key to his party's success in winning its place as the official opposition in 2000. The results of this 75 study support Johnston's finding that party leaders are an element in voting decisions but they stand in the periphery in the vote decisions for Canadians as a whole. The leaders competing for the lion's share of the conservative vote divided the opposition to the Liberal Party paving the pathway to a third consecutive victory for Jean Chretien. It seemed that a compelling story could be told based on the ideology-driven analysis of the conservative party leaders, but quite early in the statistical analysis process, it became clear that the fight waged by the Alliance and the Liberals was the centrepiece of the election. I believe that the political mudslinging between these two parties had the potential to land the Conservatives more support from frustrated voters than it would have received under less contentious conditions. The opportunity seemed to be there for the taking but it seems the Tories did not capitalize on the situation and, as stated earlier, the Conservative Party suffered a decline in both its seat share and popular vote in 2000 over 1997. Clark took over a party that was in poor shape after its 1993 crash and subsequent loss of Jean Charest after the 1997 election, and to his credit, the Tory leader managed to improve his standing within the campaign and contributed positively at an especially crucial time for his party, retaining the minimum 12 seats required to renew its official party status. The election was largely a story about two parties - the Liberals and the Canadian Alliance - and leadership mattered little to the outcomes that arguably matter the most, that is, which party would form the government and which would become the Official Opposition. The dwindling significance of the Bloc Quebecois and the pro-Quebec sovereignty movement that brought so much support to the party during its electoral heyday in 1993, leaves the battleground open for challengers to the runner-up position in Canadian government. Joe Clark and his Progressive Conservative Party's inability to take advantage 76 of the opportunity offered by the void in the existing Canadian political party sphere leaves the door open for the Liberal Party to continue its domination for the foreseeable future. Again, the vote spreads are too great for personalities to be of consequence in determining which parties will place first and second, but it is reasonable to assume that Canadian elections will , at the very least, return to a competitive level where personalities could matter in determining which party will form the Official Opposition. The two-leader, intra-campaign period analysis coupled with the thermometer ratings data for the five party leaders over the course of the election attest to the effects of campaigns, or more specifically that election campaigns have the ability to change the relevance of voters' judgements of leaders and move voters' opinions about party leaders. Clearly, campaigns can make a difference, but more important to this study is the fact that party leaders' perceived character and competence make impressions with the result being that they are one of a bundle of factors that voters consider on election day. Although the impact of leader personality traits is marginal overall in elections where two parties stand well above the rest of the pack, political party leaders can make a difference at the individual electoral riding level where, for example, a four percent margin separates winners from losers. With only a small number of ridings won by such close margins, leaders effects are not sufficient in this current electoral environment to alter the course determining which parties will form the government and the Official Opposition. However, leader evaluations are critical in this time where two parties in particular are effectively fighting for their political survival, that is, official party status in the House of Commons. So it can be said that the polarized party system with two parties on top, two on the bottom, and one isolated due to its ethno-regional concentration, weakens the potential effect of party leaders. 77 The assumption that voters can more ably evaluate the character of a new party leader than assess a new leader's level of competence seems reasonable. Chretien was a long-established quantity in national politics even before the 1993 election in which he first took over the Prime Minister's office, therefore in his case, it is understandable that voters evaluated his competence with relative ease. As a new leader, Stockwell Day offered voters comparatively little to go by when it came time to prospectively evaluate him on this trait measure. With a limited amount of information available to the general public about the Alliance leader, voters nevertheless were able to express judgements about Day that ultimately resulted in a net vote gain for his party, but positive leadership alone was not enough to make the difference to voters in their selection of the party to form the government in 2000. Even with the history, albeit relatively short, of the Reform Party behind it, the Canadian Alliance party label was new and untested under the leadership of Stockwell Day. Canadians were apparently not ready for a switch from an established centrist party to the largely untested Canadian Alliance Party. On average, voters gave equal consideration to both aggregated personality trait measures and the two frontrunner parties, the Liberals and the Canadian Alliance, showed the greatest leader effects after the removal of the Bloc Quebecois leader's anomalous coefficients. This study offers some support for the theory in Johnston's study of the 1997 election that suggests a relationship between large parties and strong leader effects, but survey design issues qualify this support. As a whole, the combined package of perceived competence and character has been relevant to voters since at least the election of 1988, and their impact was particularly high in the tumultuous election of 1993 that may in part be attributable to the introduction in that year of the two new regionalized parties, the Bloc 78 Quebecois and the Reform Party, that brought new energy into Canadian electoral politics. In particular, evidence from 1988 to 2000 proves that voters consistently included party leaders in their election day equations and, on average, appear to pay particular attention to leader character traits. The 2000 data suggests otherwise with competence assessments having a stronger impact than character assessments but again, the change in the CES's design seems to be responsible for the inconsistency. When combined and averaged with the data from the three prior elections, character still stands out as the stronger of the two trait batteries. Much like the situation discussed above regarding new party leaders, information is often scarce when it comes to new political parties and this leaves electors with less on which to base their voting decisions. So to compensate for the dearth of information, I speculate that leader personalities bear more significance in elections that include new and/or young political parties. The Alliance had only two elections behind it as the Reform Party where it established a solid support base in Western Canada and a rather limited but growing base outside of the region, so the party was still quite new on the scene for the majority of Canadians by the time the 2000 election rolled around. Further complicating matters, leading up to the election, the young party was engaged in a somewhat destabilizing "upheaval" that I argue swayed voters to factor in the Canadian Alliance as it would a new party in their voting equations. The elements contributing to this upheaval and subsequent "rebirth" are: (1) the renaming of the party, (2) the change of leadership, (3) its lack of a firm base of electoral support, and ( 4 ) the attempts to consolidate the conservative vote by amalgamating the party with the Progressive Conservatives in the period preceding the 2000 election. With a "new" party, a new leader, and an historical connection between voting behaviour and character 79 judgements of party leaders, the Liberal Party's campaign team made a clever tactical decision by choosing to focus its attacks on Stockwell Day's character. To further our understanding of the impact of leader personality traits (and would be of particular benefit to election campaign strategists), I suggest that consideration should be given to how voters evaluate leader personality traits. For example, how does a survey respondent measure his/her impression of candidates' traits when asked the question, "Which leader would you describe as 'arrogant'?" (survey response options: Chretien, Clark, Day, Duceppe, McDonough, none, all, don't know, refused).57 Does he/she base the evaluation on the frequency that the candidates exhibited the trait in the times that the voter had exposure to the leaders, from retrospective analyses (e.g. for incumbents) and/or from impressions based on the current election campaign? In what contexts did the leaders exhibit the traits (e.g. in a institutional setting such as Question Period or in a less controlled environment such as in the campaign period debates) and how were they framed (e.g. by political pundits, the media)? Voters' decision-making processes exist in a black box and there is no single model that can explain how they make their election day choices. Furthermore, their interactions may occur between any or perhaps all of the elements that are incorporated into the decision-making process. Such interactions may decrease the overall value of this study based on perceived leader traits, nevertheless, one can conduct an analysis incorporating the reasonable assumption that leader trait evaluations are not wholly dependent on other factors. Question "pesi7" of the 2000 Canadian National Post-Election Study. 5 8 Warren E. Miller and J. Merrill Shanks, The New American Voter, (Cambridge, M A : Harvard University Press, 1996), 417. 80 Previous research in the area of leader effects on vote decision-making in parliamentary systems found that personality factors make a contribution but their effects are limited, particularly in the current environment of federal elections in Canada where they have no substantive impact on the outcome of elections. The Liberal Party currently stands on its own with no strong alternatives to challenge it. The highly regionalized, but widening support base of the Alliance Party suggests that it needs more years of growth and presence on the Canadian electoral stage before voters see it as a viable alternative to the Liberals. A united conservative party consolidating the Canadian Alliance and the Progressive Conservatives, headed by a well-regarded leader, would most efficiently and effectively create a potential threat to the Liberal Party. However, this seems unrealistic at this point given that merger attempts to date have been unsuccessful in large part because the parties are too far apart in critical areas, therefore, an answer to the Liberal's dominance does not look to be close at hand. Leaders can make a difference in parliamentary elections where there is competition, which is not the case currently in Canada. The analysis of the impact of leaders traits on vote outcomes reveals that had the 2000 election been a race between political parties running within two percent of each other in the popular vote, the leadership effects could have been meaningful but this was not the case. To be sure, what is relevant in present-day Canada is the difference that leaders can make to parties struggling for their political survival, specifically in retaining official party status. Both the Tories and the NDP had the potential going into the election to become virtually extinct in 2000, but this research finds that positive leader effects helped to pull the parties through. Despite the current electoral conditions that lessen the differential impact of leaders on election day, political parties have used the tool of leadership change to transform their 81 parties' images. The parties with the greatest ability to form the government often respond with a change in leadership after lacklustre election bids as these are the parties that have the most to gain. Preston Manning led the upstart Reform Party through two elections before its members responded by selecting a new leader, and Stockwell Day suffered a similar demise recently as well. On the other hand, Ed Broadbent led the New Democratic Party for 15 years and contributed strongly to his party's electoral success through his last election as leader in 1988. From this last example, it seems that image changes are not critical for parties with no realistic chance of forming the government. The vote distortion effects caused by the single-member-plurality electoral system underlie the growth of regional parties such as the Canadian Alliance and the Bloc Quebecois but also contribute to the dominance of the Liberal Party. The issue has drawn public attention in recent years with proposals to modify the existing system, following the examples of successful conversions to mixed-proportional or proportional representation-based electoral systems for countries such as Australia, New Zealand, Belgium, Japan, and Switzerland. Additionally, a PR system may strengthen leader effects, but the opposite may be true depending on the mechanics of the model chosen. The Liberals owe much to the existing system for the strong majority government through which it currently rules as well its other previous victories, so for this reason it is likely that electoral reform will remain in the shadows and not advance quickly beyond the discussions of today. Alas, disproportionate voting effects will continue, and the changes that could eliminate what some may view as an obstruction to representative government in Canada do not appear forthcoming in the short term. 82 We currently live in a media and telecommunications-driven age, one that focuses its attention on personalities and fosters the idea that party leaders have an impact on election outcomes. This study, and other research conducted before it, strongly suggests that political party leaders operating within parliamentary systems do matter to voters on election day. The theories that prime ministerial candidates should possess (or should be perceived to possess) particular personality traits to have a favourable effect on their parties, and that voters predictably use personal road maps of schematic traits against which they assess leaders, provide compelling reasons for political parties to continue their relationships with campaign image consultants and strategists. Perhaps this is wishful thinking by an observer of leader effects, but could it be possible that Canadian voters will someday find themselves where their assessments of the party leaders are the ultimate election tiebreakers? Concluding Remarks In the final stages of the drafting of this thesis, both the Canadian Alliance and Progressive Conservatives selected new leaders for their parties. Jean Chretien appears to be sticking to his word - he promised during the 2000 election campaign that i f he were re-elected, he would step down after his term ended. Potential successors have been making themselves known to Canadians and are jockeying for position, readying themselves well in advance of their respective leadership conventions that will decide who will be the Liberal Party leader in November 2003. The findings of this study tell us that party leaders should be especially relevant to voters come election day. With three new leaders that may provide their respective parties with a "new leader bounce", more time for the younger parties and 83 those rebuilding to make their imprints and strengthen their relationships with Canadian voters, the next election should tell a very different and competitive story where leader effects on voting decisions do make a difference to parties other than just those struggling at the bottom. 84 APPENDIX Statistical control ethnic/cultural group dummy variables were created as follows: Table A . l . Ethnic/Cultural Group Statistical Control Dummy Variables Visible .Northern Eastern Southern Ethnic/Cultural (iroup Minoritv European Eurnpt-:iii European i until Other Canadian Australian British English Irish Israeli S i_ •Jewish New'Zealander Scottish Welsh Bahamian X Bangladeshi X Black/African X Chinese X Salvadoran X Ethiopian X Guyanese X Haitian X Indian X Jamaican X Japanese X Korean X Lebanese X Nigerian X Pakistani X Filipino X Sikh X Somalian X Sri Lankan X Tamil X Trinidadian X Vietnamese X Inuit, Metis, Aboriginal, Native X X Austrian X Danish X Dutch X Finnish X German X Holland X Netherlands X Norwegian X Swedish X 85 Table A . l . Ethnic/Cultural Group Statistical Control Dummy Variables Visible Northern Eastern Southern Ethnic/Cultural Group Minority European hurnpi'Hii European French Croatian X Czech X Hungarian X Polish X Russian X Serbian X Slovakian X Ukrainian X Yugoslavian X Greek X Italian X Macedonian X Portuguese X Spanish X French X Quebecois X Note: Excluded'Groups 86 BIBLIOGRAPHY Bartle, John, Ivor Crewe, and Anthony King. "Was it Blair Who Won It? Leadership Effects in the 1997 British General Election." Paper delivered at the Annual Meeting of the American Political Science Association, Washington, D.C., 28-31 August, 1997. Bean, Clive. "The Electoral Influence of Party Leader Images in Australia and New Zealand." Comparative Political Studies 26, no. 1 (1993): 111-132. Bean, Clive and Anthony Mughan. "Leadership Effects in Parliamentary Elections in Australia and Britain." American Political Science Review 83, no. 4 (1989): 1165-1179. Bennett, W. Lance. News: The Politics of Illusion, 3d ed. White Plains, N Y : Longman, 1996. Blais, Andre, Elisabeth Gidengil, Richard Nadeau, and Neil Nevitte. The 2000 Canadian Election Study [study on-line]. Toronto: York University Institute for Social Research, 2000, accessed August 2002; available from; Internet. . Anatomy of a Liberal Victory: Making Sense of the Vote in the 2000 Canadian Election. Peterborough: Broadview Press, 2002. . "Campaign Dynamics in the 2000 Canadian Election: How the Leader Debates Salvaged the Conservative Party." PS: Political Science and Politics 36 (2003): 45-50. . "Do People Have Feelings Toward Leaders About Whom They Say They Know Nothing." Public Opinion Quarterly 64 (2000): 452-463. Brown, Steven D., Ronald D. Lambert, Barry J. Kay, and James E. Curtis. "In the Eye of the Beholder: Leader Images in Canada." Canadian Journal of Political Science 21, no. 4(1988): 729-755. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. The American Voter: An Abridgment. New York: John Wiley & Sons, 1964. Clarke, Harold D., Jane Jenson, Lawrence LeDuc, and Jon H. Pammett. Absent Mandate: Canadian Electoral Politics in an Era of Restructuring, 3d ed. Vancouver: Gage, 1996. Crewe, Ivor and Anthony King. "Are British Elections Becoming More 'Presidential'?" In Elections at Home and Abroad: Essays in Honor of Warren E. Miller, eds. M . Kent 87 Jennings and Thomas E. Mann, 181-206. Ann Arbor: The University of Michigan Press, 1994. . "Did Major Win? Did Kinnock Lose? Leadership Effects in the 1992 Election." In Labour's Last Chance? The 1992 Election and Beyond, eds. Anthony Heath, Roger Jowell, and John Curtice, 125-147. Hants, England: Dartmouth Publishing, 1994. Elections Canada. Elections Canada On-Line. Home page on-line. Available from; Internet; accessed March 2003. Gidengil, Elisabeth, Andre Blais, Richard Nadeau, and Neil Nevitte. "Are Party Leaders Becoming More Important to Vote Choice in Canada?" Paper delivered at the Annual Meeting of the American Political Science Association, Washington, DC, 30 August to 3 September, 2000. Graetz, Brian and Ian McAllister. "Party Leaders and Election Outcomes in Britain, 1974-1983." Comparative Political Studies 19, no. 4 (1987): 484-507. . "Popular Evaluations of Party Leaders in the Anglo-American Democracies." In Political Elites in Anglo-American Democracies: Changes in Stable Regimes, eds. Harold D. Clarke and Moshe M . Czudnowski, 44-64. DeKalb, IL: Northern Illinois University Press, 1987. Hayes, Bernadette C. and Ian McAllister. "Gender, Party Leaders, and Election Outcomes in Australia, Britain, and the United States." Comparative Political Studies 30, no. 1 (1997): 3-27. Jackson, Robert J. The General Election of2000: A New Government for the New Millennium. Toronto: Pearson Education Canada Inc., 2001. Johnston, Richard. "Canadian Elections at the Millennium." Choices: Strengthening Canadian Democracy 6, no. 6 (September 2000): 4-50. . " A Conservative Case for Electoral Reform." Policy Options 22, no. 5 (2001): 7-14. Johnston, Richard, Andre Blais, Henry E. Brady, and Jean Crete. Letting the People Decide: Dynamics of a Canadian Election. Montreal: McGill-Queen's University Press, 1992. Johnston, Richard, Andre Blais, Elisabeth Gidengil, Neil Nevitte, and Henry E. Brady. "The Collapse of a Party System? The 1993 Canadian General Election." Paper delivered at the Annual Meeting of the American Political Science Association, New York, N Y , 1-4 September, 1994. 88 Jones, Philip and John Hudson. "The Quality of Political Leadership: A Case Study of John Major." British Journal of Political Science 26, no. 2 (1996): 229-244. Kinder, Donald, R. "Presidential Traits: Pilot Study Report to the 1984 NES Planning Committee and NES Board." [study on-line]. 1983, accessed June 2002; available from; Internet. Kinder, Donald R., Mark D. Peters, Robert P. Abelson, and Susan T. Fiske. "Presidential Prototypes." Political Behavior 2, no. 4 (1980): 315-337. King, Anthony, ed. Leaders' Personalities and the Outcomes of Democratic Elections. New York: Oxford University Press, 2002. King, Gary, Robert O. Keohane, and Sidney Verba. Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton: Princeton University Press, 1994. Lippmann, Walter. Public Opinion. New York: Macmillan, 1922. LeDuc, Lawrence. Leaders and Voters: The Public Image of Canadian Political Leaders. In Leaders and Leadership in Canada, eds. Maureen Mancuso, Richard G. Price, and Ronald Wagenberg, 53-74. Toronto: Oxford University Press, 1994. McAllister, Ian. Leaders. In Comparing Democracies: Elections and Voting in Global Perspective, eds. Lawrence LeDuc, Richard G. Niemi, and Pippa Norris, 280-298. Thousand Oaks: Sage, 1996. Miller, Arthur H . "Economic, Character, and Social Issues in the 1992 Presidential Election." American Behavioral Scientist 37, no. 2 (1993): 315-327. Miller, Arthur H. , Martin P. Wattenberg, and Oksana Malanchuk. "Schematic Assessments of Presidential Candidates." American Political Science Review 80, no. 2 (1986): 521-540. Miller, Warren E. and J. Merrill Shanks. The New American Voter. Cambridge, M A : Harvard University Press, 1996. Nadeau, Richard and Andre Blais. "Explaining Election Outcomes in Canada: Economy and Politics." Canadian Journal of Political Science 26, no. 4 (1993): 775-791. Nadeau, Richard, Richard G. Niemi, and Timothy Amato. "Prospective and Comparative or Retrospective and Individual? Party Leaders and Party Support in Great Britain." British Journal of Political Science 26, no. 2 (1996): 245-258. ' 89 Pammett, John H. and Lawrence LeDuc. Explaining the Turnout Decline in Canadian Federal Elections: A New Survey of Non-voters [report on-line]. March 2003, accessed April 2003; available from http://wvv^ =e&textonly=false; Internet. Pancer, S. Mark, Steven D. Brown, Paul Gregor, and Stephen P. Claxton-Oldfield. "Causal Attributions and the Perception of Political Figures." Canadian Journal of Behavioural Science 24, no. 3 (1992): 371-381. Politics Canada. Politics Canada. Home page on-line. Available from; Internet; accessed March 2003. Riggle, Ellen D. "Cognitive Strategies and Candidate Evaluations." American Politics Quarterly 20, no. 2 (1992): 227-246. "Series: Election 2000." The Globe and Mail. 22 October-28 November 2000; available from; Internet; accessed March 2003. Stewart, Marianne C. and Harold D. Clarke (1992). "The (Un)Importance of Party Leaders: Leader Images and Party Choice in the 1987 British Election." The Journal of Politics 54, no. 2 (May 1992): 447-470. 


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