Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Campaigns, the media and insurgent success : the Reform party and the 1993 Canadian election Jenkins, Richard W. 1999

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

Item Metadata

Download

Media
831-ubc_1999-463621.pdf [ 19.98MB ]
Metadata
JSON: 831-1.0089288.json
JSON-LD: 831-1.0089288-ld.json
RDF/XML (Pretty): 831-1.0089288-rdf.xml
RDF/JSON: 831-1.0089288-rdf.json
Turtle: 831-1.0089288-turtle.txt
N-Triples: 831-1.0089288-rdf-ntriples.txt
Original Record: 831-1.0089288-source.json
Full Text
831-1.0089288-fulltext.txt
Citation
831-1.0089288.ris

Full Text

CAMPAIGNS, THE MEDIA AND INSURGENT SUCCESS: The Reform Party and The 1993 Canadian Election by RICHARD W. JENKINS B.A., Queen's University, 1992 M.A., Queen's University, 1994  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Political Science) We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA August 1999 © Richard William Jenkins, 1999  University of British Columbia Abstract CAMPAIGNS, THE MEDIA AND INSURGENT SUCCESS  by Richard W. Jenkins Chairperson of the Supervisory Committee:  Professor Richard Johnston Department of Political Science  It is well recognized that the 1993 election campaign catapulted the Reform party into the national political scene, but our understanding of how this was possible is quite limited. Drawing on the work in cognitive psychology on attitude change, the work on the news media coverage of elections, and the political science work on election campaigns, this thesis locates the impetus for Reform's success in the dynamic flow of information about the party that was available in television news broadcasts and voters' likelihood of being persuaded by that information. This link is developed by an analysis that makes use of a content analysis of the 1993 campaign, the 1993 Canadian Election Study, and a merged analysis of the election and news data. The Reform party began the campaign as a minor component of the news coverage of the election, but the news media coverage changed dramatically. Reform was provided with more news access than its support indicated it deserved and that coverage focused on what became a major theme of the election; the welfare state and the role of government. Coverage of Reform underwent a further change as it both decreased and focused on cultural issues during the last two weeks of the campaign. Using a two-mediator model of attitude change, the analysis shows that people who were predisposed to agree with Reform's anti-welfare state message and who were likely to be aware of the news information, changed both their perceptions of the party and increased their support for the party. Further support for the impact of the media is derived from the analysis of voter response to the second change in news coverage. The analysis suggests that campaigns do matter, but that the size of the impact is dependent upon the underlying uncertainty associated with the parties and candidates, and on the degree to which the information flow of the campaign changes. The information flow contributes to both learning and priming among people who receive and accept new information. While voters respond reasonably to new information, the outcome will depend on what information voters are given and what information actually reaches the habitually unaware segments of the population.  ii  TABLE OF CONTENTS ABSTRACT  ii  LIST OF FIGURES  v  LIST OF T A B L E S  vii  ACKNOWLEDGEMENTS  viii  CHAPTER 1 : INTRODUCTION  1  The Reform Insurgency  6  Campaigns  8  The Media and Elections  20  Attitude Change and Persuasion  28  A Method for Uncovering Campaign Effects  33  Plan of the Book  35  C H A P T E R 2 : THE ELECTION C A M P A I G N A N D THE V O T E  38  Setting the Stage for the Campaign  39  The Ideological Character of the Reform Party  45  The Campaign  52  Data  65  Discussion  71  C H A P T E R 3 : T H E C H A N G I N G INFORMATION F L O W  73  Balance and Party Access  74  Data  80  The Path of the Vote and Reform Coverage  81  Which Came First: Public Support or Media Attention?  89  Discussion  95  C H A P T E R 4 : GETTING THE R E F O R M M E S S A G E OUT Issues in the News  98 99  Issues in the 1993 Election Campaign  103  Dynamics of Reform News Coverage  118  Conclusion  129  CHAPTER 5 : LOCATING REFORM  131  Locating Political Parties  132  Measuring Party Locations  136  Locating the Parties in 1993  137  Getting it Right? Individual Determinants of Party Location Early in the Campaign  150  The Referendum Campaign and Awareness of Reform's Ethno-Linguistic Position  154  Is it Knowledge or Projection?  158  Conclusion  160  C H A P T E R 6 : PERSUASION, PRIMING A N D L E A R N I N G  162  Activating Preferences: Priming or Learning?  164  The Stability of Public Attitudes: the Lack of Persuasion  169  Evidence for Priming  179  Reconsidering Priming  190  A Longer Time Horizon: Preston Manning and Anti-Welfare State Mobilization  195  Conclusion  197  C H A P T E R 7 : A M E D I A EFFECTS M O D E L OF INSURGENT SUCCESS  200  The Campaign and Individual Responsiveness to Television Coverage  201  Key Independent Variables  202  A n Approach to Understanding Individual Responsiveness to Campaign Information  208  Credibility on the Deficit  212  The Evolution of Reform Support  222  A Coverage Model of Reform Support  231  Conclusion  244  CHAPTER 8 : CONCLUSION  248  How Campaigns Matter?  248  Television News and the Information Flow of Campaigns  254  Insurgent Party Success  262  Conclusion  268  R E F E R E N C E LIST  271  APPENDIX A : 1992-93 C A N A D I A N ELECTION STUDY V A R I A B L E S  285  APPENDIX B : POLITICAL A W A R E N E S S S C A L E  289  APPENDIX C : IDEOLOGY V A R I A B L E S  :  APPENDIX D : JUSTIFICATION FOR ESTIMATION M O D E L S  300 303  LIST O F FIGURES  Figure 2-1. Party Support in the 1990s  40  Figure 2-2. Reform's Share of National Vote Intentions: 1990-94  44  Figure 2-3. Party Shares of Vote Intentions Outside of Quebec  54  Figure 2-4. Campaign Events and Support for the Reform and Conservative Parties  56  Figure 3-1. Partisan Dynamics of Television News Coverage  83  Figure 3-2. Popular Support for and Media Attention to Reform  86  Figure 3-3. Normalized Reform Coverage and Intentions  87  Figure 4-1. The Distribution of Issue Coverage  104  Figure 4-2. Dynamic Coverage Changes: Jobs, Social Programs, and Culture  109  Figure 4-3. Public Perceptions of Most Important Problem by Week of Campaign  111  Figure 4-4. Public Concern for Jobs, Social Programs and the Deficit  112  Figure 4-5. News Attention and the Importance of Social Programs  113  Figure 4-6. News Attention and the Importance of Jobs  114  Figure 4-7. Key Changes in Reform Issue Coverage  120  Figure 4-8. Exchange Between John Beck and C B C Reporter on October 13, 1993  126  Figure 5-1. Self-Reported Knowledge of Party Leaders  138  Figure 5-2. Unwillingness to Rate Parties by Week of Campaign  139  Figure 5-3. Unwillingness to Rate Party Leaders by Week  140  Figure 5-4. Mean Perceptions of Party Locations on Selected Issues  144  Figure 5-5. Party Locations on French Canada  146  Figure 5-6. Party Location on Aid to Minorities  146  Figure 5-7. Perceived Credibility on Deficit  147  Figure 5-8. Don't Knows on Deficit Party Location Question  148  Figure 5-9. Effect of Awareness of Manning's Opposition to Charlottetown on Perceptions of Reform's Position on French Language  155  Figure 5-10. Normalized Changes in Reform Support and Deficit Credibility  159  Figure 6-1. Stability of Welfare State Attitudes.  170  Figure 6-2. Public Support for Cuts to Government Programs  172  V  Figure 6-3. Stability of Attitudes toward the French Fact  174  Figure 6-4. Stability of Racial Minority Questions  175  Figure 6-5. Reform Vote Intentions and Thermometer Rating During Campaign  180  Figure 7-1. Path Diagrams of Key Hypotheses  201  Figure 7-2. Evolution of Deficit Credibility by Awareness and Predispositions  213  Figure 7-3. Reform Intentions by Awareness and Attitudes Toward Welfare State  223  Figure 7-4. Reform Rating by Awareness and Attitude toward the Welfare State  224  Figure 7-5. Reform Rating by Awareness and Attitudes toward Minorities  228  Figure 7-6. Reform Vote Intentions by Awareness and Attitudes toward Minorities  229  Figure 7-7. Effect of Anti-Minority Attitudes for Reform Support Among High Aware, Anti-Welfare State Voters Figure B - l . Reform Deficit Credibility by Media Exposure and Attitudes  230 296  Figure B-2. Media Use versus Political Awareness as Discriminator of Changes in Reform Credibility (Anti-Welfare State Rs Only)  297  Figure B-3. Interaction between News Exposure and Awareness for Anti-Welfare State Respondents (Reform Credibility)  298  vi  LIST OF TABLES Table 2-1. Key Events in the 1993 Election Campaign  55  Table 2-2. Type of Television News Coverage  70  Table 3-1. Granger Causality Test  90  Table 3-2. A n Error-Correction Model of Reform Attention and Intentions  94  Table 4-1. Temporal Distribution of Issue Coverage by Week  107  Table 4-2. Partisan Distribution of Issue Coverage  117  Table 4-3. Issues Associated with the Reform Party by Week  119  Table 5-1. Frequency of Nonresponse to Party Location Questions (Weeks 1-3)  141  Table 5-2. Locating Reform Early in the Campaign (Weeks 1-3)  152  Table 5-3. Knowing Manning's Position in 1992 and Locating Reform in 1993  157  Table 5-4. Granger Test of Relationship between Reform Credibility and Support  160  Table 6-1. Campaign Stability of Ideological Scales  177  Table 6-2. Evidence for Priming  182  Table 6-3. Intra-Individual Changes of Reform Rating  186  Table 6-4. The Effect of Priming Within Awareness Groups  189  Table 6-5. Priming Among those Aware of Reform's Position on Deficit  192  Table 6-6. Priming Among those Aware of Reform's Position on Minorities  194  Table 6-7. Changing Feeling Thermometer Ratings of Manning  195  Table 6-8. Ideological Considerations and Evaluations of Manning, 1992 and 1993  196  Table 7-1. News Coverage and the Re-Evaluation of Reform's Deficit Credibility  215  Table 7-2. Reform Deficit Credibility Simulations for Different Levels of Coverage  217  Table 7-3 Cross-Sectional and Longitudinal Effects of Credibility on Support  221  Table 7-4. Reform Rating and the Activation of Anti-Welfare State Attitudes by Reform Coverage  232  Table 7-5. Reform Rating Simulations for Different Levels of Coverage  235  Table 7-6. Activation Effects of Ethnic Coverage on Reform Support  237  Table 7-7. Reform Rating Simulations based on Ethnic Coverage Model  239  Table 7-8. A Two Issue Model of Reform Support within Awareness Groups  243  Table B - l . Two Factor Solution for Self-Reported Media Exposure  292  Table B-2. Individual and Social Bases to Political Awareness  294  Table D - l . Justification for Two-Mediator Model of Deficit Credibility  305  Table D-2. Justification of Two-Mediator Model for Welfare State Activation Model... 306 Table D-3. Justification of Two-Mediator Model of the Activation of Attitudes about Minorities  307  ACKNOWLEDGMENTS The work on this thesis was supported by a doctoral fellowship from the Social Sciences and Humanities Research Council of Canada and a University Graduate Fellowship from the University of British Columbia. A number of people and institutions deserve thanks for providing the data. In particular, Kate Morrison of the National Media Archive was gracious in providing me access to the content analysis data of the 1993 election. In addition, Bob Burge provided help with accessing the Environics data that is used in Chapter 2. O f course, none of the individuals or institutions recognized here bear any responsibility for the results or analysis. Parts of this thesis were presented at a number of conferences in the previous three years. Papers were presented at a conference on Campaign Effects at U.B.C. (June 1997); the Annual Meeting of the Midwest Political Science Association (April 1998); the Annual Meeting of the Political Methodology Group (July 1998); and, the Annual Meeting of the Canadian Political Science Association (June 1998; June 1999). Without the helpful advise and criticism of all those people who shared panels, discussed the papers, and listened attentively, this project would not be possible. While the project was conceived as a single piece of work, the opportunity arose to publish some of it as I went along. A much simplified version of Chapter 7 will appear in an undergraduate reader, Political Behaviour: Theory and Practice in a Canadian Context, edited by Joanna Everitt and Brenda O'Neil (Oxford, 2000). A slightly modified Chapter 3 is forthcoming in Political Communication under the title, "How Much is Too Much? Media Attention and Popular Support for an Insurgent party." A significantly different version of Chapter 7 is currently being considered for publication in an edited book, Capturing Campaign Effects, edited by Richard Johnston and Henry Brady (University of Michigan Press, n.d.). A work such as this would not be possible without the fellowship of a number of colleagues at the University of British Columbia. In particular, Patrick Fournier and Stuart Soroka have provided both essential opportunities for academic exchange and much needed relaxation in between periods of hard work. A number of faculty too numerous to mention have also been helpful both in terms of my intellectual development but also in terms of easing my transition from student to colleague that I am about to make. I would also like to acknowledge the intellectual debt that I owe to the recent Canadian Election Study teams and in particular my supervisor Richard Johnstoa As a student I could not ask for a better mentor than Richard Johnston. He was supportive while never allowing me to settle for half measures. Lastly I owe a considerable debt to my family who have always been supportive of my continued studies even if they rarely understand what it is I do or why I do it. I could not have completed this project without the love and assistance of my partner Amanda James. Amanda showed remarkable patience as she watched me toil these four years and her understanding of what it takes to get done was a blessing.  C H A P T E R 1: I N T R O D U C T I O N The Campaign as a Site of Political Change What the political campaign did, so to speak, was not to form new opinions but to raise old opinions over the thresholds of awareness and decision. Political campaigns are important primarily because they activate latent predispositions. Lazarsfeld, Berelson, & Gaudet (1948: 74). Election campaigns are political contests. The outcome may often be predictable, but like all contests, it may be determined by the actions of the participants as the contest unfolds. While the conventional view is that campaigns do not generally matter (German and King 1993; Holbrook 1993), the presence of significant campaign effects in recent Canadian elections (Johnston et al. 1992; Johnston et al. 1994) suggests that campaigns can sometimes have dramatic effects. There is no question that the campaign mattered for the 1993 Canadian election, which witnessed the Reform party rise from obscurity to capture 18.7 percent of the vote. The appropriate question is, however, not "Do campaigns matter?", but rather "How do campaigns matter?" For even in those situations where there is no significant change in vote intentions, it is hard to believe that in the absence of the campaign the election outcome would have been the same. There are three main ways in which the campaign can matter. First, the campaign can persuade voters to change their underlying attitudes about the appropriate government response to an issue with a concomitant change in party support. For example, if voters became more supportive of spending cuts then conservative parties would be advantaged. Second, the campaign can activate or prime certain kinds of underlying attitudes rather than others. Elections could be decided on the basis of a number of different kinds of  1  considerations (issues, character, economy) and the campaign can affect which considerations are most important for voters. Third, the campaign can inform voters about the parties' stands on issues or the character of their leaders allowing voters to make reasonable decisions. There is no question that the 1993 Canadian election campaign mattered. What began as a close race between the Liberal and incumbent Conservative parties turned into a relatively easy Liberal majority win. The key dynamic was the mid-campaign collapse of the Conservative party and the surge of Reform support in the non-Quebec electorate. Reform ran no candidates in Quebec, where the key race was between the Liberals and the Bloc Quebecois (BQ). Reformfinishedthe campaign with 18.7 percent of the vote and 53 seats in the House of Commons. How did Reform manage a mid-campaign rise in the polls? On the issues, Reform clearly staked out the most conservative positions of all the parties on both cultural questions (multiculturalism, immigration, and French-English relations) and role of government questions (size of the deficit and welfare state) before the campaign began. Given Reform's relative newness to the political landscape, before the campaign it was only on cultural issues that one would expect voters to be knowledgeable. The referendum campaign on the Charlottetown Accord in 1992 provided an opportunity for voters to understand Reform's locate on cultural matters. During the election campaign voters did learn the party's location on the deficit and welfare state and it seems clear that Reform's success can be linked to its positioning on the welfare state and deficit issues. As Johnston et al. (1994) observed, "only as Reform's anti-deficit commitment became clear did the party break out of its ethno-religious base" (15). It appears therefore that learning and priming of attitudes about the welfare state were the key to Reform's success. Sorting out which of these processes was at work is, therefore, of  2  central concern. Despite the importance of attitudes toward the welfare state, one also needs to consider the role of cultural attitudes. Did voters learn anything about Reform's positions on these issues? Did the campaign serve to remind voters of what they already knew? As we will see, coverage of Reform in the television news did emphasize the party's position on these issues late in the campaign when, in fact, Reform support actually declines. If a campaign can matter in different ways, then understanding campaign effects must also consider two subsidiary questions. How does the news media cover election campaigns? And, how do voters incorporate new information into their perceptions and evaluations of parties? While the latter follows from the fact that we would not expect voters to be passive in reacting to the information in the news, the former takes the content and media's role in the production of news seriously. In other words, sorting out persuasion, priming and learning requires an understanding of the generation and reception of campaign messages. The news media are a critical pivot in understanding campaign effects and processes. The news is obviously a key determinant of the availability of information or cues about the political conflict for voters, but the media are also important because of their independence. The news is a product of the interaction between political elites, media actors, and the audience (Cook 1996; Just et al. 1996). While news coverage is neither solely determined by candidate action nor by independent decisions by the media, the effect of the media can shape the content of the news in two ways. First, the media can focus attention on some things rather than others (gatekeeping). Second, the media can tell stories from a particular perspective (framing). The dramatic change in vote intentions across the 1993 campaign suggests that the information provided to voters changed. As Converse (1962) and Zaller (1989) have previously found, the flow of political information about the parties will be critical to observing  3  campaign-induced changes in party support. An analysis of the coverage of the parties during the campaign suggests that just such a change occurred. More importantly, perhaps, is the evidence that news decisions were independent of actual changes in the political competition. At the beginning of the campaign, one of the fundamental questions was whether the news media were going to treat the Reform party as a serious contender for political power? After ignoring the party for the first couple of weeks of the campaign, the media did treat the party as a serious player in the campaign drama. The media began to increase the amount of press the Reform party received and this increase in attention coincided with an emphasis on Reform's position on the deficit and welfare state. Later, coverage of Reform focused on the party's position on cultural issues like multiculturalism and immigration. These changes in available cues are the likely reason for the nature of Reform's successful mobilization. For campaigns to have effects voters must respond to the information that campaigns make available. As Converse recognized some time ago with respect to floating voters, the likelihood of a voter defecting to another party is conditional on the information flow of the campaign and the voter's motivation to attend to that information (Converse 1962). It is therefore necessary to conceive of the impact of media cues in terms of a model of attitude change. Building on the early work on attitude change and the recent work by Zaller, a twomediator model of media influence is developed here. Media influence is understood to be a function of both the likelihood that the voter receives the message and the likelihood that the voter accepts it (McGuire 1969; Zaller 1989; Zaller 1992; Ansolabehere and Iyengar 1995). The analysis based on the two-mediator model reveals substantively important individual differences in response to the campaign. Not all voters learned Reform's position on the welfare state and not all voters were equally more likely to support Reform as the campaign  4  evolved. Voters responded to the availability of particular kinds of cues about Reform in a manner consistent with their underlying attitudes on those issues and their awareness of the cues in the news. The Reform success and campaign dynamics are placed in the context of both a multiparty competition and a two-party competition for conservative voters but the analysis presented here is not meant to be an explanation of the 1993 election outcome. To the extent that the factors which account for Reform's rise also account for the other dynamics then some inferences about the election may be possible, but the point is not to focus on which factors made Reform successful. Instead, Reform is used to demonstrate the macro- and microfoundations to campaign change. In fact, it is insurgents, like Reform, which offer the possibility for observing a campaign process because their support is less contingent on history and partisan loyalties, except to the degree to which mobilization must break through the inertia of party commitments and institutional norms. Another way to think about it is that Reform's presence disrupted the existing equilibrium and allowed for the observation of the interaction between parties, the media, and voters. This is most apparent in the analysis of the relationship between Reform's mass support and its attention in the television news. Theoretically there are good reasons to believe that causation runs in both directions such that coverage and support tend to establish an equilibrium. In the absence of an insurgent, the interested researcher is forced to observe the relationship and speculate on its underlying causes, but is given little insight into the political process that generates it. Reform's presence forces the system to adapt and seek out a new equilibrium, which can be observed and identified.  5  One further qualification is that the emphasis in this study is on the news media, particularly television, rather than advertising or other direct communications between candidates and voters. This limitation is possible for two reasons. The first is that the Reform party ran no major advertising campaign so there were almost no pro-Reform messages in the paid media. The second is that the media are the primary source of information for voters and insurgents seem to be particularly dependent on the news media. For example, others have observed that the key to Perot's successful campaign in 1992 was the news press he received rather than his paid advertisements or appearances on daytime talk shows and Larry King Live (Zaller with Hunt 1994; Zaller and Hunt 1995). Reform's success in 1993 should be no different. The Reform Insurgency Reform's insurgent qualities are easily established. The prevailing view is that Reform was created as a reaction to the failure of the Conservative party to fulfill the "conservative" agenda that was promised in the 1984 election campaign. The party thus began as a direct challenge on the right. The challenge began and was centred in the West, particularly Alberta, where disenchantment with the Conservative government was particularly strong. By 1993, the Reform party was running candidates in all provinces except Quebec. While clearly capitalizing on disenchantment from the political process and traditional parties, Reform has been unique in carving out distinct and even extreme positions on a number of issues. During the campaign, Reform released a very ambitious deficit reduction plan, pledging to reduce the deficit to zero in three years, and for the most part acknowledged the implications in terms of reduced public spending on social programs. In contrast, the Conservative party was proposing a much more modest five year plan that sought to maintain  6  social spending. The tendency to challenge the political elite is also clearly evident in Reform's position against the Charlottetown Accord, which represented the "elite" consensus of the time. This position reflected the party's antipathy toward recognition of Quebec's distinctiveness and federal policies of ethnic and linguistic tolerance reflected by multiculturalism and bilingualism. While Reform managed to win a by-election and an election for an Alberta Senate seat in 1989, popular support was low for most of the period leading up to the 1993 election. Reform's breakthrough was a campaign phenomenon. Pre- and early campaign polls showed Reform trailing all the major parties including the New Democratic Party (NDP), but Reform finished in third place overall. For the first two weeks of the campaign, Reform was an inconsequential player in the electoral game. What little press Reform did get focused on its uniqueness and contrast to the emergence of the Bloc Quebecois (BQ) in Quebec. As the Conservatives faltered, Reform was the main beneficiary, ending the campaign with its 18.7 percent of the vote. The campaign thus produced a large Liberal majority, an opposition party - the BQ - committed to splitting up the country, and the virtual annihilation of two of Canada's traditional parties, the NDP and Progressive Conservatives. The Reform party is an interesting phenomenon from a number of perspectives. From a party system perspective, the emergence of the Reform party and the Bloc Quebecois suggest a realignment of Canada's party system (Clarke and Kornberg 1996; Johnston et al. 1996a). From a historical perspective, Reform is the latest in a long history of parties formed in Western Canada with a regional identity that have competed at the federal and provincial levels (Flanagan and Lee 1991; Lipset 1950; Macpherson 1953; Morton 1950; Sigurdson 1994; Young 1969). Reform's populist credentials, such as they are, also invite an analysis of the  7  party from the perspective of its orientation to the system and voter alienation (Harrison 1995; Harrison et al. 1996; Harrison and Krahn 1995; Laycock 1994; Patten 1993). Others have sought to come to terms with the broader historical development and ideological appeal of the party (Archer and Ellis 1994; Flanagan 1995; McCormick 1991).  1  While these perspectives inform the general context of Reform's 1993 campaign, they tell us little about how Reform managed its electoral mobilization, especially given the apparent importance of the campaign. The analysis undertaken here places the emphasis on the campaign as the site of considerable dynamic change in knowledge about and support for the Reform party. The formation and historical positioning of the party before the campaign began are taken as part of the background context and discussed more fully in Chapter 2. Campaigns The study of campaigns has undergone a renaissance in recent years. A recent conference held at U.B.C. (1997) devoted its entire deliberation to exploring the question of whether and how campaigns matter, and to the question of how to study them. Numerous books and articles have also sought to offer evidence on the question (Finkel 1993; Finkel and Schrott 1995; Gelman and King 1993; Holbrook 1996). In part the recent effort has emerged out of a paradox. Journalistic and some academic analyses often claim that the outcome of an election was the result of some element of the campaign: advertising; debate performance; or campaign strategy. In addition, political parties and candidates exert considerable effort in trying to influence the outcome and this effort would be irrational if it did not matter. Despite the persuasiveness of some of these claims, they run counter to many explanations of election  1  This list is not meant to be exhaustive and authors often contribute more than one understanding of the Reform party. In addition, there has also been a number of non-  8  outcomes which stress factors in place before the election. The conventional view is that campaigns do not matter (Bartels 1992; Gelman and King 1993; Holbrook 1996), but there are good reasons to believe that this claim is overstated and may be more a product of the emphasis on presidential elections, which are the object of most studies, rather than campaigns in general (Alvarez 1997b). The following two sections review the "minimal effects" claim and the Canadian evidence before turning to a discussion of a general model of campaign effects.  A Consideration  of the "Minimal  Effects  " View of  Campaigns  The standard claim of no effects is based on the ability to predict election outcomes in the United States with remarkable accuracy months before the campaign begins which implies a minimal role for the campaign in structuring electoral choices (Gelman and King 1993; Holbrook 1996). In fact, a considerable literature has emerged around the question of developing an accurate forecast of presidential elections before they take place. Forecasting models vary considerably both in terms of the independent variables included in the analysis and the dependent variable being predicted. For example, some predict the two-party share of 2  the vote while others predict the share of the vote in the states. Almost all of the models include economic indicators, pre-campaign measures of opinion (trial heat polls; government approval; or partisan strength), and some indicator of incumbency to predict the outcome (Rosenstone 1983; Abramowitz 1988; Campbell 1992; Lewis-Beck and Rice 1984; LewisBeck and Rice 1992). One of the simplest models is supplied by Abramowitz (1988) using only pre-campaign (November) popularity, GNP  and a variable indicating two-term  academic treatments of the Reform party and its leader, Preston Manning (Dobbin 1991; Manning 1992; Sharpe and Braid 1992)  9  presidency. Repeated estimations of a model based on all of the elections but one successfully predicts the excluded election vote with an average error of only 1.1%. Although the models are less successful predicting elections than explaining them, Rosenstone (1983) and Campbell (1992) using two different models were able to account for 93% and 85% respectively of the variance of state presidential votes in recent elections. It appears therefore that "Election outcomes and voting behaviour are easily explained with just a few variables, none of which are related to the campaign" (Holbrook 1996: 43). Given the reasonable success of forecasting, the possible role for campaigns is said to be quite limited. The conditions in place before the campaign begins determine who will win the election. In a version of this argument, Holbrook (1996) argues that for every election there is an equilibrium outcome that the campaign moves toward. While there can be fluctuations during the campaign, the campaign will only result in a minimal amount of movement from the equilibrium point by election day. The claim is not that campaigns don't matter; they matter, however, only to the extent that they help ensure that the equilibrium outcome emerges. Effects, such as they exist, are understood in terms of the inevitable result of campaign activity and campaign shifts are not viewed as fundamental alterations in the likely outcome. Despite the importance of pre-campaign forces in his own findings, Holbrook does allow for the possibility that the campaign will affect the outcome. In those situations when the national context does not heavily favour one candidate (i.e. close elections), the campaign can matter. Finkel also recognizes a potential for the campaign to matter even if presidential elections are rarely characterized by this kind of change. But if some campaigns matter while  2  In addition to the social science attempts at forecasting discussed here, forecasting models  10  others do not, then it becomes important to understand how conditions are able to predict electoral results so well and whether the equilibrium is truly stable. That is, what does it take for a campaign to produce an unpredictable result? It is conceivable that campaigns do not inevitably produce a given result. This is the lesson that can be derived from the U.S. primary experience (Battels 1988; Brady 1993) and recent Canadian experience (Johnston et al. 1992; Johnston et al. 1996b). Campaigns may either change people's minds, the dimension of choice, or perceptions of the candidates thus offering the possibility of multiple rather than unique equilibria. The election context may powerfully shape the strategic landscape that parties face going into the campaign, but thenown activities can be instrumental in shaping which of a number of outcomes emerges. From the multiple equilibria perspective, news media coverage, party strategy and events in the campaign can influence which equilibrium emerges whereas from the single equilibrium perspective, the campaign brings voters to the equilibrium. The general approach to explaining the predictability of presidential elections involves an emphasis on the activation function of campaigns. A campaign does not convert or persuade voters but activates and reinforces pre-existing attitudes and predispositions. Campaign effort can remind voters about what a party stands for and thereby motivate respondents who share similar concerns to vote for that party. The discourse of campaigns may be more about interpreting contemporary political issues into party politics than about trying to change people's minds. That said, what makes it possible for activation to be the primary story for most voters most of the time but not the story in some campaigns, especially in multi-candidate competitions? A consideration of the evidence suggests that activation is more likely in  and predictions more generally are made by the news media and campaign strategists. 11  presidential elections because of the structure of the choice and the minor chance that the campaign will reveal new information about the candidates. The first systematic study that directly tested whether campaigns mattered, among other things, was the first of the Columbia studies. Lazarsfeld and his colleagues using a panel study found that very few people changed their minds during the campaign (1948). Between May and October of 1940, only 14% of voters were converted from one vote intention to another. Since not all of the conversions were in the same direction, the net effect of the campaign on the outcome was quite small. More important were other types of campaign changes. The authors concluded that the campaign was largely concerned with activating or reinforcing pre-existing preferences rather than changing people's minds or conversion. The Michigan studies which followed didn't look for campaign change (Campbell et al. 1960; Campbell et al. 1966). In these studies the emphasis is on long term party attachments which were activated by political campaigning. In fact, the concept of the "normal vote" which refers to the distribution of partisan identifiers and therefore the relative advantage of one party over the other is the natural extension of the minimal effects conception of the importance of campaigns (Converse 1966). The normal vote does not preclude the possibility of campaign effects since the normal vote is a benchmark or starting point for understanding the outcome. Nevertheless, the notion of an underlying partisan advantage that is the basis for the normal vote is based on the idea of a fundamental stability of electoral results. Although party identification continues to be used to indicate long term and stable loyalties, there is considerable evidence that suggests that party id is susceptible to short-term (Fiorina 1981), even campaign (Allsop and Weisberg 1988), forces. A strong  association between party identification and the vote is not necessarily evidence of minimal campaign effects. There is some contemporary support for the idea that campaigns continue to activate group and partisan identifications. From this perspective a campaign serves to reinforce and remind voters of their party loyalties and the reasons for such loyalties (Finkel 1993). In one formulation of this effect, campaigns are said to matter to the extent that they provide information necessary for voters to express their enlightened  preferences  (Gelman and King  1993). That is, voters are provided with the means to figure out what the election is about and for whom they, given their attitudes and social situation, should vote for. For example, as the campaign nears its conclusion, the weights that voters attach to fundamental variables (ideology; race; income; education) are clarified (Gelman and King 1993). In other words racial minorities figure out which candidate is closer to them as a group. As Finkel concluded about recent presidential campaigns; "what occurred ... was not minimal in the sense of simple reinforcement of preferences, but predictable movement by the electorate toward casting votes in accordance with these underlying conditions and their accompanying political predispositions" (1993:18). Campaigns can also activate a third variable; the respondent's perceptions of incumbent performance. This approach is consistent with the fact that prediction models in the United States are based on variables such as economic growth and presidential popularity. It is also supported by evidence concerning the strong link between economic performance and electoral support for the president (Erikson 1989; Markus 1988). For example, in the 1992 U.S. presidential election the campaign reminded voters about the poor economic performance  13  under President George Bush, contributing to his eventual defeat (Hetherington 1996; Just et al. 1996). Obviously neither the partisan activation nor the incumbency explanations pertain to primary campaigns. While it could be objected that primary campaigns are by definition not electoral contests, if the goal is to generate a general theory of campaigns, then any model about the individual-level processes of campaigns should be able to account for the dynamics of primaries. There does appear to be significant campaign effects in the non-partisan primary campaigns (Battels 1988). Most obvious are the findings concerning media-induced momentum affects for primary candidates (Battels 1988; Shapiro et al. 1991). In a sense, primary season offers the potential for multiple equilibria. One cannot predict who will emerge at the convention as the party's choice based only on information available before the primaries begin.  Canadian  Evidence Given the multiple-party nature of Canadian races, it may be more likely that one will  observe campaign effects. Even with the three competitive parties before 1993, the potential always existed for voters to move from the Liberals to the NDP if exposed to the right stimulus. In fact, it is often argued that party identification is weaker in Canada than in other countries thus offering a greater pool of weakly attached voters who could be moved by the campaign (Clarke and Stewart 1984; Clarke and Stewart 1985; Johnston 1992). In addition, 3  the short election campaign means that voters can enter the campaign with little information  3  Johnston's (1992) experimental work fundamentally challenged the previously held belief that party identification is merely a substitute for vote choice in Canada and offered Canadian researchers a new question wording, which is used in the 1993 election study. The effect of the question wording innovation is to suggest that Canadian-American differences would be reduced, but not entirely eliminated. 14  about the political leaders, especially those leaders who were selected in the months preceding the election. The Canadian experience does not provide unambiguous support for the no effects thesis. The supposed impact of the debates in the 1984 campaign (Fletcher 1988; Lanoue 1991) suggest that campaigns do matter. Evidence from other election debates is not, however, supportive (LeDuc and Price 1985). Recent Canadian election studies (1988 and 1992-3), which have made use of an empirical tool - the rolling cross-sectional survey - capable of observing campaign change when it occurs have identified substantively important, non-trivial campaign effects. In the 1988 Canadian election campaign vote intentions shifted first toward the Liberals and then back to the Conservatives at the end of the campaign (Johnston et al. 1992) . The Liberal party was able to raise problems with the Free Trade Agreement (FTA) and use the Mulroney character issue to its advantage, especially in the debate, but the Conservatives were able to move voters back by a defence of the FTA and by attacking Turner's own character in the latter part of the campaign (Johnston et al. 1992). Further evidence comes from the campaign study of the 1992 Charlottetown referendum. During the campaign, support for the Accord underwent a dramatic shift toward No when Pierre Trudeau announced his opposition (Johnston, Blais, Gidengil, and Nevitte 1996b). In both cases, the evidence of campaign-induced change is convincing. If 1988 and 1993 involved clear campaign effects, the results from the 1997 election are less clear. While there are a couple of campaign events which triggered short-term changes in party support (debate and Reform advertising), the effects of both eventually decayed (Blais et al. forthcoming). There was, however, a gradual increase in Reform support (Blais et al. forthcoming) accompanied by voter learning about Reform's positions (Fournier n.d.). While  the net impact of the campaign on vote intentions was modest, the campaign did prime attitudes about French Canada and thereby put in place the conditions for strategic voting to affect the outcome (Jenkins 1999). In part, the 1988 Canadian election could be viewed from the partisan activation perspective since Turner's success was mainly among Liberal defectors (Johnston et al. 1992). c  These voters responded to real campaign events and coverage and in the absence of these events, the debate in particular, may never have re-evaluated Turner. The problem with this perspective is the campaign dynamics; voters moved to and then away from the Liberals. If partisan activation was the only thing occurring, then there would be a simple linear growth in Liberal support as Turner convinced partisans to come on side. Partisan activation cannot provide an explanation for the 1993 election. In 1993, the Reform party had only about 4% of the identifiers, but received almost 18% of the votes on election day. Perhaps more importantly, the party actually gained identifiers as the campaign progressed which further undermines the idea of party identification as a stable individual attribute (Allsop and Weisberg 1988). Party identification may serve or appear to serve as a basis for mobilization in some cases, but it cannot explain significant changes in party support or provide a basis for a general theory of campaigns. The activation of attitudes concerning incumbent performance makes sense in presidential election campaigns but it is not generalizable to multi-candidate races because it assumes that there is a clear alternative to the incumbent. In multi-party contexts reminding voters about the nature of economic conditions (or other aspects of incumbent performance) provides no information about where voters should cast their vote. Researchers interested in examining the impact of economic performance on election outcomes have been limited to  16  either relating economic conditions to incumbent vote (Carmichael 1990; Happy 1989) or to a single party vote share (Nadeau and Blais 1993;  Nadeau and Blais 1995). In 1988 anti-  Conservative voters had more than one place they could go so there is no clear prediction. Similarly, voters who moved to Reform in the 1993 election were certainly abandoning the incumbent, but moving to Reform was not the only alternative and if incumbent performance was the critical evaluative dimension, voters should have simply moved to the Liberals. The simple model of Liberal vote share proposed by Nadeau and Blais in 1993 does forecast the 1993 vote share reasonably well (45 as compared with an actual value of 41 percent), but it offers nothing in terms of an explanation for the other dynamics. Canadian experience with respect to campaign effects is broad and diverse. In at least two of the last three elections there was enough movement in aggregate intentions so as to ultimately affect who would form the government. Even when movement was smaller, there is evidence of campaign related changes. The question is how to understand the possible ways in which the campaign can matter.  A Model  of Campaign  Effects  Some elections campaigns effectively do no more than mechanically produce the outcome that could have been predicted months before the campaign began. In effect, the national context (economic conditions; international events; etc.) in which parties compete can limit the strategic possibilities of the parties. Activation as it is normally understood cannot, however, be the whole story. It does not account for observations of campaign change in Canada and abroad. What is necessary is a framework for understanding how the activities of parties and the media can affect the outcome which emerges from campaigns. The following  section develops such a framework based on the idea that campaigns involve three types of processes: (1) activation; (2) learning; and (3) persuasion. As discussed earlier, activation refers to the process which by certain pre-existing attitudes are made electorally relevant or salient. Here, activation is not limited to party id and incumbent performance but is understood more broadly in terms of the more general values and ideological predispositions of voters. Readers may recognize that in this form, activation is a form of priming (Bartels 1997b; Iyengar and Kinder 1987; Miller and Krosnick 1996) such that while there is a wide range of potentially relevant attitudes and issues only some of them become electorally relevant as a result of public concern, party strategy, and/ or news coverage. The campaign can matter because during the campaign the salience of certain issues or ideological dimensions changes or because parties chose to fight on some issues rather than others. While there is logic to party strategy, there is no question that different issue dimensions appeal to different kinds of party supporters (Johnston et al. 1992; Shafer and Claggett 1995). The fact that the 1988 election was fought over free trade and not national unity has important consequences for the outcome. The question is, is it possible to predict except perhaps when the economy is bad - the important issues or the strategies of the parties'? Learning refers to the process by which voters gain information. Many observers claim that the campaign does provide voters with the ability to make informed decisions. Although the campaign did not erase all differences, Fournier found that as the election day neared high and low aware respondents became more likely to base their vote intentions on the same considerations (Fournier 1997). The lesson of this analysis is that the campaign allows the less aware to get up to speed on the choice and is consistent with the low information rationality perspective on voter choice (Lupia 1992; Popkin 1991; Sniderman et al. 1991). Knowledge or  18  lack of it has been shown to be especially important for the success of candidates in multicandidate primaries (Brady and Ansolabehere 1989), and the logic applies equally to Canadian experience. One would be surprised to find that voters had attitudes about a party activated i f those same voters did not know where the party stood on the issues. It is therefore possible to distinguish learningfrompriming. Finally, there is the possibility of persuasion. Going back to the original studies, the expectation is that campaigns will not primarily function by changing people's minds. Nevertheless, there is some evidence that campaigns change the number of partisans and this is prima facie evidence that the campaign could matter through persuasion (Allsop and Weisberg 1988). More obviously, the 1988 Canadian campaign appears to have involved a persuasive situation. Over the campaign voters as a whole became less supportive of free trade and then became more supportive of free trade as a function of Conservative party arguments that free trade would not threaten social programs (Johnston et al. 1992). While it appears to be persuasion, there is of course an element of learning. Obviously the three processes are interrelated and when campaigns are considered through these three processes the wide range of observed effects and non-effects can be understood. Given that candidates enter the official presidential campaign period so well known because of extensive non-official campaigning and the tendency for news coverage to be balanced (Gelman and King 1993), is it any wonder that campaigns have only modest effects on the final distribution in presidential elections? Where information is less available effects appear to be of a larger magnitude. A n added benefit is that the three processes coincide with the apparent way in which parties and candidate campaign in terms of stressing issues that are likely to advantage the candidate's party (Budge 1993; Petrocik 1996).  19  The study of campaigns clearly motivates the analysis of the Reform party in the 1993 campaign. While conditions were certainly favourable for an insurgent, it could not be said with any confidence before the campaign began that Reform would manage to mobilize voters to its cause. Did the mobilization proceed through learning, priming or persuasion? The dramatic reversal of Reform fortunes during the campaign runs contrary to conventional American - wisdom, but it less surprising given recent Canadian experience and the experience of U.S. primary elections. What is necessary is to consider the two remaining elements of campaign change: the news media and the voters. The Media and Elections  To the extent that a campaign generates one outcome rather than another or even helps ensure that voters make reasonable choices, we should expect the media to be pivotal. A n election campaign in contemporary Western countries is a media event. Voters certainly learn from the mass media about election campaigns in the American context (Battels 1993) and it is media attention that drives the activation of political preferences in Finkel's (1993) analysis. Further evidence of the impact of media coverage is available from the 1988 Canadian election. Mendolsohn (1994) observed priming effects among those who reported high media exposure, and changes in a number of media variables were related to evaluations of Turner and support for the Free Trade Agreement in 1988 (Johnston et al. 1992). It is the media both paid and unpaid that serves as the primary conduit between candidates and voters supplanting the party organization and party press of previous periods (Carty 1988). The focus here is on the unpaid media, especially television, which has become the most important source of news for voters. Two questions dominate our interest in news: (1) what is the dynamic potential of news coverage? And, (2) how does the news cover  20  insurgent parties or candidates? The question of effects of the coverage is discussed in a subsequent section.  The Dynamics  of News  Coverage  Considerable research effort has been expended, especially in the U.S., on assessing the manner by which the news media cover election campaigns. What is clear from the previous 4  work is that both the amount and nature of news coverage of parties and candidates is related to the horserace. That is, who is ahead and why? The second thing is that there are good reasons to believe that the news media have an independent effect on the nature of the news. These two factors mean that as the campaign evolves changing party fortunes and changing perceptions of the campaign on the part of journalists can change the nature of the stories to which the audience is exposed. Coverage takes on its particular form because of the process that produces it. Recent work on news coverage emphasizes the interaction between politicians or candidates, the media and the public in the construction of news (Cook 1996; Just et al. 1996; Zaller 1997). While the politicians are dependent on the news media to convey their messages, they try to ensure that the message that is communicated is in their interest by staging media events and spinning news issues in a favourable light. News media personnel, however, do not view their role as mere conduits of political messages from candidates and parties; journalists also want to avoid being manipulated by politicians (Cook 1996; Kerbel 1994). Reporters are manipulated if voters are moved by rhetorical claims made by politicians and distributed by the news media  4  See for example, (Brady and Johnston 1987; Fletcher 1988; Hallin 1992; Lichter et al. 1988; Lichter and Noyes 1995; Robinson and Sheehan 1983; Semetko et al. 1991; Semetko and Canel 1997; Soderland et al. 1984; Wagenberg et al. 1988). A number of researchers have been critical of the news media because of the pattern of coverage either because it does not serve democratic ends (Patterson 1989; Patterson 1993) or because its cynical portrayal of the political process contributes to political alienation (Cappella and Jamieson 1997). 21  that are false or misleading. One of the ways to avoid being manipulated is to not accept the definition of the campaign as provided by the candidates. In fact, the desire to avoid being 5  manipulated led the media to more actively interpret the campaign for voters in the 1992 American election (Lichter and Noyes 1995). "An understanding of the news media's version of reality begins with a recognition that reporters must have a story to tell" (Patterson 1993: 96). Since the outcome of an election campaign is a party being selected to form the government, the natural story line is one that places each day's news in terms of the likely outcome. This leads to an emphasis on who's ahead and why. The dominant theme in this literature is the news media's tendency to focus on leadership, strategy, and the horse race aspects of the campaign at the expense of substantive issues (Hallin 1992; Lichter et al. 1988; Sigelman and Bullock 1991). This observation applies equally to Canada as to the United States (Fletcher 1988; Wilson 1981). Issues are by nature static (Robinson and Sheehan 1983) so it is not surprising that media coverage tends to be overshadowed by coverage of the more dramatic elements of the campaign. Even when the news media are "covering issues", they tend to do so from a strategic frame and thereby introduce the horserace into issue coverage. The 1988 news coverage about the Free Trade Agreement - a campaign that was often presented as a debate about free trade was framed in terms of leadership, credibility and the horserace rather than in terms of the appropriate trade policy for Canada (Mendelsohn 1993). This suggests that what issue information that voters receive will be strongly influenced not only by selective partisan  5  Reporters can also use other approaches. The "ad-watch" stories, in which reporters deconstruct and attempt to validate partisan ads, of recent U.S. presidential campaigns is an example of an attempt to prevent parties from manipulating the media and the public (Ansolabehere and Iyengar 1996; Jamieson 1992; Lichter and Smith 1996). Similarly, during the Canadian 1993 election the television news programs submitted some of the party claims to a "reality check." 22  emphasis on issues, but also by news decisions about what issues are newsworthy. That is, what issues fit into the reporter's perceptions about what the campaign is about. Reporters find out what is happening in the campaign and interpret these events through their own understandings of the election and their perception of the interests and needs of the audience. While reporters are influenced by their own perspective, they are constrained by the need to be, or at least the need to appear to be, objective and the general pack-like qualities of journalism. Expectations about how the parties will fare on election day based on polls or other information, will thus be critical to the process of covering parties. Zaller has labelled this the "rule of anticipated importance" because the news media allocate coverage of candidates and issues "in proportion to its marginal value for shedding light on future developments" (Zaller 1997: 22). For example, the perception that the National party was irrelevant to the outcome of the 1993 campaign likely motivated the news media to ignore the party. The amount of coverage that parties receive and not just the nature of the issue coverage is related to the horserace. While the allocation of coverage to parties is multi-faceted and taken up more fully in Chapter 3, parties tend to receive coverage in proportion to thenexpected electoral support. The media are neither likely to ignore a popular candidate nor heap coverage on an unpopular coverage for two reasons. The first is that to do so would violate norms of objectivity. The second is that candidates who are doing well are the ones who are likely to win and therefore the ones people need to know about to make rational decisions (Zaller 1997). While the news is the product of a complicated process involving media and political actors and economic interests, the important thing for our purposes is to understand how the  23  nature of coverage can change. Clearly the most important impetus for change is a change in party support. If the horserace takes on an interesting new character, then news stories will have to incorporate and deal with the new information. It is often noted that George Bush went from the frontrunner, to the troubled candidate, to the solid candidate in news reports during the 1992 primary and campaign season (Patterson 1989). This means that the qualitative nature of the news can change as reporters understand the story of the campaign to be different. As we will see, this has important implications for understanding the 1993 campaign because of the collapse of the Conservative party and the presence of a right-wing alternative in the Reform party.  Media Coverage of Insurgent Parties By their very presence new or insurgent parties are disruptive. How then does the media deal with insurgents and what are the consequences of this treatment? The news media employ newsworthiness as the critical determinant of press coverage. In particular, Zaller (1997) has argued that the media base coverage on the "rule of anticipated importance." Since insurgents are particularly dependent on news attention to convey their positions to voters, news decisions are critical for successful mobilization. Analyses of news treatment of insurgent candidates or new parties are sparse. Three possibilities suggest themselves. The first is that insurgents will either be ignored or given bad coverage because they represent unacceptable mainstream views. Robinson and Sheehan (1983) and others have noted that the media are selective in who they consider legitimate spokespeople and views (Hackett and Zhao 1998). A party that advocates the abolition of basic human rights or democracy itself may not be given news coverage. The obvious question is where the media draw the line. Given Reform's general antipathy toward the elite consensus  24  on such things as accommodation with Quebec, the media may have been biased against the party. In fact, there is no indication that Reform was punished despite its fairly extreme policy positions. The second is where the insurgent is the subject of news stories but the tone and framing of the stories emphasizes the unique aspects of the candidacy rather than the substantive importance. For example, an insurgent is likely to get coverage beyond what would be justified by his or her popular support by offering a unique or entertaining story of the campaign (Brady and Johnston 1987; Kerbel 1994). While this press is likely to be positive, it is unlikely the kind of press that persuades voters to switch their vote intentions because very little of the information may be related to the issue positions of the party relative to the rest of the electoral discourse. For example, if Reform only received coverage about its national unity position because it is interesting vis a vis the BQ, then Reform would be an interesting side story but would be unlikely to attract much voter interest as long as the issue discourse of the campaign focused on jobs and the deficit. Voters would be uncertain about the party's position on these important issues. The third is that the insurgent will get serious press attention both in terms of the amount of coverage and the content of that coverage. A new party which is stressing themes, which coincide with reporter's perception of the mass public, will more likely be treated seriously because reporters understand the party's potential for the outcome. A party can also earn media coverage by doing better than expected. In effect, the same factors that influence coverage more generally will apply to insurgents. As Zaller (1997) notes, if the media anticipates that a party or candidate is going to do well, they will give that candidate more news attention.  25  Like all candidates or parties, insurgents can expect bad press when it becomes more visible and viable. For example, in the pre-campaign nomination period, Perot's rise and subsequent decline in public support can be attributed to his coverage in the conventional press (Zaller with Hunt 1994; Zaller and Hunt 1995). Conventional press coverage of Perot resembled the "boom-and-bust pattern" that is typical of relatively unknown candidates; candidates begin with good press but as they become more electorally viable the tone of their press becomes more critical. The increase in bad press follows from a more critical news media and competitors who are more likely to see an insurgent as a threat when polls show it to be more viable (Hagen 1996). The success of insurgents is likely to be particularly dependent on the media, especially free media provided by news (Semetko 1989; Zaller with Hunt 1994). For insurgent parties, coverage is critical because voters begin the campaign lacking information about the party. Bartels' (1986) insight that voters dislike uncertainty therefore has particular relevance when considering the possible success of insurgent parties. Only insurgents like Ross Perot, who have access to large personal sums of money, are likely able to circumvent the news media and appeal directly to the public. Parties like Reform are unlikely to generate financial support without considerable public support which itself presupposes awareness in the mass public. In fact, the Reform party chose not to run an advertising campaign during the 1993 campaign leaving the news media as the only effective channel for information about the party to the mass public. It seems unrealistic that an insurgent could mount a successful mobilization effort in the absence of media exposure. It might be possible to generate a grassroots organization capable of nominating candidates but the transition from such a core organization to a party with  26  meaningful electoral support would require the news media. Again, Ross Perot's experience is instructive. While it is common to point to the numerous volunteers who managed to get his name on the ballot in all of the states, Perot failed in his attempt to circumvent the conventional news media in the 1992 American presidential campaign (Zaller and Hunt 1995). While Perot's unmediated campaign was effective, "His problem ... was that the audience for unmediated communication in the United States is simply too small to sustain a successful presidential campaign" (Zaller and Hunt 1995: 120). There is no reason to believe that things would be different in Canada, especially when an insurgent, like Reform, does not use advertising. Summary  The analytic work on news coverage of elections suggests that the news media will emphasize the strategic and horserace aspects of the campaign at the expense of issues. When issues are covered they will be interpreted within a limited number of frames, especially a strategic frame. This means that an insurgent party like Reform needs to convince the news media that it is deserving of attentionfromwithin thisframe.If the news media think the party is irrelevant to the main contest, we would expect Reform to be treated like a novelty story. On the other hand, if the media are convinced that Reform is important, either on their own or because the party is doing better than expected, coverage will increase and, be more serious and substantive. Nevertheless, as an insurgent becomes more viable it is likely to be treated more like other parties or candidates. Demonstrating that news coverage changed is therefore central to observing campaign-induced changes in Reform support in the electorate.  27  Attitude Change and Persuasion While the significant change in the proportion of the electorate supporting the Reform party as the campaign progressed can be related to changes in news coverage at the aggregate level, the goal of the thesis is to understand the microfoundations of the aggregate change. This requires an understanding of individual susceptibility to the persuasive content of news information. The study of persuasion has a fairly long history in the social sciences going back to the pioneering experimental studies of war propaganda carried out by Hovland et al. (1953). In these early studies, the authors identified the three key aspects of a persuasive situation: the communicator; the communication (the message), and the audience (Cohen 1964). Despite the extensive work on attitude change and media effects generally, a fairly systematic review of the evidence found that the media had minimal effects on attitudes; reinforcement rather than conversion was the major effect of media exposure (Klapper 1960). Both experimental and survey research came to the same basic conclusions. The minimal effects thesis has dominated the study of media effects since Klapper reached his pessimistic conclusions. Twenty-six years later, McGuire concluded "demonstrated impacts are surprisingly slight" (McGuire 1986: 233). Research into agendasetting and priming (Iyengar and Kinder 1987; McCombs and Shaw 1972; Mendelsohn 1996; Miller and Krosnick 1996), which found effects of coverage on such things as public priorities, undermined the scope of the minimal effects claim but has not fundamentally altered its applicability for campaigns. Despite the weight of past findings, it is still possible for the media to have a large impact on attitudes. The key is to account for previous findings and develop a framework for observing media effects on attitudes when they occur.  28  Of the possible explanations for the lack of large media effects (McGuire 1969; McGuire 1986), two possibilities stand out. The first is the lack of variance. Because such things as presidential election campaigns tend to provide fairly balanced partisan coverage, there is not enough variation in coverage to produce massive effects (German and King 1993; Zaller 1996b). Fortunately, as will be shown in later chapters, there is significant variation in the flow of messages about the Reform party, which should allow for the identification of media effects if they are present. The second is that media effects researchers may be modeling the media effects process in the wrong manner, either by failing to consider respondent's ability to resist messages or by assuming that the effect of media exposure is a simple linear relationship (Zaller 1996b). In order to develop a model that allows for the identification of media effects, one needs to begin with a broad understanding of attitude change. The best starting place is Ho viand's five-step model of attitude change. The five steps, which follow the presence of a persuasive message, are: attention; comprehension; yielding; retention; and, action (McGuire 1969; Zaller 1996b). Since survey data do not allow one to distinguish each stage of the process, it is necessary to reduce our understanding of attitude change. The likelihood of being persuaded by a message can be conceptualized as a function of two mediating processes: the person's likelihood of receiving a persuasive message and his or her likelihood of accepting it (McGuire 1968; Zaller 1992; Zaller 1996b). In order to observe a change in opinion, the attitude change model first requires that the respondent be attentive to the persuasive message. A respondent who has the television on while the message is being carried, but is busy talking to family members would not be attentive and therefore would not go through the other stages of attitude change. The second  29  step is comprehension, which refers to whether the respondent understands the message. A number of characteristics of the receiver and the message can affect comprehension, including the complexity of the message, the relationship of the message to personal experience, and the cognitive abilities of the receiver. Steps one and two clearly relate to the likelihood that the message will be processed and understood. If the message is not received, then there is no attitude change and the process ends. While there is some debate about the appropriate measure of news media consumption, it is clear that some respondents are more likely than others to consume and remember information they received from the mass media (Brians and Wattenberg 1996; Price and Zaller 1993; Rhee and Cappella 1997; Zaller 1996a). Once received and understood, the receiver is not necessarily persuaded by the message. Citizens are not simply ciphers responding to mass media inputs. Some messages are accepted and acted upon, while others are rejected or discounted. In part this reflects classical processes such as selective reception or the reduction of cognitive dissonance (Festinger 1957). Joslyn and Ceccoli (1996) found that the effect of news attentiveness was dependent upon political predispositions. "Political predispositions, such as affective  partisan  ties, serve as filters, leading to patterns of selective exposure, attention and reception" (Schmitt-Beck 1996). It is also true that people who have a store of information should be better able to counter-argue new information reducing the persuasive impact of any particular message. The attitude change model is clearly compatible with the activation of underlying attitudes perspective on campaigns. There is, of course, some controversy about the degree to which people have stable attitudes (Achen 1975; Converse 1964; Luskin 1987). While this literature suggests that there will be individual-level variation over time to specific questions,  30  systematic analysis of multiple questions has uncovered a fairly coherent issue space in Western industrialized countries (Flanagan 1987; Kitschelt 1995; Shafer and Claggett 1995). For example, Shafer and Claggett (1995) identify two deep factors - culture/ national and economic/welfare - in U.S. opinion data. These two factors are similar to the two dimensional representations of the issue space in recent Canadian elections which have featured Canada's version of the national question (French Canada) and an economic dimension (continentalism dominated the economic dimension for party competition in 1988 and the welfare state dominated in 1993) (Brady and Johnston 1996; Johnston et al. 1992). These general attitudes should orient voters to the parties and mediate campaign communication. The stage that is effectively left out of the two-mediator model is the memory stage. This is important because of a more general debate about how people form opinions and judgments, with memory-based models competing against so called on-line models. Zaller (Zaller 1992; Zaller and Feldman 1992) both alone and in collaboration has proposed a sampling-based model of the survey response, which treats an individual response as created at the time of the question rather than being retrieved from memory. As a challenge to this model, an on-line processing model of opinion formation was proposed (Lodge et al. 1989; McGraw et al. 1990; Rahn et al. 1994). In point of fact both models expect that voters in coming to judgment about parties and candidates will consider new information. For both approaches new information is not omnipotent, as it must compete with the weight of past exposure and judgment. Consider a voter who is generally favourable to the Conservative  31  party who is exposed to mformation that casts negative light on the Conservative party. The two different models do not generate different predictions.  6  The two-mediator model of attitude change as developed here builds explicitly on the work of Converse, McGuire and Zaller. I hypothesize that Reform's breakthrough during the campaign took place among those most likely to get the message that Reform existed and those most predisposed to accept pro-Reform messages, especially those related to the deficit on which little of Reform was known before the campaign began. While the conceptual apparatus of the two-mediator model is not original, the application of it is unique in two ways. The first is that it is used to untangle the daily process of campaign effects rather than change over broader time periods. The second is that actual media data are entered into the modeling exercise so that change in attitude can be linked with change in the actual content of the messages. A n additional aspect of the application of the two-mediator model is that it is not applied solely to strictly persuasive situations in the normal sense. The effects of media coverage are understood to be a function of the two mediators even when the process at work is primarily learning and priming. The reason for this application is that information that confirms one's existing opinion of the party is irrelevant and will therefore be ignored. Information about Reform's position on the welfare state can only affect pro-welfare state people who had the misperception of Reform's true position. Since support for Reform was  6  In fact the evidence for the on-line models is based on whether respondents can recall the reasons that produced their judgment. Since people generally do a poor job recalling the reasons for their evaluation, it is concluded that people must forget or discard the information they used once they have updated their ruruiing tally. While this evidence suggests potential projection related problems associated with various kinds of vote models, especially those which make use of likes and dislikes questions, it does not offer a different way of approaching the campaign. In fact, recent evidence suggests that while evaluation of a single figure proceeds on-line, choice processes require memory (Redlawsk 1998). 32  very low early on, there was not a large pool of potential respondents who could be influenced by the new information.  A Method for Uncovering Campaign Effects The rermining question is how to go about identifying the micro- or individual-level model of attitude change and the hypothesis about the impact of dynamic changes in information flow during a campaign. What is necessary is data on both individual preferences and news media content. Both kinds of data, must of course, be collected in such a way that dynamic changes in one can be linked with changes in the other. This suggests a combination of survey data and media content data.  7  The survey approach encompasses a variety of research designs. Researchers in the survey tradition can make use of cross-sectional surveys taken at a single time; panel surveys in which a respondent is interviewed at more than one time; and, rolling cross-sectional designs in which small samples are interviewed at different times. In addition, more than one crosssectional survey can be analyzed at the aggregate level. The cross-sectional approach allows one to compare the attitudes or behaviour of survey respondents who report high exposure to the mass media with those who report low exposure (Zaller 1991; Zaller 1992). The assumption is that people who are exposed to more news will be more likely to be affected by the coverage. Since most surveys are based on a sample collected over only a few days, the traditional survey does not enable one to observe the direct effects of changes in the media on  7  The alternative would be to use an experimental approach. While an experiment allows for a clear conclusion about the effect of media messages in the case of a media study, it sacrifices realism by removing the test subjects from their natural setting. The use of experiments in political science has become more common in recent years, especially in combination with other research methods (Ansolabehere and Iyengar 1995; Iyengar and Kinder 1987; Norris and Sanders 1998), and has even been used to simulate campaign effects (Lodge et al. 1995).  33  individuals because there is no variation in the exposure. This makes it less useful for our purposes. The two remaining types of surveys are more amenable to dynamic research designs. The panel study is specifically able to identify a change in a respondent's attitudes or evaluations between two time periods; a change which can be attributed to persuasion of some sort. A panel design is used in a number of campaign studies (Battels 1993; Finkel 1993). There is, of course, a danger that some of the difference is a product of chance or more importantly panel conditioning (Bartels 1997b). While the panel design is useful, most panels have a considerable lag between the two waves of the survey. This makes it difficult to link the data directly to the ebb and flow of the campaign. The rolling, cross-section more easily identifies trends in public opinion because the key feature of the design is that a small random sample of respondents is interviewed every day of the campaign. Since the day of interview is a random event, any chance in the aggregate opinion beyond that which could be a product of chance, can be attributed to real campaign events and changes. The public opinion data, discussed more fully in Chapter 2, is actually a combination of a campaign rolling, crosssection and a post-election panel which allows for both types of analysis (Brady and Johnston 1996). The primary use of content analysis data of the news media is to say something about the quality or nature of news coverage of politics (Lichter et al. 1988; Robinson and Sheehan 1983; Wagenberg et al. 1988). It is this literature that has identified the news media's emphasis on the horserace at the expense of issues and more recently identified a more critical role in the mediation process (Lichter and Noyes 1995). The other use to which content analysis is often applied is for drawing inferences about the relationship between coverage and  34  a political outcome. For example, after observing that John Turner, the Liberal leader in 1984, was described with almost exclusively negative descriptors, Wagenerg et al. (1988) suggest that the news decisions influenced the outcome of the election. The analysis proceeds by drawing on a number of these approaches. The analysis of news coverage draws from the literature on how the news media cover elections and focuses on the kind of coverage and the relationship between that coverage and aggregate indicators of public preferences. The dynamics of public perceptions and preferences, making the best use of the qualities of the rolling, cross-section design, are then considered. Finally, data from the content analysis is merged with the rolling, cross-section to estimate dynamic models of the impact of changing news coverage on individual voter perceptions and party support. The merging of content analysis data and public opinion data is not without precedent. A recent study of the impact of newspaper editorials on public opinion incorporated values for editorial slant with a cross-sectional survey (Dalton et al. 1998). Closer to the design envisioned here, the investigators of the 1988 Canadian Election Study added daily values for news and advertising content to the public opinion data (Johnston et al. 1992). The key difference between their study and this one is that the authors assumed that news and advertising content had the same effect on all respondents. Plan of the Book The following six chapters come to terms with the 1993 campaign changes as they relate to the Reform party. The analysis addresses three principal questions. First, how did the news media cover the Reform party? The answer established in Chapters 3 and 4 identifies the dynamic availability of information about Reform while demonstrating the causal link between party support and news coverage. Second, how did the campaign affect voters' knowledge of  35  and evaluation of Reform? Here the answer helps sort out the roles of learning, priming and persuasion as the basis for campaign change. Third, can the changing news coverage be understood as the reason for campaign changes of voter knowledge and evaluation? The answer provides the basis for understanding how the flow of information gets processed by individuals in ways that sometimes produce campaign effects. Neither is the media all powerful nor ineffective. Chapter 2 begins by providing the historical and analytical background of the Reform party and the 1993 campaign. In particular, key events and dynamics are identified along with explanations for the outcome. In addition, Chapter 2 ends with an introduction to the data (public opinion and media content analysis) used in the remainder of the thesis. Chapter 3 takes on the question of the dynamic change in Reform's share of television news coverage during the campaign. It is shown that while a small share of coverage early on in the campaign disadvantaged Reform, the party benefited from increased attention in the middle of the campaign. In addition, the chapter establishes the independent role of the news media in the attention that parties receive. Chapter 4 breaks down the coverage of the election and specifically Reform with respect to its issue emphasis. When Reform's share of news coverage changed, so did the issue discourse of the news in general and with respect to the Reform party. There is also significant evidence that the change of issue attention was partly mirrored by changes in the public's aggregate perception of the important issue of the campaign. The chapter thus establishes the character of the information flow of messages that voters received at different times during the campaign.  36  Chapters 5 through 7 emphasize the individual-level changes in Reform knowledge and intentions. Chapter 5 begins the individual-level analysis with a consideration of the individual learning of Reform positions in response to the diffusion of Reform across the campaign. In particular, one can show that the campaign was essential for the learning of Reform's position on the deficit and welfare state. Chapter 6 then considers the changing importance of issue dimensions for Reform support at different times during the campaign. In particular, the analysis examines the extent to which the campaign served to persuade voters to become more fiscally conservative, primed voters to attach greater weight to their attitudes about the welfare state, or provided the basis for voters to learn and thereby link their attitudes to their party evaluation. Chapter 7 brings the analysis together by estimating an individual model of media effects that links the content analysis data in Chapters 3 and 4 with the public opinion changes identified in Chapters 5 and 6. Chapter 8 summarizes the key findings and draws some conclusions about campaigns, media effects, and the role of the media in the mobilization process for insurgent parties.  37  CHAPTER 2: THE ELECTION CAMPAIGN AND THE VOTE History, Ideology, and Campaign Events  One or more of the currently visible alternatives may be erased as a result of a misstep. ... the victim could readily be an old party. The victim may also be the country. Whatever happens, the site for the critical events is likely to be next campaign. J o h n s t o n et al. (1992: 255) The authors of Letting  the People  Decide  understood in 1992 as they completed then-  analysis of the 1988 election campaign that the next campaign offered the possibility of fundamental change. The 1993 campaign did not disappoint. The campaign ended with the -Liberals the head of a majority government and the opposition composed largely of two new parties; the Bloc Quebecois based in Quebec and the Reform party based in the West. While this result is not surprising given the depth of the antipathy toward the Mulroney-led incumbent government, it was the campaign rather than obvious structural factors that produced this rather than some alternative outcome. In particular, the Reform party could have remained mired in low public support instead of supplanting the Conservative party as the main party on the right. Before turning to the analysis of the campaign and Reform's successful mobilization it is necessary to consider both the events leading up to the campaign, campaign events, and explanations for the outcome. How did the Reform party come to occupy an important place in Canadian politics leading up to the 1993 campaign? What kinds of people voted for Reform? What happened during the campaign that might account for the dynamics? These questions provide the background necessary to develop detailed hypotheses about Reform's success. The  38  following sections address the three questions before introducing the campaign data, both survey and media content analysis, which is used to provide some answers.  Setting the Stage for the Campaign The Reform party was formed in 1987 and contested 72 seats in the 1988 election. Current treatments of the 1988 election stress that the presence of the Free Trade Agreement as a central issue of the campaign, ensured that western voters would not abandon the Conservative party despite growing dissatisfaction with the performance of the Conservative government (Flanagan 1995). No Reform candidates were successful in the general election, but Deborah Gray was elected as a Reformer in a 1989 by-election (Beaver River). A Reform candidate was also successful in the 1989 Alberta Senate election. Increased membership and the topping of the polls in Alberta would accompany these two concrete victories. Building on these accomplishments, the 1991 Reform assembly approved a motion expanding the party to the rest of Canada except Quebec (Flanagan 1995). Figure 2-1 plots the trends in popular support for the parties as revealed by quarterly surveys by Environics. The figure shows Reform with an almost insignificant share of support in the early 1990s, which increases to a modest 10 percent share over the last quarter of 1990 and the first three quarters of 1991 before starting to decline. At this point, Reform stalls. In fact, the key story is the dismal showing of the incumbent Conservatives throughout most of their second term in office. The government had the support of a very small minority of the national electorate. The Conservatives chances clearly turn around with the resignation of Mulroney and the selection of Kim Campbell as leader in the spring of 1993. As Conservative fortunes improved leading up to the election, Reform fortunes continued to decline.  39  Figure 2-1. Party Support in the 1990s  Percentage  0 ' 2  1  3 90  4  1  2  3 91  4  1  2  3 92  4  1  2  3  4 93  Quarter/ Year Source: Environics Quarterly Survey. Percentage based on party's share of intentions including decided and leaners, in national sample. No party preference, B.Q., and other party are not represented on graph.  The Mulroney government's second Parliament may go down as the most disliked government in history. Government approval during the early 1990s was the lowest on record with less than 15% of the public approving the way the government was handling the country. Throughout the second term a number of unpopular events likely contributed to the low approval. The Conservative party had presided over the failure of the Meech Lake Accord, the formation of the Bloc Quebecois, and the public rejection of the elite-supported Charlottetown Accord. In addition, the Conservative government introduced the unpopular Goods and Services Tax (GST). Adding to the woes caused by these crises and policy decisions, the country fell into a recession. A study of the 1988-1990 period showed that the drop in government approval in the early period was associated with economic conditions (Clarke and  Kornberg 1992) rather than the constitutional troubles or the GST, but these events combined with the failure of the Charlottetown Accord didn't help matters for the governing party. Brian Mulroney came to symbolize everything that was wrong about the Conservative government. Despite this, Mulroney had in fact accomplished a notable feat in leading the Conservative party to victory in 1984 and again in 1988. Both victories were made on the backs of two regions - Quebec and the West - which do not form a natural alliance (Johnston et al. 1992). The Conservative party has a history of doing poorly in Quebec while strong in the West and has therefore institutionalized some of the regional conflict between the two regions. Between 1984 and 1988, government approval dropped significantly. Of course, it is dissatisfaction with the government at this time that led to the formation of the Reform party. The Conservative party recovered from low poll showing in the lead up to the 1988 campaign, but voters supported the Conservatives in spite of Mulroney's character. The 1988 election study revealed that Mulroney was neither considered trustworthy, compassionate, caring nor moral in 1988 (Johnston et al. 1992). In February 1993 Mulroney resigned leaving a new leader to rebuild the Conservative party. After winning a closer than expected leadership race with Jean Charest, Campbell who was a minister in the Mulroney government had the summer to put her stamp on the party. As Figure 2-1 indicated, under Campbell's leadership the Conservative party did rebound in the polls. The Conservative rebound is in part attributable to the good press she received coming out of the Tory leadership convention (Mendelsohn and Nadeau 1998). Despite the rise in the polls, Mendelsohn and Nadeau show that the media coverage that Campbell received benefited her because other party leaders were ignored and the coverage tended to focus on her personal background and ability as a campaigner rather than issues and leadership qualities. Once the  41  election campaign began, when the nature of the coverage changes, it was not clear whether the Conservative party could escape the Mulroney legacy. Throughout the period a minor story is the continuous decline in popular support for the New Democratic Party (NDP). For most of its history (Young 1969) the NDP has gained enough votes to be a Parliamentary player but not enough to seriously challenge for the government. Despite the small chance that it would ever form the government, the party continued to fight a national campaign as if it was a viable alternative to the other parties. After the 1984 election and with John Turner the leader of the Liberals, the party strategy appeared to have paid off. As the parties entered the 1988 election, the NDP had a reasonable chance of supplanting the Liberals as official opposition. This was not to be and in the aftermath of the 1988 election NDP support waned. In 1990 and early 1991, the NDP was as, or more, popular than the Liberals but throughout the rest of the period the party declined fairly drastically. A number of factors probably account for the failure of the party to capitalize on the poor performance of the Conservative government. The party's support of the Accord may have alienated a core of its constituency. The unpopularity of provincial governments controlled by the NDP, Bob Rae's Ontario government in particular, probably also contributed to dwindling support which continued through the campaign. Despite the weakness of the NDP and the deep discontent about the governing Conservatives, the Reform party failed to enter the campaign with a head of steam. Reform's failure to generate momentum going into the election is more apparent when one considers the absence of a pro-Reform boost at the time of the referendum. The referendum on the Charlottetown Accord provided the Reform party with its first opportunity to generate national exposure on an important political debate and, importantly, the party took a position at odds  42  with the elite consensus. This meant that the Reform party was closer to the median position of the voters outside of Quebec than the other parties. While the No victory in most parts of the country vindicated the party's position, the evidence suggests that the Reform party and Manning personally were not particular beneficiaries (Flanagan 1995; Johnston et al. 1996b). Figure 2-2 locates the referendum on the Charlottetown Accord on the trend in Reform support from Figure 2-1. There is an increase in support for the party from the beginning to end of 1992, but the rise is small and all of it evaporates within the next three months. Rather than helping the No side, Mannning's stand against the Accord produced counterintuitive results. First, Trudeau not Manning was the key intervenor. Those people who knew that Trudeau was opposed to the Accord were more likely to oppose the Accord themselves and Trudeau's announcement of his opposition had an almost immediate impact. Second, knowing Manning's opposition to the Accord, paradoxically, helped rather than hurt the Yes side (Johnston et al. 1996b). The referendum campaign clearly failed to coalesce and expand support for the party despite the availability of the issue for mobilization. On its face, the 1993 election appeared to be the ideal opportunity for a regionally concentrated party to make an electoral breakthrough. New parties had emerged on the scene in the very regions that were central to the Conservative coalition since 1984. The proseparatist Bloc Quebecois in Quebec was capitalizing on the failure of the Charlottetown Accord, while Reform was quietly strong in the West. Pre-campaign polls showed that the Bloc was going to turn its concentrated support in Quebec into a large numbers of seats, but the possibility of a Reform breakthrough was less clear. Despite offering a clear alternative to the Conservatives on the right of the political spectrum, throughout 1993 Reform was the preferred choice of less than 10 percent of decided voters.  43  Figure 2-2. Reform's Share of National Vote Intentions: 1990-94  Percentage  Quarter/ Year Source: Environics Quarterly Survey. Percentage based on party's share of intentions including decided and leaners, in national sample.  When the votes were counted on election day, Reform had managed a considerable electoral breakthrough winning 53 of the 298 electoral seats. It is tempting to view Reform's elevation to major party status in the 1993 election in the wake of the Conservative collapse as inevitable, but there is good reason to believe that this was only one of a number of potential outcomes. While the Conservative government had been highly unpopular and the economy had yet to recover, the party entered the election period with a new leader, who was more popular than any of the other party leaders in the electorate outside Quebec. K i m Campbell had orchestrated a rise in the polls that closed the gap between her party and the Liberals while reducing Reform support. There was a possibility of a strong Conservative showing. In fact,  polls showed the Conservative party trailing the Liberals by only a few points during the early days with Reform trailing all of the other parties including the NDP. The Ideological Character of the Reform Party The formation and history of the Reform party indicate that the Reform party is associated with populism or alienation from the established parties along with distinct positions on fundamental dimensions of Canadian politics. In particular, the party took extreme positions on cultural questions such as the place of French Canada in the federation and on questions of the appropriate size of government intervention in society. Populism  It is tempting to attribute considerable importance to populism as an explanation of Reform's success. From this perspective, Reform is following a tradition of western populist movements that go back to the rise of the Social Credit and Co-operative Commonwealth Federation (CCF) parties (Flanagan and Lee 1991) . Populism in its contemporary 8  manifestation has provided a crucial starting point for Harrison's account of the Reform party (1995; Harrison and Krahn 1995; Harrison et al. 1996). It has also been central to other works on the party (Patten 1993; Laycock 1994; Barney 1996). While there are obvious reasons for thinking in these terms, populism is a difficult starting point to justify because the definition of the "other," which is essential to populism, can only be understood in terms of the party's neoliberal and authoritarian positions. Discussions of populism often begin by recognizing that the term is a contested concept that has been applied to a variety of phenomena across time and space. While attempts have  8  Those interested in academic treatments of the early populist movements in Canada should refer to Macpherson (1953), Lipset (1950), Morton (1950), and Conway (1978). 45  been made to identify and classify populist movements and to identify the causal factors associated with their emergence, there is no widely accepted definition of the term (Canovan 1981). Rather than engaging in another review of these attempts or trying to create a new definition, the focus here is on the shared aspects of the definition used to understand Reform's particular brand of populism. Drawing on Laycock and Sinclair, Harrison defines populism as "an attempt to create a mass political movement; mobilized around symbols and traditions congruent with the popular culture, which exposes a group's sense of threat, arising from presumably powerful 'outside' elements and directed at its perceived 'peoplehood"'(Harrison 1995: 5). The key is that populism requires an enemy. Conditions in the West no longer resemble the conditions in place during the wave of Western protest movements so if populism and protest are important elements of Reform it is populism of a different variety. Who is the enemy? In regional terms, the enemy for the emerging Reform party was central Canada and the political elites who ignore or tread on western interests. Consider that all observers of the Reform party identity the awarding of the CF-18 contract to a Quebec firm over the more qualified Manitoba group as the precipitating cause for the formation of the party (McCormick 1991; Harrison 1995). While "the party of the west" is one of the ways Reform has cast itself (Flanagan 1995), regional alienation has a limited value as a mobilizing force for a party that is attempting to run candidates in all of English Canada. It is not, however, regional alienation that lies at the root of populism according to many of the authors. According to Laycock (1994), the enemy is the special interests. "Special interests are identified by Reform spokespersons and internal Reform media as feminist lobby groups, native organizations, organized labour, multicultural, linguistic and  46  ethnic groups...." (Laycock 1994: 217). It is this definition of the enemy that informs the empirical work by Harrison et al. (1996) that claims to show Reform's 1993 support in Alberta to be largely a populist movement. For Reform, there is an antagonism between the public and government bureaucracies dominated by special interests (Patten 1993). The party's advocacy of direct democracy instruments, including referendums and recall, is illustrative of this line of thought, but so is Manning's own philosophy that what should be important in government is the "common sense of the common people" (Flanagan 1995: 23). Traditional parties have failed to represent this common sense. But this raises the question about the true nature of populism as an ideology. While the critique of special interests, the advocacy of direct democracy, and the rhetorical reference to the common sense of the common people makes the party appear to be more democratic than the other parties, Reform is not committed to an expansion of democratic processes. By constraining the scope of governmental decision-making through arguments about the proper role of government in society, the party is moving toward less democracy (Laycock 1994). The analysis of Reform attempts at teledemocracy also reveal its rhetorical pursuit of "the reduction of democratic participation to a series of isolated transactions in an open market where votes are currency" (Barney 1996: 401). This suggests that what appears to be evidence of populist appeal is really representative of an ideological commitment to a reduced role for government. In a recent article, Sigurdson (1994) expounds the idea that Preston Manning should not be viewed in terms of the past populist experience in the West. "Manning speaks the language of postmodernity" (Sigurdson 1994: 250) and this is new. Of course, it is a conservative postmodernism that is a response to identity claims such as feminism and environmentalism.  47  Populist rhetoric is thus the expression of a particular kind of individualism; an individualism that does not accept the "postmaterialist emphasis on individuality" (Sigurdson 1994: 273). While Manning may personally reject ideology, Sigurdson identifies a more coherent ideological basis to Reform thought that links populism with more general conservative attitudes. This ideological basis is surprisingly similar to the right authoritarianism discussed byKitschelt(1995). According to the investigators of the Canadian Election Study, Reform did manage to capture populist sentiment or alienation from the political system (Johnston et al. 1996a). Clarke and Kornberg (1996) also found that negative attitudes toward a dimension labeled participation-representation, in other words negative attitudes about national parties increased the likelihood of voting Reform. These attitudes had, however, no effect on the likelihood of voting for the Conservatives. There does appear to be a relationship between support for Reform and an antipathy toward the political system even when other attitudes are controlled. While Reform does seem to have capitalized on alienation from the political system, populism may be a problematic starting point for understanding the Reform party's development and appeal. The party's advocacy of direct democracy measures and its attack on special interests can be viewed from a theoretical position concerning the proper role of government in society. So to can its position concerning special interests and identity claims. This is not to say that populist rhetoric is not important for understanding the party. As Betz argues with respect to new right parties in Europe, "they are populist in their unscrupulous use and instrumentalization of the diffuse public sentiments of anxiety and disenchantment and their appeal to the common man and his allegedly superior common sense " (Betz 1994: 4). If,  48  however, populism is a rhetorical language of more fundamental ideological principles, it is worth considering those principles.  Xenophobia  and  Neo-Liberalism  The two ideological characteristics that stand out about Reform are its ethnocentric positions on cultural questions and its neo-liberal positions on the role of government in society. A consideration of these ideological traits and their role in the 1993 pattern of voting establishes a basis for focusing on these dimensions in the later analysis. In fact, these issues have been central to understandings of new right wing parties in Europe (Betz 1993a; Kitschelt 1995; Taggart 1995). It would unreasonable to suggest that the Reform party shares a xenophobic or racist identity with radical right wing parties in Western European countries. In a recent review of right-wing extremism in Canada, the Reform party was not included as an extreme group (Scheinberg 1997) and Canada does not appear to have the problems of West European countries. Incidents of hate-based crimes targeted at racial minorities or foreigners are not comparable to West European countries and the Reform party has neither actively promoted racist views nor been directly associated with racial violence. Of course, racists have found a natural home in the Reform party. When these have become publicly known the party has generally acted quickly to distance itself from these views and people (Flanagan 1995). A look at the policy stances of the party reveals, however, that xenophobic and ethno-cultural claims are important to the party's image. The opposition to the Charlottetown Accord and its general rejection of "special status" for Quebec is symptomatic of the party's xenophobic character in that it represents a rejection of a particular vision of the country that understands Canada as a duality or mosaic. Of course,  49  there were lots of reasons for rejecting the Accord so we should not understand it as a rejection of Quebec per se. The Reform party has further established its position on ethno-cultural questions in Canada by rejecting federal multiculturalism and bilingualism policies. Both of these policies can be understood as reflecting a pan-Canadian worldview and a more general value of a pluralistic society. The barrage of criticisms of the immigration and refugee policies of the government is further evidence of its xenophobic character. It may be that the Reform party is not as xenophobic as many of the new extreme right parties in Western Europe, but by Canadian standards Reform is an outlier on this dimension. Not surprising, the empirical evidence from 1993 shows that Reform benefited from anti-French and anti-minority attitudes. There is a clear strong effect of attitudes about minorities on the likelihood of voting Reform. The more opposed to an accommodation with Quebec and the more opposed one is to multiculturalism and immigration, the more likely one is to vote Reform (Johnston et al. 1996a). Despite Manning's personal desire to build a non-ideological party (Flanagan 1995), Reform has always been a party of the right on economic questions, reflecting a neo-liberal ideological perspective. A neo-liberal perspective is one characterized by a desire for market rather than government allocation of scarce resources leading to an overall reduction in the size of the welfare state (Kitschelt 1995). There was an overwhelming consensus among delegates to the Reform Assembly in 1992 on questions of government spending and the consensual position was on the extreme conservative end of the scale (Archer and Ellis 1994). While political elites and party activists may have recognized the party's ideological distinctiveness, the party did not focus on the deficit/ role of government dimension. It was not until the prewrit period (spring 1993), that Manning outlined the party's plan to downsize the federal  50  cabinet and reduce the deficit to zero in three years (Flanagan 1995). According to Flanagan, the plan received newspaper coverage but little attention in the electronic media and was subsequently dropped as a focus until its re-release on September 20 (1995). th  Neo-liberalism clearly emerges in the analysis of the Reform vote. "Reform nosed the Conservatives off capitalist turf, supplanted them as the most credible party of the economic right. But they did so mainly be shifting the content of the economic right, away from continentalism and the deficit to anti-welfare state appeals" (Johnston et al. 1996a: 9-10). On economic questions the Reform and Conservative parties were quite different. Those voters who  were on the deficit pole of the macroeconomic dimension and those with a continentalist  position on the scale with the same name were more likely to support the Conservative party. An antipathy toward the welfare state, however, significantly increased the likelihood of voting Reform. The macroeconomic and welfare state scales capture two sides of conservatism. Concern with macroeconomic policy is rooted in the post-World War II consensus about the welfare state, while concern about the welfare state is more reflective of a neo-liberal ideological position.  The Social  Basis  of Reform  and Party  System  Change  An analysis of the social structure to Reform support also highlights important aspects of Reform's appeal. Reform was clearly the party of the West, but supporters were also more likely to be non-European, male, and from non-union families (Johnston et al. 1996a). The gender gap in Reform support is not surprising given both that the leader of the Conservative party was a women and that Reform's positions tend on average to run contrary to positions held by the average woman. Women were just unlikely to defect to Reform from the other parties. Despite its origins on the prairies, the Reform party is not a farmer's party per se. It is  51  a party of the urban middle class. As one ecological analysis has indicated, "the party did best in affluent, rural, Protestant electorates with fewer non-English/ French speakers" (Eagles et al. 1995:23). Reform party members are largely former Conservative supporters but the argument is sometimes made that Reform is also reaching out to previously unmobilized voters (McCormick 1991; Archer and Ellis 1994). While this latter point may be somewhat true of party members, the post-election survey showed that almost all of Reform support came from previous Conservative voters (Johnston et al. 1996a). As Johnston et al. recognize, this is prima facie evidence that Reform represents a shift within an ideological family (Bartolini and Mair 1990). The Reform party garnered support from a distinct clientele. Reflecting its willingness to take unique positions on cultural questions like the place of French in Canada and policy toward minorities, Reform voting was associated with anti-French and anti-minorities positions. In addition, Reform carved out a distinct clientele on economic questions. A n antiwelfare state scale, which is closest to reflecting neo-liberal ideas, was the fundamental economic dimension that signalled out Reform voters in the analysis by Johnston et al. As expected, the party also benefited from general feelings of alienation and populism. The question is how did these post-election relationships emerge in the mass public. The Campaign  The 1993 election campaign began on September 8, 1993 and ended 46 days later on October 25. As mentioned, the 1993 campaign began with the Conservative party apparently in a close race with the Liberals (Johnston et al. 1994), but the Conservative campaign  unravelled. The unravelling left no question about the likely outcome of the campaign and certainly contributed to Reform's success. The analysis in this and the remainder of the thesis focuses on the English campaign; the campaign outside of Quebec. Reform ran no candidates in Quebec so there was no potential for the Quebec electorate to play a direct role in Reform's rise. The exclusion of the Quebec electorate is also possible because of the unique nature and often independence of the party campaigns in the two electorates (Johnston et al. 1992; Johnston et al. 1996a). Of course it needs to be remembered that for the Liberal and Conservative parties the campaign was being waged in both electorates not just the English one. Each of the parties entered the campaign with a strategy, which they modified to reflect changing events and other party strategies. Given that Chretien was not the party's biggest asset, the Liberals entered the campaign emphasizing the team around their leader and a plan. The plan, called Creating Opportunity or the Red Book, was released on September 15 with an emphasis on jobs tempered by fiscal responsibility. While the plan was only a general outline, it was costed and certainly represented a fairly clear statement of the policy direction of a Liberal government. The Conservative strategy was to rely on .their new leader and steer a moderate course on policy. The party committed itself to maintaining social programs while reducing the deficit and government expenditures. In retrospect it is easy to see that the Conservative message was not a rhetorically easy one to convey in comparison with the message of jobs and hope on the left and the dramatic approach to deficit reduction advocated by Reform on the right. The Reform strategy was originally conceived in three phases (Flanagan 1995). The first phase was to last a week and would feature voters as the driving force of the campaign  53  discourse rather than the release of Reform policies. The second phase was to focus on Reform policies and the final phase was aimed to counter voter fears that a vote for Reform was a wasted vote. The strategy provided considerable scope for flexibility. Manning; "could go to the right with Zero in Three or to the centre with the New Economy" (Flanagan 1995: 138). Zero in Three refers to the Reform commitment to reduce the deficit to zero in three years with significant cuts in public expenditures, while the New Economy position, according to Flanagan (1995) is a more centrist position that focused on the need for government and society to adjust to the changes from a resource and industrial to a service and information economy. As the campaign unfolded the Zero in Three strategy was undertaken and Reform became positioned solidly on the right.  Figure 2-3. Party Shares of Vote Intentions Outside of Quebec  50  Percentage  40  30  20  10  NDP 0 10 12 14 16 18 20 22 24 26 28 30 September  2  4  6  8  10 12 14 16 18 20 22 24 October  7-day moving average  54  Campaign  Dynamics From the election study data one can construct a clear picture of the campaign  dynamics in vote intentions for the parties. Figure 2-3 makes clear that there was almost no substantial change in Reform intentions over the first 14 days of the campaign in the nonQuebec electorate. From that point, intentions increase fairly gradually until they peaked about 10 days before election day. In the last week, intentions then fell back. While the path of Reform intentions was gradual, it appears to have been precipitated by the drops in Conservative support; around September 20 and October 12 of the campaign (Johnston et al. th  th  1994). Some of the early Conservatives losses were absorbed by the Liberal party, but Reform was a particular beneficiary, especially in the middle of the campaign. Table 2-1. Key Events in the 1993 Election Campaign Date  Campaign Event  15  Start of Election Campaign Liberals release Red Book, Creating  20  Reform releases "Zero-in-3" Deficit Plan  23  Campbell says election campaign not time to discuss social programs  25  Manning announces that Reform government would reduce restrictions on provinces in the Canada Health Act  26 3 October  Election advertising begins  8 September  Opportunity  French language debate  4  English language debate  5  Liberal Fundraising controversy. Manning proposes withdrawing from N A T O  6  Pearson Airport Controversy begins  13 14  John Beck, Reform candidate, makes racist remarks Conservative advertisement using unflattering picture of Chretien  25  Election Day  55  The description of the path of vote intentions is interesting, but it is an incomplete story of the campaign without a consideration of key events and their potential impact on the changes in vote intentions. In particular, the unravelling of the Conservative party during the campaign is at least partly linked to three events during the campaign; two gaffes and a controversy over a Conservative political advertisement. The next step is to fill in some narrative using the events temporally listed in Table 2-1. To maintain the emphasis on the dynamic changes in intentions, the events are also represented on Figure 2-4, which plots the shares of vote intentions for the two main dynamic parties, Reform and the Conservative parties.  F i g u r e 2-4. C a m p a i g n Events and Support for the R e f o r m and Conservative Parties  10 12 14 16 18 20 22 24 26 28 30 2 September  4  6  8  10 12 14 16 18 20 22 24 October  7-day moving average  56  A Troubled Campaign: The Conservatives Going into the campaign the Conservative party's biggest asset appeared to be their leader Kim Campbell. As discussed earlier, she was the most popular leader and had managed to turn Conservative fortunes around. It appears, however, that a combination of the party's strategic difficulties, the increased scrutiny of the campaign, and poor campaign decisions contributed to a campaign collapse in support. The campaign got off to a rocky start for the Conservatives and their leader Kim Campbell. On the day she announced the election day, Campbell was asked by a reporter when the unemployment rate would drop below 10 percent and her remarks were understood to mean that unemployment would not reach this point till the year 2000 (Campbell 1996). The remarks played into the definition of the campaign as provided by the Liberal party which was focusing on jobs and hope. This first gaffe was followed up by a second a number of days later. In this second case, the NDP claimed to have a copy of a secret Conservative plan to cut social programs after the election and Campbell both denied the existence of the plan and stated that an election campaign was not the right time to discuss social programs. Johnston et al. (1994) suggest that the second gaffe was important not only because it reminded voters of the untrastworthiness of the previous Prime Minister but also because it signalled Campbell's weak commitment to deficit reduction. Gaffes are, of course, constructed. Jean Chretien made a similar comment to Campbell's about his plans for social programs late in the campaign without any serious fallout. This statement did not become a gaffe for a number of reasons. The first is that Chretien was being pushed for more details in the context of a Liberal campaign that had already provided some policy detail. Campbell had provided no such detail when she made her  57  remarks. The second is that Campbell made her comment as the leader of a government that presumably had people working on social program reform. In this context, Campbell's comments looked like the Conservatives were withholding their plans and unwilling to treat the public with respect. The third reason is that Campbell's comments were interpreted as gaffes because they fit a more general pattern of negative treatment for female candidates (Kahn 1992; Kahn and Goldenberg 1991). This explanation has been considered by some researchers and some empirical evidence supports more negative coverage for Campbell compared with the other leaders (Frizzell and Westell 1994; Gidengil and Everitt 1997a; Gidengil and Everitt 1997b). According to the gender argument, Campbell was punished because she did not accept traditional, male campaign norms and practices. She was attempting to "do politics differently" by acknowledging the dire nature of economic conditions. It is difficult to evaluate the accuracy of the gender explanation for the 1993 case because more than just gender separates Campbell from the other candidates. For example, while Campbell appeared to be the Conservative party's biggest asset, the leader of the Liberal party, Jean Chretien was not particularly popular and the Liberal party strategy was to focus on the team and the platform, not the leader. As Just et al. (1996) have argued, the news media tend to reach a consensus about how to report each leader based on their perceptions of the campaign, which means that candidate coverage will be framed in terms of the narrative the journalists have in mind. Since the Conservative story was Campbell, her statements took on importance in the campaign unlike those of any other leader. This could have benefited Campbell or another leader, but Campbell never managed to articulate a message.  58  Regardless of the reason for the construction of Campbell's comments as gaffes, the comments in themselves are unlikely the driving force behind the drop in Conservative support. Conservative party support held and perhaps rose after the writ was dropped and then was clearly declining before Campbell made her statement about being unwilling to discuss social programs during the election campaign. Of course, the gaffes were just the most visible manifestations of a troubled campaign. In addition to the gaffes, the Conservative party was also hurt by its own advertising strategy. In particular, the Conservative party strategy to heighten its attacks on the Liberal leader with a an attack advertisement on the 14 of October backfired. The ad, which asked th  voters to consider the Liberal leader Jean Chretien as a potential leader of the country, used an unflattering photograph which appeared to attack Chretien's facial disability. The Liberal response to the ad was speedy. On both C B C and C T V the news reported Chretien's response to the ads; "They tried to make fun of the way I look. God gave me a physical defect. And I've accepted that since I'm a kid" (CBC News, October 15, 1993). Even Conservative candidates spoke out against the ad and called for an apology, which Campbell provided as she pulled the ads. While Tories claimed that focus groups did not interpret the ads as making fun of Chretien's face, the story in the press played on this aspect. Of course, the story was framed by the desperation which precipitated "going negative" and the fracturing of the Conservative party that the ad controversy created. Conservative candidates, including Campbell herself, were unprepared for the ad and the possible backlash. Conservative insiders claimed that internal polling showed the ad was working when it was pulled (Woolstencroft 1994), but the evidence in Figure 2-4 suggests that the ad controversy generated a final drop in support for the Conservatives.  59  As Johnston et al. note, Campbell seems central to the Conservative collapse - and the original Conservative pre-election growth, but neither of the gaffes attributed to her by the media seems decisive. It is more likely that Campbell's statements became gaffes because reporters and journalists were already discounting the Conservative party's chances. In fact, Campbell's inability to articulate a commitment to deficit reduction meant that the Conservative party basically abandoned the right, thus opening itself up to the Reform party challenge (Flanagan 1995).  The Road to Success:  The Reform  Campaign  The key strategic decision on the part of the Reform party was to focus on the distinctive policy of a zero deficit in three years. While Reform was largely a beneficiary of the campaign and its emphasis on the deficit, the Conservative party was not the only party to suffer from negative campaign events. Flanagan's treatment of the campaign signals out three events that were potentially problematic for the party. The first involved the Reform party position on health care. The second involved Manning's statement that he favoured Canada's withdrawal from N A T O . The third involved allegations of racism in the party. Despite the lack of a systematic advertising strategy and the absence of obvious shifts in intentions attributable to the debates, the Reform party managed to successfully mobilize voters as the campaign progressed. The mobilization process begins sometime after the 20 of th  September; the day that the party re-released its Zero-in-three deficit program. While the release itself is unlikely to have produced the observed shift in intentions, the announcement did affect the campaign. "The other parties attacked it so vigorously that it took on a life of its own and set the tone for the rest of Reform's campaign" (Flanagan 1995: 146). There is evidence that Reform's commitment to reduce the deficit is related to its mobilization efforts.  60  Accompanying the surge in Reform vote intentions were changes in the perceived credibility of Reform on deficit reduction and the importance of attitudes toward the deficit and welfare state for Reform vote intentions (Johnston et al. 1994). In combination with other policy positions, the Reform party clearly was positioned on the right. The first potential gaffe for Reform was Manning's revelation about Reform's health care policy. In response to a question on September 25 , Manning said that a Reform th  government would reduce restrictions in the Canada Health Act on the provinces and thus "open the door to extra billing, user fees and private hospitals" (Flanagan 1995: 151). According to Flanagan, the Reform party platform did not go this far so Manning opened himself and the party to attacks from the other parties. In effect, the statements about health care were symptomatic of positions that Reform was had already made including its willingness to reform both old age security and unemployment insurance. Both of these reforms would have reduced the role of government and thus scaled back social program spending. Despite the partisan attacks that followed the incident, there is no obvious negative effect on Reform support in the graphical evidence. The second event was Manning's announcement on October 5 that he thought Canada th  should withdraw from N A T O . This statement represented an explicit contradiction with the policy documents which support both N A T O and N O R A D (Flanagan 1995). Manning quickly acknowledged his mistake and given the low salience of foreign affairs questions, the story was quickly forgotten. On October 13 , the news media revealed that one of Reform's Ontario candidates had th  made and was standing by racist comments. The fallout of the story continued for a couple of days. It was reported that John Beck had suggested that immigrants were criminals and were  61  taking jobs away from white Canadians. Manning responded quickly by asking Beck to resign but then spent the next few days explaining how such a person managed to become a candidate. Inevitably the link, even if weakly, was drawn between Beck's interpretation of the immigration problem and the party's controversial immigration policy. While there was no other evidence that Beck was more than an anomaly, the incident did send a signal to voters that racists were the kind of people who would be supporters of the Reform party regardless of the actual policies of the party. As noted earlier, Reform had staked out relatively extreme positions on the issues of immigration and multiculturalism and these attitudes were associated with Reform voting. Of the three events the first and the last seem the most likely influences on the campaign because the second issue did not generate a controversy on which other parties or the media could focus their attention. By announcing his willingness to see health care fundamentally changed in the interests of reducing the deficit and the size of government, Reform's commitment to deficit reduction and limited government were clearly established. While this has no noticeable effect on the share of Reform support, it set the agenda for partisan attacks and media coverage. In the same light, the Beck incident may have had a more direct impact through shifting the issue discourse of the campaign to other aspects of Reform identity. Since Reform's share of the vote outside of Quebec does slide back in the last days of the campaign, there is some indication that the Beck incident mattered.  Smooth Sailing: The Liberal Campaign As mentioned earlier, the Liberals started the campaign ahead of the other parties and the net impact of the Conservative collapse was a sure Liberal win. It is, however, worth  62  noting several characteristics of the campaign and events which could have affected the outcome. The Liberals had a bold and risky strategy. Several days into the campaign, the whole Liberal platform was released instead of releasing a policy position on each day. Chretien then used the platform - the Red Book - as a prop; carrying it with him and confidently asserting that he had a plan. Of course, while the platform did suggest a general direction for government, it was often short of specifics. In addition, the campaign was well managed and appeared to run smoothly both operationally and in terms of responding to campaign events. Most notably, the Liberals responded immediately to the Conservative ad attacking Chretien. On October 5 , the Liberals were caught in a fundraising controversy. Chretien th  attended at $1000 per person fundraising dinner in Montreal. The implication that was drawn was that Chretien was selling access and influence. Despite the potential for this to have negative impacts, for the most part the issue dropped from news coverage almost immediately. The press and the other parties were more interested in the Pearson Airport issue. The Pearson Airport controversy arose basically out of nowhere near the end of the campaign. The Conservative government had negotiated a deal to lease the terminals of the airport in Toronto to a consortium in return for receiving all of the profits the consortium would invest money to revitalize the terminals. The government actually signed the contract on October 7 , which raised the ire of the Liberal and NDP parties. Complicating the picture th  were allegations of political corruption. The NDP promised to cancel the deal while the Liberals started off hesitant about their plans except to say that they would review the deal. By the end of the campaign the Liberals were sounding more and more like they would cancel the deal.  63  For the most part, the Liberals were able to run the campaign in the manner they envisioned. The campaign began with a focus on jobs and the economy which benefited the Liberals and once the Conservative party fell in the polls, the Liberals had to simply avoid hurting themselves. While there were a couple of incidents that were potentially problematic the fundraising issue and the remark on social programs discussed above, these events did not appear to fundamentally change the Liberal prospects. Debates In 1993, the party leaders took part in two debates; one in each of the two official languages. The French debate took placefirston October 3 followed by the English debate rd  on the 4 . Manning gave a brief statement in French but did not participate in the French th  debate. Although debates have furnished memorable exchanges and served as pivotal points in election campaigns (Johnston et al. 1992; Fletcher 1988; Lanoue 1991), the 1993 debates were rather uneventful. In a memorable exchange in 1984, Mulroney pointed hisfingerat Turner 9  and said, "you had a choice" - referring to Turner's patronage appointments linked to Trudeau. From the perspective of the path of vote intentions in 1993, debate performance may have helped solidify Reform support because Manning appeared as a serious leader. The failure of Campbell to score a winning attack on Chretien may ultimately mean that the debate failed to reverse the Conservative downward trend.  Voters clearly interpret debates through a partisan lens thereby reducing the effect of debate performance on the outcome of the election (LeDuc and Price 1985; Sigelman and Sigelman 1984V Thefirststudy of debates - the 1979 debates - found no effects (LeDuc and Price 1985). Subsequent studies of the 1984 and 1988 election debates have found effects (Johnston et al. 1992; LeDuc 1990). 64  Party Advertising Party advertising began on September 26 and was dominated by the Conservative th  party. It seems reasonable to speculate that advertising played only a small role in the campaign dynamics of 1993. As Figure 2-4 demonstrates, the critical moment of Conservative losses and Reform gains had already passed before advertising began. Advertising may have consolidated the shift to Reform, but even this possibility is counter-intuitive. First, Reform paid for almost no advertising, preferring to rely on a populist campaign and the news media. As a result, few viewers were exposed to pro-Reform messages in advertising. Second, the Conservative party monopolized the advertising campaign to no apparent avail. Advertising may have prevented the Conservative slide in the polls for two weeks in the middle of the campaign, but as discussed earlier, it may have precipitated the final slide. The only possibility is that anti-Reform messages may have helped voters figure out where they properly belonged and, or stopped the rise in the polls over the last week of the campaign.  Data The data used throughout the thesis comes from two primary sources. The 1992-93 Canadian Election Study, which features a rolling cross-sectional survey design, provides the data on the public's party preferences. Data about media content comes from a content analysis of television news originally conducted by the National Media Archive (Fraser Institute, Vancouver, B.C.).  65  Public  Opinion  Data  The 1992-93 Canadian Election Study is a combination of two-wave panel during the 1992 Charlottetown referendum and a three-wave panel study of the 1993 election. In the 10  election study component, each respondent is interviewed once during the campaign, and then twice after the election - a post-election telephone survey followed by a mailback questionnaire.  11  Since no new respondents are added for the post-election surveys, the number  of respondents drops as one moves from each successive wave. The campaign wave survey is the most significant feature of the study because its design produces a rolling, cross-sectional survey. Before the study began, a random sample of the national population was drawn and then parts of the sample were released on a day-by-day basis over the 45 days of the campaign study. So for every day in which they were in the field, the election study team interviewed what amounts to a random sample of Canadians. The design means that approximately 90 respondents (approximately 70 in the non-Quebec sample) were interviewed on each of the 45 days of the study and each day can be thought of as a fresh random sample. The 1992 referendum study was similar but the additional feature of importance here is that some of the people who were surveyed in 1992 were also surveyed as part of the 1993 study. This allows one to compare attitudes and behaviour of the same individuals in 1992 with 1993. In a number of places, this kind of analysis is undertaken.  The 1992-93 Canadian Election Study was supported by the Social Sciences and Humanties Research Council of Canada. The co-investigators included: Richard Johnston, Andre Blais, Elisabeth Gidengil, Neil Nevitte and Henry E. Brady. None of these individuals or institutions are responsible for the interpretation of the data provided here. 1  For more information on the design of the Canadian Election Study see the technical documentation (Northrup and Oram 1994). 66  Media Content Data The media data were originally collected and analyzed to serve the interests of the National Media Archive, which monitors media practices and is particularly concerned with questions of fairness in the coverage of different networks. The National Media Archive transcribed election stories of the two major English language daily national news broadcasts (CBC and CTV). Each transcript was then broken into codeable phrases that varied in length but captured a particular idea or reference. There are 14,327 coded bits contained in 531 election stories during the forty-five days of the campaign (September 8- October 24,1993). Each bit was then assigned a code for who spoke the identified phrase and their partisan affiliation; the party that was discussed in or targeted by the phrase; the valence of the phrase (positive, neutral, or negative); and, whether the phrase involved a discussion of the campaign (if so what aspect?) or a particular issue (if so what issue?). The coding was undertaken by two coders and a random sample of the data was checked for intercoder reliability (r = 0.87) (Miljan 1994). The Archive did not attempt to code the visuals that accompanied the spoken words in the television news. While there is reason to believe that visuals do matter, a content analysis of visuals could not be easily or practically combined with the textual content analysis. In fact, most analyses and discussion of the impact of visual messages on television suggest that the impact involves the tone rather than the content of the message. For example, a pleasing visual of a campaign event is presented with a text that points out the manipulation of the event or a party advertisement is presented with an analyst pointing out the errors. The media data are clearly not a complete record of the news coverage of the 1993 campaign. Neither newspapers, radio nor local television news coverage of the campaign were  67  coded. A reliance on television news in the analysis is justified for a number of reasons. The first is that the two national network news broadcasts have a large national reach. Given that local networks cannot afford to devote reporters to the party campaigns, most local news will either offer information about local candidates or use national stories taken from the news 12  organizations of the major networks. The second is that the news is likely to be similar regardless of the medium (Wagenberg et al. 1988). The third is that voters tend to rely on television news as their main source of information about the campaign. As a practical matter, the study of newspaper coverage is made difficult in Canada because of the lack of a truly national paper read by a large cross-section of Canada. It might be objected that Reform's regional base in the West would lead us to expect regional coverage to be more important than the national news broadcasts. While this is an intuitively likely possibility, there are both practical and conceptual reasons to emphasize national coverage. One of the practical difficulties is that television news is effectively national and local rather than regional or provincial and there is no archive of this material that is easily available. It is also true that Canadians generally get their national news from the two major network news broadcasts (Johnston et al. 1992). Finally, it should be also noted that past research has shown that coverage tends to be similar even when one compares regional newspapers with the national broadcasts (Frizzell and Westell 1989). At a conceptual level, national television news is the logical focal point for parties trying to run national campaigns (Johnston et al. 1992) and Reform was clearly trying to be more than a regional party. This  12  The role of information about local candidates in the election outcome is understudied but believed to be small. Local information about how voters in the riding will vote is the key to efficient strategic voting but the absence of local polls makes this information either unavailable or speculative. Despite the fact that voters cast votes for local candidate rather than the Prime Minister or party, the general approach is to treat voting in Canada as the expression of national party preferences (cf. Ferejohn and Gaines 1991).  68  suggests that national news is the place to focus in unlocking the relationship between the news coverage and the campaign. The data collection and analysis of the National Media Archive has been the subject of academic debate (Hackett et al. 1992; Miljan 1992). While the debate suggests potential weaknesses with the early data it largely revolved around the use and interpretation of the data. The data are being used here in a fairly limited manner and there is no reason to believe that the coding of the party being discussed aspect of the analysis would be subject to systematic biases that would produce the shifts in coverage identified here. As a validation check the author examined one half of the original transcripts and their coding to identify if just such a problem existed. According to the National Media Archive coding, approximately 66 percent of television news coverage of the 1993 campaign was about the campaign - including such things as references to the leaders, party strategy and the horserace - rather than the substantive issues (Table 2-2). This leaves the remaining 34 percent to be about issues, which is somewhat larger than the amount of issue coverage in other elections in Canada and abroad. Table 2-2 also breaks down the coverage within each of issue and campaign coverage into a number of categories. Almost a third of campaign coverage was devoted to profiles of the campaign: its organization, party promises, and advertising. Another 11% was devoted to more general coverage of the parties including their membership, candidates, and unity. Approximately 19 percent of campaign coverage was devoted to leader profile coverage. Coverage that was explicitly about polls, party popularity, or the outcome of the election was coded as horserace coverage, which represented 19 percent of campaign coverage. Another 16 percent of coverage  69  was associated with a more general coverage of voter expectations and feelings that were not explicitly horserace. The debate generated 2 percent of the coverage. Except perhaps leader coverage, there are no clear reasons why different kinds of campaign coverage data would have different kinds of effects on the relationship between issues and Reform support during the campaign. For this reason, the subcategories are not considered in and of themselves.  Table 2-2. Type of Television News Coverage Type of Coverage  Amount of Coverage  33.9% (4431)  Issue Economic Policies Job creation/Unemployment Deficit/Debt reduction Free Trade Other Economic Social Programs Culture, Ethnocentrism, Moral Issue External Affairs, Defence, Helicoptors National Unity Government Reform TOTAL  Non-Issue Coverage  TOTAL  15.5% (685) 18.3 (811) 12.9 (573) 5.0 (223) 8.0 (355) 20.2 (894) 6.4 (285) 3.2 (141) 8.4 (374) 2.0 (90) 100.0(4431)  66.1% (8648)  Campaign Profile, Organization, Strategy  29.8% (2576)  Leader Profile Horse Race: Polls, Popularity Voter Mood Party Profile Debate Other TOTAL  18.6(1607) 18.5(1599) 15.8(1369) 11.3 (974) 2.3(195) 3.8 (328) 100.0 (8648)  100% (13079)  There is considerable diversity in the kinds of issues that received coverage in the 1993 election; reflecting the lack of an organizing issue like the Free Trade Agreement. Social programs received the single highest amount of coverage, but together a diversity of economic issues was clearly important. National unity, despite the recent constitutional failures and the presence of the B Q and the Reform parjy received a small relative share of news coverage. No  70  doubt, this reflects the unwillingness of the parties themselves to talk about the issue. The news agenda and its implications are taken up in Chapter 4. Discussion There is no question that electoral support for Reform grew during the campaign. The party entered the campaign with a small relative share of the vote, but the campaign produced a dramatic change in Reform fortunes. Two prominent stories emerge from an analysis of the campaign. The first is the dramatic collapse in Conservative fortunes. The second is Reform rise in the polls. While the race was close for the first couple of weeks of the campaign, the Liberals never trailed in the electorate outside of Quebec. After the collapse the outcome was never really in doubt. There was some speculation about a possible minority government but no other party was a serious contender for government. Given that the Conservative and Reform parties were competing for the support of some of the same people, how did the Conservative collapse set the stage for the Reform rise in support? To answer this question one needs to consider the news coverage. How was the Reform party covered by the television news media and what was the relationship between that coverage and mass support? More specifically, the narrative of the campaign and the post-election analysis suggests that Reform was able to capitalize on the abandonment of the right by the Conservative party by stressing the party's position on the deficit and welfare state. Did Reform's position-taking on the role of government get reflected in news coverage? If the answer is yes, did this coverage help mobilize voters who were predisposed to support Reform? And, how? If the party's positioning on the welfare state dimension affected voter support early, what effect did the John Beck incident have on Reform support. Reform is clearly positioned  71  as an opponent of government policies with respect to immigration, multiculturalism, and bilingualism. The party also benefited from these feelings within the electorate. Were cultural issues a fundamental component of Reform's news coverage? Did the campaign affect the importance of these attitudes or was it irrelevant? It is possible that the public was aware of the party's position on cultural questions as a result of the referendum on the Charlottetown Accord and therefore would be unaffected by new information about the second issue dimension. Two issue dimensions - welfare state and cultural - were clearly important for understanding who voted for Reform on election day. The remaining five chapters address the process by which the campaign mobilized voters with these characteristics and thereby increased Reform support. How did the information flow of the campaign change the cues, especially the issue cues, which voters were exposed to? What were the effects of this information flow on the presence of priming, learning, and persuasion within the electorate?  72  CHAPTER 3: THE CHANGING INFORMATION FLOW Media Attention and Popular Support for an Insurgent Party They [reporters] may be only minimally influential, as when they accelerate trends that are already underway and likely to continue. But they may also be, at least occasionally, highly influential, as when coverage affords as essential advantage to a candidate who, if press attention had been directed elsewhere, would have fallen by the wayside. Zaller with Hunt (1994: 377).  Did Reform benefit from changing evaluations of newsworthiness or did coverage of Reform largely follow Reform support? The amount of attention which political parties and candidates receive from the mass media at election time can be influenced by an array of sometimes opposing factors. A norm of equitable attention may be applied, but in particular media systems or campaign situations such a principle may be modified or overridden by news values and even journalistic judgments about the political significance or popular support of the contenders. When minor parties or new parties are competing the tension between the norm of equitable attention and news values is especially important for two reasons: first because their coverage prospects are more uncertain and difficult to predict; and second, because they are often more dependent on media attention for popular recognition and hence for support at the polls. The evidence presented here suggests that the relative attention to Reform as compared with the other parties in the news media underwent dynamic changes. At the beginning of the campaign attention to Reform was minimal, but midway through Reform became a significant  73  player in the news drama. Since there is good reason to believe that coverage and support will be related, the analysis seeks to establish which series changed first. The evidence both establishes the causal direction for 1993 and provides empirical support for the idea that there is an underlying equilibrium between coverage and political support for a party. Balance and Party Access While not a sufficient condition for attracting electoral support, news coverage should be viewed as a necessary condition for electoral success (Zaller and Hunt 1995); in the absence of information and cues, voters are less likely to support a candidate (Bartels 1986). While a focus on attention sets aside the question of the tone of the coverage, attention is not trivial. Attention measures are indicative of the likely penetration of information about the candidate in the mass public. For new parties or candidates, media attention is particularly critical because voters will be initially unfamiliar with them. It is, therefore, essential to establish the basis for how the news media allocate news attention. In principle, the balance of partisan attention is a question of fairness (Robinson and Sheehan 1983). If a party is ignored in the news media then it has a legitimate basis for arguing that the news media are biased against it. Of course, balance cannot be equated with equal attention, since parties are different. As a result, the standard of fairness is likely that equal candidates will be treated equally. The evidence tends to support a view of the news media that places a high priority on ensuring fair attention, but there appears to be an important difference between the approach to balancing the partisan shares of news coverage in the United States compared with Britain and Canada. In the two parliamentary democracies, the media, and the electronic media in  particular, are more conscious of the attention parties receive in the news media and make precampaign determinations of what constitutes "fair" coverage. If the media in Britain and Canada are more likely to make pre-campaign determinations of how they will allocate coverage, how do they establish what is fair? The proportion of coverage a party receives in Canadian elections has historically been similar to the share of the popular vote received in the previous election, which has led one author to label the practice a convention (Fletcher 1987; Fletcher 1988). The effect of using the previous electoral standing anchors the coverage to the party system and thus gives incumbents an advantage and penalizes minor parties. A n alternative approach is adopted in Britain at least for television news coverage. Parties are allocated a proportion of the news based on the formula used to allocate free time advertising (Semetko, Blunder, Gurevitch, and Weaver 1991), which in practice produces similar results to those found in Canada. Newspapers in Britain, especially the tabloids, do not however observe this convention. A clear standard of fairness in Canada and Britain is accompanied by a conscious effort to achieve balance as decided upon. Semetko et al.'s comparative analysis of newsroom attitudes in American and British newsrooms found that television news workers in Britain place a premium on equity such that they keep track of the partisan distribution of coverage (Miller et al. 1989; Semetko et al. 1991). In contrast, American reporters and producers neither monitor nor commit themselves to balance. Although he provides no empirical support for his argument, Fletcher (1987) claims that news organizations in Canada monitor their own coverage to ensure it reflects the convention concerning the appropriate allocation of coverage. In the United States there is no clear standard for determining "fair" or "balanced" coverage and no accompanying news room commitment to monitor or enforce it. Most  75  observers have found balanced coverage in the U.S., while acknowledging some deviations (Graber 1989; Robinson and Sheehan 1983). The key is that coverage decisions will be more related to newsworthiness than balance per se. While balanced coverage is a core aspect of a fair press, balance can conflict with traditional aspects of newsworthiness such as drama, conflict, visuals, and timeliness. The lack of conscious effort to maintain balance likely follows from the relatively long election campaigns and the structuring effects of the two parties in American politics. For example, each convention produces a temporally short imbalance in the attention to the two major parties in the news media, but the overall balance is eventually restored (Holbrook 1996). Lichter et al. (1988) found that during the preconvention period "the bottom line was roughly equal amounts of coverage for the two campaigns - as long as two campaigns still existed" (1988: 11). Deviations from equitable coverage are more likely during the nomination stage or in those cases where there is an independent candidate for the presidency. During the nomination period, journalists must somehow decide whom, among numerous contenders, to cover. These decisions serve a gatekeeping function. Brady and Johnston (1987) found that coverage of the 1984 primary period was characterized by a tendency of the media to quickly - possibly prematurely - dismiss some candidates while allocating more than equitable coverage to a different candidate (Jessie Jackson) for idiosyncratic reasons. Robinson and Sheehan (1983) found that that there are four patterns of partisan distribution of coverage: (a) equal access is given to equal players; (b) also rans and minor parties are ignored; (c) equal attention is given to frontrunners; and, (d) extraordinary attention is given to anyone who does better than expected. An additional pattern is for unique candidates to attract attention for either idiosyncratic reasons or because of their perceived importance among journalists.  76  While the U.S. evidence tends to downplay a conscious effort at balance, one appears to emerge with some regularity because of the importance of news values for allocating coverage. As Hagen argues; "it seems clear that the candidates' positions in the race influences the decisions journalists make about how to cover them" (Hagen 1996: 192). Party standing is central to the classification scheme presented by Robinson and Sheehan and provides a way of understanding the much discussed notion of momentum in primary campaigns (Battels 1988). Doing better than expected can lead to an increase in the amount of press and the amount of good press relative to bad which can in turn increase voter support. The implication is that real indicators of growing party or candidate support are necessary to increase coverage, but it is not necessary for the causal process to go from support to coverage. Rather than linking coverage just to indicators of popular support, Zaller and Hunt go further in suggesting that coverage is allocated in proportion to the expected electoral standing of the parties (Zaller with Hunt 1994; Zaller and Hunt 1995). The news media use their judgment about the future importance of candidates based on such things as financial resources, organizations, and campaign themes not just their current levels of support. Zaller's rule of anticipated importance states that, "Coverage of candidates and issues should be allocated in proportion to its marginal value for shedding light on future developments in American politics" (Zaller 1997: 22). Thus the possibility exists that candidates will receive more than their share of coverage because journalists view their candidacy as important. The possibility for the amount of media coverage to have a causal impact on vote intentions is most likely when the candidate or party is relatively unknown. It is the lack of previous knowledge of the candidate that allows momentum to work. Increases in coverage that arise because of journalistic judgments should also have momentum-like effects for  77  relatively unknown contenders because a larger media presence is a cue about how others think about the candidate. Media coverage does predict actual public support for suddenly emerging presidential hopefuls in the nomination period, as seen with Carter and Hart as well as Perot (Zaller with Hunt 1994; Zaller and Hunt 1995). Zaller and Hunt find that in the pre-campaign nomination period, Perot's rise and subsequent decline in public support can be attributed to his coverage in the conventional press. It is also argued that good performances in unconventional media exposure may have helped gain the attention of the conventional press and establish that Perot was newsworthy, but were not enough to directly influence his popularity at the national level. In contrast, Kerbel argues that Perot received attention in 1992 because of the novelty of his presidential campaign rather than intrinsic newsworthiness to the point that the media were reporting on Perot when he was doing nothing and reported his commercials as if they were campaign events (Kerbel 1994). Although increased coverage of suddenly emerging presidential hopefuls is essential for their rise in popular support, increased support comes with a price. The price is more negative coverage. Hagen (1996) argues that increased negativity of media treatment of frontrunners is the result of the nature of party competition with candidates targeting their attacks at the strongest candidate. Zaller, however, argues that increasing negativity is more than a change in the behaviour of candidates; as electoral viability increases candidates will be subject to more press-initiated coverage (Zaller 1997; Zaller and Hunt 1995). The news media 1  1  Zaller identifies press-initiated criticism as a form of enterprise reporting where the journalist provides a criticism of the candidate that is not the product of statements of the candidate's opponents. As such it is different from bad press per se. The analysis reported here does not make use of the tone of the news coverage for several reasons. First, there is no measure of press-initiated coverage in the media dataset. Second, tone is a more controversial indicator of news, especially when the coding is based on fragments of the story rather than the tone of the news story. Third, a preliminary analysis of tone found that the amount rather than the direction of the news stories was more important for explaining Reform support (Jenkins 1998). 78  will provide increased scrutiny to relatively unknown candidates who are becoming electorally viable. The literature suggests that major parties can expect to receive the majority of attention in the news media, with incumbents possibly benefiting from a greater share of coverage because of an office-holding advantage. For new or minor parties looking for an electoral breakthrough capturing news attention requires one of two things: to do better than expected or to be different in a way that leads journalists to view your party as important for the country's political future. Of course, media judgments about the importance of a candidate or party are unlikely to be sufficient over the long term. If the public does not respond to the increase in coverage, one would expect the media to reduce its attention to that candidate (Zaller, 1997). News values provide an opportunity for minor parties to become newsworthy beyond what pre-campaign polls would suggest, but the relative amount of coverage a party receives should not stray too far from the party's standing in the polls. If voter support increases, that increase should show up in an increase in coverage. At the same time, increased coverage should lead to an increase in popular support as the public gets new and likely positive information about the party reflected in the increased coverage. Of course the coverage as a whole is likely to become more negative as the party become more viable. It would be counterintuitive for the media to increase its coverage of a small party when that new information was simply revealing why the party was so small and unlikely to make a break through. The intuition that a new party's share of media coverage will be closely related to its share of vote intentions is the basis for assuming that there is an underlying equilibrium between two series and that change in one series will lead to an adjustment in the other. The  79  particular dynamics of the two series during the 1993 campaign, which is discussed below, allows for the use of an error-correction method to model the equilibrium. As mentioned, the literature on media coverage in Canadian'election has not considered the possibility of dynamic, campaign changes in party attention in the media. Precampaign decisions based on either poll results or past electoral success are stressed and it has been asserted that the balance driven "pattern tends to persist with little regard to the quality or substance of the campaign" (Fletcher, 1987: 353). The lack of campaign dynamics in coverage stands in contrast to recent analyses of Canadian elections, which have identified significant dynamic qualities (Johnston, Blais, Brady, and Crete 1992). As discussed below, the 1993 campaign, in particular, involved a considerable temporal discontinuity with the Reform party emerging out of nowhere in campaign news and popular support. It would be surprising i f these campaign effects were not reflected in, or related to, media coverage. The next section reviews the data used in this chapter. The first step in the analysis is to briefly recount the details of the 1993 election campaign and, provide a descriptive and visual analysis of the pattern of coverage and intentions. The paper then considers both a causality test for the direction of the temporal relationship and an econometric model of the responsiveness of aggregate vote intentions to changes in Reform coverage. Data  The analysis proceeds using two different time series of the campaign. To reflect changes in support for Reform, the daily share of vote intentions as measured by the 1993 campaign wave survey of the Canadian Election Study is used. The aggregate survey data is compared to daily values of Reform's coverage in the television news media.  80  The measure of coverage or news attention is derived from a variable in the original coding identifying which party was being discussed in the coded material. From this variable, a daily value for Reform attention was constructed by taking the proportion of all coverage (coded items) for that day which featured a reference to the Reform party. Non-party oriented coverage, such as background pieces on issues or public attitudes were not included in the denominator. The coverage variable will, of course, contain day-to-day noise. The overall size of the news hole varies by day and the relative prominence of election news as compared with other issues and events also varies. In addition, candidates chose not to campaign some days, which affected the overall amount of "new" information available for the media to cover. These factors constrain the media's ability to balance coverage and offer the possibility of fairly random imbalance. Although using density rather than absolute magnitude of coverage controls for some of this variation, we would expect some fluctuations around an equilibrium point even if the media were attempting to balance coverage and support. The key is whether there are systematic changes in access that can be linked to similar changes in support. The Path of the Vote and Reform Coverage There is no question that the news media were faced with a complex political landscape as the 1993 campaign began. The discussion in Chapter 2 noted that while Campbell had managed to turn things around for the Conservative party in the lead-up to the campaign, Conservative fortunes were uncertain. New parties had emerged on the scene in the very regions which were central to the Conservative coalition since 1984. The pro-separatist Bloc Quebecois in Quebec was capitalizing on the failure of the Charlottetown Accord, while Reform was quietly strong in the West. Pre-campaign polls showed that the Bloc was going to turn its concentrated support in Quebec into large numbers of seats, but the possibility of a  81  Reform breakthrough was less clear. Throughout 1993, Reform was the preferred choice of less than 10 percent of decided voters. This actually underestimates their support because of the regional strength in Alberta, but it does suggest a problem for the media in allocating coverage. While the Bloc was going to generate attention in the English Canada media because of its importance to the results in Quebec and its unique character as a pro-separatist party, Reform was less clearly deserving of attention in spite of its regional strength. Given the dynamic evolution of Reform's share of vote intentions (discussed in Chapter 2), one could hypothesize that the news media played an important role in allowing the Reform party to get its message out. For the campaign period as a whole, the Conservative party received the most coverage (33%), followed by the Liberals (27%) and the Reform party (19%). With Reform's 18.7 percent of the vote (52 seats), the party received as much coverage as electoral support nation-wide. The Liberals received more electoral support than coverage, while the Conservatives lacked the electoral support (16%) to complement their media attention. Of course, the magnitude of the Conservative collapse was not expected based on the party's standing early in the campaign. The aggregate figures hide, however, important temporal dynamics in the media attention to the parties. Figure 3-1 tracks the proportion of all coverage, smoothed over 7 days, devoted to the parties over the campaign (coverage that did not refer explicitly to a party was excluded). In the early days of the campaign, the attention pattern appears to have followed the formula based upon the proportion of the popular vote cast for the parties in the previous election. The Liberals and Conservatives received over sixty percent of the coverage between them in the first ten days and the N D P was third, ahead of Reform, in amount of attention. If  82  the early campaign appeared to conform to expectations, things changed rather quickly for all of the parties. Coverage of the Conservatives surged dramatically in the wake of Campbell's fumbles. On September 26 , the Conservative party received over forty percent of all partisan coverage. th  This is, of course, only two days after Campbell denied the presence of a secret plan to cut social programs. From the high with around 30 days left in the campaign, the Conservative share of party coverage then declined in a fairly linear manner except for the brief rise with about 10 days remaining. This latter rise took place at the same time as the Conservative party was dealing with the fall-out of the negative advertisement which appeared to mock the Liberal leader's facial disability. Figure 3-1. Partisan Dynamics of Television News Coverage 50  Percentage  40  30  20  10 NDP  0 8  10 12 14 16 18 20 22 24 26 28 30  September  2  4  6  8  10 12 14 16 18 20 22 24  October  Moving 7-Day Average. Daily value is the percentage of coverage devoted to the party as a proportion of all coverage targeted at a specific party. Coverage of the BQ is included in calculation but not shown.  83  There are three periods where Liberal access to the news media peaks. The first, between September 11 and 19 covered the release of the Red Book and Liberal emphasis on jobs. After this period of intense access, coverage remained relatively low throughout the middle phases of the campaign until a brief surge with about 18 days remaining and a final surge in the last week. The second last surge likely reflects the ability of the Liberal party to generate coverage based on the Pearson Airport issue , while the last surge is most likely a 2  product of election day speculation. The N D P began relatively strong but their share quickly evaporated and they received about 10 percent of the party coverage throughout the campaign. The density of Reform coverage undergoes a significant surge around the 20 day of th  the campaign, rising until about 9 days left before election day, when the density drops significantly. The path of Reform coverage is strikingly similar to the path of Reform vote intentions. The decision part way through the campaign to increase the attention to the Reform party stands out in the pattern of coverage. It is as if the media discovered the Reform party. Were the media leading public opinion or were they merely responding to the appearance of an electoral breakthrough? The only hard evidence of party standing available to the media were the fourteen published national polls although many of them received scant attention. Given the emphasis on the horserace we would expect the media to be influenced by the polls in making decisions about allocating attention (Rhee, 1996). The key poll story emerged at the end of September when an Environics poll indicated the Conservative party at 31% of the popular vote; 6 points less than a Gallup poll that was reported only days earlier. The trend was confirmed with an Ekos poll on September 30 and a th  2  The Liberal party announced that they would review and possibly cancel the awarding of a contract to run Pearson Airport (Toronto) to a consortium which appeared to be tainted by political corruption (see Chapter 2).  84  Compass poll on October 2 that indicated the Conservatives at 25 and 26 percent respectively. From this point, the Conservatives continue to decline in the polls, while the Liberals improve on their way to an easy electoral majority. The national polls provide weak evidence that Reform was making inroads at the expense of the Conservatives. Reform remained at or below 11 percent of the vote until the 25 of September when Gallup recorded a two-point rise to 13 th  percent. On average Reform support was about 6 percentage points higher after September 30, but this did not become obvious for about a week. In the meantime, the news media had dramatically increased the attention they paid to Reform at the expense of the Conservatives. The poll results indicate the limited nature of cues about the projected outcome of the election that were available to the news media. While the media did not have access to the election study daily tracking, one can use the poll to indicate the real path of intentions as compared with the news media coverage. Reporters may have got a sense of Reform's true support from observing the campaign (i.e. attendance at meetings; discussions with campaign advisors) and the rolling, cross-section provides the best indicator of the true level of electoral support. Figure 3-2 compares the movement of Reform vote intentions in Canada outside of Quebec with the amount of attention Reform received in the television news. Both series are smoothed by taking the 7 day moving average. It appears from a comparison of vote intentions and coverage, that coverage preceded changes in the proportion of the public supporting the Reform party. Over the course of the first two weeks of the campaign, both series underwent no significant deviations. Around day 15, coverage increases and intentions appear to follow changes in coverage for the rest of the campaign. It is significant that when coverage declines at the end of the campaign, Reform vote intentions also collapse.  85  While one can get a sense of the changes from Figure 3-2, the different starting levels of each series complicates the visual evidence. To sharpen the analysis, both series were normalized on the basis of the first 14 days of the campaign by subtracting from each day's value the mean for the first 14 days of the series. Consequently, for the first 14 days, the two series vary together around the origin. Figure 3-3 shows the two series starting at September 24 . This approach reveals two substantive surges in coverage which were followed by th  declines. The declines did not wholly erase the surges so that after each decline, coverage remained higher than in the period before the surge. Vote intentions appear to mirror these changes with a three or four day lag.  Figure 3-2. Popular Support for and Media Attention to Reform Percentage  10  12  14  16  18  20  22  24  26  28  30  2  4  6  September  8  10  12  14  16  18  20  22  24  October  Values are 7 day moving average.  The first surge begins around September 26 and continues for approximately four th  days before declining almost back to where it began. This period is the immediate aftermath of  86  Campbell's second gaffe and the revelation of the decline in Conservative vote intentions. Significantly, Reform intentions do not begin to rise for several days and rise as coverage has begun to drop.  Figure 3 - 3 . Normalized Reform Coverage and Intentions 20  15  10  tentions  -5  I  24  I  26 28 30 September  I  I  2  4  I  I  I  6 8 10 October  I  I  I  I  I  I  12  14  16  18  20  22  24  Moving 5 day average. Values are normalized by substracting the mean for the first 14 days (not shown) from the value of each day before smoothing.  The second surge begins around October 6 and is sustained for almost 10 days before a dramatic drop in coverage around October 16 . As was the case with the earlier surge in th  coverage, the second surge was accompanied by an increase in vote intentions that begins a few days after the media resumed their focus on Reform. By the peak of the second surge, Reform was the partisan object of almost 25 percent of campaign coverage; a 15 point gain from where it started at the beginning of the campaign. Intentions followed coverage and peaked approximately 2 days after the coverage peak. Much of the gains in coverage and vote  87  intentions were, however, erased in a couple days as Reform ended the campaign with a smaller share of coverage and intentions. Coverage led opinion, but there is also evidence for the rule of anticipated importance. The media decided to increase their attention to Reform and voters responded. If voters had not responded, the media may not have increased their attention in such a way as to create the second surge. It appears from the decline in coverage following the first surge that the media was unwilling to continue to give Reform attention in the news if Reform's message was not affecting vote intentions. The second surge may in part reflect the fact that the news media was observing a change in Reform support. Nevertheless, the first surge and the dramatic decline in coverage at the end of the campaign were clearly exogenous to Reform intentions. One alternative possibility is that effective advertising was the principal cause of the changes in both the trends observed here. Advertising began around the time of the change in coverage but there are a number of reasons to believe that advertising was not the principle cause of the observed dynamics. The primary reason, as discussed earlier, is that Reform did not engage in an advertising campaign. This meant that Reform was limited to free time advertising (2, two minute and 4, thirty second ad plays during primetime). Voters and the media were not exposed to party advertising that was putting forward a Reform agenda. It is possible, of course, that Reform benefited not so much from its own advertising but from the attacks of other parties because the attacks were cues about where the party stood. Here the Conservative advertising strategy is relevant. The first Conservative ad that explicitly attacked the Reform party did not, however, appear until day 25 of the campaign which is well after the first surge of coverage had already occurred. If advertising mattered for the news agenda, it  88  only did so during the second surge. This follows from the fact that by the second surge Reform was more viable and therefore more likely to be a target of partisan attacks.  Which Came First: Public Support or Media Attention? The graphical evidence provides strong support for the hypothesis that coverage was leading and influencing vote intentions, which suggests that coverage was driven by media expectations that were largely divorced from real indications of popular support. Nevertheless, it would be helpful to estimate the relationship between attention and vote intentions in such a way as to be more precise in our conclusions about the causal ordering of the variables. Clearly, there is the potential for a relationship between intentions and coverage as well as its inverse. Granger Causality Test While establishing causation is never an exact science, attempts have been made to identify what it means to talk about causation when two series of temporal data are compared. The most useful for the purposes here is the definition of causality provided by Granger (1969; 1980); " A variable x is said to Granger-cause y if prediction of the current value of y is enhanced by using past values of x" (Kennedy 1997: 68). As such causality is defined in terms of the temporal ordering of the variables and, while there are situations in which temporal ordering does not mean causation (Spring department store sales do not cause spring to arrive), for the most part this definition is sufficient when combined with theoretically derived expectations. The logic of the test is to first explain as much of the variation in Y on the basis of lagged values of Y and then examine the extent to which lagged values of X can account for  89  variation in Y.  An F-test is then used to determine if the inclusion of lagged values of X  significantly improved the predictability of Y. If the F-test is significant one can reject the null hypothesis that X does not Granger-cause Y. Since causality can go in either direction, one also has to run the test to see if Y Granger-causes X . Testing the relationship between Reform coverage and intentions requires that we enter the media data so that the previous days news media coverage (t-1) is paired with the current days intentions (t). While we expect that coverage and intentions will be related, coverage does not occur until after the interviewing process has begun so they cannot be entered simultaneously.  Table 3-1. Granger Causality Test Null Hypothesis  Lag 1  Lag 2  Lag 3  Lag 4  Lag 5  Lag 6  Coverage does not Granger-Cause Intentions  9.07***  3.40**  3.16**  2.82**  2.78**  2.08*  Intentions do not Granger-Cause Coverage  1.02  1.23  1.17  0.93  0.40  0.32  DF, DF for F-Test  1741  2738  3731  4732  Approx. Threshold for Rejecting Null Hypotheses 2.8 F-Statistic(10%) 2.5 F-Statistic (5%) 4.1 3.3 F-Statistic (1%) 7.3 5.3  2.3 2.9 4.5  2.1 2.7 4.0  6,26  2.1 2.6 3.7  2.0 2.5 3.6  ***p<.001; **p<.01; *p<.05  3  For a statistical discussion of the calculation of the test see Gujarati (1995: 620-3) and Judge et al. (1988: 767-70).  90  Table 3-1 includes the results of six different Granger tests, each with a different lag structure. It is important to test different lag structures because the test is sensitive to the 4  temporal structure of the data. In other words, the effect of a change in one variable may take several days to show up in the other variable. We would expect this to be especially true of the responsiveness of coverage to changes in intentions since it may take the news media time to discover that the party's situation has changed and to alter coverage accordingly. The results show that the F value is not above the threshold necessary to reject the null hypothesis that intentions do not Granger-cause coverage for any of the six possible lag structures. We are, however, able to reject the null hypothesis that coverage does not Grangercause intentions on the basis of both the 1 and 2 lag tests. There is thus statistical support for the visual observation that coverage increased before - and therefore caused - the change in Reform share of vote intentions. In addition, there is preliminary evidence based on the lag structure of the significant F-tests that the coverage-intentions relationship has a relatively short memory. A Time-Series Model of Coverage and Intentions While the Granger causality test is supportive of a significant relationship between coverage and intentions, it does not allow one to assess the relative importance of a shift in coverage or more accurately specify the time it takes for a change in coverage to realize its full impact on intentions. One approach to the estimation problem is to make use of an "error 5  correction" estimation model, which originated in the study of economic time series but has  4  The test is run in S H A Z A M based on a procedure file developed by Diana Whistler (UBC, Revision date Feb. 16, 1994) and is available from the SFIAZAM web site, http://shazam.econ.ubc.ca/  5  Granger causality tests can often find no causal relationship even when one exists where there is cointegration (Durr 1993).  91  been successfully used in recent analyses of economics and government popularity (Johnston forthcoming; Ostrom and Smith 1993). The advantage to the error correction approach is that 6  it is able to capture both the long and short-term relationship between temporally ordered variables (Beck 1991; Durr 1993). In effect an "error correction" model is based on the notion that there exists an 7  equilibrium relationship between two series in time (equation 1); at any given level of coverage there is a corresponding level of intentions. L=KG-\  (1)  where: I is the percentage of vote allocated to Reform; and C is the Reform party's share of news coverage as a percent of all party coverage. Both variables are daily values.  Assuming that such a long-term equilibrium exists, there is always likely to be shortterm disequilibrium. Intentions may be lower or higher than we would expect based on coverage either because of a non-coverage related change or because it takes time for a change in coverage to have its impact on aggregate intentions. Rather than estimating (1), the approach in error correction is to regress change in intentions on change in coverage and an error correction term. The equation (2) thus combines both differenced variables and levels  In effect, error correction is a form of economic modeling that is based on the notion of distributed lags (Kennedy 1997) and is similar to the more advanced technique of cointegration. Cointegration is a technique that uses more than one equation to model change over time in such as way as to be agnostic about the reequilibrium process involved (Beck 1993). This raises the question of causality. As Beck notes, " I f we have causal theories, our methods should be consistent with those theories" (1993: 242). Since visual observation and the Granger test provide external information about the relationship than conforms to theoretical possibilities, the error correction strategy is more appealing. The discussion here is based on the discussion of the error correction methodology in Beck (1993) and Johnston (forthcoming).  92  variables. In contrast, traditional A R I M A models only make use of the information available in differenced variables with a loss of long-run information of the series. A / M - I = a + /3AQ-1,-2  + 5(1,-i-KCt-2)  (2)  In equation 2, P is the immediate effect of a shift in coverage on a shift of intentions. 8  The error correction term is in parentheses and is the difference between the actual and predicted values of the hypothetical estimation of equation 1 for the previous day. This means that if the difference between I and KC is positive, intentions are higher than we would expect in equilibrium. The next day, intentions should be corrected down. The size of the shift in I for a unit shift in C is captured by the size of K and the speed of the shift by 8. If 8 is small, the longer it will take for the correction to be completed.  9  In practice we cannot estimate (2) as it is and must multiply each of the terms in the error-correction term to get equation (3), which can be easily estimated. AI,,,-\ = a + pAG-i,/-2 + 8 i / - i - 8 i c G - 2  (3)  As a result, after the model is estimated, one must solve for K by dividing the coefficient for C_2 (8K) by the coefficient onI .i (8). t  t  Table 3-2 reports the results of an error correction model using the Hildreth-Liu Grid Search to remove first-order autocorrelation from the estimation. Although the traditional level of statistical significance is not met (p =0.076), the coefficient for change in coverage is in the right direction and suggestive of a positive effect on intentions that is realized immediately. Failing to meet this level is not surprising given that the relationship evident in Figure 3-3 must  8  Note that since the difference is t-1, t-2 for coverage, it is actually the next day impact of a shift in coverage.  9  The value of 8 > - l and<0. 93  burn through considerable daily noise. The equilibrium effect establishes that a unit shift of 10  coverage is associated with 0.76 percent change in the percentage of the population supporting the Reform party. The equilibrium shift takes place over about four days (8 is -0.29). The error correction estimation technique appears on its face to be a useful approach to understanding the relationship between the density of coverage about Reform and Reform intentions over the course of the 1993 election campaign. This likely represents one of the first attempts to use this econometric method for analyzing the over-time relationship between noneconomic variables, which raises the question of whether the nature of the data support the use of such a technique. At issue is the assumption that there is an underlying equilibrium relationship between the two variables.  Table 3-2. A n Error-Correction Model of Reform Attention and Intentions  Constant Immediate Effect Change in Coverage -2, t-i Error Correction Component Reform Intentions -i Coverage -2 t  t  t  b  SE(b)  2.98  (2.46)  0.26  (0.14)  -0.29 0.22  (0.13) (0.11)  Equilibrium Effect  0.76  0.46 R 2.14 Durbin Watson 45 N Estimation is by Hildreth-Liu Grid Search (AR1) 2  The only alternative is to smooth the data before the estimation, either by using a moving average in place of the actual daily value or by using a Kalmanfilterto purge the error.  94  The literature review certainly suggests that there is a relationship between party support and coverage, but is there reason to believe that the relationship qualifies as an equilibrium? Certainly, previous evidence suggests that the relationship between coverage and support is such that there is a constraint; neither series moves too far away from the other. In fact, there is likely a party-specific, coverage-support equilibrium. A party may receive more coverage than it currently has support, but the news media are unlikely to continue to devote extensive coverage to candidates or parties who remain low in the polls. The media are equally unlikely to ignore parties or candidates who are doing well because to do so would conflict with the norms of objectivity (Zaller, 1997). Although deserving of more research, an expectation for an equilibrium appears reasonable. Theoretically, the equilibrium assumption and the possibility for shocks to be media or public induced suggests that coverage and support are truly a co integrated system. The 1993 case happened to allow the use of error-correction because side information supported the claim that media decisions were the source of the shocks. Discussion  The key finding that emerges from this analysis is that media attention tended to precede electoral support for the Reform party in 1993. Despite the fact that there was no evidence that Reform had become more viable as reflected in either the published polling results or the tracking of vote intentions in the Canadian Election Study data, Reform was given an increasing share of television news space. Although we lack evidence of what the journalists and new organizations were thinking about the campaign, it seems reasonable to conclude that Reform benefited not so much because of its successes in the campaign, which  While either approach would overcome the arbitrary use of the 0.05 level, the basic 95  occurred after coverage had begun to increase, but because of where the party was located in terms of electoral competition. With the faltering Conservative campaign, Reform represented a new dimension to the horserace because it meant a conflict over the right-wing vote in Canada and because Reform was talking tough on the deficit and social programs where the Conservatives and K i m Campbell were soft. To the extent that this accounts for the news judgment, there is some support for Zaller's rule of anticipated importance. While the analysis provides the basis for an explanation of the 1993 campaign, it also offers considerable support for the argument that party coverage and party support should be understood in terms of a dynamic equilibrium. Although no research has specifically addressed this question, previous research does suggest that the two things are tied together because support is an indication of how well the party is doing and therefore its relevant  i  importance for the outcome. At the same time, parties are unlikely to be successful without coverage. The existence of a coverage-support equilibrium helps account for the findings of balance in many campaigns and the reported lack of campaign-related changes in attention to parties in Canada because the causal direction of impact runs both ways such that in most situations there is little that could happen to upset the equilibrium. Reform's presence upset the previous equilibrium because voters were uninformed about Reform so its potential support was unrealized. As the news media incorporated Reform into the story of the campaign a new equilibrium had to be established. In 1993 the impact of the dynamic equilibrium was increased support for the Reform party as the party received a larger share of television election news. While intentions adjusted to changing levels of coverage, it is unlikely that a continuous increase of news coverage  relationship so clearly evident in the graphical analysis would not be changed.  96  would continue to produce higher support for Reform. At some point, increasing levels of coverage would not be accompanied by larger levels of political support because the pool of available people for partisan mobilization would decrease. Once this point was reached coverage would decrease and level off at a new, stable equilibrium. Since voters did respond to the news attention to Reform, the media seem justified in their decision part way through the campaign to allocate greater coverage to Reform; they did anticipate, even if unconsciously, Reform support. At the same time, since the media decisions were exogenous to changes in support, media decisions about the potential electoral support of electoral competitors may have important and potentially dramatic implications for electoral outcomes. The implication of the analysis is that increased news coverage caused Reform support to rise and therefore the amount of news coverage a party receives will be critical to its support. While this may be true, it cannot be directly inferred from these results. In fact, recent experimental research found that differences in exposure or attention do not have short-term effects on party support (Norris & Sanders 1998). Without more information about the nature of the coverage (its tone and emphasis) and the voters responding to it, the actual means by which increased coverage produced a shift in support remains a mystery. Developing the individual-level implications of increased attention to Reform is the goal of subsequent chapters.  97  CHAPTER 4: GETTING THE REFORM MESSAGE OUT Reporting the Issues By making certain issues, candidates, and characteristics of candidates more salient, the media can contribute significantly to the construction of a perceived reality that voters rely upon. Weaver etal. (1992: 349) The previous chapter establishes that the election coverage significantly changed so as to increase the exposure of the Reform party and its leader Preston Manning. This change raises the possibility that by becoming a focal point for the coverage, the Reform party was at the centre of a more fundamental re-orientation of the discourse about the campaign as reflected in the news media. In particular, did Reform's increased exposure lead to an increase in the attention to issues on which Reform was strategically located? It seems unreasonable to claim that the deficit and welfare state attitudes were central to Reform's mobilization if these issues were not important in the news coverage. The first step is to use the National Media Archive data to show what issues received prominence in the news, and when they did so. This initial analysis identifies the issue agenda of the campaign and the information that was available to voters at different times. The analysis shows that jobs, the deficit, and social programs were all important at different times in the campaign. The second step is to consider the relationship between the news and public agendas. The third step is to consider the partisan distribution of news with a particular interest in Reform. Reform was clearly associated with the deficit and social programs early in the  98  campaign but later in the campaign, Reform was also associated with its positions on culture and ethnic issues. Issues in the News  Did news stories focus on some issues while ignoring others? What effect did the resulting issue emphasis have for the information that was available for voters to factor into their decisions about the candidates? What kinds of effects should issue coverage have? The media do not cover issues to educate the public. Issues become news because they allow reporters to tell stories about what is happening in the campaign. As a result issues are often incorporated into the news as part of the horserace or game aspect of the campaign. For example, the 1988 news coverage about the Free Trade Agreement - a campaign that could be viewed as a prototypical example of an issue dominated election - was framed in terms of leadership, credibility, and the horserace rather than in terms of the appropriate trade policy for Canada (Mendelsohn 1993). This means that the issues that appear in news coverage and the parties which are associated with them are likely to reflect the issue competition of the parties and how reporters are telling the story of the campaign. As the political competition changes or is perceived to have changed, the issues appearing in news stories are also likely to change. The lack of issue coverage is certainly lamented by many observers of election news coverage (Cappella and Jamieson 1997; Fletcher 1988; Leshner and McKean 1997; Patterson 1993), but this should not be lead us to ignore the dynamic possibilities and impacts of issue emphasis in campaign news. News can have effects on voter knowledge and on which issues voters consider important. Despite the lack of media emphasis on issues and public policy in election coverage, there is evidence that voters do learn from the news despite its lack of issue emphasis.  99  Significant learning of candidate positions has been found in U.S. primary (Brady and Johnston 1987) and election campaigns (Bartels 1993; Chaffee et al. 1994; Just et al. 1996). There is also an increase in awareness of party positions in the 1993 and 1997 Canadian elections (Fournier n.d.). Even a strategic frame does not prevent voters from learning about the substantive aspects of the issue. The experimental work of Cappella and Jamieson (1997) found that the frame, strategic versus issue, of the story does not affect recall of substantive issue knowledge. These findings suggest that news attention to issues, regardless of how superficial in comparison to some ideal, does matter. One of the ways that the coverage of some issues rather than others can matter is through agenda-setting. A considerable literature, mostly supportive, has developed around the hypothesis that media are able to influence the salience of issues in the mass public (Behr and Iyengar 1985; Iyengar and Kinder 1987; MacKuen 1981; Wanta 1997), including Canadian work (Jenkins 1996; Urmetzer et al. forthcoming). In both survey and experimental work it has been shown that the more attention that the news media gives to a particular issue in the news, the greater the likelihood that the public will think that issue is the most important problem for the country. A n election campaign offers one opportunity for such agenda-setting impacts. In fact, the first empirical test of the agenda-setting hypothesis involved a comparison of media and voter agendas at the end of an election campaign (McCombs and Shaw 1972). While most subsequent agenda-setting analyses have focused on non-campaign media effects, a number of studies of agenda-setting during campaigns have been undertaken (Roberts 1992; Weaver 1994; Williams and Semlak 1978). The emphasis on some issues rather than others has, of course, effects on more than just the priorities that voters attach to some issues rather than others. As the election comes to be  100  defined as about one thing - free trade for example - it becomes less about a different thing national unity and Meech Lake, with important consequences for who wins the election. In other words, the news attention to some issues rather than others can prime those issues. While the impact of the issue agenda of the election is taken up more fully in Chapter 6, it is the possibility of significant substantive effects of the media agenda that makes this analysis additionally relevant. While the agenda-setting literature tends to focus on the media as the causal force affecting public priorities, one must be careful in putting forth such a causal argument for a couple of reasons. The first is that the media are capable of "reading" public opinion and responding to it. The second is that television news is affected by the strategic behaviour of politicians and political parties. Roberts and McCombs (1994) found that the political parties largely supplied the media agenda. Parties are likely advantaged to the extent that the news media will have a hard time writing stories about issues that the politicians are not talking about. In fact, the agenda-defining role of the media in election campaigns does vary in different countries. The news media in the United States appear to be the most unwilling to allow the candidates to set the agenda for the campaign (Semetko et al. 1991) and the American news media is becoming more aggressive in substituting its agenda for that of the politicians (Lichter and Noyes 1995). Contrast this with the British news media who are more willing to allow the politicians to set the agenda by providing them with a greater opportunity to state their case in their own words (Semetko et al. 1991). In other words, less mediation and more time is the norm. In Canada, there is the suggestion that the media allow the parties to set the issue agenda (Fletcher 1987; Fletcher 1988). " A more detailed examination of the ebb and  101  flow of attention to particular issues would simply underline the tendency of the media to permit the parties to set the issue agenda"(Fletcher 1987: 361). Since the media made decisions about which parties deserved attention, the media may have also influenced the issue character of the coverage in 1993. The battle over the issue agenda reflects the inherent tension between the news media and politicians. Politicians clearly desire the opportunity to convey their message with the least mediation or journalistic comment and thus stress issues that are beneficial to themselves, campaign in ways that are convenient for the news media, and construct campaign events so as to maximize good visuals. The news media, however, do not view their role, especially in the United States, as simply the broadcaster of candidate messages. The news media have an interest in producing dramatic and interesting stories, but also in preventing themselves and the public from being manipulated by politicians seeking to define the agenda (Semetko et al. 1991). It is for this reason that recent analyses of the news media decisions focus on the news as the product on an interaction between politicians, the press, and the public. A good example of the dynamic interaction of the public, the media and the parties is the 1992 U.S. presidential election. While the Democratic campaign slogan 'It's the economy stupid" came to define the U.S. election and does reflect the actual impact of economic considerations on the vote, how the economy mattered is the subject of academic discussion. Hetherington (1996) argues that the media portrayed the economy as more negative than real economic indicators suggested its true state during the 1992 campaign. Importantly, this coverage had negative consequences for the Republican party. In contrast, Just et al. argue that Bush's attempt to shift the agenda to other issues was ineffective because of the public concern with economic conditions (1996). It appears to be the case that the economy mattered in 1992  102  because the Democrats pushed the issue, the media thought the issue was the story of the campaign and the issue resonated with the public. It would be surprising for an issue to have a large impact solely because one of the actors viewed it as important. While substantive issue coverage is not the norm for election coverage in Canada or abroad, the evidence does indicate that the issue coverage that is provided does increase voter knowledge and thereby potentially impacting on the election outcome. In fact, since issues are used as the basis for news stories an analysis of the issue coverage of an election can be particularly revealing about how the news media and politicians understood the meaning of the election. The key here is that the media and politicians are both able to influence the agenda by stressing, or covering some issues rather than others. Issues in the 1993 Election Campaign General Observations about Issue Priorities of the News Media While a vast majority of election coverage in the past and in the 1993 case was horserace coverage, with its focus on who is winning and the strategies and decisions aimed at success on election day, the remaining 34% still represents a significant amount of coverage. Of course, even this nominal issue coverage is often presented implicitly in terms of the horse race. Nevertheless, an analysis of the issue coverage over the campaign provides an important perspective on how the media constructed the campaign agenda for the voters. After briefly providing a general overview of issue coverage, the analysis turns to the temporal and partisan dynamics of issue coverage. The correspondence of the public and media agendas is also considered.  103  Figure 4 - 1 . The Distribution of Issue Coverage  25  Percentage  20.2 20  18.3 15.5  15  12.9  10  8.4 6.4 3.2  Economic Policies Deficit/Debt Other Economic Culture/Ethnic National Unity Unemployment Free Trade Social Programs External Affairs Govt Reform Note: All non-issue coverage is exluded  Figure 4-1 provides an outline of the salience of ten different issues during the campaign. Some of the categories are self-explanatory such as the deficit and unemployment categories. The other economic category is a residual category which contains coverage related to the Goods and Services Tax, income tax fairness, agriculture, privatization, and labour relations among other things. The social programs category contains general references to social policy spending and more specific program concerns such as Social Assistance; health care; day dare; education; and unemployment insurance. The culture/ethnic issues category concerns coverage of such issues as immigration and multiculturalism; aboriginal issues; general moral issues; abortion; gay rights; and crime. The label captures the fact that these issues all share a sense of the kind of Canada that people live in. Since the category is 104  dominated by references to immigration and multiculturalism, these issues are often referred to as being about ethnocentrism. Government Reform refers to coverage related to doing politics differently or changing the democratic process to facilitate more frequent opportunities for citizen participation. Given that Canada was emerging from a recession, it is not surprising that news coverage tended to focus on economic matters. The unemployment rate was approximately 11 percent and while the Conservative government had tried to reign in the deficit, the deficit continued to grow every year of their term. Almost fifty percent of the news coverage involved economic policy directly and was distributed largely within three categories: economic policy, unemployment, and the deficit. Although economic policy dominated news coverage in 1993, a non-trivial amount of coverage was devoted to social programs. In fact, one fifth of all issue coverage pertained to social programs. This is in fact an underestimate of the total impact of social programs since issue coverage of the deficit is related to social spending. A commitment to reducing the deficit raises the question and challenge of how much social spending, especially on programs Canadians value, will be reduced. Despite or perhaps because of the failure of the 1992 Constitutional referendum, national unity remained on the sidelines in the news media. While national unity was off the agenda in 1988 (Johnston et al. 1992), the conditions in 1993 actually favoured the emergence of the issue. The Liberal party seeking to build a pan-national coalition and the Conservative party trying to hold itself together had no interest in raising the salience of national unity, but this is obviously not true for the other parties. The BQ, committed to separatism, clearly focused on the unity question, while the Reform party represented the opposite extreme on the  105  question. The clear difference of positions offered the potential for news stories if the parties chose to fight on the unity issue. Perhaps recognizing a public weary of the national unity debate, the news media may have contributed to the limited importance of the issue by playing it down. Other issues received minor amounts of coverage. Unlike in 1988 and despite the recent expansion of the Free Trade Agreement (FT A) to include Mexico, free trade only received 5% of the issue coverage and not significantly more than external affairs and defence.  1  Culture and ethnic issues also received a small share of overall coverage. Temporal Dynamics The first step in disaggregating the story in Figure 4-1 is to analyze the issue coverage at different times during the campaign. If Reform was truly important for the campaign then as Reform gained in news attention or visibility, there should be a change in the overall issue agenda of the campaign. Table 4-1 breaks down the issue coverage by week of the campaign to isolate the key temporal dynamics. Dynamic changes of three of the issue categories (jobs, social programs, and culture) are then graphically represented on Figure 4-2. Although emphasis on economic issues waned by the end of the campaign, the campaign began with an economic emphasis. The five economic categories (economic policies; job creation; deficit; free trade; and other economic) contained 75 percent of the issue coverage in the first week of the campaign, but only 58 percent in the last week. In effect, issue coverage became more diverse by the end of the campaign. National unity, new politics,  Much of the coverage in the external affairs, defence category was in fact related to public spending on the EH-101 helicopters rather than broader questions of foreign and defence policy.  106  defence, and cultural issues emerged as minor, if not significant, issues in news media coverage. Issue coverage during the last week of the campaign coverage was diverse; no particular issue emerged as decisive. Job creation, the deficit, and social programs received about the same amount of coverage during this final week, which stands in decided contrast from the dynamics of coverage during the campaign. According to the allocation of issue coverage, the deficit never managed to assert itself as the dominant issue, but together with social programs managed to define one aspect of the campaign. In fact, what stands out in Table 4-1 are three changes in the issue agenda of the campaign.  Table 4-1. Temporal Distribution of Issue Coverage by Week  10-19 Sept.  Week of 20-26 27 Sept. 3 Oct. Sept.  4-10 Oct.  11 - 17 Oct.  18-24 Oct.  TOTAL  Economic Policies Job creation/Unemployment Deficit/Debt reduction Free Trade Other Economic Social Programs Culture/ Ethnic External Affairs/ Defence National Unity Government Reform  26.0 23.3 14.7 10.1 5.6 10.8 0.7 2.0 6.5 0.4  17.3 26.9 10.9 6.8 8.4 18.3 3.6 0.6 6.1 1.1  7.0 10.4 14.1 6.0 4.8 41.7 4.0 4.2 7.1 0.6  11.2 14.4 14.8 4.7 11.2 25.6 5.6 2.4 10.1  8.1 14.5 11.0 1.9 8.1 21.8 17.0 3.7 12.3 1.7  18.1 12.9 14.8 0.9 11.3 11.2 7.5 6.4 10.4 6.6  15.1 17.5 13.2 5.1 8.1 20.6 6.5 3.2 8.7 1.9  Total Percent N  100.0 769  100.0 799  100.0 645  100.0 465  100.0 781  100.0 778  100.0 4228  107  The first temporal change is the large drop in coverage about job creation and unemployment, and economic policies between weeks two and three; a drop of 17 and 10 percentage points respectively. In fact, Figure 4-2 shows that the proportion of issue coverage devoted to jobs and unemployment dropped by over 20 percentage points between September 8 and 16. In part this likely reflects the controversy about the unemployment rate that occurred when Campbell kicked off the campaign. Nevertheless the magnitude of the drop suggests that while the campaign could have potentially been oriented around jobs, the issue seem to fall out of media emphasis. The second change is the dramatic increase in the coverage, to 42 percent of all coverage, associated with social programs that began around September 20 . Of course, two th  events may have driven the increased importance of social programs. It is the middle of week 2 (September 23 ) that Campbell makes her comment that the campaign is too short to discus rd  social programs and two days later that Manning makes his comment that a Reform government would reduce Canada Health Act restrictions on the provinces. The daily tracking in Figure 4-2 reveals two peaks in the social program coverage: September 27 and October 11. While the former can be accounted for in terms of the events of the time, the latter peak is somewhat of a mystery. There is no particular controversy that triggers the latter surge and the most reasonable explanation is that the surge reflects attacks on Reform's position by other parties. It is in week 4 (October 4-10 ) that the Liberal and Conservative parties focused th  greater attention on the Reform party reflecting the party's improved viability at this time. Despite the second rise, social program coverage undergoes a gradual decline back to where it began by the end of the campaign.  Figure 4-2. Dynamic Coverage Changes:  Jobs,  Social Programs, and Culture  Percentage C o v e r a g e  40  20  0  I—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i i i i i i r~ 8 10 12 14 16 18 20 22 24 26 28 30 2 4 6 8 10 12 14 16 18 20 22 24 September  October  Moving 7-day average.  The third change is the increasefrom6 percent of week 4 coverage to 17 percent of week 5 coverage associated with culture and ethnic issues. While the daily tracking indicates that coverage of this issue started to increase as early as October 5, this is largely an artifact of the moving average approach; October 8 is thefirstsurge in coverage that is sustained. As will become clear, the increase is largely associated with the Reform party. In particular, the issue appears to have been sustained on the agenda because of the John Beck controversy and the questions that were raised about Reform's immigration policy which followed the Beck revelation on October 13. Briefly, a Reform candidate, John Beck, made several racist statements that were captured by and shown on the major television networks. The controversy appeared to have sparked increased attention to cultural policy, especially with  109  respect to Reform. The surge of coverage about cultural questions is taken up again throughout the chapter. Clearly, the campaign underwent a redefinition of sorts. A discourse organized around economic issues, especially jobs and unemployment, gave way to a more diverse agenda. Social programs dominated coverage in the middle of the campaign in a way that suggests that the whole meaning of the electoral choice for voters could have been changed. Nevertheless, the discourse about social programs was not sustained till the end of the campaign. Rather than  reftHTiing to jobs and employment, the issue discourse grew more diverse at the end of the campaign reflected in the increased coverage devoted to cultural questions that emerged over the last 14 days. Agenda Setting Changes to the issue agenda in the news media are not without their consequences. While a complete test of the agenda-setting hypothesis is beyond the scope of this analysis, it is useful to consider how news coverage reflected the priorities of the mass public and whether the changes in the news emphasis on issues coincided with changes in the public agenda. Figure 4-3 breaks down the answers to an open-ended question, "what do you think is the most important issue in the campaign," by week. Individual responses were coded such that the resulting categories are compatible with the news media coding. While there are general similarities in the public and news agenda, there are noticeable differences of issue emphasis. The public was clearly preoccupied with jobs and unemployment in the 1993 election campaign. Jobs was the modal category and was the most important problem for about 35 percent of the public. While the jobs issue was present in the media coverage, it was more dominant in the public compared with the news agenda. Only in the first week was jobs the  110  modal category of news issue coverage. This finding is not particularly striking given the unemployment rate remained high and the public's natural bias toward being concerned with jobs and the economy (Jenkins 1994). After jobs, the next most frequent responses were the deficit and social programs followed by the economy and taxes. Not surprisingly only a small percentage of the public felt that national unity was the most important problem. Here the public and the news media basically agreed; the media coverage rarely touched on national unity despite the presence of the B Q .  2  Cultural issues were also rarely the most important issue  for voters. Figure 4-3. Public Perceptions of Most Important Problem by Week of Campaign  510-19 Sept nm20-26 Sept M27 Sept-3 Oct M4-W Oct • 11-17 Oct H18-24 Oct  50  25  National Unity  Deficit Jobs  Taxes Other Economic  Culture Social Programs  Note: NAFTA and Election Issues not included in figure but part of calculation.  2  •  Evidence has previously demonstrated that the salience of national unity in the mass public is particularly related to news coverage, which suggests that had the media increased its attention the public may have responded (Jenkins 1994). ill  Dynamically a number of things occur to the issue priorities of the public. Although it remains a top priority, jobs declines in importance. In its place is an increased importance of the deficit, social programs, and taxes. The increased importance of taxes is related to the question of the appropriate role of government and therefore related to the increased importance of the deficit and social programs. Unfortunately the content analysis data of news coverage does not adequately distinguish between taxes and other economic concerns which prevents a comparison of the two trends. Figure 4-4 identifies the dynamic changes in the three most important issues which are subsequently compared to the amount of coverage in the news media of these same issues. Figure 4-4. Public Concern for Jobs, Social Programs and the Deficit Percentage Public Concern  40  20  Social Programs 0 10 12 14 16 18 20 22 24 26 28 30 2 4 6 8 10 12 14 16 18 20 22 24 September October Moving 7-day average.  112  There is more than a passing resemblance between news attention to social programs and the public's belief that social programs are the most important campaign issues. A n approximately 15 percentage point change in coverage is followed closely by an approximately 10 percentage point change in the proportion of the public identifying social programs as most important. Figure 4-5 places the two series on the same graph with different scales in order to identify the relative change and the temporal ordering of the change in the two series. At least in the moving average approach, the issue agenda of the news media changes before the public agenda with the public responding to the first increase in news media emphasis while being less affected by the second peak in social program coverage. In addition, when social program coverage dropped off at the end of the campaign public concern with the issue did not return to its pre-rise level. Figure 4-5. News Attention and the Importance of Social Programs Percentage Public Concern  Percentage Coverage  30 15  15  0  0  10 12 14 16 18 20 22 24 26 28 30 2 4 6 8 10 12 14 16 18 20 22 24 September  October  Moving 7-day average.  113  The relationship between news attention and public importance of jobs and unemployment is also significant if less dramatic (Figure 4-6). After devoting a plurality of coverage to the jobs question in the first couple of days, the news media reduced its importance overnight. From September 15 onward, the jobs issue is then a stable proportion of coverage with a slight downward trend. The same basic pattern is evident in the mass public. It takes, however, fourteen days for the public importance of jobs to realize its full decline. This is consistent with a public gradually changing its evaluation of the most important issue in response to the diffusion of information about other issues that opened up when the media reduced its own emphasis on jobs and employment.  Figure 4-6. News Attention and the Importance of Jobs  60  Percentage Public Concern  Percentage Coverage  40  40 20  20  0 10 12 14 16 18 20 22 24 26 28 30 2 4 6 8 10 12 14 16 18 20 22 24 September October  Moving 7-day average.  114  There are only two changes in the news agenda that are not mirrored in the public agenda. The first is the lack of a change in the amount of coverage of the deficit. While the deficit rises in importance for the public, there is no increase in coverage of the deficit. Since a concern with the deficit is related to a desire to reduce social program spending, the increased public concern may reflect the increased attention to social programs. Voters opposed to social program spending are likely to believe that it is the deficit which is more important and therefore might quite reasonably respond to the news discourse about social programs by identifying the deficit as more important. The second difference is the increased attention to cultural issues in the news during the last two weeks of the campaign, which is not evident in the public agenda. While Figure 4-3 suggests a smali increase in the importance of cultural questions, it is very small and overall less than four percent of the population cited this issue as the most important. It is not surprising that the public agenda did not change in response to increased news attention to cultural questions. In the absence of an immigration related crisis, it would be surprising if voters thought an issue with very small direct impacts on day-to-day life would be more important than economic concerns.  3  Three categories of the public agenda, jobs, the deficit, and social programs, all undergo significant changes. Consistent with changes in coverage, jobs became less of a priority for the public after the first week of the campaign, and there was a significant midcampaign boost in the importance of social programs. In addition, the public became more  3  The agenda-setting literature has recognized that some issues are more susceptible to media influence than others (Jenkins 1994; Yagade and Dozier 1990; Zucker 1978). For example, unemployment is particularly resistant to agenda-setting (Behr and Iyengar 1985; Jenkins 1996; MacKuen 1981). The likely reason is that voters care so much about this issue anyway that they can monitor the situation themselves and do not rely on agenda-setting kinds of cues. 115  concerned with the deficit as the campaign evolved. Since real-world conditions remained static, there is support from these observations that the media played an agenda-setting role during the campaign. Of course, change in media coverage is a necessary but not sufficient condition for observing change in public attitudes. Changes in coverage did appear to reflect and potentially cause real changes in the issue discourse of the campaign as reflected in public perception data. This is significant since too wide a divergence would make the hypothesis that coverage affected attitudes unreasonable. Partisan Dynamics While the general analysis of news media issue coverage provides a broad picture of the public's exposure to issues, campaigns get constructed in such a way that parties come to be associated with particular issues. Table 4-2 breaks down the issue coverage by partisan object. Partisan object is based on the coding of which party was being talked about in the coded news. A party could thus be the associated with an issue either because it is talking about its position, defensively or aggressively, or because another party or a journalist is talking about its position on the issue. For example, Campbell defending the government's plan to purchase helicopters would be considered a Conservative partisan object, while Campbell attacking Reform's deficit plans would be considered a Reform partisan object. Because some of the news coverage reflects statements that characterize the public as a whole or provide background information about the issue, some of the coverage did not have a partisan object.  116  Table 4-2. Partisan Distribution of Issue Coverage Party Associated with Issue in Media PC Liberal NDP Reform Economic Policies Job creation/Unemployment Deficit/Debt reduction Free Trade Other Economic Social Programs Culture/ Ethnocentrism External Affairs/ Defence National Unity Government Reform  16.0 18.0 17.9 3.4 9.2 24.4 2.6 6.1 1.1 1.4  25.4 25.9 8.5 6.8 11.5 11.5 3.7 3.4 2.6 0.7  9.0 18.1 6.7 18.7 10.0 29.8 3.0 1.0 0.7 3.0  11.2 8.7 13.7 0.3 4.0 35.5 16.7 0.4 4.6 5.0  BQ 8.7 4.0 4.6 2.5 4.6 4.0 3.4 1.5 66.3 0.3  No Party Total 12.5 23.7 15.8 4.7 6.2 15.1 8.0 3.1 9.3 1.6  15.5 18.3 12.9 5.0 8.0 20.2 6.4 3.2 8.4 2.0  N  1140 940 299 744 323 929 4431 Coverage associated with other parties (e.g. Green party) are included in the total category but not represented in the table (n=56). For the most part, parties are associated with the issues with which one would expect them to be associated. The NDP pattern is that the party was significantly more likely to be associated with free trade than the other parties. The NDP election strategy was to focus on this issue, but despite its efforts, the party failed to shape the media agenda and as the previous section indicated, this emphasis did not resonate with the public. Given its position on the Left in Canada, the NDP was also associated with social programs and not associated with the deficit. The Liberals were most associated with economic policy, especially with respect to job creation and unemployment reflecting the title of the Red Book, Creating Opportunity. The Liberals were less associated with the deficit and social programs than the Conservative and Reform parties. The Conservatives were associated less than the Liberals with economic policy and job creation but significantly more with the deficit and social programs. Naturally, almost all of the coverage associated with the Bloc was related to the national unity issue.  117  The Reform party stands out, however, from the other parties. Thirty-five percent of Reform coverage was associated with social programs; 11 and 24 points more than the Conservative and Liberal parties respectively. Reflecting its preoccupation with the deficit, Reform was also more associated with the deficit than it was with unemployment. Reform was also distinct in the large proportion of coverage associated with the cultural/ ethnic issue category reflecting its unique position on multiculturalism, immigration and crime. These associations are consistent with Reform's ideological underpinnings discussed in Chapter 2. Dynamics of Reform News Coverage General Trends Given the temporal and partisan characteristics of issue coverage, it is useful to consider how the issues associated with Reform changed over time. Table 4-3 breaks down the issue coverage associated with Reform by week of the campaign in terms both the absolute and relative distribution of issue coverage. In this case, the sample of coded material includes any situation where Reform, either as the target of partisan attack, the object of a journalist or anchor statement, or the spokesperson of a quoted or paraphrased statement, was paired with an issue. The basic pattern is consistent with the Reform column in Table 4-2, which restricted the analysis to those statements for which Reform was the object, but considering both types of association increases the number of cases. During the first couple of weeks the Reform party received little coverage and what coverage there was tended to be consistent with the general news agenda. A third of the statements had to do with job creation and unemployment, suggesting that Reform was not having a distinct impact on the issue focus of the campaign. A l l of that changes as the campaign entered week 3 (September 27-October 3). Reform began to be associated with the  118  deficit issue and social programs. In both the third and fourth week of the campaign (October 4-10), more than half of the statements associated with the Reform party concerned social programs as compared with 4 and 9 percent respectively in the first two weeks. The change is particularly evident in the change in the absolute number of times the Reform party was associated with social programs in the news (second half of Table 4-3 and Figure 4-7).  Table 4-3. Issues Associated with the Reform Party by Week Total  WEEK  A. Percentage within Week Economic Policies Job creation/Unemployment Deficit/Debt reduction Free Trade Other Economic Social Programs Culture/ Ethnic External Affairs/ Defence National Unity Government Reform Total Percent B. Absolute N within Week Economic Policies Job creation/Unemployment Deficit/Debt reduction Free Trade Other Economic Social Programs Culture/ Ethnic External Affairs/ Defence National Unity Government Reform Total N  10-19  20-26  Sept.  Sept.  24.1 33.3 14.8 1.9 3.7 3.7 3.7 0.0 9.3 5.6 100 13 18 8 1 2 2 2 0 5 3 54  27 Sept.3 Oct.  28.7 23.0 23.0  4-10  11-17  18-24  Oct.  Oct.  Oct.  8.0 9.2 4.6 0.0 0.0 2.3 100  5.0 7.9 10.1 0.0 3.6 58.3 2.2 0.0 11.5 1.4 100  8.8 6.4 15.2 1.6 1.6 56.0 6.4 2.4 1.6 0.0 100  9.3 6.6 11.0 0.0 4.0 31.3 26.9 0.0 6.2 4.8 100  14.0 2.7 16.0 0.7 6.7 22.0 28.0 0.0 1.3 8.7 100  12.5 9.7 14.1 0.6 4.5 33.9 15.3 0.4 5.0 4.0 100  25 20 20 1 7 8 4 0 0 2 87  7 11 14 0 5 81 3 0 16 2 139  11 8 19 2 2 70 8 3 2 0 125  21 15 25 0 9 71 61 0 14 11 227  21 4 24 1 10 33 42 0 2 13 150  98 76 110 5 35 265 120 3 39 31 782  1.1  119  The changes in Reform coverage can only partially be explained by Reform's own activities. It is at the end of the second week that Manning speculated about the implications of Reform's position on health care. Since it occurred early in the second week, the announcement of Reform's Zero in Three deficit reduction plan on September 20 did not th  appear to be the direct stimulus for the increased attention to these issues. Although it remained an important dimension of Reform issue coverage, the dominance of social programs was not sustained through the final two weeks of the campaign. The relative measure indicates that social programs declined in importance in the second last week, while the absolute measure suggests that the number of times Reform was associated with social programs in a week did not change until the last week of the campaign. The fall-off in social program coverage was made up by the emergence of cultural issues. Figure 4-7. Key Changes in Reform Issue Coverage Number of Mentions  100  1  i  10-19 Sept  27 Sept-3 Oct 20-26 Sept  11-17 Oct 4-10 Oct  18-24 Oct  Weekly value is the number of times in which the Reform party was associated with the issue in the television news as either the speaker of the coded material or the object of someone else's comments.  120  Both in a relative and absolute sense, the Reform party was more directly associated with culture and ethnicity issues in the last two weeks of the campaign compared with the previous four weeks. These issues - immigration, multiculturalism, moral issues (abortion, gay rights), crime, aboriginal issues - represented one quarter of the issue coverage associated with Reform during the last two weeks of the campaign. Figure 4-7 captures the shift in dramatic fashion. As the absolute amount of coverage changed, so did the issue cues that were available to the public. Cueing the Deficit and Welfare State It is clear from a general perspective on the issue coverage of the campaign that the Reform party was associated with a clear anti-deficit message that acknowledged the consequences of that commitment for social program spending. In contrast, the Conservative party staked out moderate ground. By advocating clear and unique positions the Reform party was open to partisan attacks which reinforced the cues that Reform was committed to a reduced role for government in society. In this way, coverage of social programs should have contributed to voter understanding of Reform's distinct position. It will be recalled from Chapter 2 that the Reform party position was to reduce the deficit to zero in three years. This was to be accomplished by significant reductions in program spending including programs which are considered sacred cows by many politicians. This was most clearly evident when Manning suggested that universal medicare could be reformed. Those attentive to the news had the opportunity to pickup on Reform's commitment to deficit reduction and reduced social spending. Preston Manning: Is it possible to cut the deficit and to balance the budget without reforming transfers to individuals and provinces? Doesn't somebody have to bite that bullet? There's not — you can't do it by defence cuts and wish cuts (CBC, October 4, 1993).  Kevin Tibbies: And Manning maintains that medicare is not the only social program in need of an overhaul. He points to unemployment insurance, saying it should be run like any other insurance program, with no government funding. He admits that may likely lead to higher premiums and lower benefits (CBC, October 7,1993). The clear position-taking on the deficit and its consequences is clearly different from that of Campbell and the Conservatives. As discussed in Chapter 2, Campbell was unwilling to acknowledge that a substantial reform of social programs would be necessary to meet deficit reduction targets. In the news media, Reform's plan to reduce the deficit in three years received a considerable amount of criticism from two sources. The first source was non-partisan. In several stories, the media reported that economists did not believe that a three year target was credible because of the effects on the economy of reducing expenditures so drastically. The clearest evidence of media-induced criticism of the zero in three plan is a story that ran on October 14 about the deficit promise that fell under a segment called "Reality Check." This series sought to critically examine various claims by the parties throughout the campaign. In this segment, an economist states that "Three years strikes me, frankly, as much too fast, at this point in the economy, near a take-off, just too fast. We don't need to do it that fast. And if we try to do it, we're going to hurt ourselves, possibly badly" (CBC, October 14,1993). As the reporter notes, radical deficit reduction would likely lead to an increase in unemployment and a decline in consumer confidence, which would lead us back into a recession. The reality check segment was a clear critical approach to Reform's plan, but there were a number of other similar examples. Four days earlier, Pamela Wallin began her interview with Preston Marining by focusing on the deficit plan.  122  Wallin: A lot of economists in this country have looked at your plan, they have looked at the growth projections in this country, they have looked at the unemployment rate, and they have said that you would push us back into a recession and put thousands more on the unemployment rolls (CBC, October 10,1993) Manning dismissed the idea that the cuts were high enough to damage economic growth, but the stories tend to be critical of the approach. So while the media spent considerable amount of time on the deficit as a campaign issue, Reform's specific plan came under scrutiny for being too extreme. The second type of criticism was partisan. Partisan opponents of Reform certainly challenged the Reform approach in the same manner as the economic elites shown in news media coverage, but they spent more effort focusing on the consequences of the deficit reduction plan for social programs and unemployment. "Zero deficit in three years. What will that mean? It will mean zero growth. That will mean zero jobs. And that will mean zero hope" (Jean Chretien, C T V and C B C News, October 9 ). Conservatives also tried to attack Reform th  positions on social programs and the deficit by portraying his cuts as drastic and ill-conceived. Illustrative of these attacks is a Conservative ad which featured a magician performing a trick on a stage with the following voice-over: Preston Manning thinks that governing is an easy trick. Just pick a spot and start cutting: health care, unemployment insurance, old age pensions. You can make governing look pretty simple if you are prepared to cut and run. From the middle of the campaign onward, the Conservative party distributed its partisan attacks equally between the Reform party and the Liberals, attempting to attack the frontrunner, while fighting off the Reform insurgency. The Liberals also started aiming attacks at the Reform party after the half way point and the social program and deficit issue were key points of attack.  123  Reform's deficit plans came wider considerable scrutiny, but there is good reason to believe that the criticism served to activate underlying preferences as well as driving voters away from Reform. Consider that Reform support continues to grow even during the period of most sustained social program coverage. The expert criticism of the economic consequences of the plan may have undermined Reform's message, but it should be remembered that grasping this critique required considerable sophistication given that it conflicted with the common sense notions that were central to the Reform message. The partisan criticisms that focused on the implications for social programs were easier to understand but they could only undermine respondents who held a pro-welfare state orientation towards politics. Since Manning was willing to concede much of the social program implications of his deficit program, the partisan attacks should have limited the movement to Reform among moderate small " C " conservatives, but it was unlikely to move people who had already decided to support the Reform party. Early on NDP leader, Audrey McLaughlin, seized on Manning's remarks about health care and suggested that they would determine the outcome of the election. McLaughlin may have been right, but only in the sense that a news discourse that emphasized welfare state issues offered Reform a primary role in the campaign. As the only advocate of fundamental change, Reform demanded attention. Cueing Racism and Intolerance The late campaign shift in the news to feature Reform's position on cultural affairs can be in part linked to the coverage of the John Beck affair that took place on October 13 . It will th  be recalled from Chapter 2 that John Beck was a Reform candidate in Toronto who was asked to resign from the party on October 13 after Manning was challenged by protesters with th  charges that Beck had made racist comments. The controversy appeared to signal a change in  124  the way parties attacked Reform and how the media covered the party. A reading of the television news transcripts suggests that the importance of the John Beck controversy went beyond the facts of the specific case. Earlier discussion establishes that the general increase in news coverage of cultural and ethnic issues began around October 8 , which is several days before the Beck controversy. th  Nevertheless, the issue generated considerable news attention itself and was a highly visible example of the kind of issues with which Reform was associated. Of all of the cultural issue coverage during the latter two weeks, 50 and 88 percent respectively were either about immigration or multiculturalism. It appears that in the latter two weeks of the campaign Reform became more associated in the news with its positions on multiculturalism and immigration. The television news certainly made much of the Beck incident. The exchange between Beck and a C B C reporter (Figure 4-8), which appeared in the first C B C story about the controversy revealed that Beck clearly believed what he was saying and had not just made an inappropriate comment that he regretted. The C T V story on the 12 contained no quotes from th  Beck and just showed a man challenging Manning with Beck's statements. On the next day, however, Beck is quoted as saying, "I feel that the gentile people are under attack" (CTV, October 14 ). th  It should be acknowledged that the Reform party acted quickly to distance itself from Beck, but the controversy was transformed from a single event to symbolize something about Reform. In effect, Beck was evidence for what many believed all the time; that Reform members were extremists. While Manning could deal with Beck, he had more trouble explaining that Beck was an exception. Over a of couple days, numerous references were  125  Figure 4-8. Exchange Between John Beck and CBC Reporter on October 13,1993  John Beck You have $150,000 on a guy coming into buy a citizenship in Canada, to create a job, fine, he's bringing something to Canada, but what is he bringing? Death and destruction to the people. Paul Adams You said the things we printed? John Beck Yes, I did. Paul Adams Did you say those things? John Beck Yes, I did. Paul Adams At best, they're inflammatory. John Beck At best, they're the truth, and I stand on the truth. I feel that if an immigrant comes into Canada and gets a job for $150,000, he is taking jobs away from us. Paul Adams Who is us? John Beck The gentile people. Paul Adams The white people? John Beck The white people.  made to a 42 page questionnaire that potential candidates were required to fill out in order to weed out candidates who were not compatible with the party. Beck claimed that he filled out such a questionnaire, but the party was unable to confirm it. Manning also suggested that "this was one of these weaker ridings that was desperate to get a candidate" (CBC, October 14) as a rationale for how Beck slipped through the screening process. Regardless of the accuracy of  126  these explanations, their presence on the news over a couple of days transformed a single incident into a much longer controversy. Because Beck was identified with a controversial policy area, the story from the start was about more than a racist individual. It has perhaps always been true that the Reform party's positions on immigration, multiculturalism, and crime could be labelled ethnocentric and even racist which suggests a similarity to other new right parties (Betz 1993b; Betz 1994; Kitschelt 1995). A number of Reform's immigration positions are certainly controversial. The party advocates a drastic reduction in the number of immigrants and tougher measures to fight illegal immigrants and refugees. According to Manning, "the Charter ought not to apply to people who get to Canada illegally" (CBC, October 13). At the same time, Manning has tried to distance the party from racism and intolerance. On the news, Manning is seen making the following statements: We have made every effort to ensure that particularly our constitutional positions, our immigration positions, and our cultural positions are not only utterly free of racist connotations but are anti-racist. This is the reason we support the constitutional model that talks about the equality of all citizens rather than tying entitlements to racial considerations or linguistic ones, or gender ones (CTV, October 13). We've done everything to try to make sure we're, there's nothing racist in the basic positions of the party and we're prepared to take steps against people with those views that try to infiltrate the party. We've got to I think just keep saying that over and over again until people believe it (CTV, October 15). While he states he is trying to distance the party from racism, Manning uses the language of exclusion. Identity-based claims are clearly incompatible with the Reform vision. One can imagine that in the wake of the Beck comments, Manning's assurances would appear less credible, but more importantly, the fact that Manning had to continually stress the party's position on racism meant that he could not stick to his deficit and anti-welfare state message.  127  For the remainder of the campaign, the Reform party was clearly vulnerable on the immigration issue and it appears that others exploited this vulnerability. On October 15, CTV reported that the Council of Canadians asserted that the immigration policies of the Reform party were racist. A short story on McLaughlin's attack on Reform's immigration policy appeared on October 19. On the same day, both networks ran stories about Reform's response to recent partisan attacks. In both stories, Manning is attacking the other parties for practicing "the politics of fear" and lying about the Reform party stands. In both stories, however, there is a reminder of the controversy and new coverage that at least some people think Reform is racist. Both stories refer to the presence of protesters outside a Reform rally and the C T V story shows protesters chanting, "Racists, sexists, anti-gay; Reform party bigots go away." In effect, the report that large crowds are attending Reform meetings is overshadowed by the story which is framed in terms of Manning's response to other party attacks on Reform policies, especially immigration and French Canada policy. There is a danger that one makes too much of the John Beck controversy. The party acted quickly to control the damage and there was no other information in the media to suggest that this was anything but an isolated incident. In addition, the Reform party stands on immigration, multiculturalism and French Canada were not new. Earlier in the campaign there was some modest coverage of Reform's position on cultural questions and some early indications of potential problems. On October 1, a news story about Reform reported very briefly on a protest outside a Reform rally in which the protesters were chanting, "Immigrants in! Reformers out!"(CBC). A n incumbent Conservative MP, Dorothy Dobbie was also in the news for her criticisms of the Reform party on cultural questions. First on C T V on October 4 , th  it was reported that Dobbie, "called him [Manning] a dangerous man who would impose his  128  Christian fundamentalist values on the rest of the country." Later she asked whether "there is room for people of other - of other persuasions, of other faiths [in Reform's Canada]?" (CBC, October 11,1993). It appears, therefore, that as Reform became more viable its cultural policies became a more important issue for the election. Although national unity did not figure into much coverage, the French Canada card and culture more generally were played by both the Liberals and the Conservatives at the end of the election. With two strong regional parties, the Liberals spent the last several weeks trying to ensure a majority by stressing their national credentials. "It is divisive, he [Chretien] said for Reform to run candidates everywhere except Quebec. Reform's language policies are divisive too, he said. Interviewed in Toronto later, Chretien warned opposition to bilingualism could split the country apart" (Denise Harrington, C B C , October 9, 1993). At times, Campbell also pointed to Reform's failure to run nationally. Conclusion The rise in coverage of Reform documented in the previous chapter corresponded with a change in the issue agenda of the campaign and the issues associated with the Reform party. Reform's presence as an active player in the construction of the news meant the public received greater exposure to Reform's position on the role of government in society. As is clear from both the qualitative and quantitative evidence, it was not just that voters were exposed to the Zero-in-Three deficit plan. They were exposed to the implications and the ideological-basis to the Reform commitment to a reduced deficit. It seems reasonable to conclude that Reform was aided rather than penalized by the attacks of the other parties, who framed their attacks in terms of threats to social programs. Reform rose in the polls throughout the period when Reform was  129  most associated with social programs and a significant portion of that coverage was partisan attacks. The campaign also exposed voters to information that allowed them to associate Reform with its ethnocentric positions on cultural questions. In the last couple weeks of the campaign, apparently assisted by the John Beck controversy, the Reform party became associated with its positions on cultural questions. Specifically, the party was attacked and defended its position on multiculturalism and immigration. The shift was large enough to show up in the temporal distribution of issue coverage of the campaign so it was not just an ephemeral campaign event. It is, of course, striking that this change in coverage took place at the same time as the media were reducing Reform's coverage in the news media and as Reform's share of vote intentions was declining. One can conclude from this correspondence that the news media recognized the implications of Reform's cultural positions for its potential support. The issues in the news do appear to be fundamentally associated with the way the campaign is going for the parties. As the media decided to increase their attention to Reform, they began to tell different stories about the campaign; stories that involved the issues of social programs and the deficit. Possibly because Reform had become more viable and/ or because other parties viewed this as a vulnerable area, the issue discourse shifted a second time. Since attitudes about cultural accommodation and attitudes about the welfare state can be quite independent the two shifts in coverage may have had dynamic effects on voter knowledge about and voter support for the Reform party.  130  CHAPTER 5: LOCATING REFORM Party Issue Positioning and Voter Learning ... what voters know or don't know of the candidates and the policy issues in any election is largely determined by the information, and the clarity of that information, presented to them during the presidential campaign. Alvarez (1997a: 25).  There is no doubt that news decisions fundamentally altered the information environment of the 1993 campaign. The Reform party became more visible in news coverage and that coverage was characterized by an emphasis on the welfare state early on and cultural issues in the latter weeks of the campaign. Did this changing coverage matter? In other words, did the coverage actually teach voters about Reform? To demonstrate the impact of changing coverage on the information that voters held about the Reform party, the analysis proceeds in four steps. The first step is to compare voter uncertainty about parties, leaders, and issue positions using self-reported measures of knowledge and the patterns of non-response to questions. This will establish the degree to which voters learned about the Reform party during the campaign. The second step is to analyze the dynamic evolution of party placements across the campaign. The level of uncertainty is replaced with a concern with the implications of learning for perceptions of party locations. The third step is to estimate a series of regression estimations to identify the individual correlates of voter perceptions of Reform. The fourth step is to compare knowledge of Reform during the Referendum campaign with knowledge during the election campaign to further identify campaign learning. 131  Locating Political Parties Since it is knowledge of party stances that is pursued in this chapter, it is useful to briefly consider previous findings with respect to campaign learning and the process by which voters impute locations to parties. While there are individual-level processes associated with knowledge of party locations, a significant aspect of locating parties is related to the context of the election and the flow of information. One of the potential influences on the ability of voters to make accurate identifications of candidate issue stands are the actions of the candidates themselves. In particular it has been suggested that candidates intentionally attempt to obfuscate; that is, they avoid taking firm policy positions and rely largely on appeals to broad, shared goals (Page 1976; Shepsle 1972) or valence issues rather than position issues (Budge 1993). The benefit of making vague, as compared with direct, issue statements, is not, however, universally recognized in recent empirical research. Voters penalize parties for uncertainty (Alvarez and Franklin 1994; Battels 1986), which suggests that any benefits from avoiding issue positions could be lost. While there may or may not be strategic benefits of obfuscating on issues, candidate action does appear to either foster or prevent issue clarity (Campbell 1983; Pomper 1972; Alvarez 1997a). As Franklin (1991) notes, contrary to expectations, "candidate campaign strategies have significant effects on the clarity of voter perceptions" (1210). Voters are more likely to know the location of a party on an issue that the party has taken both a distinct and vocal position. If context matters, it is not the only factor in the knowledge voters haVe about parties. It is common to observe or lament the lack of information about political affairs among the mass public (Bennett 1995; Converse 1964; Delli Carpini and Keeter 1991; Luskin 1987) and  132  Canadians are no different from other mass publics (Fournier n.d.). Despite the lack of factual knowledge as revealed by these studies, in recent years the focus of considerable research effort has been on understanding how voters reach decisions based upon minimal information and even rational ignorance (Popkin 1991; Sniderman et al. 1991). While it remains an understudied area in electoral studies, there are some limited studies of how voters locate parties. Three processes have been identified as important determinants of where and if voters are able to locate parties: party and ideological cues; issue importance for respondents; and projection. The flow of campaign information provides a number of different kinds of cues that allow voters to infer locations for the parties. For example, poll results and endorsements are said to provide crucial information about the spatial locations of the parties (Lupia 1992; McKelvey and Ordeshook 1986). Of course, the range of possible cues is quite large. Ideological and party cues like liberal or right-wing can influence inferences about where a candidate stands on issues because "placing a candidate in one or more of these categories allows the voter to use the beliefs related to that category to infer specific candidate characteristics" (Conover and Feldman 1989: 916; Rahn and Cramer 1996). Even candidate characteristics such as gender, race or religion (Popkin 1991; Sniderman et al. 1991) can cue voters to assign a party to one position rather than another. While the reliability of various shortcuts has been challenged (Bartels 1996), it is a reasonable expectation that voters would generally be able to locate the parties relative to each other given past experience and the structuring effect of party identification. Given the complexity of political affairs, one way to manage information is to specialize. This raises the possibility that each respondent has knowledge about a number of  133  specific issues rather than knowledge about every issue. On any one issue there will be a corresponding group of citizens who are concerned with and therefore monitor the situation and candidate positions. Converse (1964) labelled this phenomena "issue publics" and others have identified empirical support for the existence of issue publics (Elkins 1993; Krosnick 1990b). We then expect those respondents who care about an issue to be better able to locate the parties on that issue. In fact, the more one cares about an issue the greater one perceives the differences between the parties (Krosnick 1988; Krosnick 1990b). If one cares about an issue, then one values even small differences. One of the easiest pieces of information that voters can use to infer issue locations for a party is their own location. Rather than the perception of the party's position determining the utility for that party, it is possible that voters who like a party will locate that party closer to themselves. Research into this possibility does suggest that survey respondents factor in their own evaluations of how much they like a party when they locate that party on the issues (Conover and Feldman 1982; Johnston et al. 1998; Judd et al. 1983; Krosnick 1990a; Markus and Converse 1979; Page and Jones 1979). These effects can emerge either because voters "project" their own position onto candidates they like to avoid cognitive imbalance or because voters judge candidates in self-referential terms (Conover and Feldman 1989). Projection or rationalization effects as they are variously labeled add a wrinkle to our understanding of voter perceptions of the parties since any change in perceptions may be as much a product of changed evaluations as of learning. The same factors related to where voters locate parties are also relevant to whether or not a respondent is even willing to impute a location to the parties. Respondents can answer don't know to questions about where the parties stand on various issues. Although respondents  134  are generally capable of identifying where parties are located on the issues, there is considerable willingness to acknowledge that one is unaware of the party stances. Wright and Niemi (1983) attempt to predict whether a respondent will impute a party location based on the respondent's characteristics, including media use, interpersonal conversation, strength of partisanship, the intensity of affect for the candidate being located, and group membership. Each of these factors represents a different basis for inferring locations: media use and interpersonal characteristics capture exposure to campaign information; strength of partisanship captures the availability of partisan or ideological cues; intensity of affect captures projection related inferences; and, group membership captures group-based cues. No single variable is the most important suggesting that there are multiple ways that voters can make a location inference (Wright and Niemi 1983). While the model is reasonably good at predicting whether a respondent will impute a position or not, the failure of some predictions and the variation across candidates in the importance of the independent variables reinforces the importance of context for location perceptions. Most of the literature on locating parties, reflecting its American study area, focuses on inferences about the location of relatively unknown candidates who represent one of the two established parties. History and short cuts certainly provide respondents with a reasonable basis to generalize about the issue commitments of established parties even when the candidate is relatively unknown, but the public should have more difficulty identifying the positions of a new party. New parties lack an existing base of party loyalists and must also counteract the lack of information that voters have about their issue location. As a result, they are often more dependent on the exposure they receive in the mass media (Zaller and Hunt 1995).  135  Measuring Party Locations The 1993 National Election Study includes six questions that ask respondents to locate the parties relative to each other on a five point scale. The six questions concern the following issues: place of French in Canada; power for unions; aid for racial minorities; women's equality; closeness to the United States; and, deficit reduction. The question wording varies depending upon the issue being asked about, but the questions share a common framework. Respondents are asked to identify each of the parties' positions on a five point scale (-2 to 2), where the middle category is the status quo (e.g. same as now). For example, one of the questions used throughout the analysis asks how much each of the parties wants to do to promote the French language in Canada (much more, somewhat more, same as now, somewhat less, or much less). The deficit location question is slightly different, but the scale is the same; respondents are asked what effect a (Conservative/ Liberal/ NDP/ Reform) party win would have on the size of the deficit. While the question asks specifically about the deficit, the question has broader implications. The deficit can be reduced three major ways: increased taxation; economic growth; or, reductions in expenditures. Given the antipathy to increased taxes and the high speed of economic growth that would likely be required to increase revenues, it seems reasonable to expect that voters would understand the latter as the principal means of deficit control. As such, this question can also be understood as a question of credibility. The smaller a respondent believes that the deficit will be the higher the credibility the party has both as a deficit fighter and more generally as a party committed to the reduction of government spending and social programs. As Chapter 4 makes clear, the deficit is intrinsically tied to government spending and programs in the political discourse. The evidence does indicate that  136  respondents had little difficulty locating the parties on this dimension suggesting that it is tapping a fundamental aspect of political conflict. Locating the Parties in 1993 Trouble Locating Parties One expects respondents to have considerable trouble locating the Reform party and Preston Manning, especially at the beginning of the campaign. Of course, one would expect that the ability to locate parties would increase as the campaign progresses. The following analysis demonstrates the scope of the uncertainty attached to Reform compared with the other party positions in three ways. The first step is to compare respondent's perceptions of their own knowledge of the party leaders. The second step is to consider the frequency of a nonresponse to questions that asked respondents to rate the parties and leaders of the parties. Finally, the uncertainty about party locations is considered. Evidence from self-reports about how informed respondents were of the party leaders shows that most of the public felt aware or very aware of Chretien and Campbell, but only minorities felt this way about Manning and McLaughlin (Figure 5-1). Both the minor party leaders suffered from uncertainty earlier on and both became better known at about the same rate as the campaign progressed. The close relationship between self-perceptions of McLaughlin and Manning are somewhat surprising. It is true that this was McLaughlin's first campaign and she seemed to have a small pre-election profile so one might have expected some voter uncertainty, but given the lack of coverage of McLaughlin and the NDP one would not expect the campaign to lesson respondent's uncertainty about her. Of course, asking respondents to evaluate their knowledge of candidates may produce a gradual increase over the campaign as a function of voters having become more certain of their vote preference.  137  Figure 5-1. Self-Reported Knowledge of Party Leaders  Weekly Mean  1 '  10-19 Sept  27 Sept-3 Oct 20-26 Sept  11-17 Oct 4-10 Oct  18-24 Oct  Self-reported knowledge is derived from question asking respondents to rate their own knowedge on a four point scale from "Not at all informed" to "Very informed." The four item scale was transformed into a five point scale (0-4) with "Don't know" responses coded at the neutral point.  Evidence of the actual behaviour of respondents suggests that Manning was the key beneficiary of the campaign. During the campaign survey, respondents were asked to rate both the parties and the leaders on a 100 point scale. This is not a cognitive ly difficult task since it does not require the respondent to justify or explain his or her evaluation. In Figure 5-2 and Figure 5-3 the percentage of the respondents who were unable to rate the parties and leaders respectively is plotted by week of the campaign. For the most part, respondents were able to answer the questions with less than 5 percent giving the "don't know" response. In both figures the only exception is Reform. The frequency of "don't knows" for the question rating both the Reform party and Preston Manning declines across the campaign by about five percentage points. There is no substantive variation in the likelihood of saying don't know for  138  the other parties or candidates. In particular, there is no change in the NDP or McLaughlin trends suggesting that the earlier growth in self-reported knowledge of these objects is not reflected in this more concrete measure of awareness.  Figure 5-2. Unwillingness to Rate Parties by Week of Campaign Percentage Don't Know  10-19 Sept  27 Sept-3 Oct 20-26 Sept  11-17 Oct 4-10 Oct  18-24 Oct  Weekly value is the percent of the sample w h o answered "don't know" to the party feeling thermometer question.  A further demonstration of the uncertainty associated with party locations comes from the frequencies of "don't know" to questions of the party's stands on issues evident in Table 61. Also included is the nonresponse to the question that asked the respondent for his or her position on the issue (self-placement). The nonresponse to the self-placement is important because of the ability of respondents to locate themselves but also because those people who were unable to locate themselves were not asked to locate the parties. For this table, only those  139  respondents who were interviewed during the first three weeks of the campaign were included to identify what people knew based on as little as possible campaign experience. The frequency of nonresponse to the issue questions varies considerably from 50 percent who could not impute a Reform position on closeness to the United States to a low of 8 percent who could not impute a Conservative position on the deficit. There appears to be two kinds of variation: across parties and across issues.  Figure 5-3. Unwillingness to Rate Party Leaders by Week Percentage Don't  Know  10  10-19 Sept  20-26 Sept  27 Sept-3 Oct  4-10 Oct  11-17 Oct  18-24 Oct  Weekly value is the percent of the sample who answered "don't know" to the leader feeling thermometer question.  While sizable proportions of the public were unable to locate the major parties, this was particularly true for the questions about the issue stands of the Reform party. In every case, more respondents answered that they "did not know" the Reform party's stand on the issue  140  compared with the other parties. About 40 percent or more of respondents were unable to identify Reform's position on aid to racial minorities, promotion of the French Language, power to unions, and closeness to the United States. For the Liberal, Conservative, and NDP party locations the average is significantly smaller; often by 10 percentage points or more.  Table 5-1. Frequency of Nonresponse to Party Location Questions (Weeks 1-3) Party Asked About Conservative Liberal  Self Placement  Reform  Aid to Minorities  15.6% (1443)  43.8% (1202)  35.8% (1202)  33.9% (1203)  37.3% (1202)  Support for French  19.1 (1443)  39.0 (1160)  28.2 (1163)  32.2 (1161)  44.6 (1157)  Union Power  22.4 (1446)  38.6 (1116)  29.5 (1122)  25.6 (1122)  23.5 (1119)  Women's Equality  11.1 (1443)  38.5 (1400)  26.6 (1404)  24.1 (1404)  26.4 (1401)  Closeness to U.S.  11.3 (1447)  50.1 (1278)  24.3 (1279)  19.1 (1280)  32.2 (1274)  Deficit  Not Asked  25.4 (1439)  9.0 (1446)  8.4 (1443)  14.8 (1439)  NDP  Cells represent the frequency of the "Don't Know" response for the self-placement or the party location question on that issue. Those people who did not answer the self-placement were not asked to place the parties. N in brackets. N varies reflecting changes in the refusal category and the response to the self-placement.  Respondents found it easier to locate all of the parties on the deficit question compared with the other issues, which may in part be related to the different style of question. Despite the fact that it was easier to locate the parties on this issue, the gap between the proportion of don't know responses for the Reform and other parties is both absolutely and proportionally larger for the deficit than the other issues. As will become clear, the campaign served to significantly reduce this gap.  For the most part, respondents supplied a nonresponse to questions on which one would expect greater ignorance. The gap between the number of don't knows for the Reform party compared with the major parties, especially on the deficit, represents a fundamental difference in the uncertainty attached to Reform party positions. Not only did voters shy away from imputing issue locations to Reform, but they were more likely to recognize that they had no opinion about how positive or negative they felt about the Reform party and Preston Manning. The campaign did manage to reduce uncertainty about the party and leader so we should expect that the same process at work for Reform's issue stands, the dynamics of which are taken up in the following section. Where the Parties Stand The locations of the parties in issue space, especially with respect to attitudes about French Canada and the deficit, has already been the subject of some analysis (Brady and Johnston 1996; Johnston, Blais, Gidengil, Nevitte, and Brady 1994) and discussed briefly in Chapter 2. In this section, the aggregate perceptions of issue stances are considered before examining the temporal changes in where voters located the parties. The analysis reveals that the Liberal and Conservative parties were clearly perceived as similar on most issues; Reform started off distinct on ethno-cultural questions; and, the campaign mattered for the evolution of Reform's position on the deficit. Figure 5-4 provides a visual representation of the campaign wave perceptions of where the parties were located on the issues. Each party is represented by the mean. Since the questions were asked in the campaign wave of the survey, the mean represented on Figure 5-4 is also an average across the campaign. The analysis will turn later to the dynamics of these perceptions. Those people who did not know where to locate the parties were assigned a  142  neutral value since not knowing is effectively an inability to distinguish the party from the status quo (Brady and Johnston 1996). This convention is continued in the graphical analysis 1  throughout this chapter. For the most part, the Liberals were perceived as a slightly left of centre party. Desiring to do more for racial minorities, French, women, and unions while adopting a more nationalist position on Canada-United States relations and allowing the deficit to grow. None of this portrait of the Liberals is particularly surprising. On each of these questions, the NDP was perceived to be more left or nationalist except on the question of the French language reflecting its ambiguous and strategic problem with the issue (Johnston et al. 1992). What is striking is the relative closeness of the Liberal and Conservative parties in the eyes of the public. The public may be misguided about the real differences but on the whole, the Conservative party was perceived as almost as pro-minorities, pro-French, and pro-women as the Liberals. Possibly most important was the low credibility of the party on the deficit; over a quarter of respondents thought the deficit would get bigger under a Conservative government. Only on union power and continentalism was the party perceived as clearly to the right of the Liberals. Of course, continentalism had faded in importance since the 1988 election and the question of unions was not a salient basis of conflict. It appears that in the eyes of the voters, the Conservative party was rooted in the centre, thus leaving its right flank vulnerable to a new party.  1  The assumption is that when differentiating parties Tor anything else) people begin at the status quo. They then ask themselves, "is the party different from the status quo?" If the answer is yes, they then ask whether how far the party is from the status quo. I assume that the cognitively difficult question is the first and it is this question that leads some people to say they do not know. The decision between somewhat and a lot more should not create a situation that provokes a don't know response. 143  Figure 5-4. Mean Perceptions of Party Locations on Selected Issues Perceptions of French Language Location REF  PC LIB Do More  Do Less  -1  NDP  1—*  -0.75  1  -0.5  1  0  -0.25  0.75  0.5  0.25  Perceptions of Aid to Minorities Location REF  PC  Do More  Do Less l  -1  LIB NDP  -0.75  1  -0.5  r 0  &i  -0.25  0.25  0.75  0.5  Perceptions of Women's Equality Location REF  PC LIB  NDP Do More  Do Less  -1  -0.75  -0.5  0  -0.25  0.25  0.5  0.75  Perceptions of Union Power Location LIB  REF PC  NDP  Have More  Have Less  •1  -0.75  -0.5  0.25  -0.25  0.5  0.75  Perceptions of Closeness to the United States Location NDP  PC  REF LIB  Closer  Farther Apart  -1  -0.75  -0.5  I  -0.25  0.25  I  0.5  T  0.75  Perceptions of Size of Deficit PC LIB  REF Smaller  -1  NDP Bigger  -6-  -0.75  -0.5  -0.25  0  0.25  0.5  0.75  144  Throughout the campaign, Reform was perceived as distinct on the question of promoting the French language and aid to minorities. Almost half of the respondents who offered an opinion (46%) felt that Reform wanted to do less for racial minorities compared with only a small fraction who believed the Conservative party wanted to do less (11%). On the question of promoting the French language, Reform was particularly distinct. Over 70% of those who offered an opinion felt Reform would do less; over half of which expressed the most extreme position that the party wanted to do much less. Although it would become more credible on the deficit as the campaign progressed, Reform was not particularly credible on the deficit during the first three weeks. The party was perceived to be at the same place as the Conservatives on the question of union power. On the question of closeness to the United States, or continentalism, Reform is perceived to share with the Liberal party a slightly nationalist or anti-free trade position. This perception is clearly inaccurate given that the Reform party is not opposed to freer trade. Clearly Reform was distinguished on ethno-cultural questions not conservative positions generally. While more people viewed the party as wanting to do less for women compared with the other parties, it did not occupy an extreme position on this question; the mean is at the neutral point. To demonstrate the distinctness of the Reform party on the question of promoting the French language and aid to minorities during the campaign, Figure 5-5 and Figure 5-6 show the smoothed daily mean for each series centred at the neutral position (same as now) for the Liberal, Conservative and Reform parties. While there is some movement in each series across the campaign such as the more extreme and distinct position of Reform on the question of aid to minorities that emerges at the end of the campaign, the important thing to note at this point is the distance between the Reform party and the other  145  F i g u r e 5-5. P a r t y L o c a t i o n s o n F r e n c h C a n a d a  Daily mean  -1  I—i—i—i—i 10 12 14 16 18 20 22 24 26 28 30  2  4  6  September  8  10 12 14 16 18 20 22 24  October  Moving 7-Day Average. "Promoting the French language in Canada; How much does the [Liberal/ Conservative/ Reform] party want to do?" F i g u r e 5-6. P a r t y L o c a t i o n o n A i d t o M i n o r i t i e s  Daily mean  0.5  i  0.25  0  -0.25  -0.5  |—i—i—i—i—i—i—i—i—i—i—i—i—i 10 12 14 16 18 20 22 24 26 28 30 2 4 6  September  i i i i i i i r 8 10 12 14 16 18 20 22 24 !  October  Moving 7-Day Average. "Does the [Liberal/ Conservative/ Reform] party want to do much more, for racial minorities, somewhat more, about the same as now, somewhat less, or much less?"  parties on each of these questions stays about the same. The distinctiveness was clearly a precampaign phenomenon. The campaign did not change the fundamental relationship between where the Reform party was located relative to the others. While the movement on ethno-cultural questions is relatively small across the campaign, there is significant movement during the campaign on the question of Reform's deficit credibility. Credibility is measured by the location question discussed above. Most observers would have identified the Reform party as the party most committed to deficit reduction so this location question can be thought of as a measure of the credibility of the party in reducing government expenditures. Figure 5-7 presents the daily mean, averaged over 7 days, for the three major parties' perceived credibility on the deficit.  Figure 5-7. Perceived Credibility on Deficit Daily Mean Rating  i r i i i i i i i i i i i i i i i i 10 12 14 16 18 20 22 24 26 28 30 2 4 6 8 10 12 14 16 18 20 22 24 September October Moving 7 Day Averages  147  Figure 5-8. Don't Knows on Deficit Party Location Question Proportion Don't Know  10 12 14 16 18 20 22 24 26 28 30 2 4 6 8 10 12 14 16 18 20 22 24 September  October  Moving 7-day average. Daily value is the proportion which did not know what effect a party winning would have on the size of the federal deficit.  Reform is the key to Figure 5-7. At the start of the campaign, the Reform party was perceived to be as credible as the Conservative party on deficit reduction. By the second week of the campaign, however, the Conservative party had lost much of its early credibility on the deficit while Reform's credibility surged. So despite Campbell's attempt to sell her brand of fiscal conservatism with a heart, voters did not in the end see the party as any more credible on deficit reduction than the Liberals. The Liberal party was never considered a credible party on the deficit and the public's perception of its position never underwent a re-evaluation. Much of Reform's early evaluation on the deficit was certainly susceptible to reevaluation. Twenty-eight percent of respondents interviewed in the first ten days of the campaign were unable to say what effect a Reform win would have on the size of the federal deficit. For comparison, the proportion who did not know what effect a Liberal or 148  Conservative win would have was ten and nine percent respectively. Figure 5-8 graphs the proportion of "Don't Knows" for the four parties. Whereas the campaign clearly reduced the uncertainty associated with the Reform party, it only marginally reduced the uncertainty for the NDP's credibility and had no effect on perceptions of the Liberal and Conservative parties. Between 5 and 10 percent of respondents were unable to locate the two major parties on the deficit issue, which is stable across the campaign. By the end of campaign, only 15 percent of respondents gave a don't know to the Reform location question, which is higher than the nonresponse for the other parties but only half the size of the nonresponse at the beginning of the campaign. It appears voters learned about Reform. Summary As the 1993 election campaign began, there was considerable uncertainty about Reform's issue positions and about Manning himself. Nevertheless, on ethno-cultural questions the party was perceived as distinct from the other major parties. One could even suggest that they were extreme except that voters in English Canada were closer to Reform than the other parties on the question of support for French (Brady and Johnston 1996). As the campaign progressed, the party continued to be perceived as extreme on ethno-cultural questions but there was little evidence that the public became more likely to impute an extreme position. On the question of the deficit, the underlying uncertainty that existed at the beginning of the campaign appeared to have contributed to an aggregate perception of Reform that was close to the Conservatives. During the campaign voters re-evaluated their perceptions in such a way as to distinguish the party clearly on the deficit-role of government dimension.  149  Getting it Right? Individual Determinants of Party Location Early in the Campaign The discussion at the beginning of the chapter suggests a number of variables that should be related to where someone will locate the parties. People are likely to use their own attitudes about the issue, their party identification, and their feelings about the parties in the imputation process. In addition, one would expect other variables related to the acquisition of information to be related to party locations. This section considers the relationship between these variables and voters imputations of Reform's location on French Canada and the deficit. Again, the analysis is restricted to the first three weeks of the campaign to limit the influence of the campaign on perceptions. The effect of the campaign is considered more fully in Chapter 7. The analysis proceeds by using Ordinary Least Squares regression to estimate a party location model for each issue. hk = a + frAi + fcR* + p [// + $*{Rik * U,) + psP; + peA + e 3  (1)  where: Lik is the location respondent / imputes to Reform on issue k (French or deficit); A; is respondent / score on standardized awareness measure; Rjk is location respondent / imputes to him/herself on issue &;  2  U ; is respondent /'s feeling thermometer score for Reform; Pi is a series of party identification dummy variables; and, Dj is a series of demographic variables including region, education, and income. Since the dependent variables both vary between positive and negative with a true neutral point, the model assumes that there exists a "true" location for the party that is off-centre. This  2  For the French issue, the respondent's position is identified using a comparable question to the party location question. Unfortunately the 1993 election study did not ask voters to situate themselves on how much the deficit should be cut. To represent voter's own position, the anti-welfare state index is used in its place. 150  assumption is not difficult to justify given the discussion of Reform's issue stands in Chapter 2 and 4; Reform staked out the most anti-deficit position of the parties in 1993 and was the only party to oppose the Charlottetown Accord. Awareness is included because those people generally aware of political information should be better able to locate the true locations of the parties, to the degree to which a party can be said to have a true location. Party identification and demographic variables are entered to capture group and party inferences. Party identification should be used carefully in any treatment of the 1993 election because it is not truly exogenous to the campaign, but since the analysis is based on the first half of the campaign, it is unlikely to be affected by the campaignrelated changes in Reform identification. The respondent's own attitude is included to capture the fact that someone who is more anti-welfare state, may perceive party locations differently than those who are more moderate or supportive. Feelings toward the party in question are included to capture one aspect of the projection hypothesis and because the more one likes a party the more one is also likely to know about it. Also included is an interaction between a person's own attitude and their feelings toward the party. This is an attempt to recognize that a person's own attitudes about an issue will be more influential if one likes the party. Admittedly this is a generally unsatisfactory way of identifying projection effects. The model proposed by Markus and Converse (1979) is better suited and has in fact been used to identify projection effects in the 1993 election study (Johnston, Fournier and Jenkins 1998). The problem is that the absence of a comparable measure (one in the same metric as the party location question) of a respondent's own position on the deficit makes the calculation of the Markus/ Converse model impossible.  151  Table 5-2. Locating Reform Early in the Campaign (Weeks 1-3) Deficit Credibility Awareness Voter Location (Welfare State; French Canada) Feelings toward Party Voter Location * Feelings Party Identification Conservative Liberal NDP Reform Demographics Education Income Woman Atlantic West  Intercept  0.711 *** (0.135) -0.426 (0.239)  0.006 * (0.003)  0.015 ** (0.005) 0.077 (0.078) 0.048 (0.077)  0.269 * (0.110) 0.203 (0.126)  0.032 *  (0.015) -0.060 (0.039) 0.015 (0.062) 0.120 (0.092) 0.064 (0.065)  -0.776 *** (0.171)  0.16 Adj-R2 0.90 SEE 1000 N * p<.05; ** p<.01; and *** p<.001. Standard error in parentheses. Significant coefficients in bold face.  French Canada -1.076 *** (0.148) -0.110 (0.057)  0.003 * (0.002)  0.003 * (0.001)  -0.166 *  (0.084) -0.136 (0.085) -0.065 (0.119) -0.192 (0.140)  -0.030 (0.017) -0.014 (0.043) 0.074 (0.068)  0.222 (0.101) -0.037 (0.071) -0.006 (0.160) 0.12 0.90 885  152  The results of the estimations for the deficit and French Canada placements are included in Table 5-2. As expected, awareness plays a strong role in locating Reform on both issues but especially French Canada. It seems that the politically aware distinguished the Reform party from the others because they had more information about it, but that information was more available for Reform's position on Quebec. In addition, post-secondary education also correlated with placement of Reform. As expected there is some effect of voter's own ideological position and his or her rating of the party being evaluated on where the party is located. Generally speaking, the more you like the party and share an anti-welfare state orientation, the more credible you think the party is on deficit reduction. Similarly, those who both like the party and are themselves antiFrench identify that party considerable more anti-French. This is modest support for the idea that voters pull parties they like closer to their own position on the issue. Those people who like the party are also more likely to perceive the party to be extreme (i.e. unique relative to the other parties). Surprisingly, there is no independent effect of voter location independent of the 3  interaction with feelings. The analysis reveals only minimal effects of party identification on perceptions of party locations. Imputations for both issues are basically unaffected by party identification variables except that N D P identifiers perceive Reform to be more extreme on the deficit than non or other party identifiers and Conservative identifiers perceive Reform to be more anti-French than others. Similarly, there are no systematic regional or gender-based differences in where the party is perceived controlling for feelings and the respondent's own position.  3  Since on average the Reform party is more extreme (off-centre) on both issues compared with the other parties, it is likely that the more you like the party the more you will increase the distance between Reform and the other parties. 153  The Referendum Campaign and Awareness of Reform's Ethno-Linguistic Position Between 1987 and 1993 Manning and the Reform party did not only stress ethnic and cultural issues, but the Meech Lake Accord and the subsequent Charlottetown Accord gave the public an excellent opportunity to associate the Reform party with a distinct position on unity and cultural questions. Government had clearly left the Conservatives identified as a party of the centre - almost indistinguishable from the Liberals. If the distinctiveness of Reform during the early stages of the 1993 campaign was the product of information received before and during the 1992 referendum campaign, then there should be a relationship between what people knew in 1992 and their subsequent opinions. Fortunately the referendum-election study panel provides one way of testing the hypothesis that voters learned Reform's position on French Canada during or before the referendum campaign. During the referendum campaign, respondents were asked whether Manning had taken a position on the Accord. Responses to questions of this type were used by Johnston et al. (1996a) to show whether Manning was visible during the campaign and to assess the impact of his intervention. We should expect that those people who were aware of Manning's anti-Accord position would be more likely to know in the early weeks of the election campaign that the Reform party was likely to do much less for the French language i f it formed the government. It may also be true that people would understand the party's antiAccord position as an expression of its ethno-cultural position more generally since the Accord raised questions of Canadian identity.  154  Figure 5-9. Effect of Awareness of Manning's Opposition to Charlottetown on Perceptions of Reform's Position on French Language Daily Mean  10 12 14 16 18 20 22 24 26 28 30 2 September  4  6  8 10 12 14 16 18 20 22 24 October  Moving 7-Day Average. Knowledge of Manning's Position on Charlottetown Accord is based on 1992 survey.  Before estimating the relationship, Figure 5-9, graphically simplifies the relationship between respondents knowledge of Manning's opposition to the Accord measured during the 1992 referendum campaign and perceptions of Reform's position on promoting the French language. Clearly, those people who were aware of Manning's opposition were significantly more likely to rate the party as distinct and extreme. Only 21% of respondents who knew Manning was opposed to the Accord when they were interviewed in 1992 did not know whether the party would do more, less or about the same to promote the French language. If the public learned Reform's anti-French position as a result of the increased salience of that issue during the failed Meech Lake Accord and subsequent referendum campaign on the Charlottetown Accord, then those respondents who correctly identified Manning's opposition  155  to the Accord during the referendum campaign should be more likely to perceive the party as anti-French. To test this proposition equation 1 was re-estimated with the addition of a coefficient representing whether the respondent correctly identified Manning's opposition to the Charlottetown Accord when he or she was interviewed in 1992. The sample is thus limited to respondents who were interviewed during the referendum campaign and the first three weeks of the election campaign. The basic pattern to the identification of Reform's issue positions 4  remains the same and the results support the proposition that knowledge of Reform's position can be traced to information learned before or as a result of the Charlottetown referendum. Knowing Manning's opposition to the Accord when surveyed the previous year is associated with a change of -0.30 in the perception of the party's position on support for the French language (Table 5-3). Controlling for this specific piece of information does not reduce the effect of awareness.  5  Manning's opposition coefficient is a clear link back to the referendum period and suggests that everyday politics did nothing to fundamentally alter the perception of the party on ethno-linguistic lines. Any information learned between 1992 and 1993 maintained the fundamental difference between those who knew and those who did not  In this estimation strategy the referendum wave is treated as a single point of time even though the respondents were interviewed randomly over the campaign. Since knowledge of Manning's opposition grew throughout the campaign, there is measurement error in the "knowledge of Manning's opposition" variable. The nature of the measurement error should serve to suppress the size o f the coefficient leading us to underestimate the effect of knowing Manning's opposition. 5  Since the sample size drops from 835 to 358 I am hesitant to make too much of the fact that the awareness coefficient increases in magnitude and the fit of the model is better. 156  Table 5-3. Knowing Manning's Position in 1992 and Locating Reform in 1993 Deficit Credibility  French Canada  Aware of Manning's Position on Accord in 1992  0.126 (0.116)  -0.261 * (0.124)  Awareness  0.702 ** (0.245) -0.440 (0.393) 0.006 (0.004) 0.015 (0.008)  -0.839 *** (0.252) -0.036 (0.101) 0.008 ** (0.003) 0.003 (0.002)  0.121 (0.131) -0.004 (0.133) 0.160 (0.180) 0.106 (0.221)  -0.142 (0.139) -0.106 (0.140) -0.037 (0.185) -0.157 (0.233)  0.060 * (0.025) -0.049 (0.065) 0.205 * (0.102) 0.297 (0.155) 0.110 (0.111)  -0.096 *** (0.026) 0.095 (0.068) 0.003 (0.108) 0.259 (0.167) -0.051 (0.118)  -1.169 *** (0.295)  0.072 *** (0.268)  Voter Location (French; Welfare State) Feelings toward Party Feelings x Location Party Identification Conservative Liberal NDP Reform Demographics Education Income Woman Atlantic West  Intercept  Adj-R SEE N * p<.05; ** p<.01; and *** p<001. Standard error in parentheses. Significant coefficients in bold 2  0.17 0.90 358  0.17 0.87 298  face.  157  know Manning's opposition when interviewed in 1992. Further evidence that knowledge of Manning's position is not a general awareness of the party's issue priorities is provided by the replication of the model with Reform's deficit position as the dependent variable. Knowing Manning's position on the Charlottetown Accord in 1992 was unrelated to perceptions of the party on that issue. Is it Knowledge or Projection? The evidence based on the cross-sectional analysis in Table 5-2 suggests that both projection and awareness are related to where the Reform party is located. The question is whether the clarification of Reform positions during the campaign, especially on the deficit, is a function of learning. It could be argued that some or all of the movement in the perception of Reform locations reflects the fact that Reform becomes better liked. That is, the causal relationship between knowledge of Reform positions and support for Reform is reversed. Those people who like Reform are more likely to locate the party closer to themselves. While projection is a potential influence on party location, there are a number of reasons for believing that the evidence for Reform is actually learning. Firstly, it is not clear why only one issue location measure undergoes a fundamental change as Reform becomes liked by a greater proportion of the electorate. If projection was a significant influence, then perceptions of Reform's position on French Canada should have moved as voters who previously disliked Reform perceived the party to be far away from their true position. There is no evidence that as Reform became better liked that its mean location shifted very much.  158  Figure 5-10. Normalized Changes in Reform Support and Deficit Credibility  10 12 14 16 18 20 22 24 26 28 30 2 4 6 8 10 12 14 16 18 20 22 24 September  October  Moving 7 day average. Both lines are normalized by subtracting themean for the first 15 days from the daily value score. Credibility is measured on five point scale while Reform support is 100 point feeling thermometer.  Secondly, it appears to be the case that in the aggregate perceptions of Reform's credibility changed before support for Reform changed. Rather than using Reform share of vote intentions, the analysis uses the daily mean feeling thermometer scores for the Reform party since this is the standard approach to identifying a respondent's utility for a particular party. The normalized curves, placed on different axes reflecting the different scales, suggest that the public became more aware of Reform's deficit position before it changed its evaluation of the Reform party (Figure 5-10). The results of a Granger causality test, reported in Table 54, are indicative of such a temporal ordering. The evidence suggests that in the aggregate, credibility changes occurred before changes in Reform support as reflected in our ability to reject the null hypothesis that credibility does not Granger-cause Reform support. According to the Granger test there is no statistically significant reciprocal effect of changing support on  159  intentions. If the changed perception of Reform's position was largely a product of projection then one would expect support to change before any change in the ability to locate the party. Table 5-4. Granger Test of Relationship between Reform Credibility and Support Null Hypothesis  Lag 1  Lag 2  Lag 3  Credibility does not Granger-Cause Rating  2.24  2.18  4.53** 3.26** 2.98** 2.81*  Rating does not Granger-Cause Credibility  1.02  0.32  1.35  1.28  1.23  1.03  2738  3~735  4~32  5/29  6~26  2.5 3.3 5.3  2.3 2.9 4.5  2.1 2.7 4.0  2.1 2.6 3.7  2.0 2.5 3.6  DF, DF for F-Test  i~4i  Lag 4  Lag 5  Lag 6  Approx. Threshold for Rejecting Null Hypotheses F-Statistic (10%) F-Statistic (5%) F-Statistic (1%)  2.8 4.1 7.3  *** p < .001; ** p < .01; * p < .05 Projection is likely an important part of the story of the precise location of where Reform is located, which raises concern in any estimation that makes use of voters' perception of the candidates as a predictor of party support. Estimates of projection here are, however, conservative estimates since they do not attempt to take into account the obvious causal impact of location on support. Nevertheless projection is unlikely to be the driving force across time in the changed perception of where the Reform party stood on the deficit question. Instead, reflecting the importance of awareness for where the Reform party is located, learning in response to new information is the most likely explanation. Conclusion In the years preceding the 1993 election the Reform party was perceived to be extreme on the question of the place of Quebec in and the cultural character of Canada. While it may have wished to moderate its position, it certainly wanted to appeal to a broader ideological  160  group of Canadians. This meant establishing itself as a party committed to a reduced role for government in the lives of its citizens. The evidence is consistent with the idea that the Conservative party had left itself vulnerable to such a new right appeal on both the cultural and role of government dimensions. Before the campaign started, Reform had established itself as the party of the right on the cultural pole. It was perceived as both anti-French and anti-minorities. The evidence suggests that those who learned Manning's opposition to the Charlottetown Accord in 1992 were more likely to recognize this distinctiveness. During the election campaign the party also managed to establish its distinctiveness on the role of government dimension. In this case, it appears that it was the increased coverage part way through the campaign that allowed the party to gets its anti-deficit message out to the politically aware and anti-welfare state group of the public. This hypothesis is taken up more comprehensively in Chapter 7. Despite Reform's electoral breakthrough in 1993, the evidence suggests that knowledge of its issue positions was limited to the politically aware. It was the politically aware who viewed the party as distinct on ethno-linguistic matters and the deficit during the early weeks of the election campaign. There is thus good reason to believe that uncertainty about Reform among less aware respondents may have prevented them from voting Reform. The only thing that increased the likelihood of knowing Reform's distinct French Canada position was previous learning of Manning's opposition to the Charlottetown Accord. While this reinforces the importance of the media in the process by which new parties emerge, it also suggests that new parties need to overcome the limited reach of the media attention they get. In fact it may be necessary for parties to be associated with a highly visible series of events in order to get its message out.  161  CHAPTER 6: PERSUASION, PRIMING AND LEARNING The Dynamic Impact of Issues for Reform Support .. .by priming some considerations and ignoring others, television news can shift the grounds on which campaigns are contested. Priming may therefore determine who takes office - and with what mandate - and who is sent home. Iyengar and Kinder (1987: 121). What effect did the campaign have on Reform support and the considerations that voters brought to bear on their electoral choice? This chapter considers the changing ideological basis to Reform support during the campaign. The evidence provides convincing evidence that the news coverage discussed earlier changed the basis of Reform support at different times with implications for who would support the party. It also, however, develops the role of priming as compared with learning as separate processes that contribute to the activation of underlying issue dimensions in campaigns. In chapter 2 a number of ideological factors were identified as contributors to Reform voting. Reform was said to have benefited from alienation and its populist leanings as well as its anti-French, anti-minorities, and fiscally conservative positions. Of these three ideological factors related to Reform voting, the evidence points to the conclusion that the campaign mattered because of the activation of welfare state and deficit attitudes (Johnston et al. 1994). It was on this issue that learning of Reform's deficit position occurred. Certainly coverage, especially that of Reform, focused on the deficit issue which likely helped to activate these predispositions. Since coverage also emphasized Reform's position on minorities and cultural  162  accommodation in the latter period of the campaign, we need to assess how this coverage impacted on the importance of attitudes about minorities for Reform support. There are three ways that the information in the news media and the campaign more generally could have increased the importance of issues for Reform support. The first, persuasion, is that voter could have changed their positions on the welfare state and other issue dimensions. Persuasion could have moved voters closer to Reform's position on the issues and changed the underlying relationship between attitudes and evaluation. The second, priming, is that attitudes toward the welfare state and later cultural questions could have become more salient as the result of the campaign. The third, learning, is that the campaign provided voters with information about the party's stand on the issues that enabled voters to link their underlying attitudes to their party evaluation. The evidence does not support the persuasion hypothesis. While the results confirm that attitudes about the welfare state and macroeconomic policy became more important for Reform as the campaign evolved, the public as a whole did not become more fiscally conservative. Early in the campaign, voters who were ideologically compatible with Reform in the sense of holding fiscally conservative attitudes on the deficit and social programs were not as strongly pro-Reform as they would come to be at the end of the campaign. It is also true, however, that ethnocentric attitudes also became more important in the latter weeks of the campaign. Despite the apparent relevance of the priming hypothesis to these changes, the evidence suggests that the growing importance of the welfare state is more a product of the diffusion of issue information about Reform than priming per se. Priming rather than learning is more important, however, for unlocking the rise of importance of attitudes toward minorities.  163  Activating Preferences: Priming or Learning? The prirning hypothesis began as an explanation for fluctuations in presidential approval and has been more recently applied to the electoral context. Similar to agendasetting, prirning is generally considered a second order media effect. While the media are not able to change people's minds, they are able to influence political judgments, usually presidential approval, by extensive coverage of one policy area rather than another. Those issues that receive large amounts of coverage are likely to be particularly strong determinants of a person's overall judgment. The prirning hypothesis expects that the more attention the media pay to a particular issue area, the more citizens will incorporate what they know about that domain into their overall judgment. We know this to be the case from experimental work (Iyengar et al. 1984; Krosnick and Kinder 1990) and survey evidence (Edwards III et al. 1995; Krosnick 1988; Krosnick and Brannon 1993) on presidential approval. In one experiment, researchers found 1  that those subjects exposed to stories on energy policy "weighed their attitudes toward Carter's energy policy performance more heavily in their evaluations of Carter's overall performance than those who watched no such stories" (Miller and Krosnick 1996: 85). Similar effects were observed across time in survey research. Those people interviewed after the Iran-Contra revelations were more likely to factor their attitudes about the Contras and government policy in Central America (Krosnick and Kinder 1990). This obviously helps account for apparently stunning changes in aggregate support for the President in the United States. For example, President Bush's impressive approval rating during the Gulf War evaporated as voters turned their attention to other matters.  1  For a review see Miller and Krosnick (1996) 164  Evidence for priming is usually in the form of a change in the magnitude of a coefficient on an issue or attitude from time t to t+1. At time t the issue is not factored into the person's judgment while at t+1 it is or vice versa. The change in the coefficient across time is possible because the range of relevant considerations in any choice situation exceeds the cognitive abilities of most if not all citizens. One of the easiest way to simplify the choice is to focus on that information which is accessible in memory or is at the top of one's head (Zaller 1992) . The accessibility of some attitudes rather than others is ripe for media and candidate 2  influence. In fact, it is an emphasis in the media that is usually said to cause certain issues to be primed. Evidence for priming in election campaigns is more limited. Iyengar and Kinder (1987) provide some limited experimental support for priming of economic assessments and candidate qualities. Bartels (1997a) found consistent evidence of priming of economic perceptions in the campaigns between 1980 and 1992, but only modest issue priming largely confined to 1988. This latter finding could, of course, be attributed to the generally weak evidence of campaign effects in presidential elections. There is some evidence that campaigns activate existing political predispositions in both American (Finkel 1993; Gelman and King 1993) and German elections (Finkel and Schrott 1995), which is consistent with a primingbased model of campaign effects. Mendelsohn (1994) found evidence of priming in the 1988 Canadian election. In this case, voters were primed to increase the importance of leadership rather than issues. Blunt et al. (1998) argue for priming effects in a study of a 1991 Senate election. In this case a relatively unknown Democrat was able to overcome a significant gap in  2  The priming literature tends to emphasize memory-based models rather than on-line models in conceptualizing the priming process. Such an emphasis is not necessary since on-line models would also predict that voters' on-line tally would be affected by increased attention  165  vote intentions by supposedly emphasizing the health care issue. The notion of prirning in this latter case is weaker in form than the standard presentation because it does not link changes in campaign communications to changes in attitude importance and does not consider other hypotheses. Despite the lack of systematic empirical evidence for prirning in campaigns, prirning is the kind of change process that may be more likely than direct forms of persuasion for a number of reasons not the least of which is that it is consistent with how we understand the rhetorical efforts of candidates in campaigns. Party platforms contain, "assertions of the importance of various problems" (Budge 1993: 47) and this would suggest that the campaign is not about communicating spatial location, but rather an attempt to inform voters about the appropriate basis for choice. In election campaigns, "politicians prime issue to provide people with reasons for supporting them" (Johnston et al. 1992: 5). Prirning also provides the conceptual starting point for understanding Kennedy's 1960 use of issues and polling data to develop an electoral image (Jacobs and Shapiro 1994). In recent study of the 1992 American Presidential election, Petrocik (1996) observed that: candidates have distinctive patterns of problem emphases in their campaigns; election outcomes do follow the problem concerns of voters; and the individual vote is significantly influenced by these problem concerns above and beyond the effects of the standard predictors. This is not to say that parties and candidates, or even the news media, are able to singularly prime certain issues rather than others. The nature of campaigns may reduce the likelihood of distracting voters from a salient public issue if that issue would benefit the other parties. For example, Bush was forced to address the economy during the 1992 election in  to new issues.  Lodge et al. (1995) provide evidence that a respondent's on-line tally  166  spite of his preferences (Just et al. 1996) because of public concern with the issue and media attention to the state of the economy (Hetherington 1996). Parties and elites may collude, however, in keeping an issue off the campaign agenda if it would split both of the parties (Johnston et al. 1992). The issues that emerge during the campaign, as discussed in Chapter 4, are the product of the interaction between candidate strategies, news decisions, and the public. Prirning is also consistent with recent literature on low information rationality. Since voters lack the time or desire to be fully informed, they must find ways to make decisions based upon the information that is readily available (Popkin 1991). It follows then, that voters will use issues that are more salient in the news media than those which aren't. On non-salient issues, voters would be required to invest time and effort to determine where the parties are located. Despite the intuitive resonance of the prirning notion, it may be that more than priming is occurring in many of these cases. Persuasion or learning could both equally produce a shift in the weight voters attach to a consideration between two time points. By changing the distribution of voter opinion on an issue and changing the relationship between attitudes about that issue and party support, persuasion could effectively produce the appearance of a stronger association between the consideration and the judgment at a later time. Similarly, voters who lack information about a party's true position are unlikely to weight that issue highly in their overall judgment. If respondents learn the party's position between time 1 and time 2, his or her judgment at the second interview could reflect a stronger correspondence between attitude and judgment. One of the conditions for voters to be able to attach a large weight to an issue in their evaluation is knowledge of the candidate's stand on  undergoes information-related changes in an experimental study. 167  that issue. If this condition is not met, then the weight voters attach to an issue can change over time as the person is provided with information about the candidate's position. B y emphasizing some issues rather than others, the media may be doing no more than communicating the correct location of the candidates. In fact, the observed implications of learning are so similar to priming, that some of what is labelled as priming could be nothing more than learning. While the net effect will be to increase the importance of the "primed" issue, it is not clear that this should be labelled priming. In many ways, priming researchers are aware that learning is an important aspect of the priming process. The effect of the news is such that, "Coverage of a particular problem provides new information that is accessible by its recency. Coverage may also provoke viewers to recollect what they already know about the problem" (Iyengar et al. 1984: 779). A n election related experimental study does confirm that voters were able to learn about candidates from news but did not assess the impact of learning on the size of the priming effect (Iyengar and Kinder 1987). As such, discussions of priming often conflate two different coverage induced effects into a single political process. The importance of learning is implicit in the discussion of the role of knowledge in facilitating or resisting priming messages. It is hypothesized that, "the more knowledge about politics one has, the more resistant one should be, because knowledge helps one to withstand influence" (Miller and Krosnick 1996: 84). Those with more knowledge are presumed to have more crystallized calculi, but it may be that they already are aware of the location of the parties and therefore do not gain new information from the news media. The knowledgeable from this perspective are not resisting priming messages. There is specific evidence that priming was lower in the experimental studies among those people who were previously informed about the  168  area being primed (Iyengar et al. 1984). The researchers do not distinguish between increasing the accessibility of judgments that the respondent already has and a more general process of information distribution that allows voters to link their attitudes about a issue to their evaluation. Media emphasis on some problems rather than others has been linked to changes in candidate evaluations, especially presidential evaluations. The correspondence of increased media attention and greater weight given to attitudes about the emphasized issue in the public's overall judgment of presidential performance has been labeled priming. The label captures the basic notion that the public is responding to media emphasis because media emphasis increases the accessibility of that issue, but it does not clearly identify what accessibility means. The discussion here suggests that it would be useful to distinguish between a media impact that affects the weights attached to issues and a media impact that affects accessibility by providing the information necessary for incorporating issue judgments into political evaluations. For the remainder of this discussion, the former is understood as priming while the latter is considered learning. The Stability of Public Attitudes: the Lack of Persuasion One of the ways the campaign could have mattered was by changing people's minds. That is, people could have been persuaded to change their opinions on fundamental questions of public policy. This kind of change is not unprecedented. During the 1988 election, Canadians changed their opinion about free trade as a response to changing information. While one should be open to the possibility of attitude change, the 1993 campaign lacked this kind of pivotal judgment issue. People could have also changed their opinions in order to bring their attitudes into line with their evaluations of parties (Macdonald and Rabinowitz 1997).  169  Figure 6-1. Stability of Welfare State Attitudes Proportion  0.5  0.25  0 10-19 Sept  27 Sept-3 Oct 20-26 Sept  11-17 Oct 4-10 Oct  18-24 Oct  Reduce deficit: Proportion willing to cut deficit even if it means cutting programs. Health care fees mean less waste : Proportion who think that user fees will reduce waste rather than decreasing access to services. Same Services: Proportion who feel that services should be provided to all people even if they can afford the service themselves.  Attitudes toward the Deficit and Welfare State The Canadian Election Study includes a number of questions about fiscal matters and attitudes about the welfare state. Figure 6-1 graphically represents the change in opinion on three central questions. In each case respondents were asked to choose a statement that most closely represents their views. To determine the commitment of voters to deficit reduction, voters were asked to chose between: "We must reduce the deficit even if that means cutting programmes" and "Governments must maintain programmes even if that means continuing to run a deficit." Since each option has an undesirable cost, the choice reflect the voter's own underlying tradeoff. Perhaps surprisingly, a majority of voters (about 60%) viewed deficit reduction as a higher priority than the maintenance of government spending.  170  It is often noted that universal health care and social programs are essential aspects of the Canadian political culture and psyche. Two questions in the campaign wave of the survey tapped commitment to universal access. In the first, respondents were asked whether government services should be provided to all citizens even if they can afford it themselves. Presumably, those who desire a small government will not support the government acting as a general service provider; preferring government services to be targeted, and private interests to provide the services to those who can afford it. In the second, respondents were asked about the impact of charging people a fee to access health care. Here the choice was between saying that an access fee would reduce waste and that an access fee would mean that some people would not get health care they needed. Canadians were split on both questions. While a bare majority of decided respondents felt that a fee would reduce waste, only a bare minority supported providing government services to all people. The striking, and most important, aspect of Figure 6-1 is the lack of any systematic change in attitudes during the campaign. On all three questions, the proportion providing the respective option stays about the same. If the campaign had an effect, it did not do so by changing people's minds in a systematic manner. Any movement of individuals in one direction was off-set by a corresponding movement in the opposite direction which suggests that all such movement was random. One can conclude therefore that the campaign did not persuade voters. Further evidence of the stability of attitudes comes from the results of a question that asks respondents how much seven policy areas should be cut. The lines in Figure 6-2 represent the proportion of the respondents who said that they are in favour of having the government cut the policy area "a lot." Attitudes about spending cuts are sensitive to the policy domain, but  171  Figure 6-2. Public Support for Cuts to Government Programs  50  Percentage  40  30  20  10  0 | — 10-19 Sept  50  1  20-26 Sept  1  27 Sept-3 Oct  r 4-10 Oct  11-17 Oct  18-24 Oct  Percentage Foreign Aid  40  30  20  10  10-19 Sept  20-26 Sept  27 Sept-3 Oct  4-10 Oct  11-17 Oct  18-24 Oct  Lines represent the percentage of the public who want the spending category cut a lot.  172  again the striking thing is the degree of stability. Of the seven policy areas only two have a significant temporal change. For the welfare policy area, there is approximately a 10 point drop between the week of October 11 and the week of October 18. For the education area, which is largely a provincial responsibility, there is a decline in the willingness to cut from the beginning to the end of the campaign. The Cultural Dimension: Attitudes toward French Canada and Ethnic Minorities Cross-cutting the economic dimension, is a cultural one. While the French fact in Canada dominates the cultural question, the multicultural nature of Canada raises the salience of other issues, as was the case with Reform's issue coverage in the latter period of the campaign. Fortunately, the campaign wave of the election survey includes a number of questions that allow one to observe any changes in public attitudes toward racial minorities and the place of French in Canada. Attitudes toward the French fact are captured by two questions. The first is a question asking respondents about the importance of having a Prime Minister that speaks French. Clearly choosing a Prime Minister who does not speak one of the official languages would represent a clear signal to Francophones about their place in Canada. The proportion that answered that it is "not very important" or "not important at all" is represented by the dark line Figure 6-3. The second is a question about how much should be done for Quebec that was actually asked in two different ways: half of the sample were asked, "How much do you think should be done to promote French?"; and, half were asked, "How much do you think should be done for Quebec?" The proportion willing to do more on the version they were asked is represented by the grey line on Figure 6-3.  173  On neither question about the French fact is there any dynamic movement. There is a slight increase in the size of the public which does not think it is important for the Prime Minister to speak French well, but the increase is small and virtually disappears by the end of the campaign. There is also no change in public willingness to do more for Quebec and the French language. The question wording experiment did work, with respondents asked the French language question somewhat more likely to do more, but it has no impact on the stability of the trend in attitudes toward Quebec. Figure 6-3. Stability of Attitudes toward the French Fact  0.4  Proportion  " '  Not important PM Speaks French  0.3  Increase support for Quebec/ French Language  10-19 Sept  27 Sept-3 Oct 20-26 Sept  » "*"  11-17 Oct 4-10 O c t  18-24 Oct  Important PM Speaks French: Proportion who feel that it is important that the Prime Minister speaks French Increased Support for Quebec/ Language : Proportion who think that more should be done for Quebec.  There are three questions available to tap attitudes about ethnic minorities. The first asked about immigration levels, which clearly relates to racial minorities given the racial background of contemporary immigrants to Canada. Respondents were asked, "Do you think Canada should admit more immigrants or fewer immigrants?" The proportion supporting  174  increased immigration is represented on Figure 6-4 by the grey line. The second question was a general question about government support for racial minorities. Respondents were asked, "How much do you think should be done for racial minorities?" The broken line on Figure 6-4 represents the proportion that wants more to be done for racial minorities. The third question concerned aboriginal rights. Respondents were asked to choose between the following statements: "Aboriginal people should have the right to make their own laws" or "Aboriginal peoples should abide by the same laws as other Canadians." The proportion giving the former (pro-aboriginal rights) position is represented by the dark line. Figure 6-4. Stability of Racial Minority Questions  0.4  Proportion  Increase Aid to Minorities  0.3 Aboriginals Should Make Own Laws  Allow More Immigrants  0  11-17 Oct  27 Sept-3 Oct  10-19 Sept 20-26 Sept  4-10 Oct  18-24 Oct  Increase Aid to Minorities : Proportion who think that more should be done for racial minorities. Aboriginals Make Own Laws: Proportion who think that aboriginals should be able to make their own laws. More Immigrants: Proportion who think that Canada should admit more immigrants.  Not surprisingly, the results of the three questions indicate only modest support for doing more for racial minorities, for increasing the number of immigrants admitted, and for 175  allowing aboriginal peoples to be able to make their own laws. The key story is not the conservatism of Canadians on these questions but the continued stability of attitudes across the campaign. While there appears to be slightly more campaign change in attitudes about racial minorities, the week to week changes are very modest. There is a net drop from week 1 to week 6 of 4 percentage points in the proportion desiring to increase aid to minorities but the trend is not obvious; the proportion increases in the middle. Similarly, while the willingness to admit more immigrants appears to decline, this is largely the result of the high first week value.  The Temporal Stability ofIdeological Scales The underlying stability of attitudes on economic and cultural questions is what allows for the creation of scales of underlying attitudes about the welfare state, macroeconomic policy, ethnic minorities, and French Canada even though many of the questions that go into the scale were asked during the campaign wave of the survey (see Appendix C). The anti-welfare state scale taps attitudes toward social programs and the size of government with a high value indicating an opposition to the welfare state. The macroeconomic policy scale is based on questions that aim to reflect the respondents' attitudes toward the traditional tradeoffs between jobs on the one hand, and inflation, the deficit, and taxes on the other. A high value on this scale reflects a preference for lower deficits, inflation control and lower taxes. Attitudes toward French Canada reflected in an anti-French scale and is based on answers to two questions: attitudes about how much the government should do for Quebec, and attitudes about the importance of having a Prime Minister who speaks French. Attitudes toward minorities are measured by an anti-minorities scale which is based on a number of questions that ask respondents for their views on ethnic minorities, multiculturalism and immigration.  176  Table 6-1. Campaign Stability of Ideological Scales Anti-Welfare  Macroeconomic  Anti-French  Anti-Minorities  Canada  State 1. Mean September 10-19  0.40  .40  0.47  0.65  September 20-26  0.39  .40  0.47  0.67  September 27-October 3  0.39  .40  0.49  0.68  October 4-10  0.39  .38  0.49  0.68  October 11-17  0.41  .41  0.49  0.68  October 18-24  0.39  .41  0.47  0.67  Total  0.40  .40  0.48  0.67  September 10-19  0.23  .22  0.24  0.25  September 20-26  0.24  .23  0.25  0.24  September 27-October 3  0.23  .23  0.27  0.22  October 4-10  0.23  .22  0.26  0.24  October 11-17  0.23  .22  0.27  0.24  October 18-24  0.23  .23  0.26  0.24  Total  0.23  .22  0.26  0.24  September 10-19  509  449  515  510  September 20-26  457  391  463  458  September 27-October 3  460  403  462  460  October 4-10  380  332  381  374  October 11-17  421  383  434  426  October 18-24  492  431  499  488  Total  2720  2389  2755  2717  2. Standard Deviation  3.N  The campaign could have shifted the mean (e.g. the public as a whole could have become more anti-welfare state) or the variance (e.g. the public as a whole could have become 177  more polarized) of the scales. While the analysis of individual items focused on major changes in the direction of opinion, the analysis of the scales also considers the how the shape of the distribution changed over time. To test for the possibility of changes to the public's underlying attitudes the campaign was divided into 6 one week periods. Within each period the mean and standard deviation for each of the four issue scales is provided. The results are provided in Table 6-1. Consistent with the analysis of individual survey questions, for none of the three ideological scales is there any evidence of campaign movement in either the mean or the standard deviation. Like their components, the scales do not undergo any significant campaign-movement in either mean or variance, which allows them to be used as an indicator of a person's underlying attitude regardless of time of interview.  3  Summary The evidence here suggests that persuasion or conversion was not a significant campaign process in 1993. The public as a whole was divided on the question of the appropriate role of the state, but it did not become more anti-welfare state as the campaign neared its conclusion. The same holds for public attitudes on cultural questions. Canadians may not reveal themselves to be particularly liberal on most questions of ethnicity and language but there is little in the way of evidence that Canadians became more conservative during the campaign.  3  There is a large and important debate about the stability of public preferences going back to Converse (Converse 1964; Achen 1975; Luskin 1987; Zaller 19921 If each individual was asked the questions at different times, their responses may have differed reflecting measurement error and the salience of some considerations over others. Since there is little aggregate movement, it seems reasonable to use the questions as tapping fairly fundamental and long standing attitudes. 178  Evidence for Priming Before attempting to sort out the extent to which what appears as priming is really learning, it is necessary to establish that the weight of different considerations changed over time. One approach would be to create an interaction between the ideological consideration and day of interview. While this would identify a linear growth it might miss important temporal discontinuities. The approach used here is to break down the campaign into periods and estimate the role of a number of issue dimensions within each period. The campaign naturally breaks down into three distinct periods: the period of no change in Reform intentions; the period of sustained positive growth in Reform's share of vote intentions; and, the levelling off period. The three periods roughly correspond to the two week intervals of the campaign used here. There are two dependent variables that could be used in the regression estimations: a dummy variable representing whether the respondent intended to vote Reform; and, the party evaluation variable that asked respondents to rate the party on a hundred point scale. The former approach has been used in all previous analyses of the change of issue importance (Johnston et al. 1994). The approach taken here is to use the latter and estimate the equation with OLS for two reasons. The first is that a dichotomous dependent variable is best dealt with using estimation procedures such as Logit or Probit, which are more difficult to interpret. The second is that using vote intention dummies introduces potential problems of interpretation because of the lack of variance in the dependent variable. As the respondents become more likely to report a Reform intention, the variance of the variable increases, which can affect the statistical relationship between the variables of interest and Reform support. The use of the  179  Reform evaluation variable reduces this problem. The third is that party evaluation scores are related to the vote while being less likely to be influenced by strategic considerations. Figure 6-5 shows that changes in support for Reform as measured by feeling mean thermometer score are mirrored in changes to Reform's share of vote intentions. The only cost of using party evaluation is the significant number of people who were unwilling to rate Reform in the early weeks of the campaign. Effectively these people are excluded from the analysis.  Figure 6-5. Reform Vote Intentions and Thermometer Rating During Campaign  50  Percentage Vote Intentions  Daily Mean Rating  30  45 20  40  35  10  0 10 12 14 16 18 20 22 24 26 28 30  2  4  6  8  10 12 14 16 18 20 22 24  September  October  Moving 7 day average. Note: Those who refused or did not know were treated as missing.  180  There are two hypotheses about the relationship between underlying attitudes and judgments about Reform. The first is that as the campaign progressed attitudes about the economy and welfare state became more important (eg. were more strongly associated) for judgments about Reform. Both previous studies (Johnston et al. 1994) and the evidence concerning learning and issue coverage suggest this is likely. The second is that the emphasis on cultural questions increased the importance of anti-minority feelings for evaluations of Reform. To test these hypotheses, the respondent's rating of the Reform party is regressed in each of the three periods on five issue variables and a series of demographic controls. Four of the independent variables are the ideological scales - anti-welfare state, macroeconomic policy, anti-French, and anti-minorities - that were discussed earlier. Each of these scales varied between 0 and 1. In addition, the respondent's attitude about the corruption of public officials is entered to capture the populism and alienation dimension of Reform support.  4  Table 6-2 reports the results of the OLS estimation for each of the three time periods. As expected, the results confirm that there is a stronger relationship between ideological predispositions and evaluations of Reform in the latter periods. Both goodness of fit statistics show improvement across the models, and there are more significant coefficients as time goes by. More importantly, there is support for both priming hypotheses. Welfare state and broader conservative economic attitudes along with attitudes toward ethnic minorities become more important at the end of the campaign.  4  This is ultimately an unsatisfactory way of controlling for the impact of populism and alienation. While it is possible to construct a broader index of alienation (Johnston et al. 1996a), the cost of doing so is too high because of the loss of a substantial number of cases (the necessary items were only asked in the mailback portion of the survey). 181  Table 6-2. Evidence for Priming Sept 10-26  Sept 27 - Oct 10  14.98***  10.73 *  (4.10)  (4.24)  (4.04)  0.78  16.71***  15.33***  (4.12)  (4.37)  (4.05)  13.00***  7.38  7.71 *  (3.80)  (3.77)  (3.51)  9.48*  5.36  14.39***  (3.76)  (4.27)  (3.96)  -0.89 (0.61)  -0.90 (0.64)  -0.24 (0.59)  -0.33 (0.45)  0.84 (0.48)  0.51 (0.45)  -0.60  0.15  -1.45  (1.20)  (1.24)  (1.14)  0.84  -1.15  -4.16*  (1.82)  (1.84)  (1.79)  -3.66  -2.89  -8.00**  (2.92)  (2.77)  (2.82)  6.05**  5.45**  4.47**  (2.00)  (2.08)  (1.90)  21.20*** (5.08)  16.94** (5.24)  16.34** (5.17)  Adj-R  0.09  0.10  0.18  SEE  22.60  22.13  22.29  664  603  680  Anti-Welfare State Macroeconomic Policy Anti-French Canada Anti-Minorities Politicians are Corrupt Demographics Education Income Woman Atlantic West  Intercept  2  N  Oct 11-24 22.17***  * p<.05; ** p<.01; and, *** p<.001 Standard Error in parentheses. Significant coefficients in bold face.  182  In the first period, evaluations of Reform are oriented around the two key dimensions: attitudes about the welfare state and French Canada. By the second period, the effect on Reform evaluation of moving along the welfare state dimension decreased by several points but the relationship between economic predispositions and evaluation of Reform is strengthened by the importance of macroeconomic attitudes to Reform intentions in this period; 16.7 compared with 0.78. In the third period, October 11 to 24, the welfare state coefficient becomes larger while the coefficient on macroeconomic attitudes remains high. Being extreme on the anti-welfare state and macroeconomic scales (a value of 1 on each), was worth 15 points in the first period and 40 in the last period; a difference of 25 points on the hundred point scale. It appears that as voters became more likely to give Reform their vote intention and a higher party evaluation score, the respondent's score was more likely to be linked with their underlying attitudes on the welfare state and macroeconomic policy. If the campaign primed economic concerns, it appeared to suppress the importance of attitudes toward French Canada in the middle of the campaign. During the second period, the coefficient on French Canada attitudes is almost one half its first period magnitude and is not significant. By the third period attitudes toward French Canada were significant but had not undergone any significant increase. There is, however, support for the idea that the coverage of Reform's immigration and multiculturalism positions reminded voters at the end of the campaign of the party's position on these questions. Those opposed to doing more for ethnic minorities were significantly more likely to rate Reform higher in the first and last weeks. This is especially true in the last week. Since much of Reform's growth in support occurs in the second week, it appears that this growth was largely among people who were opposed to the welfare state and wanted a lower deficit regardless of their position on cultural issues.  183  In the last two weeks the Reform party resembled the neo-liberal, right-authoritarian parties of Western Europe (Kitschelt 1995). Surprisingly, there is no effect of alienation on party evaluation scores in any of the periods. While this reveals a discrepancy between the results reported for vote by Johnston et al. (1996a), there are two possible explanations for this finding. The first is that the indicator of alienation is so crude as to explain the null findings. The second is that those who think that politicians are corrupt are unlikely to rate any party, even Reform, high even if they rate Reform higher in relative terms than the other parties. Individual-Level Evidence for Priming Before developing and challenging the priming explanation for 1993, it is worth tackling the question of priming from an alternative angle. That is, from the perspective of individual-level change across the campaign. To this point, the analysis of priming took place through the utilization of the properties of the rolling, cross-section design of the CES. If the relationship between independent and dependent variables in one sample is different than in another, then one can assume that the same difference would be observed if one was comparing the same individuals at different times. In this section, the post-election measure of Reform 5  party evaluation is also used to determine whether the activation of ideological predispositions accounts for the difference between the respondent's pre- and post-election rating of the party. The relationship between pre- and post-election evaluations will clearly depend on the time the respondent was interviewed during the campaign so the approach here is to continue to distinguish the respondent's time of interview in terms of the three two week periods. The respondent's post-election evaluation of the Reform party is then regressed on the four key  5  In fact, comparing the same individual introduces a conservative bias because of the tendency for respondents to consider their previous answer when asked the same question at a latter time. 184  ideological variables (anti-welfare state; macroeconomic policy; anti-French; and, antiminorities) discussed above and the respondent's campaign evaluation. Including the campaign evaluation should control for the impact of ideological variables on Reform rating at the time of the campaign survey. Any further impact of predispositions on the difference is evidence of priming. For example, if the priming story about cultural questions is correct, then those people interviewed before these issues were salient in the campaign will not have factored their underlying position on these issues into their original evaluation. Their post-election evaluation will, however, reflect the priming of this new information producing a significant coefficient for the anti-minorities scale. The evidence presented in Table 6-3 supports this expectation. While the best predictor of a respondent's post-election evaluation is the respondent's campaign evaluation, there is clear evidence of priming for both economic predispositions (attitudes about welfare state and macroeconomic policy) and cultural attitudes (anti-minorities and anti-French). Those people interviewed during the first two weeks of the campaign were likely to evaluate the Reform party higher in the post-election survey as demonstrated by the magnitude of the intercept and the difference between their pre- and post-election ratings can be explained by the fact that attitudes about the welfare state increased in importance. It is the second period, September 27th to October 10th, which is the most interesting. Here again the anti-welfare state coefficient is significant suggesting that the post-election rating was affected by the respondent's underlying attitudes toward the welfare state over and above the impact that occurred at the time of interview. This coefficient is partly picking up a residual effect of the time of interview because of the size of the period and the magnitude of  185  the shift in evaluation vvithin it. The key, however, is the magnitude and significance of antiminorities attitudes in this period. A respondent who takes an extreme anti-ethnic, antimulticulturalism stand is almost 11 points more favourable toward Reform at the end of the campaign as compared with the middle.  Table 6-3. Intra-Individual Changes of Reform Rating  Anti-Welfare State Macroeconomic Policy Anti-French Canada Anti-Minorities Campaign Rating Demographics Education Income Woman Atlantic West  Intercept  Adj-R2 SEE N  Oct 11-24  Sept 27 - Oct 10  Sept 10-26  7.154 *  8.127 * (3.665) 5.648 (3.631) 4.623 (3.388) 5.091 (3.354)  4.307 (3.011) 1.247 (3.001)  (3.378)  8.383 * (3.498) 0.818 (2.979)  5.216 * (2.562) 4.529 (2.940)  11.187 ** (3.421)  0.471 ***  0.628 ***  0.603 ***  (0.035)  (0.033)  (0.028)  0.087 (0.399)  0.110 (0.384)  -0.346 (0.330)  0.228 (1.064) -1.233 (1.617) -1.870 (2.612)  -0.593 (0.973) 0.354 (1.457)  -0.038 (0.837) -0.571 (1.315) -0.497 (2.069)  -5.238 * (2.180) -0.407 (1.657)  2.463 (1.774)  17.711  ***  10.725  2.774 * (1.377)  *  14.243 ***  (4.517)  (4.183)  (3.743)  0.29 20.12 667  0.44 17.69 611  0.51 16.36 683  * p<.05; ** p<.01; and *** p<.001 Standard error in parentheses. Significant coefficients in bold face. 186  Not surprisingly, those people interviewed during the final two weeks of the campaign were less likely to change their evaluation of the Reform party in any systematic manner. There is evidence that French Canada attitudes become more important for the post-election evaluation than the campaign evaluation. Perhaps this is a response to an election which produced an opposition party that was committed to breaking up the country. The comparison of individual pre- and post-election ratings has strengthened the claim that cultural and welfare state issues became more important over the last two weeks of the campaign than they had been previously. Those people interviewed in the middle of the campaign did factor these issues into their post-election evaluation but not in their campaign evaluation. Awareness as a Mediator of Priming Effects As discussed earlier, researchers into pruning argue that one should expect priming to be larger among the least aware because the high aware are more resistant to new information (Miller and Krosnick 1996). This assumes, of course, that the chronically unaware of political events get the prirning information, while the aware have that information available. In the context of an election campaign where an insurgent party is competing, it may be the high aware rather than the low aware who are most impacted by new information. Fortunately, one can test the prirning hypothesis for two different kinds of respondents: those with high political awareness compared with those with low political awareness. Parts A and B of Table 6-4 provide the evidence to support the claim of differential impact. Although the coefficients are made unstable by the smaller sample sizes, the differences are apparent. The coefficients for the high aware respondents are basically similar to the ones for all respondents grouped together except that the priming of economic issues is  187  more apparent in the middle of the campaign and is captured by changes to the macroeconomic coefficient rather than the welfare state one. The pattern suggests that high aware respondents factored both welfare state and French Canada attitudes into their evaluation of the party early, but relied more on economic issues in the middle of the campaign before reverting to the first period pattern in the last period. Of course, added to the mix in the last period is the striking importance of attitudes toward ethnic minorities and moral conservatism. While ideological predispositions are important aspects of the evaluations of high aware respondents, they are basically irrelevant for low aware respondents. The only variable that is related to party evaluation for this group is welfare state attitudes in the last period. It appears that to the extent that role of government attitudes were primed for low aware respondents, the priming took longer and was less effective. The results confirm the increased correlation between Reform support and economic ideology as the campaign neared its conclusion. In addition, the importance of French Canada attitudes declines relative to the first period and there is a significant increase in the importance of ethnocentrism for Reform support in the latter period. These findings are consistent with the general pattern of news coverage identified in earlier chapters. During the middle of the campaign, Reform was most likely to be associated in the news media with social programs and the deficit, but at the end of the campaign the party was trying to defend its cultural positions leaving less of an opportunity for it to talk about the role of government in society.  188  Table 6-4. The Effect of Priming Within Awareness Groups Sept 10-26 A. HIGH AWARENESS  =  _  =  =  Sept 27-Oct 10 =  =  _  =  ^  Oct 11-24 =  =  =  =  ~  =  =  Welfare State Attitudes  18.40 **  12.56 *  16.04 **  (6.15)  (5.70)  Macroeconomic Policy  (5.95) -2.21 (5.83)  23.96 ***  17.95 ***  (6.00)  (5.36)  19.18 **  16.62 **  17.20 **  (5.71) 7.07 (5.19) -1.05 (0.84)  (5.36) 2.45 (5.69) -1.37 (0.89)  27.38 ***  23;23 **  24.90 ***  (7.21)  (7.20)  (6.97)  0.11 23.72 375  0.16 21.35 296  0.26 21.84 337  9.42 (5.82) -0.86 (6.27) 6.64 (5.12)  0.50 (6.32)  24.42 ***  French Canada Ethnic Minorities Politicians are Corrupt Intercept  Adj-R SEE N  2  B. LOW AWARENESS Welfare State Attitudes Macroeconomic Policy French Canada Ethnic Minorities Politicians are Corrupt Intercept  Adj-R SEE n  2  10.69 (5.61) -0.46 (0.90)  14.76 * (7.20) -3.72 (5.60) 9.58 (7.12)  (5.23)  21.57 *** (5.39) -0.65 (0.82)  (6.09) 11.49 (6.61) -1.31 (5.06) 1.51 (6.46)  0.21 (0.98)  0.36 (0.90)  10.74 (8.51)  17.51 *  (7.85) 0.07 21.19 277  0.09 21.39 248  0.16 21.62 290  18.22 *  (8.63)  * p <.05; **p<.01; *** p<.001; significant coefficients in bold face. Standard errors in parentheses. Note: Estimations included variables to control for demographics (coefficients not shown).  189  In contrast to previous priming evidence, the results identified here suggest that awareness actually facilitates rather than inhibits priming. Voters must be exposed before they can be affected and it appears that in the electoral context, it cannot be assumed that voters will be exposed to enough issue cues so as to be affected by the issue agenda of the campaign. This is a significant finding since the use of salient issue cues is regarded by some as a means of low information rationality (Popkin 1991). Reconsidering Priming The evidence in Table 6-2 and 6-3 appears to be the kind of evidence that is presented in support of a priming thesis, but it is worth considering whether such a change necessarily means that priming has occurred. The fact is that a coefficient change in importance may reflect the fact that voters are better able to locate a party's position on an issue in one time period as compared with another. If voters are unaware of a party's position on an issue, they have no basis other than general partisan cues for linking their predispositions with thenevaluation. The true test of priming is if voters who are knowledgeable about party positions change the weight they attach to issues emphasized by the news media and political parties. To test whether the results in Table 6-2 are truly evidence of a priming effect, two separate analyses are undertaken. The first replicates the priming analysis estimation for those aware of Reform's deficit position and the second replicates it for those aware of the party's position on ethnic minorities. If priming is the most important aspect of the change in the coefficients then it should be evident within these subgroups. Since the subset of respondents who were aware of Reform's position on the deficit is quite small, especially early in the campaign, the analysis makes use of only the four independent variables which were the most important in the previous analysis: dropping attitudes concerning corruption.  190  To distinguish those who knew from those who did not know Reform's position on the two questions, a crude indicator was developed from the responses to two questions about Reform's policy positions. For knowledge of the party's deficit position, the question on the effect of a Reform win on the size of the deficit was used. Those respondents who indicate that a Reform win would have the effect of a lower deficit (either somewhat or much) are coded as knowledgeable. For knowledge of the party's position on cultural questions, those people who said that Reform wanted to do less (either somewhat or much) to a question about how much the Reform party wanted to do for racial minorities were coded as knowledgeable. Since Reform was clearly against multiculturalism and the application of the Charter of Rights to immigrants, Reform clearly wanted to do less. Table 6-5 reports the results for the analysis of prirning among those knowledgeable about the Reform's deficit position. Evidence for the prirning thesis is contradictory. When welfare state and macroeconomic attitudes are considered together it appears that economic issues did not become more important in the latter stages of the campaign, but the kind of economic attitude did change perceptively. In the first period (September 10-26), welfare state attitudes are more important than French Canada attitudes for Reform support. In the latter two periods the coefficient on welfare state attitudes is cut in half from its first period magnitude. This drop is, however, made up by the broadening of the economic conservative appeal of the Reform party reflected in the large and significant macroeconomic policy coefficient.  6  Consistent with earlier evidence, Table 6-5 also suggests the increased importance of attitudes toward ethnic minorities, especially in the latter period.  6  If the macroeconomic policy variable is not included, thus allowing welfare state attitudes variable to capture all of the influence of the economic attitudes, there is no large shift in the weight attached to welfare state attitudes. 191  Table 6-5. Priming Among those Aware of Reform's Position on Deficit  Welfare State Attitudes Macroeconomic Policy French Canada Ethnic Minorities Demographics Education Income Woman Atlantic West  Intercept  Adj-R SEE N  2  Sept 10-26  Sept 27-Oct 10  Oct 11-24  38.65 *** (7.21) -8.11 (6.58) 19.41 ** (7.08) 10.15 (6.50)  20.97 *** (6.09) 13.60 * (5.83) 22.35 *** (5.26) 9.96 (5.78)  22.10 *** (5.87) 16.38 ** (5.43) 15.73 ** (4.93) 19.75 *** (5.50)  -1.14 (0.71) -4.63 * (2.19) -1.45 (3.23) -6.44 (6.78) 3.43 (3.32)  -0.05 (0.65) -3.24 (1.75) -0.06 (2.55) -5.69 (4.24) 2.07 (2.86)  -0.51 (0.61) -5.08 ** (1.64) -1.93 (2.61) -5.14 (4.34) -0.58 (2.59)  37.29 *** (8.71)  25.81 *** (7.60)  32.43 *** (7.04)  0.21 21.29 201  0.22 18.95 240  0.23 20.21 300  * p <.05; **p<.01; *** p<.001; significant coefficients in bold face. Standard errors in parentheses. Given that anti-welfare state attitudes are an important correlate of party evaluation in the early period and that the weight of economic attitudes does not increase across the three periods the campaign did not prime as much as it informed. In other words, the small impact of the campaign on the size of the coefficients in Table 6-5 cannot account for the changes in  192  the coefficients in Table 6-2. As people moved into the group that knew Reform's deficit position, they weighted their underlying attitudes about the deficit and welfare state in a manner consistent with those people who knew Reform's position in earlier weeks. When one considers those people who did not know Reform position (evidence not shown), there is weak evidence of priming. Even though these people did not know Reform's position on the issue, they increased the weight of welfare state and macroeconomic attitudes slightly. This effect is likely the result of an imperfect measure of knowledge of Reform positions since it is awkward that voters who do not know the party's position are affected.  7  If it is hard to make a case for priming of the deficit, can such a case be made for the emergence of minorities attitudes? It seems more likely because we know from Figure 5-6 that there was less change in the perception of Reform's location on this question. Thus the increase in the coefficient in Table 6-2 is more likely to reflect respondents giving this issue greater weight rather than learning Reform's true position. To test the priming hypothesis, the regression analysis is re-estimated for those people who correctly located Reform's position on minorities. Table 6-6 provides some support for the priming thesis. In the first period, the more ethnocentric the respondent the higher the rating of Reform. In the second period, however, the coefficient is cut in half and is no longer statistically significant. By the final period, attitudes toward ethnic minorities are again important. The second to third period shift is persuasive evidence of priming even with the relatively small number of cases. The priming hypothesis is assisted by the knowledge that Reform was clearly associated with its  7  Since the question asked about the effect of a Reform win on the size of the deficit rather than about Reform's position on reducing the deficit or cutting social programs, there is good reason to expect some error. 193  Table 6-6. Priming Among those Aware of Reform's Position on Minorities  Welfare State Attitudes Macroeconomic Policy French Canada Ethnic Minorities  Sept 10-26  Sept 27 - Oct 10  Oct 11-24  24.25 * (9.48) 9.83 (8.48) 15.91 (8.16) 18.18 * (7.49)  18.05 * (8.55) 34.40 *** (7.84) 17.29 * (7.88) 5.25 (8.54)  27.34 *** (7.07) 21.17 ** (6.76) 24.89 *** (6.45) 16.87 ** (6.26)  -0.50 (0.91) -1.95 (2.56) -0.55 (4.03) -13.32 (7.94) 0.39 (3.93)  -0.35 (0.96) -2.23 (2.43) -5.04 (3.78) -7.70 (8.46) 0.63 (3.86)  0.18 (0.72) -1.90 (2.14) 1.74 (3.26) -7.55 (5.17) -3.57 (3.25)  11.64 (10.70)  16.03 (10.84)  5.83 (8.47)  Demographics Education Income Woman Atlantic West Intercept  0.28 0.16 Adj-R 21.05 24.63 SEE 141 178 N * p <05; **p<.01; *** p<.001; significant coefficients in bold face. Standard errors in parentheses. 2  0.32 22.33 223  "minorities" policies in the latter period. It is not obvious how to explain the initial drop between period one and two. The likely explanation is that respondents who liked Reform early on were a small but fairly unique and ideologically committed group. Reform became more popular on the backs of its fiscal policies, which meant that respondents liked Reform in spite of its position on cultural questions during the second period. Once the news media began to emphasize the party's position on cultural questions, respondents who knew the  Reform position reassessed their support for Reform. Chapter 7 provides support for this conclusion. Those who did not know Reform's position on the minorities issue were unaffected by their own attitudes about minorities at all points of the campaign. It appears, therefore, that knowledge is a prerequisite for priming. In those situations when we expect voters to lack information, it is not prirning that is likely to matter but learning. A Longer Time Horizon: Preston Manning and Anti-Welfare State Mobilization The 1993 campaign provided information about Reform's positioning on the economic right that was unavailable at the beginning of the campaign. In fact, one would expect that it was also unavailable in 1992. If this hypothesis is correct, there should be evidence of an increase of importance in the overall evaluation of Reform of attitudes about the welfare state and macroeconomic policy between 1992 and 1993. While it would be advantageous to compare thermometer scores for the Reform party between 1992 and 1993, the Referendum wave of the study did not ask these questions so thermometer ratings of Manning must be used instead. Table 6-7. Changing Feeling Thermometer Ratings of Manning Mean  N  Post Referendum Rating  40.8  777  Campaign Wave Rating  47.9  1985  Post-Election Rating  53.5  2081  Certainly Manning becomes more popular during this time as revealed by the mean thermometer ratings (Table 6-7). After the Referendum the mean rating was merely 40.8, 195  whereas after the election it had risen to 53.5 out of one hundred. To test the relationship between issue mobilization and leader evaluation, the importance of an individual's position on ideological scales for his or her rating of Preston Manning is estimated for the post referendum and post election waves of the survey (Table 6-8).  Table 6-8. Ideological Considerations and Evaluations of Manning, 1992 and 1993 Po st-Referendum Welfare State Attitudes Macroeconomic Policy  17.02  2.76 (3.95)  10.05 11.47  *  13.02  Politicians are Corrupt  (3.83) 0.36 (0.63)  Intercept  21.66  Adj-R2 SEE N  11.61  ***  (2.11) **  7.22  ***  (1.88)  (3.86) Ethnic Minorities  ***  (2.08)  (4.05) French Canada  Post-Election  ***  11.45  ***  (2.01) -0.07 (0.32) ***  30.80  (3.17)  (1.67)  0.05 23.05 657  0.11 19.87 1886  ***  * p<.05, ** p<.01, and *** p<.001 Standard Error in parentheses. Significant coefficients in bold face. It is clear that Manning's feeling thermometer score in the post-referendum wave is largely explained in terms of attitudes about culture and national identity including, French Canada, and minorities. While these attitudes remain important in 1993, evaluations of Manning are also driven by attitudes toward macroeconomic policy and the welfare state. Surprisingly, Manning does not benefit from alienation at either wave of the survey despite his populist claims. These results suggest that the campaign did manage to re-orient the public's  196  thinking about Maiining in such a way as to incorporate questions of economic ideology into his leadership rating. Of course, they do not change the conclusion that it was learning rather than priming that allowed this activation to occur. Conclusion The 1993 campaign mattered. It mattered because the public was exposed to information about the Reform party that allowed them to bring their evaluation of Reform into line with their own positions on the important ideological questions of the campaign. While there is no support for the persuasion hypothesis for 1993, there is evidence for both priming and learning. There is also empirical evidence that the distinction between priming and learning is useful. Previous evidence using vote intentions during the campaign established that attitudes toward the welfare state became more important as a determinant of Reform support during the campaign. The evidence here suggests that this is not simply the product of Reform increasing its share of vote intentions. Given the change of importance of economic attitudes, a reasonable interpretation is that the Reform emphasis on the deficit and social programs primed voters to weight their economic predispositions more heavily than their other attitudes. Two other possibilities exist. The campaign could have changed voters' minds about the appropriate role of government in society thus shifting the distribution of voters to the right on economic matters and consequently closer to Reform. Alternatively, the increased importance could be a function of the flow of information that allowed voters to learn Reform's true position on the deficit; something most voters were not aware of early on.  197  The campaign did not produce a systematic change in the underlying predispositions of Canadians on the welfare state or deficit. A change in the distribution of attitudes cannot then be an explanation for the increased importance of the welfare state or for the rise of Reform in the polls. As was demonstrated elsewhere, voters were persuaded that Reform was a credible party on the deficit (Johnston et al. 1994; Jenkins 1998). In effect, voters learned Reform's true position. Learning appears to be more of the story than priming. Among those who knew Reform's deficit commitment, there was no significant increase in the importance of economic attitudes. If learning is the key to understanding the increased importance of welfare state attitudes, it is prirning that is responsible for the increased importance of cultural attitudes. The late campaign rise in importance of cultural attitudes among those who were knowledgeable of Reform's position on this dimension is prime facie evidence of priming. Voters were aware of Reform's distinctiveness before the campaign began. In fact, the 1992 Charlottetown referendum campaign may have provided this information given that attitudes about minorities and French Canada were important for evaluating Manning in 1992. The conceptual distinction between learning and prirning is supported by the empirical evidence presented here. This suggests that by emphasizing some issues rather than others the media and parties are doing more than simply increasing the salience of those issues for voters with a concomitant increase in the importance of those same issues for evaluating candidates and presidential performance. Sometimes the coverage provides voters with information that they either did not know or had forgotten, while at other times it does serve to focus them on information they already knew.  198  In the election context these two processes - learning and priming - are fundamentally different because of the implications that they have for understanding the outcome. Although the scope for learning is likely limited by the weight of historical events and conflicts between established parties in most elections, the magnitude of voter ignorance about politics should not be underestimated. In addition, learning is consistent with a democratic process of decisionmaking as long as the scope for learning is not so large that voters learn only one of a number of potential things. Priming is, however, more likely to the extent that parties are able to focus their efforts on increasing the importance of an issue. The implication is that an equally plausible, but different, outcome would emerge if different issues were primed. The notion of a mandate is problematic under these circumstances (Clarke et al. 1991; Johnston et al. 1992; Mendelsohn 1998; Jacobs and Shapiro 1994). The evidence also supports the view that the aware are more likely to be affected by the campaign. Either through learning or priming, the weight of considerations among high aware voters was more likely to change as a function of the campaign. Further evidence in this regard is presented in Chapter 7, but clearly awareness did not serve as a basis for resisting the effects of issue emphasis in the news. High aware voters increased the weight they attached both to anti-welfare state and anti-minority attitudes.  199  CHAPTER 7: A MEDIA EFFECTS MODEL OF INSURGENT SUCCESS Exactly as common intuition would suggest, mass communication is a powerful instrument for shaping the attitudes of the citizens who are exposed to it, and it exercises this power on an essentially continuous basis. Zaller (1996b). The previous four chapters have established a number of important points about the relationship between news coverage of the campaign and Reform's rise in the polls. Independent decisions on the part of the news media changed the accessibility of Reform's key messages at different times during the campaign. As accessibility increased, coverage tended to reflect a discourse about Reform's position on social programs and the deficit. Unfortunately for Reform coverage began to drop and emphasize the party's position on cultural questions in the last couple of weeks. Coverage appeared to have served its purpose; people became more confident about their own knowledge of Reform and more clearly identified Reform positions at the end of the campaign, especially on the deficit. In fact, coverage appeared to be more important for clarifying the party's position rather than persuading voters to change their attitudes or prirning the welfare state issue. There was, of course, evidence that cultural issues were primed at the end of the campaign. Is the news media responsible for the campaign changes identified in Chapters 5 and 6? Did the character of the issue coverage identified in Chapter 4 have an impact or did issues play no role in the learning and mobilization of Reform supporters? Is the news media omnipotent or do individual characteristics mediate the impact of news decisions? These questions go to the heart of the issue of how voters respond to election campaigns and this chapter seeks to provide answers by merging daily media readings of Reform coverage,  200  identified in Chapter 3 and 4, with the rolling cross-section wave of the 1992-3 Canadian Election Study. The Campaign and Individual Responsiveness to Television Coverage Figure 7-1 provides a visual representation of the key hypotheses being considered in this chapter. In both path diagrams a diamond is present to indicate that coverage does not automatically lead respondents to change their opinion. As discussed in Chapter 1, the influence of persuasive messages should be mediated by the person's likelihood of receiving a persuasive message and his or her likelihood of accepting it (Dalton et al. 1998; McGuire 1968; McGuire 1986; Zaller 1992; Zaller 1996b). One would not expect respondents who lack either the motivation to attend to or the cognitive ability necessary to process the news to be influenced by the mass media. Similarly, those people who share underlying issue positions that are incompatible with Reform should not respond to Reform coverage in the same way as those available for Reform cues.  Figure 7-1. Path Diagrams of Key Hypotheses  Hypothesis 1: -•  Deficit Credibility  r  Reform Support  Reform Support  The first hypothesis identified in Figure 7-1 identifies two causal processes that affect Reform support. Coverage of Reform, mediated by the respondent's awareness and underlying position on the welfare state,firsthas an effect on perceptions of Reform's credibility on the  201  deficit. Changing perceptions of Reform then have an effect on Reform vote intentions. Both the coverage-credibility perception and credibility-support relationships are statistically analyzed here. From this perspective, the key role of the campaign is in signalling to voters Reform's true location. The second hypothesis is a reduced form of the first causal model. Coverage has an effect directly on support. Coverage can have this effect because it activates - makes politically relevant - respondent's underlying attitudes as in the case of priming or because voters learn where to locate the parties. Chapter 6 demonstrated that by the end of the campaign, attitudes about the welfare state and about minorities became more strongly associated with Reform support compared with earlier in the campaign. The question is whether increased coverage is the causal impetus for the increased importance and whether particular kinds of individuals are more susceptible to the change than others. Key Independent Variables  Reception  In the survey tradition, the search for media effects has tended to rely on self-reported media consumption questions to differentiate those who receivefromthose who do not receive media messages. While a large body of research has focused on the relationship between media exposure and knowledge (Chaffee and Frank 1996; Weaver 1996), a number of studies have argued that self-reports are not particularly useful in understanding media effects and differentiating voter learning (McLeod and McDonald 1985; Robinson and Levy 1986). Political knowledge performs better as a measure of the respondent's likelihood of receiving media messages than self-reports (Price and Zaller 1993; Zaller 1996b). A brief review of the evidence supports using knowledge as an indicator of reception.  202  The measurement difficulties associated with self-reported media use as compared with political knowledge should not be underestimated. There are good reasons to believe that survey questions of this type are particularly prone to measurement error. People make mistakes because they do not normally monitor media use, and research supports the conclusion that respondents often overestimate their media surveillance (Zaller 1996a). In addition, responses to self-reports capture inattentive media consumption, such as the situation where one has the T V news on while performing some other tasks. Political knowledge measures based on factual questions are not, however, likely to be biased in these ways. Because the questions ask people to answer questions of fact, "respondents cannot credibly over-report their propensity to receive information from the media" (Zaller 1996a). The empirical evidence that compares political knowledge and self-reported consumption or exposure is unequivocal. In a survey-based study by Price and Zaller (1993) respondents were asked whether they heard or read news about a recent event (e.g. Oliver North; death penalty case; Jim Bakker trial) and then asked to recall something about that story. A political knowledge scale based on factual questions was a more robust predictor of news awareness than self-reported newspaper or national television exposure scales. This conclusion was confirmed in more recent work on the ability of political knowledge to reflect news reception and act as a basis for observing media effects (Zaller 1996b). Further support comes from an experimental study that showed that political sophistication was a better predictor of learning than self-reports (Rhee and Cappella 1997). On the basis of this 1  1  In this case, sophistication, which was measured by questions that tied knowledge to ideological structure, was used rather than factual knowledge per se so the results are not strictly comparable. 203  evidence, there appears to be a general audience for news which is "quite well stratified by preexisting general levels of general political knowledge" (Price and Zaller 1993: 152). Of course, political knowledge distinguishes news reception because it is more than a measure of exposure to the news. Political knowledge questions are often used as an indicator of political sophistication (Luskin 1987) and to the extent that political knowledge is an indicator of deeper cognitive ability, political knowledge should be understood as a measure of how well the news is processed. Rhee and Cappella (1997) conclude on the basis of their experimental findings with a political sophistication measure (see note 1) that, political sophisticates are not simply more intense consumers of news but rather deeper processors of news. They bring different motivations to news consumption and they carry away knowledge more readily integrated into an already elaborate knowledge store (Rhee and Cappella 1997: 230). So while those who consume more media messages in absolute terms are likely to be more knowledgeable, there is likely to be slippage between exposure and learning because learning from the media is dependent on more than exposure (Neuman et al. 1992). So knowledge distinguishes people on the basis of "attending to, comprehending, and retaining news" (Price and Zaller 1993:134) and this is closer to the likelihood of receiving political communication than self-reported exposure. For this reason a knowledge scale is used here as a measure of awareness and as an indicator of the likely reception of new information. The details of the scale construction can be found in Appendix B . The appendix also contains a minor comparison of the effectiveness of self-reported media exposure and knowledge as a mediator of news reception. It must, of course, be recognized that political knowledge is not universally recognized as a measure of reception. Miller and Krosnick (1996) view knowledge as a measure of the amount of information in memory and separately measure exposure and attention. In their  204  actual conceptual scheme knowledge is used as a measure of resistance to media priming and persuasive messages generally. In experimental studies, knowledge performed its expected resistance role (Iyengar et al. 1984) but knowledge was unrelated to priming in one survey study (Krosnick and Kinder 1990) and actually facilitated priming in the other (Krosnick and Brannon 1993). Miller and Krosnick (1996) attribute this reversal in the survey studies to the delay between exposure and the survey intervention that leads to a large decay in the priming effect among low knowledge respondents. This explanation points to the problem of conceptualizing the role of knowledge in the process of media effects. It appears to be the case that knowledge indicates how well the information is integrated into prior knowledge; the higher the prior knowledge the greater the remembering  of the priming stimuli. But, the  authors expect knowledge to act as a basis for resistance to priming cues because people with more knowledge have the capacity to counter-argue new information. Both processes are likely at work and depending on the research question being investigated and the timing of the survey intervention, one may be more relevant than the other. There is no question that for some purposes it might make sense to differentiate simple exposure from comprehension. When considering an appropriate measure of reception to mediate news effects, it does not matter so much whether knowledge has an effect because it discriminates between those who are exposed to the news or because it discriminates who processes the news. The key is that some voters are less likely to be affected by the news media because they do not receive the information contained in the stories.  Acceptance Not everyone will be affected by the news so the two mediator model requires that there is an adequate means by which to identify individual predispositions. Since both anti-  205  welfare state and anti-minorities attitudes are key to understanding the dynamic evolution of Reform support, they are used here as an indicator of the likelihood that the respondents will accept the messages. For example, the hypothesis is that those who are predisposed to oppose the welfare state will be more receptive to cues about Reform and will both be more likely to identify Reform's deficit commitment and to support the Reform party. A n alternative approach would be to use party identification as a mediator (Zaller 1992), since it provides an indicator of the respondent's underlying political orientation. There are two problems with this approach. The first is that party identification is a general orientation which would not allow for the possibility that particular issues were made politically relevant. The second is that the Canadian party system was clearly in a state of flux if not transition as evidence by the formation of the BQ, the Reform party and the National party. It is unclear what party identification means in these situations. The fact is that party identification is not a stable attribute of respondents to the 1993 election study (see Johnston et al. 1996a). The number of identifiers increases from 4.1% of the week 1 sample to 6.9% of the week 6 sample. While party id continues to have its traditional meaning for established parties, a Reform identification is more clearly related to recent, even campaign stimuli. Party identification is, therefore, not exogenous and would bias the estimation if included in the equation.  While one could theoretically get around this problem by positioning people on the postelection measure of party identification, this would be unsatisfactory because in the postelection wave almost 9% of the sample identified with Reform. Nevertheless, party identification simply does not capture the notion of party loyalty. Including party identification would improve the fit of the models and reduce the impact of the ideological scales but would not affect the conclusions about the dynamics of coverage and would come at a cost of bias and mispecification. 206  News Coverage and Opinion Change If the attitude change model lays out a set of expectations about the likelihood that a respondent will change his or her mind, the key to observing that change will be significant changes in the information flow of the campaign. That is, those likely to receive political information must be exposed to different messages at different times during the campaign. With respect to the Reform party, this minimal condition certainly applies. Not only did the amount of news attention that Reform received change, but the kinds of issue cues also underwent a change in a manner consistent with a re-evaluation of Reform's credibility and the aggregate changes in Reform intentions. The principal coverage data that is merged with the survey data is a variable that indicates the frequency of coded items that involved the Reform party - as either the spokesperson or as the object of someone else's speech - for the days preceding the respondent's interview. In particular, the analysis of the effect of coverage on credibility makes use of this data. It will be recalled that Chapter 3 demonstrated that attention to Reform in the news media changed before aggregate vote intentions changed which suggests a causal relationship. It also happens to be the case that the changes in attention also reflect the changes in the availability of information about Reform on the deficit and social programs. Attention is a good surrogate for more complex measures of the availability of Reform cues. To test more specific hypotheses and to compare the effects of different kinds of coverage, the analysis also makes use of some indicators of issue coverage. The obvious candidates for influence are the two changes in Reform issue coverage identified in Chapter 4: social programs and culture/ ethnic. During the third and fourth weeks of the campaign, over half of the television news associated with the Reform party concerned social programs. Social  207  program coverage declines as a proportion of Reform coverage in the last two weeks, as issues of culture (immigration, multiculturalism) dramatically increase. A n Approach to Understanding Individual Responsiveness to Campaign Information This chapter considers the evidence of campaign change in terms of two general methods. The first method is graphical and relies on the idea that to the extent that the hypotheses about campaign change are reasonable there should be evidence of particular kinds of individual-level changes during the campaign. The second approach is statistical. Here, fairly simple statistical models are considered that both directly link changes in coverage to opinion and control for the impact of variables not considered in the graphical treatments. A Graphical Approach The two-mediator model of media effects provides a set of hypotheses about who would be most likely to respond to campaign information. For example, the theory expects that the greatest change will occur among those who are most likely to both receive and accept Reform cues. The clearest way to demonstrate campaign change is with a picture. In the analysis that follows the public is broken down into four different groups based on their crude positioning on each of the two mediators (awareness and predispositions). A weekly value of the dependent variable for each of the four groups is plotted to show the evolution of campaign changes. For example, the awareness and welfare state scales were divided at their respective means to create four groups based on the interaction between awareness and welfare state attitudes. The expectation is that those predisposed to accept anti-welfare state messages would be more likely to re-evaluate Reform on the deficit. Since awareness of messages is a necessary but insufficient condition for learning or persuasion, awareness should magnify the  208  effect of the respondent's welfare state predisposition. Those respondents with a pro-welfare state orientation are unlikely to be persuaded even if they are sufficiently aware to get the message. To the extent that the pictures reveal substantively important campaign changes that can be visually compared with the patterns of Reform coverage, there is prima facie evidence that supports the two-mediator model of campaign effects. While the evidence from the pictures is convincing, it could be argued that some other variable could be the real cause of the change; that the distinction between the four groups is crude and artificial; or, that coverage is not the cause of the campaign movement. These potential criticisms encourage the use of statistical techniques but the general approach is quite conservative and the results are so striking as to make the criticisms unlikely.  Estimation  Strategy The analysis in the previous chapter reduced the campaign into three temporal periods  and estimated separate static regression models for each period. Any day-to-day variation within the periods is averaged out. While this approach allowed for the identification of period-effects, one needs a truly dynamic model to capture the impact of changing media coverage on Reform support. The model proposed here incorporates both dynamic and static components. The static components are the variables normally thought relevant to Reform voting (attitudes about minorities, the welfare state and French Canada; awareness; and, demographic variables). The dynamic component is provided by entering into the equation a value for television coverage that corresponds to the news during the days preceding the day the respondent was interviewed.  209  To allow for both recent coverage and some memory from recent days, the media content for any given day is the average of the previous three days of coverage. This is consistent with the evidence in Chapter 3 that coverage took between three and four days to have its full impact on vote intentions realized. While entering coverage with different lags would be preferable, this would make stable estimates hard to identify because of the collinearity such variables would introduce. The coverage variable differs depending upon the dependent variable being estimated and the test being considered. For some estimations, the coverage variable is a global measure of the total number of times the Reform party was mentioned regardless of the content of the coded material. At other times, the content variable is a measure of the frequency of a particular Reform issue mention (i.e. social programs or culture) as discussed in Chapter 4. A l l coverage variables are measured in terms of the absolute frequency of mentions of the coded variable rather than as a density measure as was the case in Chapter 3. While a density measure controls for the size of the news hole and therefore provides a good indicator of news media gatekeeping, when considering the effect of information on the public the absolute rather than the relative amount of information is more relevant.  3  The approach adopted here is to compare a simple model of media effects with a more complex model that attempts to represent a two-mediator process of media effects in a single equation. In the simple coverage model (equation 1), an increase in coverage between t and 4  t+1 shifts the intercept but does not change the fundamental relationship between the other independent variables. Coverage has the same effect on all respondents.  3  4  In fact, as one would expect, the two variables are highly correlated (Pearson's R - 0.85). Compare this with the work by Zaller (1992; 1996b) who models media effects with more than one equation.  210  Y  = a + p.C/ + fiiR,• + $iA,+  PtD;  + e  W  X = a + p.Cr + p C< • M, + p C , « ^ , + P 4 / I , « A/, + p C , • Mi« 2  = a + $iC,+  Y  fiiA,-  3  M,+  fiiC-  5  M,>  A,+  ^AMI + $sR,+  Ps/?* +  $iM,  + p £>, + e 8  (2) (3)  P«D,e  where, Y is the dependent variable ; C is the average Reform coverage in the previous 3 days (some models involve all coverage of Reform while others rely on particular issue coverage); M is the respondent's position on the issue indices used to indicate the likely acceptance of pro-Reform information (anti-welfare state or anti-minorities depending on model) standardized to vary between 0-1. R is the respondent's position on the underlying issue indices (anti-French Canada and sometimes anti-minorities or anti-welfare state) standardized to 0-1; A is awareness index standardized; and, D is a series of variables to account for a respondent's education, income, and region of residence. 5  Equation 2 provides a fully specified delimitation of the two-mediator model. The hypothesis is that the three-way interaction of coverage, attitudes, and awareness will be the substantively important one, but statistically the three lower order interaction terms must also be included to generate efficient and unbiased statistics. The problem with including four interaction terms is the considerable collinearity that is introduced into the estimation. In practice this means that the coefficients from equation 2 are uninterpretable. Equation 3 represents one way of addressing the collinearity problem. Rather than including all four interaction terms, the model only includes the awareness  coverage  x awareness  x attitude  x attitude  and the  terms. In effect, the excluded lower-order interaction terms  are constrained to be zero. While clearly a less than optimal solution, there is evidence to suggest that estimating the models using equation 3 does not significantly bias the results. In Appendix D, the results of estimating equation 2, along with a number of other models in which different lower-order interaction terms are dropped, is presented. The results show that  5  Two different dependent variables are used in this chapter: Reform's deficit credibility (2,+2), a and, Reform thermometer rating (0-100). 211  constraining the two interaction terms to be zero has a marginal impact on the magnitude of the three-way interaction term. Credibility on the Deficit The surge in Reform credibility during the campaign described in Chapter 5 fits closely with the results from the prirning analysis in the previous chapter. This is further evidence that "Only as Reform's anti-deficit commitment became clear did the party break out of its ethnoreligious base" (Johnston, Blais, Brady, Gidengil, and Nevitte 1996b: 15; Johnston et al. 1994). Reform's credibility gains during the campaign are potentially important in demonstrating the process by which welfare state attitudes became important to the Reform vote on election day because the credibility gains reflect a learning process. Does the general pattern of credibility gains reflect a simple diffusion of information or is there evidence for individual resistance to information about Reform's deficit position. Figure 7-2 shows that credibility gains do come from voters who were predisposed to think the issue important. Those respondents who fall below the mean on the welfare state scales did not re-evaluate Reform's credibility throughout the campaign period. The two pro-welfare state groups, move toward the neutral point but their net gain at the end of the campaign is less than one quarter of a point. In addition, there is no apparent difference among those who were likely to be exposed to information about Reform's position compared with the chronically unaware.  212  Figure 7-2. Evolution of Deficit Credibility by Awareness and Predispositions Mean Credibility  10-19 Sept  Sept 27- 3 Oct 20-26 Sept  11-17 Oct 4-10 Oct  18-24 Oct  The awareness and welfare state scales were divided at their respective means to create the four groups (n=679; 409; 542; 647).  O f course, the gains among the anti-welfare state respondents were themselves confined to a subset of the population. Awareness mediated the effect of coverage. While both awareness groups view Reform as more credible in weeks 2 and 3, and increase at about the same rate, the credibility gains are only sustained in the high aware group. In fact, in the last two weeks of the campaign, the high aware, anti-welfare state respondents are the only ones who have correctly identified Reform as distinct on the question of the deficit.  213  A Media Effects Model of Reform Credibility The re-evaluation of Reform's credibility by politically aware respondents who were ideologically predisposed to be concerned with or at least give priority to the deficit is prima facie evidence of a campaign and a media effect. There is no compelling reason for high aware, anti-welfare state respondents to change their perception of the party's position except as a response to the flow of information. The next step is to test whether highly aware respondents responded to a quantifiable campaign stimulus. In this case, the surge and latecampaign decline in the amount of Reform coverage is a likely candidate. Table 7-1 provides OLS estimations of equations 1 and 3 with respondents' evaluation of Reform's credibility as the dependent variable. The evidence from the simple coverage model estimation clearly establishes that respondents who were interviewed when Reform was the subject of more news coverage were more likely to view the party as credible on the deficit. Awareness has a strong effect on perceptions of credibility as does an anti-welfare state x awareness interaction orientation. This suggests that people who were likely to have information and those people who were likely to know and accept the information were more likely to correctly perceive Reform's deficit commitment. Anti-welfare state attitudes in and of themselves were not enough. Not surprisingly, education and residence in Western Canada are also associated with perceiving Reform to be more credible on the deficit. Both of these are likely capturing aspects of awareness not reflected in the awareness scale. Income has a smaller but reverse impact.  214  Table 7-1. News Coverage and the Re-Evaluation of Reform's Deficit Credibility  Coverage of Reform  Simple Coverage Model  Two-Mediator Coverage Model  0.003 ***  0.000 (0.001)  (0.001)  0.012 **  Coverage x Awareness x Anti-Welfare State Awareness x Anti-Welfare State  0.916 ** (0.341)  (0.005) 0.456 (0.384)  Awareness  0.647 ***  0.645 ***  Anti-Welfare State  (0.163) -0.045 (0.186)  (0.163) -0.046 (0.186)  Demographics Education  0.032 **  0.031 **  Income Woman Atlantic West  Intercept  Adj-R SEE N  2  (0.011)  (0.011)  -0.062 *  -0.061 *  (0.028) 0.020 (0.044) 0.099 (0.064)  (0.028) 0.023 (0.044) 0.099 (0.064)  0.170  ***  0.170  ***  (0.046)  (0.046)  -0.597 ***  -0.507***  (0.114)  (0.119)  0.10 0.95 2103  0.11 0.94 2103  p<05; ** p<.01; and *** p<.001 Standard Error in parentheses; significant coefficients in bold face. Although coverage is significant in the simple model, the effect of changing coverage is fairly small. The coverage variable, which is the average number of Reform codes during the previous three days, varies between 9 and 115 although most of the days have values less than  215  50. Given this variation, a difference of 30 coded phrases, which is associated with a difference of 0.09 in perceptions of credibility, is a reasonable benchmark for evaluating the impact of coverage on opinion. Since the scale of the credibility measure varies between -2 and +2, this is not persuasive evidence that the coverage was decisive. The estimation of the two-mediator model suggests that the simple model is not a good indicator of the effect of coverage on opinion. As expected, the main effect of coverage is reduced to insignificance when an interaction term is added to capture the possibility that only anti-welfare state respondents who got the message re-evaluated Reform during the campaign. The remaining coefficients in the model are unchanged by the inclusion of the interaction except the coefficient on the two-way interaction between awareness and welfare state attitudes which is lower. This is not surprising given that the graphical evidence suggests that the interaction between awareness and welfare state attitudes only occurs in the second half of the campaign. The interaction emerges because of the change in the availability of information. 6  Estimating the impact of a change of coverage on perceptions is more difficult with the two-mediator model because the impact of a change in a single variable cannot be read off the table. To get around this problem a series of simulations were run with a hypothetical respondent who has average characteristics. He is a middle income, high school educated male living in Ontario. From this starting point, values of coverage, awareness and anti-welfare state attitude are manipulated.  ' This insight is supported by statistical analyses (not shown). If one divides the campaign into periods and estimates the model used here without coverage, the anti-welfare state by awareness interaction is not significant in the early periods but is significant m me latter periods While this finding could be used to constrain the mteraction to be zero and thereby reduce the collinearity in the model, the conservative approach and more appropriate approach is to leave the model as fully-specified as possible. 216  Table 7-2. Reform Deficit Credibility Simulations for Different Levels of Coverage'  Coverage =10  Coverage =40  Difference  Perfect awareness (1.0) Better than average awareness (0.75)  0.48  0.72  0.24  0.23  0.41  Average awareness (0.50)  -0.03  0.10  0.19 0.13  Less than average awareness (0.24)  -0.28 -0.53  -0.21  No awareness (0.0)  0.07 0.02  -0.51  Reform Credibility Prediction for High Aware Respondent  0  Coverage =10  Coverage =40  Difference  Very anti-welfare state (1.0) Somewhat anti-welfare state (0.75)  0.42  0.58  0.15  0.30  Neutral (0.5) Somewhat pro-welfare state (0.25)  0.19  0.42 0.27  0.12 0.08  Very pro-welfare state (0.0)  -0.05  0.12 -0.03  0.05 0.02  0.07  simulations are based on two-mediator model coefficients from column 2 of Table 7-1 assuming the hypothetical respondent is a high school educated (education =4); middle income (income=2), manfromOntario. the hypothetical anti-welfare state respondent for these purposes is one who scores one standard deviation above the mean for the distribution of the scale (0.63). the hypothetical high aware respondent for these purposes is one who scores one standard deviation above the mean for the distribution of the scale (0.70).  a  b  0  Consider first the respondent who is opposed to the welfare state. For the sake of argument, that person has a score on the anti-welfare state index which is one standard deviation (actual score equals 0.63) above the mean for the sample. The predicted values for that respondent's perception of Reform credibility given the two different coverage levels and varying levels of awareness are provided in Table 7-2. Also included is the difference between the two credibility values. Thefirstthink to notice is that changing coverage by 30 - the same value as above - is more substantively important when one considers the respondent's likelihood of accepting and receiving that information. For the respondent who is anti-welfare 217  state, but not extreme by any means, a change from 10 to 40 produces a 0.24 difference if the respondent is fully informed but only a trivial 0.02 difference if that person is generally ignorant of political affairs. In the second part of Table 7-2, a high aware respondent - one who is one standard deviation more aware than the mean - is used to demonstrate the impact of varying the respondent's attitude toward the welfare state on the size of the media effect. The pattern mirrors the earlier results. Increasing Reform's media exposure by 30 points has a larger impact on those people who are predisposed to accept deficit reduction holding awareness constant. Different predictions could be generated by manipulating either the respondent's awareness, level of opposition to the welfare state or the amount of Reform coverage that the respondent was exposed to in the days before the respondent was interviewed. Obviously, the effects would be larger if more extreme coverage values were chosen since a difference of 30 represents a moderate change given the campaign variation. The Reform party began the campaign about as credible on the deficit as its nearest rival the Conservatives, but for all of that, a significant proportion of the public was not informed enough to say what effect a Reform win would have on the deficit. If Reform could get its message out, it could significantly improve its credibility. Two weeks into the campaign, the media began to devote increased attention to the Reform party and that coverage appears to be a plausible explanation for the surge in credibility at least among the politically aware and anti-welfare state respondents.  7  7  It should be noted that the early fiscal controversy that precipitated the Conservative first drop in support also took place about the time when the public was beginning to re-evaluate Reform's deficit position.  218  Perhaps surprisingly, the graphic and statistical evidence indicates that pro-welfare respondents regardless of their level of awareness did not learn, in the sense of attributing high credibility to Reform on deficit reduction. Hypothetically learning that a party's position is far from one's own has no cost in terms of affecting one's judgment about which party to support. It is possible to acknowledge Reform's anti-deficit commitment while simultaneously being unconcerned about the deficit. Nevertheless, it is still possible that those unsympathetic to deficit reduction will not incorporate new information into their evaluation of Reform. Consider that the most important individual detenninant of credibility is a person's ideological position on the issue in question. This is consistent with the idea that the more one cares about an issue the greater one perceives the differences between the parties (Krosnick 1990; Krosnick 1988). Respondents with conservative attitudes as reflected by their position on the welfare state dimension are more likely to perceive party differences (Johnston et al. 1994) and should therefore perceive Reform to be more credible. Alternatively, it may be that other partisans who  are likely to be less opposed to the welfare state received a more partisan message that  Reform's Zero-in-Three strategy would not work. Certainly the wording of the question assumed that a lower deficit was a good thing so other partisans had an incentive to question Reform's ability to realize its goal. For whatever reason, respondents who were unsympathetic to a reduced role for government did not change their rnind about Reform's credibility.  Reform's Credibility  and Changing  Support  for  Reform  The remaining step in establishing that coverage mattered is to consider the relationship between the change in perceptions of Reform credibility on changing support for the Reform party. It is possible, however unlikely, that voters learned Reform's position but that this did not increase Reform support. To test this relationship a pooled, cross-sectional model of  219  R e f o r m ratings that incorporates the respondents p o s i t i o n o n R e f o r m ' s c r e d i b i l i t y (crosssection effect) as w e l l as the m e a n c r e d i b i l i t y score for the party o n the day w h i c h the respondent w a s interviewed (longitudinal o r c a m p a i g n effect) is e m p l o y e d . G i v e n that a p r o welfare state respondent should not be affected b y either process interaction terms are added. T h e results i n T a b l e 7 - 3 c o n f i r m these expectations. A t any t i m e d u r i n g the c a m p a i g n , the more credible a n i n d i v i d u a l v i e w s the R e f o r m party, the more l i k e l y he or she is to support the party c o n t r o l l i n g f o r the aggregate e v a l u a t i o n o f R e f o r m . S i n c e the m a i n effect o f c r e d i b i l i t y is s m a l l , a p r o - w e l f a r e state respondent w h o perceive the party to be credible is o n l y m o d e s t l y m o r e l i k e l y to l i k e the R e f o r m party. S i n c e a n t i - w e l f a r e state attitude range b e t w e e n 0 and 1, the impact o f a change o f one unit o f c r e d i b i l i t y varies between 2.8 i f the respondent is p r o - w e l f a r e state to almost 13 points i f the respondent is a n t i - w e l f a r e state. There is also evidence that the time o f interview matters. That is, respondents w h o are interviewed w h e n R e f o r m is perceived b y the p u b l i c as a w h o l e to be m o r e credible are m o r e l i k e l y to support the R e f o r m party. A g a i n , the impact is dependent u p o n the respondent's u n d e r l y i n g attitudes w i t h p r o - w e l f a r e state respondents being i m m u n e from the trend t o w a r d p u b l i c understanding o f R e f o r m ' s p o s i t i o n . In fact, a p r o - w e l f a r e respondent is l i k e l y to have l o w e r rating o f R e f o r m as p u b l i c understanding o f R e f o r m ' s p o s i t i o n increases. W h e n the p u b l i c as a w h o l e understands R e f o r m ' s mistakes.  8  true nature, p r o - w e l f a r e state voters m a k e f e w e r  A l t h o u g h the l o n g i t u d i n a l effect appears to be stronger g i v e n the size o f the  unstandardized coefficients, it is actually w e a k e r than the cross-section effect since the d a i l y m e a n o n l y increases b y a p p r o x i m a t e l y 0.3 points across the c a m p a i g n .  8  S i n c e some p r o - w e l f a r e state voters w i l l support R e f o r m despite its p o s i t i o n o n s o c i a l programs, not a l l o f their support for R e f o r m is the result o f mistakes.  220  Table 7-3 Cross-Sectional and Longitudinal Effects of Credibility on Support Reform Support Cross-Sectional Effect of Credibility Credibility on Deficit Credibility on Deficit x Welfare State Longitudinal Effect of Credibility Daily Mean Credibility on Deficit Daily Mean Credibility x Welfare State Awareness Anti-Welfare State Attitudes Anti-French Canada Attitudes Anti-Minorities Attitudes Demographics Education Income Woman Atlantic West  Intercept  Adj-R2 SEE N  2.799 ** (0.995) 9.515 *** (1.979) -12.721 * (5.969) 32.381 ** (12.457) -2.171 (2.291) 10.971 *** (2.517) 8.337 *** (2.020) 12.713 *** (2.202) -0.173 (0.266) -0.437 (0.663) -0.798 (1.047) -4.594 ** (1.544) 4.787 *** (1.085) 25.135 *** (2.904) 0.21 21.32 1920  * p<05; ** p<.01; and *** p<.001 Standard Error in parentheses; significant coefficients in bold face.  While the link between credibility perceptions and Reform support as demonstrated is clear, there is the possibility of individual rationalization effects where respondents perceive the party to be more credible because they like the party. Since coverage plays a significant activation role for perceptions of Reform credibility and, as will be demonstrated, for Reform support, rationalization only makes it difficult to sort out the causal direction between credibility perceptions and support. The presence of rationalization does not change the argument that the campaign mattered for activation or that media coverage was an exogenous cause of the campaign changes.  Summary The path diagram is a reliable representation of how the campaign mattered for Reform support. News coverage provided a subset of voters, defined by awareness, the information necessary for them to identify Reform's position on questions of the role of the state and the importance of deficit reduction. Pro-welfare state respondents effectively ignored the information but anti-welfare state voters used the information to change their perception of the party and to evaluate the Reform party in light of these perceptions.  The Evolution of Reform Support While the credibility results are confirmation for the role of the campaign in convincing voters that Reform was committed to deficit reduction and a reduced role for government in society, we are also interested in more direct effects of the campaign on Reform support. Using two different indicators of Reform support - share of vote intentions and mean party evaluation score - the following analysis provides a graphical demonstration of the usefulness of the two mediator model for understanding the response of anti-welfare as well as antiminorities voters to the campaign.  222  Mobilization ofAnti-Welfare State Voters The graphical analysis begins with a breakdown of Reform vote intentions (Figure 7-3) and then proceeds to compare it with the breakdown of Reform thermometer evaluations (Figure 7-4). The breakdown is based on the four groups used in Figure 7-2.  Figure 7-3. Reform Intentions by Awareness and Attitudes Toward Welfare State Percent with Reform Intention  50  25  10-19 Sept  20-26 Sept  Sept 27- 3 Oct  4-10 Oct  11-17 Oct  18-24 Oct  The awareness and welfare state scales were divided at their respective means to create the four groups (n=681; 411; 545; 649).  For the first three weeks of the campaign there is no substantively important change in Reform's share of vote intentions. The most aware, anti-welfare state group of respondents is more likely to have a Reform vote intention compared with their least aware counterparts, while the pro-welfare state groups reveal low probabilities of Reform support. Almost all of the change in Figure 7-3 takes place among the high awareness, anti-welfare state group. 223  Between week 3 (September 27-October 3) and Week 5 (October 11-17), the proportion of this group supporting the Reform party rises from 25 to almost 50 percent. Surprisingly, after peaking at almost 50 percent in the 5 week of the campaign, the high aware and anti-welfare th  state group underwent a significant trend away from the Reform party in the last week. Over the last two weeks, the low awareness, anti-welfare state respondents become more likely to vote Reform, but the magnitude of their increase is small.  Figure 7-4. Reform Rating by Awareness and Attitude toward the Welfare State Mean Thermometer Rating  55  Low Awareness, Pro-Welfare State — L o w Awareness, Anti-Welfare State -•High Awareness, Pro-Welfare State — H i g h Awareness, Anti-Welfare State  10-19 Sept 20-26 Sept The awareness and welfare state scales were divided at their respective means to create the four groups (n=681; 411; 545; 649).  The same basic pattern emerges when respondent's evaluation of Reform is substituted for Reform intention as the dependent variable (Figure 7-4). While the pattern is consistent with the pattern observed for vote intentions, the difference among the anti-welfare state  224  respondents between the high and low aware groups is not as stark. This difference is at least partially an artifact of the large numbers of missing cases in the early weeks of the campaign. As we saw in the last chapter, as the campaign grew there was a dramatic decline in the proportion of respondents who were unwilling to rate the Reform party. Since low aware people are most likely to say don't know, the mean evaluation for the low aware group in the early weeks is drawn from the most aware of the chronically unaware, which should inflate the daily mean for the low aware in the early weeks. In both analyses, the gap between the high and low awareness groups is small or nonexistent by election day, despite the large gaps at different times during the campaign. Fournier (1997) argues that the campaign reduced the differences between high and low aware respondents in terms of the considerations that were brought to bear on the choice. While this confirms Fournier's conclusion, it suggests that awareness mattered for how respondents got to their final vote and suggests that it is not inevitable that campaigns reduce the gap. Among anti-welfare state voters, high aware respondents responded both positively to the campaign early on and negatively over the last week of the campaign, whereas low aware respondents continued to move toward Reform right up till election day. What explains the movement away from Reform among the high aware, anti-welfare state respondents? Given that this is the group that is most likely to receive new information from the media it seems reasonable to suggest that this change was also induced by the media coverage in the last two weeks of the campaign. There were, however, two kinds of information available in the last week that could hurt Reform's electoral prospects. The first kind of information is information about expectations. Since it had become clear that the Liberals were going to win the election, voters who learned this information  225  could have been motivated to vote strategically by casting their vote with the Conservatives. Outside Alberta, a Reform vote had the potential to contribute to the splitting of the right thus allowing voters' least preferred party to win. While strategic calculation may have been involved, it is unlikely that information about the likely outcome was the principal cause of the shift away from Reform for two reasons. Since the strategic environment was complex, there was no obvious party to move toward on the right of the political spectrum, especially in Ontario. In some ridings the Reform party was the most likely to defeat the Liberals while in others it was the Conservative party. In addition, the fact that the same pattern emerges with both dependent variables indicates that it was information about issues or leadership. Strategic information should have only influenced the vote intention variable and not the party evaluation score. The second kind of information is issue based. As Chapters 4 and 6 demonstrated, the Reform party was associated with cultural issues in the l