UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Information and interests in the absence of partisan endorsements : efficient decision-making in the… Hooper, Graeme Andrew 2006

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

Item Metadata

Download

Media
831-ubc_2006-0504.pdf [ 3.44MB ]
Metadata
JSON: 831-1.0092711.json
JSON-LD: 831-1.0092711-ld.json
RDF/XML (Pretty): 831-1.0092711-rdf.xml
RDF/JSON: 831-1.0092711-rdf.json
Turtle: 831-1.0092711-turtle.txt
N-Triples: 831-1.0092711-rdf-ntriples.txt
Original Record: 831-1.0092711-source.json
Full Text
831-1.0092711-fulltext.txt
Citation
831-1.0092711.ris

Full Text

I N F O P v M A T I O N A N D I N T E R E S T S IN T H E A B S E N C E O F P A R T I S A N E N D O R S E M E N T S : E F F I C I E N T D E C I S I O N - M A K I N G IN T H E B R I T I S H COLUMBIA REFERENDUM ON ELECTORAL REFORM  By Graeme Andrew Hooper B . A . (Hons), The University of British Columbia, 2004  A THESIS SUBMITTED IN P A R T I A L F U L F I L L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF M A S T E R OF ARTS  in  THE F A C U L T Y OF G R A D U A T E STUDIES  (Political Science)  T H E UNIVERSITY OF BRITISH C O L U M B I A August 2006  © Graeme Andrew Hooper, 2006  ABSTRACT  This thesis examines how voter decision-making was affected by the absence of partisan endorsements in the 2005 B.C. referendum on electoral reform. In particular, it is hypothesized that parties' positions of neutrality were sufficiently obscure to have been missed by most voters, with those voters who incorrectly inferred a position of support or opposition using their feelings towards the particular party to determine their alignment with the perceived party position. The adverse effects of such mistaken inferences should be mitigated by increasing information, with higher-information voters capable of spreading decisions across complex measures of interest. Two such measures are proposed. First, voters could rely on their feelings towards parties, as the single transferable vote (STV) was generally accepted to be bad for the large parties. Second, voters could rely on the likelihood that their vote would be wasted under the current electoral system, as STV was generally accepted to reduce wasted votes. Analysis of survey data confirms that a majority of respondents missed the party positions of neutrality, with even high-information respondents demonstrating little ability to pick up on the correct position. Further, respondents who incorrectly inferred a position of support or opposition from one of the major parties used their feelings towards that party to determine their alignment with the perceived position. While the data confirms that increasing information mitigates these adverse effects, it does not support the hypothesis that such occurs because of reliance on complex measures of interests. While other highinformation voters were able to activate these interests, high-information voters who were using incorrectly inferred partisan endorsements showed little or no capacity to do so.  ii  T A B L E OF CONTENTS  ABSTRACT  ii  T A B L E OF CONTENTS  iii  LIST OF T A B L E S  v  ACKNOWLEDGEMENTS  vi  DEDICATION  vii  C H A P T E R I: Overview and Summary 1.1: Introduction 1.2: Research Question: 1.3: Overview  1 1 2 4  C H A P T E R II : The Role of Information in Efficient Decision Making: A Survey of the Literature 7 2.1: Introduction 7 2.2: The Role of Information 8 2.3: Complex Versus Simple Processes: 12 2.4: Why Some Simplifications are Liabilities for Efficient Decision-Making 15 2.5: Conclusion 19 C H A P T E R III: Is No Position a Good Position? Potential Partisan Endorsements 3.1: Introduction 3.2: Gordon Campbell and the B.C. Liberals 3.3: Carole James and the NDP 3.4: Are Sour Grapes Green? Adriane Carr and the Green Party 3.5: Conclusion  21 21 21 26 29 32  C H A P T E R IV: Measurements and Hypotheses 4.1: Introduction 4.2: Expected Effects of Party Neutrality 4.3: Data and Methodological Issues 4.4: Conclusion  33 33 33 38 42  C H A P T E R V : Partisan Endorsements and Complex Interests: Empirical Evidence 5.1: Introduction 5.2: Mixed Signals and Poor Decision-Making 5.3: The Role of Information in Overcoming Mistakes 5.4: The Role of Interests in Overcoming Mistakes 5.5: Conclusion  43 43 44 52 54 59  C H A P T E R VI: Concluding Remarks and Implications for Literature  61  iii  BIBLIOGRAPHY  64  APPENDIX: Data, Measurements and Select Methodological Concerns  68  LIST OF TABLES Table 5.1: Perception of the Liberal Position  44  Table 5.2: Perception of the NDP Position  44  Table 5.3: Logit models for Probability of Voting for STV  48  Table 5.4: Conditional Impact of NDP Feelings on Yes Vote  49  Table 5.5: Conditional Impact of Liberal Feelings on Yes Vote  49  Table 5.6: Conditional Impact of NDP Feelings on Yes Vote at Different Levels of Information  53  Table 5.7: Conditional Impact of Liberal Feelings on Yes Vote at Different Levels of Information  53  Table 5.8: Logit Results for Yes Vote and Interests  55  Table 5.9: Logit Results for Yes Vote and Interests with Endorsements  56  Table 5.10: Conditional Impact of Political Partisan Interests on Yes Vote at Different Levels of Information 58 Table 5.11: Conditional Impact of Wasted Vote Interests on Yes Vote at Different Levels of Information 58  v  ACKNOWLEDGEMENTS  I would like to extend my gratitude to Dr. Fred Cutler for his assistance with this project. I would also like to thank the researchers responsible for the collection of the B.C. Referendum Study data set for permitting me to use the data in this work: R. Kenneth Carty, Fred Cutler, Richard Johnston, Mark Warren, Andre Blais, and Patrick Fournier.  vi  DEDICATION  For Mike Meade.  CHAPTER I Overview and Summary  1.1: Introduction On May 17 , 2005 British Columbia (B.C.) voters were asked whether or not the th  province should switch electoral systems. A potentially historic change, the question likely came to many as a surprise. Held concurrently with a provincial election, the referendum registered surprisingly low levels of public awareness in the polls throughout the campaign period. Whether they were distracted by the election, deterred by the relative complexity of the proposed alternative, the British Columbia Single Transferable Vote (STV), or simply overcome by the lack of debate amongst prominent public figures, British Columbians did not seem to know much about STV. In this context, it seems only natural to question the efficiency of the decisions being made. Beyond the absence of information, there were a number of characteristics to the 2005 referendum that also make it a candidate for further study. On the one hand, the referendum's agenda setter was a surprisingly non-partisan body, the Citizens' Assembly on Electoral Reform (CA), which was composed of randomly selected citizens engaged in deliberative democracy. The effects of this agenda setter on voter decision-making have already been the subject of compelling research, however, and the present research 1  aims to explore a different, though related, characteristic of the referendum. Indeed, perhaps the second most notable unique trait of the B.C. referendum is at issue here: the absence of partisan endorsements. The C A design required that political parties remained out of the body's decision-making process on electoral reform, but nothing demanded  Fred Cutler and Richard Johnston, "The B.C. Citizens' Assembly as Agenda Setter: Shaking Up Voter Choice," forthcoming. 1  1  they remain out of the referendum debate that ensued. Yet this is precisely what each of the major political parties decided to do. The negative effects of this decision, as well as how voter could overcome them, are the primary subject here.  1.2: Research Question: The present research effort specifically aims to answer the following question: what happens to efficient decision-making on complex policies when clear partisan endorsements are absent? Political science literature has long promoted the partisan endorsement as a mechanism for voters to bypass informational requirements for decision-making on complex policies. Yet, in the B.C. case, both endorsements and information were absent. How did voters cope with these absences? The literature has overlooked, perhaps because of its rarity in parliamentary democracies, what happens when such endorsements are unavailable, and the case of the B.C. referendum provides an opportunity to fill this gap. Answering the research question first demands an understanding of precisely what efficient decision-making entails. It is common to associate the word 'efficiency' with some concept of maximizing output and minimizing input. Indeed, this was common within how authors such as Anthony Downs, or Arthur Lupia used the term, for each of whom the key output was decision-making, while the input was information. According to Downs, collecting information should only occur until its marginal return equals its marginal cost; to proceed further would be inefficient. For Lupia, voter efficiency is 2  often achieved through eliminating the need for encyclopedic information about a given  Anthony Downs, Economic Theory of Democracy (New York: Harper and Row Publishing, 1957)215.  2  2  decision by relying on some type of simplification (these are discussed in more detail in second chapter). In either case, efficiency requires using what little information voters have in a manner consistent with how one might expect them to behave if they possessed more. For the present research, a similar approach to efficiency is taken. Here, efficiency is about using what a voter has, particularly personal characteristics, to achieve decisions that are objectively consistent. Specifically, efficient decision-making demands that voters use things such as feelings towards the major parties, personal interests, or information, in a manner that is logically consistent. A voter who likes the N D P and dislikes the Liberals, for example, should use these personal characteristics, absent other strategic considerations, to determine that he or she should vote for the N D P rather than the Liberals; to do otherwise would be logically inconsistent, and thus an inefficient use of his or her personal characteristics. Yet, determining whether or not a particufar choice is logically consistent presents a number of methodological dilemmas when applied to survey data. The logical consistency of any given decision, which is very likely unique to each individual, is nearly impossible to track across multiple respondents with limited survey questions. One way around this, though imperfect, is to ignore issues of individual consistency, and focus instead on aggregate behavior. By doing so, researchers can examine how groups of individuals with common characteristics behave, and observe the impact of particular differences. Using this approach, efficient decision-making can be understood as how groups, categorized by certain relevant characteristics, use those characteristics to determine support or opposition for a particular policy. Deviants from this group  3  behavior may still be behaving in a manner that is logically consistent; what is important is that the reasons why such deviations from the group behavior occurred are uncovered. Where deviations are related to incorrect inferences, in this case about partisan endorsements, the decision is assumed to be inefficient. The following research uses this 3  idea of efficient decision-making to understand how voters dealt with the unique characteristics of the 2005 referendum. As will be demonstrated towards the end of the thesis, the absence of clear partisan endorsements caused some voters, even the relatively well informed, to steer away from this understanding of efficient decision-making.  1.3: Overview Following this introduction, the second chapter will provide a general overview of how political science literature has approached the role of information in voter decisionmaking. This will involve first exploring the early and modern skeptics, who questioned the capacity of less informed voters to achieve efficient decisions. Alternative approaches will highlight the partisan endorsement (also known as the partisan cue) as a means for voters to bypass informational requirements. Central to this is an understanding of how decision-making processes vary depending on an individual's level of information on a given topic. Specifically, as information increases so does a voter's ability to spread decisions across more complex considerations. The final portion of the chapter will look at why these simplifications sometimes fail.  For an example of this approach, see: Larry M . Bartels, "Uniformed Votes: Information Effects in Presidential Elections," American Journal of Political Science 40, no 1, (February 1996): 194-230. 4 3  The third chapter will focus on uncovering what exactly happened in B.C., promoting an understanding of how each party's history with electoral reform provided potential cues to voters. Despite these potential cues, the chapter will show that parties never adopted positions of endorsement, generally remaining quiet about STV. The expected effects of these decisions, and any methodological issues associated with later attempts to test them, are the subjects of a brief fourth chapter. Here, four hypotheses are presented. The first two of these hypotheses relate directly to the absence of partisan endorsements. First, the absence of parties from the debate on STV is expected to make their official position of neutrality hard to access, leading to a significant portion of voters mistakenly inferring a position of support or opposition. Second, building on the works reviewed in the second chapter, it is expected that these voters made noticeable mistakes in how they used their feelings towards the parties in question to make decisions. The second two hypotheses relate to how high-information voters might have been able to overcome the effects of these mistakes. Building on the theoretical work presented in the second chapter, it is expected that voters who possessed considerable levels of information about STV were able to spread decisions out, incorporating complex interests specific to the policy. Two interests are presented— partisan feelings and the likelihood of having one's vote wasted under the current electoral system— each in the form of a hypotheses. Before concluding, the chapter briefly discusses the data source, key measures, and excluded variables. The fifth chapter tests each of these hypotheses. The first two hypotheses are confirmed: a significant portion of voters mistakenly inferred a partisan endorsement, with those who did using feelings towards the party in question to orient themselves  5  towards the inferred position. While making mistakes appears to occur regardless of information about STV, the impact is most dramatic amongst low-information voters. In other words, low-information voters appear relatively incapable of overcoming mistakes in party positions. While this hints at support for the hypotheses regarding whether or not high-information voters can spread decisions across more complex interests to over come these adverse effects, attempts to confirm either of the interest hypotheses fail. Voters with high information do demonstrate a relative capacity to use complex interests to determine support, but not to the point of overcoming the effects of getting party positions wrong. This conclusion is interesting in and of itself: not only did the absence of party endorsements deprive low-information voters of one of their most useful simplifications, the obscurity of the position cost high-information voters the ability to use more sophisticated interests to determine their position on STV. The thesis concludes with a discussion that ties the present research findings back into the broader literature. Here, skepticism of the role of heuristics in non-stereotypical electoral conditions will be confirmed, combined with a suspicion of the capacity of voters to use information to overcome such conditions. The conclusions will also suggest a cautious approach to the use of non-partisan debates. As will be shown throughout the thesis, voters' lack of information alone is not the only cause of faulty electoral decisionmaking; a lack of clear partisan endorsements shares some of the burden and therefore deserves attention by democratic theorists.  6  C H A P T E R II The Role of Information in Efficient Decision Making: A Survey of the Literature  2.1: Introduction  How important is information for voters to make efficient decisions? This question, or some variant of it, has fueled prominent debates within political science for the past four decades. Early works suggested that voters with less information had difficultly developing consistent beliefs, while latter work proposed various means through which voters could bypass informational requirements. Yet, in theorizing how such bypassing was possible, political scientists have long enlisted the partisan endorsement as a key tool for voter decision-making. What does this mean for a referendum where informational requirements were reasonably high, information was reasonably low, and partisan endorsements were almost entirely absent? The following chapter explains how political science literature has dealt with the role of the partisan endorsement in enabling voters to make efficient decisions without low levels of information. This begins with a discussion of early skeptics who questioned the capacity of voters to form consistent belief systems without significant amounts of information. From there, an overview of how voters could get around these demands, particularly by simplifying complex decisions into relatively simple ones, will be presented. This will lead to an understanding of the decision-making processes of high and low information voters are unique, with the increases in information enabling a spreading of decisions across more complex considerations. In other words, individuals with more information relying on fewer simplifications, and thus may be less susceptible to making mistakes. The final section, however, cautions against any optimism in 7  assessing voter decision-making, with the dangers of simplifications at any level of information being discussed. Through all of this, the reader will be guided to questions about the role of the partisan endorsement, information, and complex measures of interest, in achieving efficient decision-making. These questions will be essential to uncovering the expected impact of an absence of partisan endorsements in the B.C. referendum, to be discussed in throughout the remainder of the thesis.  2.2: The Role of Information Individuals' capacities to meet the demands of democratic citizenship have long been questioned. Works as early as Schumpeter's Capitalism, Socialism and Democracy have championed skepticism that the conditions set up by classical democratic theory are likely outside the reach of all but the most informed of the populace. If this is the case, are voters able to find ways around the informational requirements, or are they left incapable of making efficient decisions? On the one hand, evidence that voters with less information fail to achieve consistent opinions comes from Converse's seminal article on the nature of belief systems. There, Converse asserts that 'constraint,' or the interdependence of 4  ideas/beliefs, is essential to understanding how belief systems are formed, with sophisticated belief systems expected to contain relatively more constraint than less sophisticated ones. Converse suggests that, at a minimum, such constraint can be achieved through logic alone: i f I support lower taxes, I should not also support higher spending, as one logically excludes the other. Yet, Converse contends, with support from  Philip E. Converse, "The Nature of Mass Belief Systems," Ideology and Its Discontents, ed David E. Apter, (New York: Free Press, 1964) 207.  4  8  interview data, that voters with lower levels of information seem to miss even these logical connections, and their capacity to express consistent opinions is limited across temporal and ideological realms. The gloomy conclusions are verified in later work by Delli Carpini and Keeter, who find not only support for Converse's suggestions regarding a lack of consistent opinion formation, but also record startlingly low levels of information overall.  5  This may be the case, but even Converse accepts that there are some means for voters to get around the excessive requirements of information to achieve constraint. Most notably, Converse suggests that constraint can occur without effort through socialization. This comes as the result of elites — individuals capable of influencing mass communication— packaging certain ideas together: by picking up one idea, individuals often unintentionally pick up another. This is often used by elites to link ideas that have no inherent logical connection (such as advertisements that link buying a certain product to a consumer's happiness), but in doing so enables individuals to form opinions with relative consistency in the absence of the necessary information. Thus, individuals might not know why two ideas go together, but they certainly know that they do.  6  This process has often been called 'cue-taking,' and has long been considered one of the fundamental ways voters get around informational demands. Within the realm of politics, the most useful cues are often partisan in nature. Works as early as Graham  Michael X . Delli Carpini and Scott Keeter, What Americans Know about Politics and Why it Matters (New Haven: Yale University Press, 1996). While in agreement with their core findings, Converse would later argue that Delli Carpini and Keeter had actually understated the lack of political knowledge, citing "the easiness of the items making up these authors' test" of political knowledge as the reason for their underestimation. See: Philip E. Converse, "Assessing the Capacity of Mass Electorates," Annual Review of Political Science 3 (June 2000) 333. Converse (1964) 212. 5  6  9  Wallas's Human Nature in Politics recognized the fundamental role of political parties in reducing the informational requirements of voters. For Wallas, political parties bridged the gap between the complex environment that people live in and the desire for continuity. By presenting such continuity across elections, parties enabled voters to use already formed opinions to tackle new tasks, reducing the informational requirement.  7  Later empirical work supported this simplifying role of political parties, particularly when democracy demands voters develop positions on complex policies, such as is the case in referenda. As Campbell et al put it in The American Voter, such complexities "increase the importance of relatively simple cues to evaluate what cannot be matters of personal knowledge." Such cues can take numerous forms, though in a 8  referendum context the partisan endorsement, whereby political parties make clear statements of support or opposition, is the most pertinent. Reliance on such endorsements does not take any large amount of information, the authors assert, but instead can be achieved by mere association of a particular position with the party label.  9  In a more recent work, Lupia and McCubbins go even further, suggesting that when such endorsements are made readily available, acquiring additional information becomes irrational. This is premised on the assumption that acquiring information requires exerting effort, and such effort, being in limited supply, carries with it an opportunity cost. If voters are able to get to the same 'reasoned-choice' with less information that they would achieve with more information, then acquiring information  Graham Wallas, Human Nature in Politics (New Brunswick: Transaction Books, 1981). See also: Downs. Angus Campbell et al, The American Voter (New York: John Wiley & Sons, 1960) 128. Ibid, 128-129.  7  8 9  10  entails a certain cost but affords no certain benefit.  This takes earlier partisan  endorsement arguments to their extreme: not only is relying on less information sufficient for making good decisions, it is actually preferable.  11  Yet, Lupia and McCubbins recognize that a voter's acceptance of a partisan endorsement requires persuasion, and a voter's willingness to be persuaded should not be taken for granted. Rather, the authors suggest that there are particular conditions when voters will act on endorsements, outside of which they can be expected to simply ignore them. Applying a signaling model, the authors determine that voters must perceive the cue sender as knowledgeable on the subject of the endorsement, and sharing common interests with the voter. Of course, while these conditions may lead to persuasion, the perceptions must be reasonably accurate for low-information rationality to occur (that is, to avoid being tricked). Earlier, empirical work by Lupia suggests that these strict conditions produced by their formal model can actually be softened. Testing the role of interest group endorsements in a California insurance referendum, Lupia found that voters actually accepted endorsements if they simply saw the sender as likeable or possessing a  In later work, Lupia introduces the term of 'competent choice,' or the decision that voters would make if provided full information, seemingly in place of the earlier 'reasoned choice.' See: Arthur Lupia and Richard Johnston, "Are Voters to Blame? Voter Competence and Elite Maneuvers in Public Referendums," Referendum Democracy: Citizens, Elites, and Deliberation in Referendum Campaigns, ed. Matthew Mendelsohn and Andrew Parkin (Toronto: MacMillan/St. Martin's Press, 2001) 191-210. In some respects, this mirrored an argument presented by Downs, who suggested that people who care most about which party wins a given election, that is those who have strong preferences, have the least need for information. See: Downs 238-259. 11  11  reputation for honesty.  The findings confirmed earlier formal modeling done by Sobel,  which demonstrate the importance of reputation in facilitating trust. Much in line with 13  Wallas, Sobel's work supports the notion that partisan endorsements function effectively because voters are able to rely on their past experiences to interpret them.  2.3: Complex Versus Simple Processes: Within the works reviewed thus far has run a theme of bypassing informational requirements by reliance on simplification, notably the partisan endorsement. Such endorsements represent the most widely cited form of heuristics, or cognitive learning tools through which an individual can simplify decisions. But does this actually take place, and i f so, how do the decisions reached compare to those reached by the relatively informed? Sniderman et al's Reasoning and Choice presents an elaboration of the unique considerations that different information classes of voters rely on in decision-making. While partisan endorsements are chief among these considerations when it comes to political decision-making, there are also a number of others, such as poll results or candidate appearance. Sniderman et al suggest that by focusing on a single type of 14  Arthur Lupia, "Shortcuts Versus Encyclopedias: Information and Voting Behavior in California Insurance Reform Elections," The American Political Science Review 88 (March 1994): 63-76. Joel Sobel, " A Theory of Credibility," Review of Economic Studies 52 (Issue 4, 1984): 557-573. Richard R. Lau and David P. Redlawsk, "Advantages and Disadvantages of Cognitive Heuristics in Political Decision Making," American Journal of Political Science 45, (October 2001): 951-971. 12  13  14  12  considerations (often the liberal-conservative spectrum), minimalists oversimplify decision-making, and thereby overlook the consistency of less informed voters.  15  Perhaps at the risk of oversimplification themselves, the authors present two unique heuristics. First, voters who have sufficient information to identify the liberal or conservative position, and also identify themselves with one of those groups (and necessarily not the other), may utilize the 'likeability heuristic' Here, in an effort to achieve consistency, voters align themselves with the position of their preferred group, and demonstrate a fair degree of competence in doing so. Alternatively, voters who lack the necessary information can rely on affect-driven heuristic, motivated by feelings towards groups rather than identification with ideology.  16  Of additional importance, Sniderman et al provide a useful test of the Miller and Shanks multiple-stage voting model, differentiated by levels of political sophistication. Within this model various factors that can influence an individual's decision-making are compartmentalized into a number of different considerations, or stages, separated by proximity to the vote. Furthest away from the vote, individuals consider their stable social or economic characteristics, or their policy related predispositions. Closest to the vote, individuals consider their impressions/evaluations of candidates or the political parties. The ordering is intentional and purposeful: stages further away from the vote influence stages closer to the vote, with voters working, perhaps less than consciously, through each stage to determine decisions.  17  Paul M . Sniderman et al, Reasoning and Choice: Explorations in Political Psychology (New York: Cambridge University Press, 1991). Sniderman et al. Warren E. Miller and J. Merrill Shanks, The New American Voter (Cambridge: Harvard University Press, 1996) 189-211. 15  1 6  17  13  Sniderman et al make a minor adjustment to the ordering of these stages, but retain most of the substance in their testing. What stands out about their results, however, is the discovery of noticeable differences between voters of low and high levels of information. The results demonstrate that voters with less information rely disproportionately on those considerations that are closest to the vote, such as evaluations of candidates or parties. At the other end, as voters increase their information they also increase their capacity of spreading decisions across more distant stages, including complex considerations such as evaluations of policy based on predisposition and interests. While the authors point to the conclusions as evidence of how voters can 18  achieve low-information rationality, it also confirms the suspicion, alluded to earlier, that low-information voters are relatively vulnerable to shocks in the informational environment that surrounds decision-making by over relying on fewer considerations. Sniderman's more recent work adjusts to appreciate the vulnerability of voters, particularly to the choices of political elites. Reacting to what he suspects was an implicit, and probably unmerited, agency afforded to voters in his earlier work, Sniderman points to the role of political elites in shaping how decisions are processed. This is accomplished through the manipulation of 'choice sets,' or the perceived options available in making a decision. In the context of a referendum, it is easy to imagine how elites accomplish this by selecting the wording of the ballot question, though Sniderman suggests a number of ways that are relatively subtle. In particular, elites may shape the debate surrounding the  For similar results, see: Lau and Redlawsk. 14  choices, using devices such as the organization of alternatives, to determine how voters think about what is at stake and what meaning to provide to particular decisions.  19  2.4: Why Some Simplifications are Liabilities for Efficient Decision-Making  Thus far, the literature reviewed has uncovered a tendency for low-information voters to simplify decision-making by relying on relatively few considerations, which may lead to greater vulnerability. Yet, is there evidence to support such vulnerability? In this concluding section, I will review research that questions the viability of heuristic use, particularly for low-information voters. While such voters are the most likely to demand simplification through heuristic use, much empirical evidence contends that they are also the most susceptible to making mistakes in its application. In one effort to verify the conclusions of low-information rationality arguments, Bartels instead found further support for the contention that information matters. His results showed that rather than behaving similarly, well-informed members of particular groups, such as ethnic or religious groups, used information to drive themselves away  Paul M . Sniderman, "Taking Sides: A Fixed Choice Theory of Political Reasoning," Elements of Reason: Cognition, Choice, and the Bounds of Rationality, ed Arthur Lupia et al (New York: Cambridge University Press, 2000). The manipulation of choice sets was the strategy of the federal Liberal party during the 1980 Quebec referendum on seccession, whereby promises were made that altered how voters perceived what they were voting for, and more importantly against; no longer was the referendum one between seccession and the status quo, but between seccession and an ambiguous promise of 'renewed federalism,' an important factor for many Quebecoise. See: Lawrence Leduc, "Opinion Change and Voting Behavior in Referendums," European Journal of Political Research 41, (2002): 719. Arguably, when Carole James suggested an alternative referendum on M M P i f STV failed, noted in third chapter, she was also attempting to manipulate choice sets, though it is not expected she was as successful as Trudeau. 19  15  from the positions of less informed members.  While Bartels' conclusions were reason  to be suspicious of low-information rationality, he seemingly overlooked the central feature of works like Lupia and McCubbins or Sniderman et al: rather than insisting that low-information voters do achieve efficient decision-making, such authors only asserted that they can do so, and listed the appropriate conditions. Chief among them was the use of heuristics; while Bartels had demonstrated that information drove group members apart, it did not explicitly test i f low-information group members who utilized heuristics were able to achieve similar decisions as high-information group members. Lau and Redlawsk paid closer attention to the role of heuristics, but also failed to verify the low-information rationality hypothesis. Using a mock-election campaign in an experimental research setting, the authors modestly hypothesize that "the use of cognitive heuristics generally will be associated with higher quality decisions."  The results are  surprising: rather than helping low-information voters, heavy reliance on heuristics by less informed voters actually inhibits efficient decision-making. In addition, the authors find that even the decisions of well-informed voters are subject to the circumstances of a campaign. When non-stereotypical information appears (such as candidates who do not easily fit within existing assumptions about party behavior), the use of heuristics that depend on stereotyping leave even the well-informed voters making poor decisions. With regards to endorsements, similar findings appear in Johnston et a/'s work on the 1992 Canadian referendum. There, comparable to the 2005 B.C. referendum, voters were asked to decide the fate of a relatively complex policy, in this case a package of constitutional amendments that included senate reform, federal-provincial division of  2 0 2 1  Bartels 194-230. Lau and Redlawsk 955. 16  powers, and the recognition of Quebec as a distinct society. This diversity proved to be the Achilles heel of the accord, with voters having an easier time finding something they disliked about the package than accepting it in its entirety. One interesting feature of the referendum was the role of former Prime Minister Pierre Trudeau, who denounced the accord. For Trudeau, who had spent many of his years as Prime Minister chasing constitutional amendments, the position outwardly appeared as a break from his stereotypical position. While a majority of voters managed to pick up on his position (not surprising, given Trudeau's prominence and the coverage of his statements), voters who mistakenly thought Trudeau favored the accord positively related their feelings towards Trudeau with their support for the accord. Thus, rather 22  than using heuristics to effectively simplify decisions, such voters actually used their feelings to undermine efficient decision-making. The contrast between these types of findings and those predicted by lowinformation rationality approaches can at least in part be attributed to the seemingly incomplete transplanting of heuristics into political science literature. As Kuklinski and Quirk observe:  "Ironically, political scientists have borrowed the concept of heuristics from psychology while overlooking its main significance in that literature. Viewing heuristics as rational strategies for dealing with ignorance, political scientists have stressed how they enhance competence... For the most part, cognitive  Richard Johnston et al, The Challenge of Direct Democracy: the 1992 Canadian Referendum (Montreal and Kingston: McGill-Queen's University Press, 1996) 140. 17  psychologists look at heuristics differently. They see the use of heuristics as automatic, unconscious, and frequently dysfunctional."  According to Kuklinski and Quirk, poorly adapted cognitive capabilities are to blame for the often dysfunctional nature of heuristics, particularly in political decisionmaking. Reviewing a number of works in modern psychology, the authors suggest that the usage of various heuristics is built into human genetic circuitry through evolution; since this evolution took place over millennia of human history very dissimilar to the demands of modern democracy, it is only in rare circumstances that these heuristics prove useful in political decision-making.  24  This does not mean that heuristics are never expected to lead to low-information rationality, but it does point out a number of key obstacles. For one, citizens are often called upon to answer very complex questions. While political parties have been suggested to be a means to simplifying these choices, the circumstances surrounding a 25  choice does not always provide useful party endorsements. In such settings, heuristics are expected to be of little use. Second, the environment within which voters make decisions often provides little information, and political interests often manipulate the information that is available. Thus, acquiring balanced information takes considerable effort. This presents a problem because informed decision-making in a democracy is a collective good, thus meaning there are no incentives for individual voters to undertake any great  James H . Kuklinski and Paul J. Quirk, "Reconsidering the Rational Public: Cognition, Heuristics, and Mass Opinion," Elements of Reason: Cognition, Choice, and the Bounds of Rationality ed. Arthur Lupia et al, (Cambridge: Cambridge University Press, 2000) 166. Ibid, pp. 165. See, for example, Sniderman.  2 3  2 4  18  effort to acquire information, or even to use the information they have in thoughtful ways. Finally, politics rarely provides clear feedback about the accuracy of heuristics, meaning opportunities for correction of future application based on past experience, even if a citizen was interested enough to attempt it, are sparse. In the end, the authors 26  conclude that we should only expect heuristics to facilitate efficient decision-making "when a task is intrinsically simple, when helpful capabilities or dispositions are hardwired, or when institutions or other environmental conditions promote competence."  27  2.5: Conclusion The demands placed on voters in modern democracies are often onerous, and interest is generally limited. This imbalance is often overcome through reliance on simplifications, narrowing complex decisions to a limited number of considerations. The less-informed voter, however, narrows decisions noticeably further than the informed one, and in doing so creates both the potential for low-information rationality and risks making unfortunate mistakes. Much error stems from the nature of information surrounding the vote; notably, where the endorsement of political actors is nonstereotypical, or otherwise unclear, voters who normally rely on partisan heuristics are often stunted in the capacity to recover. While the less-informed voter exhibits a greater tendency to make mistakes when over relying on fewer considerations, high-information voters are also expected to make mistakes when non-stereotypical conditions arise. In the following chapter, the nature of partisan endorsements specifically and information more  2 6 2 7  Kuklinski and Quirk, 167-169. Ibid, pp. 167. 19  generally will be examined in the B.C. referendum. Drawing on the works reviewed here, a number of hypotheses will be generated to better understand the nature of decisionmaking around STV.  20  C H A P T E R III  Is No Position a Good Position? Potential Partisan Endorsements  3.1: Introduction  If any one thing will stain how the widely venerated the C A experience is remembered, it is the uninspired nature of the referendum campaign that followed it. Information was sparse, the proposal was reasonably complex, and political parties stayed quiet. For their part, Gordon Campbell and the Liberals had been pushed to this position by ideas of the virtues of a non-partisan approach to electoral reform well before the process began. Once established, this idea extended naturally to a campaign position of neutrality on the proposed reform. The NDP, while making some statements about their preferences during the process, followed suit, declaring the party as neutral upon the announcement of STV. The Green party took a more nuanced approach, to say the least. Leader Adrianne Carr first ardently supporting reform, then vociferously opposing STV, and finally agreeing to remain neutral only to have the vast majority of her candidates endorse the 'Yes' vote. While these positions had different motivations for each party and leader, they all promoted a general silence around the referendum debate. The purpose here is to demonstrate that this was indeed the case, enabling later hypothesizing about the expected effects of an absence of clear and consistent partisan endorsements. 3.2: Gordon Campbell and the B . C . Liberals  Given the party's history, it is difficult to say with any certainty whether Liberal party voters should have supported or opposed electoral reform on May 17th. More than any other party, the provincial Liberals have had their fate determined by the electoral 21  system of the day, for better and for worse. It was, after all, a Liberal dominated coalition government that was defeated in the 1952 election shortly after implementing B.C.'s only major experiment with electoral reform. The chosen electoral system, the single transferable ballot, was intended to prevent the Canadian Commonwealth Federation 28  (predecessor to the NDP) from forming government, but backfired, facilitating the rise of the relatively unknown Social Credit party. After receiving a majority in the 1953 election, the Social Credit government would revert to FPTP, leaving the Liberal party subject to its disproportional effects for most of the next 40 years. The experiences left 29  the Liberal party and its supporters ample reasons to both desire change and be wary of it. Of course, the elections of this period were likely an abstraction for most voters in 2005, but the mixed signals that period implied for electoral reform were to be paralleled in the decade before the Liberals returned to government. From 1991 to 2001 the Liberal party served as official opposition during an era of N D P governments that most Liberal supporters saw as utterly disastrous. Plagued by scandals, the era reinvigorated interest in reforming the electoral system to encourage  While called the 'single transferable ballot' at the time, it was actually an alternative vote system, and should not be confused with STV; the former worked within single member districts, while STV worked exclusively with multi-member districts, ranging in size from two to seven. Accordingly, the only vote transfers under the single transferable ballot would be after the candidate with the fewest votes had been eliminated (assuming no candidate won on the first count), with that candidate's votes being recounted based on each ballots second preference. While this would also occur under STV when no candidate could be elected, votes would also be transferred at a fractional value when candidates were elected with a surplus of votes. For a history of the short lived electoral system, which was actually proposed by the Social Credit premier, W.A.C. Bennet, while he was a member of the Conservative party (also wiped out in the 1952 election), see: J. Terence Morley et al, The Reins of Power: Governing British Columbia (Vancouver: Douglas & Mclntyre, 1983) 86-87; David J. Mitchell, W.A.C. Bennett and the Rise of British Columbia (Vancouver: Douglas & Mitchell, 1994). 2 9  22  greater balance and prevent future monopolies on power.  For Liberal voters, this focus  was intensified by the results of the 1996 election: widely expected to win, the Liberals failed to get more seats than the incumbent NDP government despite having a larger share of the popular vote. This sparked an interest within the Liberal party to demand reform, though leader Gordon Campbell, in keeping with the populist nature of B.C. politics, took an interesting route. Soon after the election, Campbell adopted the rhetoric that would carry him to, and indeed through, the C A process: electoral reform should be in the hands of the citizens, not political parties. The sentiment led to the 1999 promise 31  at a Liberal party Annual General Meeting that a Liberal government would initiate a non-partisan 'citizens assembly' to explore electoral reform, a promise later added to 32  the Liberal platform for 2001. In dramatic contrast to the 1996 results, however, the Liberal party would pull off the largest landside in B.C. history in the 2001 election, winning over 97% of the seats with 58% of the vote. While Campbell would stick to the earlier promise for a citizens' assembly, it was the results of the 2001 election, in which Liberals now reaped the benefits of the disproportional effects of FPTP, which became the new focal point in the  See, for example, Mike Milke, Barbarians in the Garden City: The B C N D P in Power (Victoria: Thomas and Black Publishers, 2001). In particular, Gordon Gibson's foreword to the book explicitly links the policies of NDP in the 1990s to the need for electoral reform. See comments in: Lori Culber, "Coalition seeks electoral reform in B.C.," The Vancouver Sun 21 Oct. 1997: B I . William Rayner, British Columbia's Premiers in Profile: the Good, the Bad, and the Transient (Surrey: Heritage House, 2000) 270. In his memoirs, B.C. political commentator Rafe Mair takes credit for giving birth to the idea of a citizens' assembly in the summer of 1999 with an ad hoc group of friends. This included Gordon Gibson, who would go on to serve as the architect for the C A , and Nick Loenen, a prominent advocate of reform. See: Rafe Mair, Rafe: A Memoir (Madeira Park: Harbour Publishing, 2004) 219-222. 3 1  3 2  23  pro-reform movement. While voters opposed to the NDP may have had reasons to call for changes during the late 1990s, by the time the 2005 election rolled around, it was the Liberal party that was being cited as the most obvious example of the dangers of excessive control under FPTP. In all likelihood, the prominence of this later message overrode concerns about the 1996 results, now almost a decade old, and left Liberal supporters unsure about change. Yet, if the mixed experiences of the Liberal party left supporters confused about electoral reform, the party's messaging during the C A process and later campaign provided no additional help. While the C A was a minor feature in the 2001 election, the fulfillment of the campaign promise would take a decidedly non-partisan route. When Gordon Gibson, who had been retained to recommend how the C A should be established, structured and run, provided his proposal to the Liberals, one of Campbell's only amendments was to the selection of the C A members. Rather than using the standards of selection used for juries, as Gibson suggested, the Liberals would declare that any candidate or representative of a candidate in the previous two municipal, provincial, or federal elections should be explicitly excluded from membership on the C A .  As the  odds of such persons being selected under Gibson's method were minimal, the change was likely inconsequential. Nonetheless, it reflected a consistency with the rhetoric Campbell had adopted in the 1990s: i f changes to the electoral system were to be undertaken, it should be the citizens, not the parties, to decide.  See British Columbia Citizens' Assembly on Electoral Reform, Making Every Vote Count: the Case for Electoral Reform in British Columbia: Technical Report, (Victoria, Queen's Printer: 2004) 113. 24  Campbell and his government's messages stuck to this notion of the C A as nonpartisan throughout the C A process. Of the three major parties, the Liberals were the only party that did not have their leader, or even a representative, make a presentation or submission to the C A about electoral reform. On numerous occasions, Campbell and other members of cabinet reiterated the intention of leaving the decision to the C A members to decide, intending to avoid any perception that the process was tainted by political involvement. As a natural extension of this message, it thus came as little 34  surprise when Campbell announced that he and the Liberal cabinet would not take a position on the actual recommendation. While private members were welcome to take a 35  position, cabinet and the premier were staying quiet. Arguably, few Liberals broke this silence during the campaign. In one exception, Gordon Gibson, noted earlier as the architect of the C A process, did write a number of articles during the campaign that supported a Yes vote on STV. Yet, these articles focused on Gibson's role in the C A process, not his political past as a Liberal M L A , making connections between his position and the Liberal one unlikely. Alternatively, 36  former deputy premier Christy Clark took a public position opposed to STV. Clark had earlier announced that she would not be running in the 2005 election, which provided her the liberty to speak more freely about STV. Yet, her positions were never linked to the  See, for example, comments in: "Citizen's assembly 'wonderful' experience" Prince George Citizen 29 Nov. 2004: 5 For Campbell's remarks, see: Tiffany Crawford, "Citizens' Assembly ends with praise," Times - Colonist. 28 Nov. 2004: A.5. To be fair, Gibson was once a prominent Liberal, though that was during an era when 'prominent Liberal' was an oxymoron. Gibson was the leader of the Liberal party from 1975 to 1979, a title that came to him by default when he was the only Liberal elected in the 1975 election. More recently, Gibson ran for the leadership of the B.C. Liberals in 1993, placing second to Campbell. From 1993 on, however, Gibson remained relatively removed from the political scene, working as a research fellow at the Fraser Institute. 3 5  6  25  Liberal party, and she played no official role in the Liberal campaign. While voters likely had an easier time linking Clark to the Liberals than Gibson, neither represented an obvious Liberal endorsement of either the 'Yes' or ' N o ' vote. Further, each of these, along with statements made by a small number of Liberal private members, was overshadowed by a surprising solidarity from the cabinet around 37  the 'no-position' position. For his part, Campbell made almost no statements about the details of STV, generally sticking to comments regarding the C A rather than its recommendations. This may have signaled to voters that the Liberal party supported the change, though it was not entirely clear that was the case. Vaughn Palmer, a respected political columnist, attempted to draw this conclusion, but based his argument on the look in Campbell's eyes when he spoke about the C A .  3 8  While Palmer may have been correct  about Campbell's support, the approach spoke volumes for the tight lipped nature of the Liberal campaign: even veteran political observers were reduced to speculation. The average voter, it is fair to assume, should have been equally perplexed.  3.3: Carole James and the N D P Unlike the provincial Liberal party, the provincial N D P is tied through its constitution to its federal counterpart, and the two share a significant overlap in 39  Jeff Bray, the incumbent M L A for Victoria-Beacon Hill, took a position in favor of STV, while Walt Cobb, the incumbent M L A for Cariboo South, took a position against STV. Vaughn Palmer, "Other Liberals not as high as Campbell on STV," The Vancouver Sun 11 May 2005: A3. New Democratic Party of British Columbia, The Constitution of the New Democratic Party of British Columbia (2005). Last viewed August 2006: http://bc.ndp.ca/upload/20060517145555 constitution2005.pdf 3 8  3 9  26  membership and organizers.  This has important implication for policy around electoral  reform, as it was the federal NDP that first actively pushed for the adoption of some form of proportional representation (PR). Championed by NDP Member of Parliament (MP) Lome Nystrom on the eve of the dissolution of parliament for the 2000 election, N D P leader Alexa McDonough would join Nystrom in calling for the appointment of a special committee to consider adopting PR in 2001. When Jack Layton replaced McDonough 41  in 2003, beating out Nystrom, among others, the NDP would adopt an official pro-PR position for the 2004 election. When that election returned a Liberal minority, Layton made N D P support conditional on Liberal movement towards P R .  42  While PR never materialized during the short lived Liberal minority, the positioning of the federal NDP served as an important backdrop to the electoral reform efforts underway in B.C. Following the rise to prominence of the issue within the federal NDP, provincial N D P leader Carole James made a presentation to some C A members outlining her party's stance on electoral reform. The presentation was cautious, steering away from explicitly endorsing one system over another. There was a definite preference indicated for a change to the status quo, and in hindsight of her later comments, noted below, the presentation seemed to hint at a preference for a Mixed Member Proportional  Donald Blake et al, Grassroots Politicians: Party Activists in British Columbia (Vancouver: U B C Press, 1991) 59-62. F. Leslie Seidle, "Electoral System Reform in Canada: Objectives, Advocacy and Implications for Governance," Canadian Policy and Research Networks (Oct. 2002) 9. Henry Milner, "First Past the Post? Progress Report on Electoral Reform Initiatives in Canadian Provinces" Policy Matters 5 (September 2004): 9.  4 1  4 2  27  (MMP) system, though she never explicitly stated so. Rather, James mirrored the Liberal sentiment, encouraging the C A to make up its own mind.  43  Beyond these messages indicating support for PR, NDP supporters had relatively obvious reasons to want change. The results of the 2001 election, discussed above, had devastated the party, reducing their narrow majority to just two seats. To add insult to injury, the party was deprived Official Opposition status because at least four seats were necessary to qualify as a party. Even Campbell recognized that the lopsided results of the 2001 election were a reason to at least consider change. Combined with similar sentiments of disgust for the actions of the Liberal majority government that Liberal supporters had felt during the N D P reign, NDP voters may have had a relatively easier time linking their feelings towards the party to support for electoral reform. The strength of this connection, however, was again unaided by the official messages of the party once the C A final decision was made. Though appearing at times uneasy towards STV, James followed the Liberal lead of taking a position of neutrality, avoiding the strategically messy option of opposing the publicly lauded C A . James' messages during the post C A period did occasionally hint at a preference for M M P , particularly evident in her remarks after the C A announced it had selected S T V ,  44  and  again much later into the campaign when she proposed a second referendum be held on  For a summary of James' presentation, see: British Columbia Citizens' Assembly on Electoral Reform, Presentation Summary, last viewed August 2006, http://www.citizensassembly.bc.ca/resources/kamloops_presentations/James.pdf For example, see remarks in: Scott Deveau, "Green's Reimer not big on Citizens' choice," Vancouver Courier 27 Oct 2004: 25.  4 4  28  M M P i f STV should fail to pass.  Yet the overall message of NDP neutrality on the  subject dominated. One possible perceived break from this position was former NDP premier Dave Barrett's denouncement of STV. Barrett's statement, however, was personal, and did not come from the NDP. More importantly, in what almost appeared to be a co-coordinated effort, former Social Credit premier Bill Bennett released a statement to press denouncing STV on the same day. Media coverage made light of the fact that Barrett and Bennett made odd bed-fellows: Barrett had defeated Bennett's father's government in 1972, and Bennett himself defeated Barrett's government in 1975. Thus, rather than signaling that the NDP opposed STV, the message carried by media was more that politicians from either side of the political spectrum disliked the proposed system. Thus, as was the case 46  with the Liberals, it is unlikely that voters had much in the way of NDP endorsements to help simplify decisions.  3.4: Are Sour Grapes Green? Adriane Carr and the Green Party  47  Prior to the C A process, Green voters could rely on a relatively consistent proreform message from the party and its leaders. In the mid-1990s, Stuart Parker, then leader of the Green party, co-founded the Electoral Change Coalition of British Columbia (ECCO-B.C.) along with Julian West. ECCO-B.C. promoted the idea of a referendum  Vaughn Palmer, "May 17 vote on electoral reform not necessarily the last word, James says," Vancouver Sun 6 Apr 2005: A.3 See, for example: Doug Ward, "Ex-Premiers Barrett, Bennett say no to STV," The Vancouver Sun 6 May 2005: B I . The pun, "Are Sour Grapes Green?," is borrowed from political commentator Sean Holeman's blog entry on the topic of Adriane Carr's response to the C A ' s announcement of STV. See: Sean Holeman, "Are Sour Grapes Green?," Public Eve Online. 24 Oct. 2004; last viewed August 2006: http://www.publiceveonline.com/archives/000347.html  4 5  4 6  4 7  29  whereby citizens could select a new system from a range of choices, and did not explicitly endorse one system over another. Adriane Carr, when she replaced Parker in 48  2000, would go a number of steps further, creating Free Your Vote B.C. to pursue a citizen initiative under B.C.'s Recall and Initiative Act. Perhaps out of necessity given the requirements of the Act , Carr used the process to explicitly lobby for the adoption of 49  M M P . While the initiative was unsuccessful, understandably given the requirements of the Act, it raised awareness about electoral reform and served to firmly label Carr and the Green party as supportive of change. With the initiative behind her by the time the C A was underway, Carr took easily the most active role in promoting change through the C A process of any of the three main party leaders. Unlike James, Carr did not hold back on her party's particular preferences, stating on numerous occasions that M M P was clearly the best option for British Columbia. Carr provided the C A with the draft legislation for M M P that she had developed during the initiative process, and even outlined reasons why the C A should reject S T V .  50  Some C A members appeared resentful of Carr's enthusiasm, with some  Though largely inactive since the establishment of Fair-Vote B.C., ECCO-B.C.'s webpage can still be viewed, which provides the organization's mission statement. See: Electoral Change Coalition of British Columbia, last viewed August 2006: http://www.mala.bc.ca/~westi/ECCO/ The Act allows a citizen to introduce a legislative proposal, through the Chief Electoral Officer and a Select Standing Committee, into the Legislative Assembly, but requires that a draft piece of legislation be provided; this meant Carr had to pick a system as part of the initiative process. British Columbia Citizens' Assembly on Electoral Reform, "Presentation Summary," last viewed August 2006: http://www.citizensassemblv.bc.ca/resources/victoria presentations/Carr.pdf 4 9  5 0  30  accusing her and the Green party of bombarding the public submission process with duplicate M M P proposals.  51  Yet, i f all of this provided Green supporters with relatively clear signals that the party supported reform, Carr's response to the C A ' s selection of STV would only serve to muddle it. With an obvious preference for M M P , Carr chided the C A for ignoring the overwhelming majority of public submissions, describing the day she heard about the STV decision as one of the hardest of her life. Most dramatically, Carr surprisingly 52  called for the Green party to establish and support a ' N o ' campaign. Many observers were shocked by her response, with political columnist Les Layne noting that Carr looked "like someone poised to win the lottery, but already complaining about the colour of the money." The Green party membership was similarly perplexed by Carr's position, with 53  growing anger towards her for seemingly dooming a proposal from which the party clearly stood to gain. The anger boiled over at the party's 2004 Annual General Meeting, and as a result Carr and the Green party would fall into line with the other two parties, adopting a position of neutrality and encouraging individual candidates and constituencies to take their own position. Unlike the other parties, however, Carr's earlier statements, and 54  See comments in: Bill Tieleman, "Greens, Liberals Drove Assembly to STV," Georgia Strait 9 Dec 2004: 15. See comments in: Cindy E. Harnett, "Leader urges Greens to keep the faith," Times Colonist, 21 November 2004: B I . Les Layne, "Carr finally gets her wish, but she's still not happy," Times Colonist 26 October 2004: A10. For a copy of the press release announcing this position, last viewed August 2006, see: Green Party of British Columbia, "Making Every Vote Count," last viewed August 2006: http://www.greenparty.bc.ca/frames/frame503.html. Unfortunately, the Green party has removed Carr's original press release against STV from their webpage; efforts to receive a copy from the party proved unsuccessful. 5 1  5 2  5 3  5 4  31  their scrutiny from the media, would follow her during the campaign. Ironically, the comments would stand in sharp contrast to the majority of Green candidates who publicly promoted a Yes vote, leaving Green supporters with mixed messages about the party's 55  position. Thus, while Green supporters did not have any official endorsements to consider, they could have been reasonably expected to pick up some message of support or opposition.  3.5: Conclusion Despite a number of potential cues within each party - Campbell's supportive statements towards the C A , James' promotion of M M P - no clear endorsements from any of the major parties emerged. This meant that the debate around STV was run relatively separate from the debate on the provincial election, which ran concurrently. In this context, it is worth recalling the works presented in the second chapter, which suggested voters could often use partisan endorsements to simplify decision-making. If partisan endorsements were unavailable, it is fair to assume that voters simply ignored this piece of information? Or were voters left guessing where each of the parties stood? The work done so far leaves a number of questions about how the absence of partisan endorsements, discussed above, affected voter decision-making. In the chapter that follows, these expected effects will be drawn out.  Late in the campaign, Carr herself estimated that 95% of Green candidates had taken up the Yes position. See comments in: Glenn Bohn, "Green's won't endorse, but like electoral reform," Vancouver Sun 3 May 2005: A5. 32  C H A P T E R IV Measurements and Hypotheses 4.1: Introduction  In the second chapter, the partisan endorsement was uncovered as a primary tool used by voters to overcome low levels of necessary information. In the third chapter, the B.C. referendum was understood to lack such endorsements, despite each party having a number of potential endorsements. If these conclusions are accurate, it makes sense to proceed by questioning what the expected impact the absence of partisan endorsements in B.C. had on efficient decision-making. The chapter thus begins with the exploration of four testable hypotheses, the first two dealing with how the absence of endorsements negatively affected efficient decision-making, and the second two focusing on means through which voters, particularly those with more information about STV, could overcome these effects. The latter half of the chapter then turns to questions of data sources, measurement issues, and rationale for why certain variables or considerations are excluded from later statistical models. This will lay the foundation for the analysis work that will comprise the fifth chapter.  4.2: Expected Effects of Party Neutrality  Party neutrality likely had a number of effects on the capacity for voters to make efficient decisions, though the focus here will be primarily on predictable mistakes. Perhaps the most obvious mistake relates to voters' abilities to pick up on the position of neutrality from each of the main political parties. This was the case because, intentional or not, taking no position actually proved an effective means to avoid talking about STV altogether. Indeed, despite being the subject of a referendum, STV was not once  33  mentioned in the televised leaders' debate, and with the exception of James' announcement that a separate referendum could be held on M M P , none of the three main parties issued press releases on the subject of STV during the election campaign. In effect, this meant two separate information drives had to be launched, with media having to sacrifice attention to one to provide coverage of the other. In a competition against a general election, even a lackluster one at that, a referendum on a complex electoral system often lost this battle for media time. This means that information about STV was sparse, and even the information about party positions of neutrality was likely drowned out by the election coverage. Given this lack of clarity, it is possible that a significant portion of voters made mistakes about party positions, assuming either positions of opposition or support coming from either of the main political parties. The first hypothesis follows: Hypothesis 1: A significant portion of voters mistakenly inferred a party endorsement from one of the main political parties.  For such voters, mistakes in party endorsements should translate into predictable mistakes in how feelings towards the party in question are used to determine support or opposition for STV. Recalling how voters who mistook Trudeau's position on the Charlottetown Accord used their feelings towards Trudeau, we should expect voters who thought the Liberals opposed STV to behave differently from voters who thought the party supported it. In either case, voters should use their feelings towards the party to orient themselves to the inferred endorsement, with voters who like the Liberals coming into alignment with the position, and voters who dislike the Liberals moving away from  34  the position. We expect the same to occur with N D P voters. This is an application of what Sniderman et al referred to as the iikeability heuristic,' where the only information a voter requires to make a decision is the perceived position of a particular group, and feelings towards that group. As the data used in the later analysis, discussed below, provides measures of each of these types of information, this logic can be presented as a testable hypothesis. The second hypothesis thus follows: Hypothesis 2: Voters who mistakenly inferred party endorsements used their feelings towards the party in question to determine their alignment with the perceived position of that party.  These first two hypotheses bring us to questions about the role of information in overcoming the adverse effects of making mistakes in partisan endorsements. Recalling the work of Sniderman et al, and Miller and Shanks, reviewed in the second chapter, voters with higher levels of information about STV should be able to mitigate the effects of these mistakes by spreading decisions across more complex considerations. Such information levels should measure factual knowledge about STV, with the content of the actual measure discussed in the Appendix, in order to activate specific complex considerations. This occurs because complex considerations often include predispositions relevant to the policy in question, in this case STV; it is only by increasing information about STV that voters can become capable of linking their interests to their decisionmaking. It thus follows that some attention must be paid to the nature of the system in order to appreciate what types of interests should be looked for.  35  Developed by a statistician and in use sparingly throughout the world, STV is designed to maximize voter choice and minimize wasted votes, all the while producing relatively proportional results. To do so, however, the process is a bit abstract: ballots must enable voters to rank candidates in larger, multi-member constituencies, a quota for being elected must be calculated based on the number of M L A s to be elected in each constituency and the number of ballots cast, first preferences must be counted, candidates can be excluded, certain votes are transferred at fractional values, and so on. This speaks to the complexity of the choice required of voters, though such is not the explicit purpose here. In terms of interests, there were a number of features of the proposed system that should have guided voters; given data and time constraints, only two are selected. The first relates to how partisan feelings should be used. Partisan feelings are understood to capture how strongly a voter feels towards one party relative to his or her average feelings towards all parties. Acting on the assumption that voters with strong 56  preferences for one party over another should also have a strong preference for which party controls policy output, we can hypothesize how such partisan feelings should have influenced decision making on STV. STV was widely expected to reduce the likelihood of majority governments, replacing them with coalition governments. As coalitions would reduce any one party's ability to monopolize policy output, than voters with strong partisan feelings should be relatively less supportive of STV than voters with weak partisan feelings. In other words, as a voter's preference for one party over another increases, his or her support for STV should decline. We should expect this to occur  A precise measure is presented in the Appendix. 36  primarily within those voters who have a relatively large amount of information about STV, as such information would be necessary to appreciate the impact of switching electoral systems on the major parties' abilities to form majority governments. Accordingly, it is hypothesized: Hypothesis 3.1: Voters decreased their support for STV as the strength of their partisan feelings increased, but only at relatively high levels of information.  Second, by decreasing the number of wasted votes, STV should have differently affected voters whose preferred candidate was almost certain to lose compared to voters whose preferred candidate is almost certain to win. By using vote transfers, STV 57  enabled the substantial reduction in wasted votes. This lead some critics to scoff that STV was for losers, benefiting candidates that have little or no chance under the current 58  electoral system. While this critique may have been aimed at fringe parties, it equally applied to candidates from either of the major parties running in constituencies where they had little or no chance of winning. A n N D P voter in West Vancouver-Capilano had considerably more to gain from STV than an N D P voter in Vancouver-Mount Pleasant, while the reverse was true for Liberal voters.  59  A precise measure is presented in the Appendix. Ted Colley, "Viewpoint: STV is for losers, so I'm glad the vote failed," Now 25 May 2005: 8. Representing lopsided constituencies from either party in the 2005 provincial election, the Liberal candidate won West Vancouver-Capilano with 68% of the vote compared to the N D P candidate's 18%; the NDP candidate in Vancouver-Mount Pleasant won with 64% of the vote compared to the Liberal candidate's 21%. See: Elections B C , "2005 General Election Statement of Votes: Summary of Results by Electoral District," last viewed August 2006: http://www.elections.bc.ca/elections/sov05/resultsbved.pdf 5 8  5 9  37  It is thus hypothesized: Hypothesis 3.2: Voters decreased their support for STV as their vote's likelihood of being wasted decreased, but only at relatively high levels of information.  4.3: Data and Methodological Issues While testing each of these hypotheses will be the purpose of the fifth chapter, it is appropriate to briefly mention here the source of the data, key variables and concerns regarding that testing. The 2005 B.C. Electoral Reform Referendum Study Data forms the primary source of data used. The study consisted of interviews of over 2500 respondents, conducted from mid January until May 16 , the day before the election. th  Respondents were asked questions related to the referendum. The data was modified for 60  this thesis, using respondents' postal code information to add data about the constituency level results for the 2005 provincial election. This information is essential in calculating the expected likelihood of wasted votes, used in hypothesis 3.2 above. Among the survey questions, respondents were asked whether or not a particular party, or its leader, has taken a public position on STV. Respondents were also asked to estimate their feelings towards each of the main political parties using a 0 to 100 scale. As noted above, the answers to these questions are essential to testing each of the first two hypotheses. Further, respondents were asked a series of factual questions about the proposed reform, with a composite measure of'information' about STV being formed from these answers. This information scale, the specifics of which are discussed in the Appendix, is used throughout the fifth chapter. Such factual information about STV could  See: Cutler and Johnston. 38  have been substituted with more general information about politics, which the survey also recorded, or some composite that includes both types of information could be formed. Yet the nature of hypothesis 3.1 and 3.2 demands that any analysis incorporates a measure of information that is specific to the proposed policy, STV, and thus the more direct measure is preferred. Most importantly, the survey asked respondents the question on the referendum ballot: "Should British Columbia change to the BC-STV electoral system as recommended by the Citizens' Assembly on Electoral Reform?" This forms the primary dependent variable of interest, as the study is focused on why voters decided to support or oppose STV. As can be expected, a significant portion of respondents admitted they did not know, and are thus excluded from all models presented in the analysis chapter. These excluded individuals are disproportionately those with lower levels of information about STV. This represents a constraint of an independent variable, however, and does not present specific concerns for causal inference. The survey also asked respondents more 61  generally whether or not they knew anything about the referendum. Over half of respondents admitted that they did not; however, such respondents were potentially included in the models as they were still asked about their vote intentions. Nevertheless, over 60% of these respondents were self excluded by declining to answer this latter question. Attempts to separate the remaining individuals out posed too great a threat to the overall sample size, and no surprising differences in the nature of effects were accomplished in doing so. As such, these individuals remain in the sample.  This issue is elaborated in the discussion of the information measure, found in the Appendix. 39  In testing each of the hypotheses in the following chapter, the Green party will be excluded. This is appropriate for a number of reasons. First, neither of the complex interests is expected to apply to Green party voters, for obvious reasons. Even Green candidates that are relatively competitive to other Green candidates are still largely uncompetitive when compared to the Liberal or NDP candidates. Accordingly, Green voters never reached the point where their votes ceased to be wasted (that is, no Green candidate came close to being elected in the 2005 election). The very nature of Green support, therefore, means that no significant variation exists on the measure for wasted votes, making conclusive results difficult. Second, the adverse effects of STV on party control are disproportionately borne by parties that elect members. Thus, while Green voters are expected to demonstrate variation on this second independent variable, the causal mechanism that makes it of interest (that is, partisanship predisposing voters to disliking policies that hurt parties) is not expected to be present. Each of these concerns supports exclusion of the Green party from explicit consideration when testing the third hypothesis. As the Green party position was also the most mixed, with Carr vociferously opposing STV and local candidates clearly supporting it, it is unlikely that the researcher can comfortably conclude that voters who said the Green party either supported or opposed STV got that position wrong. This makes testing either of the first two hypotheses difficult. The analysis will also avoid any direct treatment of the C A ' s influence over voter decision-making. This may alarm some readers, as the voters' attitudes towards the C A have been demonstrated to be a surprisingly important heuristic in the referendum. In a paper using the B.C. Electoral Reform Referendum Study Data, Cutler and Johnston  40  found that the C A ' s role as agenda setter shaped how voters oriented themselves towards STV. Voters appeared capable of replacing information about STV for information about the C A , using the latter to determine their support for STV. If this was the case, why leave considerations associated with the C A out of the present analysis? To put it bluntly, because the links found by Cutler and Johnston, between attitudes towards the C A and support for STV, were strong enough to wash out other, more subtle relationships. Removing these variables, specifically attitudes towards the C A , enables the examination of these more subtle relationships, such as the link between complex interests and support for STV. While this may be the case, it also opens the analysis to considerable risk by failing to control for an important variable. Such a failure means that the relationship being captured, in this case, the one between the complex interests and support for STV, may actually be caused by a spurious relationship with the excluded variable. That is, it may be that attitudes towards the C A are causing support for STV while also causing the measures of complex interest. If this were the case, excluding attitudes towards the C A would provide misleading results that interests had actually caused support for STV. Yet, sustaining an argument that attitudes towards the C A caused either of the complex interests being examined here is difficult. Partisan feelings, particularly those towards the Liberals and the NDP, should be unaffected by attitudes towards the C A , as each party publicly lauded the C A . Even if this were not the case, it would be more probable to suggest that partisan feelings, which presumably existed for most voters prior to the C A , influenced attitudes towards the C A . Perhaps more obviously, the notion that attitudes towards the C A caused the likelihood of an individual's vote being wasted is  41  highly improbable. The likelihood that a vote is wasted is determined by the chances of a voter's preferred party in that voter's constituency. This is clearly independent of feelings towards the C A . Therefore, if neither of the complex interests being examined were caused by the excluded variable, attitudes towards the C A , than the exclusion should pose no serious risks for causal inference.  4.4: Conclusion A number of hypothesized effects have been presented. The first two relate specifically to how the absence of partisan endorsements, as discussed in the third chapter, adversely affected voters' efficient decision-making. The second two pursue the expectation that voters with more information about the policy in question, in this case STV, should be relatively less impacted by any such adverse effects. In particular, voters with more information should be able to spread decisions across more complex measures of interests, such as their partisan feelings, or the likelihood that their vote will be wasted under the status quo electoral system, mitigating the dangers of making mistakes about partisan endorsements. With a number of core methodological issues also confronted in this chapter, the fifth chapter may focus entirely on testing these hypotheses.  42  CHAPTER V Partisan Endorsements and Complex Interests: Empirical Evidence  5.1: Introduction  The present chapter will utilize data from the 2005 B.C. referendum study to test the hypotheses presented in the previous chapter. I will begin by exploring the availability of party positions. In line with the first hypothesis, data suggests party positions were sufficiently obscure for most voters to have missed them, even voters with high levels of information. Further, a considerable number of voters appear to have 62  gotten party positions wrong altogether. Such voters, as hypothesized, demonstrate a tendency to use feelings towards the party in question to achieve consistency with the perceived positions, either increasing support as their feelings increase towards a party thought to be in favor of STV, or decreasing support as the feelings increase towards a party thought to be opposed to STV. The final hypothesis that high-information voters are relatively capable of overcoming false signals by relying on more complex heuristics, fails. This in itself is an interesting conclusion: even high information voters suffered poor decision-making if they mixed up the party signal. This occurred despite having the political partisan consideration and the wasted vote considerations at their disposal, which other high-information voters used with impressive efficiency. This confirms that information enables spreading of decisions across more complex considerations, but also cautions against the promotion of the partisan endorsement as the panacea for informational requirements in decision-making.  Through out this chapter, voters are divided into categories of information. For an overview of how this division is accomplished, see the Appendix.  43  5.2: Mixed Signals and Poor Decision-Making  Survey data confirms the first hypothesis's suspicion about the obscurity of party signals generally, and the significant portion of voters who mistakenly inferred a partisan endorsement specifically. Table 5.1 and 5.2 present results for respondents when asked about the position taken by the major political parties, where Table 5.1 shows the perceived Liberal position and Table 5.2 shows the perceived N D P position. Across both tables, most respondents either missed the signal of neutrality, admitting they did not know, or providing the wrong answer. Perhaps counter-intuitively, information proves to be of little use in increasing respondents' ability to get the correct signal. While highinformation respondents are twice as likely to get the right answer for the Liberals, and almost twice as likely for the NDP, they are also more than twice as likely to get the wrong answer. Indeed, i f looking at just those respondents who provided an answer, lowinformation respondents were actually more likely to give the correct response than either the medium or high-information category. Clearly, many respondents were guessing.  Did the Liberal Party take a position? Yes No (correct answer) Don't Know  Did the NDP take a position? Yes No (correct answer) Don't Know  Table 5.1: Perception of the Liberal Position Level of Information High (%) Medium (%) Low (%)  201 (28.2) 157(22) 355 (49.8)  180(12.4) 213 (14.6) 1,062 (73)  Table 5.2: Perception of the NDP Position Level of Information Medium (%) Low (%)  145 (20.4) 132(18.5) 435 (61.1)  131 (9) 191 (13.2) 1,130 (77.8)  141 (30.3) 138 (29.7) 186(40)  High (%)  109 (23.5) 106 (22.8) 249(53.7)  Yet, high-information voters reveal a tendency to increase their accuracy as the campaign carries on. For example, taking into consideration only those respondents 44  surveyed in May, the final month of the campaign, 36% of high-information voters provide the correct response while 24% continue to provide the wrong response when asked about the Liberal position (the remaining 40% did not know; results not shown), with similar results for the N D P position. Moderately-informed voters maintained the tendency to provide the wrong answer more often than the right answer through all stages of the campaign, while low-information voters remained slightly more likely to provide the right answer than the wrong answer. Overall, by considering all respondents from May, we can expect that at least 20% of voters went to the polls with the wrong signal from the Liberal party. Similar results are found for the NDP, though respondents remain less likely to pick up the NDP position than the Liberal position, with an estimated 17% of voters going to the polls with the wrong signal from the NDP. These estimates of 20% and 17% for the Liberal and N D P positions, respectively, are conservative for a number of reasons. First, they consider the party positions separately; the numbers are considerably higher when considering how voters did at getting the right position for both parties. While 36% of high-information voters were able to get the correct Liberal position in May, and 27% were able to get the correct N D P position, only 22% were able to get both. For all information levels, only 13% of respondents got both party positions correct in the May sample, while 27% got at least one signal wrong. Second, the percentages are considerably higher if respondents who provided a response of 'don't know' are excluded. The prevalence of such responses, particularly since the survey design did not provide the option to respondents (instead requiring them to provide it on their own), speaks to the general hypothesis that party positions were 45  unclear and overshadowed by the election campaign.  While it is impossible to know  whether or not voters made decisions about party positions by the time they went to the polls, the data suggests that respondents who answered 'don't know' to questions about party positions were approximately 8% less likely to turnout to vote. If the respondents 64  who claimed they did not know the party positions are less likely to have actually turned out, and if respondents who provided an answer other than 'don't know' were roughly evenly split between right and wrong answers, it is fair to assume.that the number of actual voters who had a mistaken position is again higher than suggested. Of equal importance is how voters used party signals, correct or not. Table 5.3 presents two logit models, the first testing for the effect of the perceived N D P position on the referendum vote, interacted with feelings towards the NDP, and the second testing for the effect of the perceived Liberal position on the vote, interacted with feelings towards the Liberals. Before turning to a more readily interpretable analysis of the Table's findings, it is worth first making two initial observations. First, the figures show a distinct effect for N D P feelings relative to Liberal feelings, in the directions hinted at in the third chapter. As respondents increase their support of the NDP, they also increase their  The total number of 'don't know' responses may have been reduced had the survey design allowed for probing questions from the interviewer to elicit responses, though such may have only served to introduce randomness to answers. See: Howard Schuman and Stanley Presser, Questions and Answers in Attitude Surveys: Experiments on Question Form, Wording, and Context (New York: Academic Press, 1981) 114. Results for this specific finding are not shown, but are based on the collapsing of those respondents who said they were 'very likely' to vote and respondents who said they were 'somewhat likely' to vote into one category, those who are expected to turnout, and respondents who said they were 'somewhat unlikely' to vote and 'very unlikely' to vote into another, those who are not expected to turnout. As is the case in most surveys on the subject, the data considerably over predicts turnout, at approximately 90% for those who provided a response to the party position questions, and 82% for those who provided a 'don't know' response. 6 4  46  likelihood of supporting STV, while the reverse is true for feelings towards the Liberals. Further, the nature of the correlation is much stronger for feelings towards the N D P than for feelings towards the Liberals. This too was expected, as Liberal voters faced a much more complex association between Liberal support and support for STV, given the party's history noted in chapter three, compared to NDP supporters.  65  Second, the findings of Table 5.3 support the general idea that getting party positions wrong had significant, clearly perverse effects on voting behavior. Each of the interactions between feelings towards a party and the perceived position of that party, particularly the wrong position, is in the direction one would expect, and all are statistically significant with the exception of the perception that the Liberal party was opposed to STV. The failure of this latter interactive effect may be a further manifestation of the earlier observed relative complexity of the connection between Liberal feelings and support for STV. The data suggests that voters who thought the Liberals were opposed to STV relied on their feelings in manner indistinguishable (statistically) from voters who knew the party had taken no stance. In other words, feelings towards the Liberal party were useful, but strength was not clearly heightened by thinking the Liberals opposed STV. Nonetheless, the results show that for the three other incorrect positions, respondents used their feelings about parties in a manner both predictable and different from other voters to determine support for STV.  The differing strength of association could alternatively be treated as additional support for a hypothesis suggested by Blake et al that B.C. N D P supporters demonstrate greater cohesion around particular policies relative to right of centre party supporters, though such is not of specific importance here. See: Blake et al. 47  Table 5.3: Logit models for Probability of Voting for S T V  Coefficient (Standard Error) Variable  Model 2: Party .21  Model 1: Party = NDP .6  Party position: don't know Party position: favor Party position: opposed Feelings towards Party Don't knowXfeelings FavorXfeelings OpposedXfeelings Knowledge about STV Constant  (.27)  (.26)  -.46  -.76**  (.42)  (.34)  1.76*  .54  (.68)  (.53)  .02*  -.01**  (.01)  (.00)  -.01  .00  (.01)  (.01)  .03**  .03*  (.01)  (.01)  -.04*  -.00  (.01)  (.01)  1*  1.07  (.16)  (.16)  -.66*  .33  (.24)  (.22)  Number of Observations 1384 Pseudo R 0.06 * Significant at the 0.01 level, two tailed testing ** Significant at the 0.05 level, two tailed testing 2  |  1384 0.04  To appreciate the substantive significance of these findings, tables 4.4 and 4.5 are presented to show what the results tell us about the expected probability of voting for STV under certain conditions. Table 5.4 presents the conditional effect of feelings 66  towards the NDP for different perceived positions of the NDP on the referendum question. The table demonstrates the expected difference in the probability of STV  To do so, CLARIFY, a program developed by Michael Tomz, Jason Wittenberg, and Garry King, is employed. CLARIFY enables Stata users to implement the techniques described in: Garry King et al, "Making the Most of Statistical Analyses: Improving Interpretation and Presentation," American Journal of Political Science 44 (April 2000): 341-355. 48  support, comparing voters who were within the 5 percentile on feelings towards the N D P (those who did not like the N D P at all) and voters in the 95 percentile (who really th  liked the NDP), all else equal. This means that a value of -.34, for instance, indicates that voters who believed the NDP opposed STV and were in the 5 percentile on feelings th  toward that party are 34% less likely to support STV than voters who also thought the N D P opposed it, but were in the 95 percentile on feelings toward the party. th  Table 5.4: Conditional Impact of NDP Feelings on Yes Vote  NDP Position N D P Opposed  Effect of change from 5 percentile to 95 percentile for feelings towards the N D P on probability of Yes vote -.34  (.18) N D P Position Unknown  . 15  (.06) N D P Took No Position  .28  (.09) N D P in Favor  .59  (.09)  Table 5.5: Conditional Impact of Liberal Feelings on Yes Vote  Effect of change from 5 percentile to 95 percentile for feelings towards the Liberals on probability of Yes vote th  Liberal Position Liberals Opposed  -.21  (.14) Liberals Took No Position  -.17  (.08)  Liberal Position Unknown  -.09  Liberals in Favor  .31  (.05)  49  th  For the bulk of respondents who said they did not know the party position, the shift from the 5 percentile position to the 95 percentile position, from anti- to proth  th  N D P , resulted in a 14.7% increase in the likelihood of supporting STV. This speaks to the general finding: feelings towards the NDP are associated with support for STV, with such voters relying on their feelings towards the party in the expected manner. Interestingly, respondents who got the party position correct were better equipped to use their feelings towards the N D P to determine support for STV, with an expected increase of 27.9%. At first glance this finding appears puzzling, as the party signal in question portrayed no information about support or opposition for STV. Yet it is likely that the voters who picked up the N D P signals correctly were more likely to have picked up other N D P messages as well. This would enable them to more readily translate feelings towards the N D P into feelings towards STV by borrowing from other known party positions, such as those noted in the third chapter. More dramatic, however, is the comparison between respondents who got either of the wrong signals. For those who thought the NDP supported STV, a shift from the 5  th  to 95 percentile resulted in an expected increase of 58.6% in the likelihood of th  supporting STV. This represents more than double the expected increase for voters who got the signal right. Yet thinking the NDP opposed STV actually reversed the positive association between feelings for the N D P and support for STV, with the shift from the 5  th  to the 95 percentile actually resulting in a 34.1% drop in the likelihood of support. Thus, th  while N D P voters were generally expected to support STV, the nature of the association was conditional on the position they perceived the NDP as taking. When N D P voters got the position wrong, as was the case with Trudeau's position on the Charlottetown  50  Accord, and inferred a position of opposition, they were actually expected to drop their support for STV as their feelings towards the party increased. Looking at the conditional effect of feelings towards the Liberal party on the vote, conditional on perceptions of the Liberals' stance, similar findings emerge. As was discussed earlier, the connection between Liberal feelings and support for STV is weaker than for N D P feelings. For those respondents who did not know the Liberal position, the shift from the 5 percentile to the 95 percentile position, anti- to pro-Liberal, results in th  th  an expected drop in the likelihood of supporting STV by 9.4%. Anti- and pro-Liberal respondents who got the position correct, however, were 16.6% apart. This is precisely the flip side of the finding on feelings towards the NDP: respondents with the right signal appear better equipped to link feelings towards the party to support for STV. Again, the most interesting comparison comes when looking at the two categories of respondents who got the wrong signal. Respondents who thought the Liberal party opposed STV had a more dramatic association between their feelings and opposition towards STV, with an expected drop in the likelihood of support by 21.4% when moving from the 5 to 95 percentile position. As suggested earlier, this position is not th  th  particularly different from the expected change for voters who got the position correct. At the other end, however, respondents who thought the Liberal party supported STV show a more dramatic difference; having this perceived position actually reverses the trend of a negative correlation between feelings towards the Liberal party and support for STV, with the difference between the 5 to 95 percentile, anti- to pro-Liberal, resulting in an th  th  increase in the likelihood of supporting STV by 31.3%. Those who liked the Liberals and thought they supported STV were powerfully moved toward support.  51  This tells us a fair bit about how voters used mistaken party signals. While voters who did not know the party position used their feelings towards the party in predictable and arguably logical ways, voters who got positions wrong behaved differently; for these latter individuals, their position on STV was determined by connecting their feelings towards the party in question to the perceived position. Thus voters who thought a party favored STV used their feelings towards that party quite differently from voters who thought a party opposed STV, with either behaving generally in a manner distinct from voters who got positions correct or simply did not know party positions. What was the role of information in all of this? Increasing information, after all, proved to be of little assistance in reducing the errors about party positions. Does this mean we should expect all information classes of voters to be as susceptible to the effects presented in Tables 5.4 and 5.5?  5.3: The Role of Information in Overcoming Mistakes The third hypothesis was that voters with higher levels of information could spread decisions across more complex considerations, thereby reducing the impact of any given mistake. If this was the case, at a minimum we should expect the translation of mistakes in party positions into mistakes in voting behavior to be weaker as information increases. In other words, the coefficients in Table 5.4 and 5.5 should be noticeably smaller for respondents with higher levels of information. To investigate this, Tables 5.6  52  and 5.7 recreate the figures presented in Tables 5.4 and 5.5, respectively, now separating voters with high and low levels of information.  67  Table 5.6: Conditional Impact of NDP Feelings on Yes Vote at Different Levels of Information 68  Effect of change from 5 percentile to 95 percentile for feelings towards the N D P on probability of Yes vote th  NDP Position N D P Opposed N D P Position Unknown N D P Took No Position N D P in Favor  th  Low Information -.37  High Information -.28  (.19)  (.16) .11  .16  (.06)  (.04) .25  .3  (.09)  (.08)  .64  .48  (.09)  CO  Table 5.7: Conditional Impact of Liberal Feelings on Yes Vote at Different Levels of Information  Effect of change from 5 percentile to 95 percentile for feelings towards the Liberals on probability of Yes vote th  Liberal Position Liberals Opposed Liberals Took No Position Liberal Position Unknown  Low Information -.24  High Information -.16  (.16)  (.11)  -.17  -.13  (.08)  (.06)  -.11  (.06) Liberals in Favor  th  .35  (•l)  -.13  (.06) .24  (-07)  Tables 5.6 and 5.7 are probability tables generated from the Logit results presented in Table 5.3. Here C L A R I F Y is used to compare the expected effects of different inferred partisan endorsements for respondents with either high or low information. For Tables 5.6 through 5.9, low information is determined by setting information to 0 in Logit simulations; high information is determined by setting information to 1 in Logit simulations; this represents the lowest and highest possible levels of information, respectively. See the appendix for a fuller description of the measure. 8  53  Clear across both tables, the gulf between respondents who got positions wrong and those who got positions right is minimized by having more information. In other words, the effects of mistakes in party signals are not as severe for high-information voters. Getting the false signal that the N D P opposed STV, low-information voters were about 9% more likely to oppose STV than high-information voters when comparing those in the 5 and 95 percentiles; getting the false signal that the NDP was in favor of STV, th  th  low information voters were pushed 16% more strongly toward support of STV than high information voters at the two extremes of NDP feelings. Similar, though slightly less dramatic, results are presented for the Liberal model in Table 5.7. Across both models, the influence of mistakes in party endorsements is most obvious for low-information respondents. While high-information respondents still translate mistakes in party endorsements into mistakes in decision-making, they do so less drastically than their low-information counterparts. How such voters did so shall be the subject of the final section, with particular attention paid to the role of interests in determining support or opposition for STV.  5.4: The Role of Interests in Overcoming Mistakes  This brings us to the third general hypothesis that high-information voters are able to overcome the effects of incorrect party signals by relying on more complex interests. Specifically, recall hypothesis 3.1: Voters decreased their support for STV as the strength of their partisan feelings increased, but only at relatively high levels of information; and 3.2: Voters decreased their support for STV as their vote's likelihood of being wasted decreased, but only at relatively high levels of information.  54  The use of each of these interests may be readily tested using the measures for partisanship and wasted votes, each of which is discussed in the Appendix. The results from a Logit model are presented in Table 5.8. The results confirm the suspicion that voters with more information use more varied considerations in their electoral decisions. Supporting the use of the political partisan consideration, the results show that as partisan feelings increase, support for STV drops, but only as information increases. Similarly, though a weaker effect, as the likelihood of a voter's vote being wasted goes up, support for STV increases, but again, only as information increases.  69  Table 5.8: Logit Results for Yes Vote and Interests  Estimate (std. err.)  Variable Partisan  -.01*** (.01)  Wasted Vote  -.00 (.00)  Knowledge about STV  1.76* (.28)  Partisan*Knowledge  -.03*  Wasted*Knowledge  -.01***  (.01) (.01)  Constant  .38 (.14)  N Pseudo R 1 * Significant at the 0.01 level, two tailed testing ** Significant at the 0.05 level, two tailed testing *** Significant at the 0.05 level, one tailed testing 2  1395 0.06  The latter finding only passes standard measures of statistical significance, the 0.05 level, when using single tailed testing. Using two tailed testing, the results are nonetheless significant at the p > 0.065 level, narrowly missing standard measures of statistical significance. Relying on single-tailed testing is of little concern as the predicted direction was found, and there is little reason to suspect that the effect should have been in the other direction; to suggest that voters should have decreased their support for STV as their expected benefit from STV increased would be implausible. 55  Though they are interesting in their own right, the results do not exactly test the hypothesis of interest. The results show that voters overall demonstrate an ability to use the political partisan and wasted vote considerations as their information increases, but do not show whether or not high-information voters who got party positions wrong were able overcome their mistakes by spreading decisions to these more complex considerations. To do so, Table 5.9 recreates the model used in Table 5.8, now separating voters who got at least one party position wrong from voters who either did not know the party position or correctly asserted that parties had taken no position.  Table 5.9: Logit Results for Yes Vote and Interests with Endorsements Estimate (std. err.) DK/Right Party Position Wrong Party Position -.01 -.01  Variable  Partisan  (.01)  (.01)  Wasted Vote Knowledge of STV Partisan* Knowledge Wasted* Knowledge Constant  -.02  -.00  (.02)  (.00)  1.38**  1.9*  (•51)  (.34)  -.00  -.04*  (.01)  (.01)  .01  -.02**  (.01)  (.01)  .7  .27  (.27)  (.16)  493 | 0.04 Pseudo R * Significant at the 0.01 level, two tailed testing ** Significant at the 0.05 level, two tailed testing  N  2  |  902 0.06  Surprisingly, the results disconfirm the hypothesis. For high-information voters who got party positions wrong, neither the political partisan interest nor the wasted vote interest meets standard tests of statistical significance. The political partisan 56  consideration is in the direction predicted. Nonetheless, the expected effect is considerably smaller for voters who got the positions wrong than for voters who got party positions right, suggesting that the application of the endorsement heuristic overwhelms other considerations. This is particularly true for the wasted vote consideration, the impact of which is reduced to nil among high-information voters who got party positions wrong. The importance of this finding is best drawn out by looking at the behavior of high-information voters without the mixed signal, for whom the complex considerations were used with surprising efficiency. Table 5.10 and 5.11 provide readily interpretable results for this conclusion. As was the case with Table 5.4 and 5.5, the results measure the probability of supporting STV, ranging from 0 to 1. In the case of Table 5.10, however, the results look at the expected effect of moving from the 5 percentile position th  to the 95 percentile position on the measure of partisan feelings. In other words, the th  table shows the expected difference between non-partisan respondents and highly partisan respondents at different levels of information. Similarly, Table 5.11 presents the expected difference between 5 percentile respondents and 95 percentile respondents on th  th  the measure of wasted votes. The 5 percentile respondents are those who vote in th  constituencies where their preferred party is almost certain to be a regular winner, while 95 percentile respondents prefer a party that is a perpetual loser in their constituencies. th  57  Table 5.10: Conditional Impact of Political Partisan Interests on Yes Vote at Different Levels of Information = = = _ _ =  Effect of change from 5 percentile to 95 percentile for partisanship (feelings towards the favorite party) on probability of Yes vote th  Information level Low Information  th  -.09  (.08)  -.33  Medium Information  (.05) High Information  -.47  (.08)  Table 5.11: Conditional Impact of Wasted Vote Interests on Yes Vote at Different Levels of Information  Effect of change from 5 percentile to 95 percentile for likelihood of having a wasted vote on probability of Yes vote th  Information level  Low Information  th  .02 (.08)  Medium Information  .16  (.06)  .23  High Information  (.08)  Looking at the results for Table 5.10, the role of information in activating the political partisan consideration amongst voters who did not mistake party signals is evident. While the model estimates that low information voters will decrease their expected likelihood of supporting STV by about 9% by moving from non-partisan position to the extreme partisan position, the large standard error (relative to the coefficient) is a solid clue that this effect is far from certain. Medium information respondents, however, respond to the shift from non-partisan to extreme partisan with surprising efficiency, dropping the expected likelihood of supporting STV by 32.7%. For their part, high information voters display the greatest efficiency in applying the political 58  partisan consideration, with a shift from non-partisan to highly partisan leading to a 47.3% drop in the expected likelihood of supporting STV. Table 5.11 provides similar results for the wasted vote consideration. Again, the model predicts an effect for low information voters that is small, particularly when considered with the relatively large standard error. Medium information voters demonstrate a much clearer effect, though again not as dramatic as high information voters, for whom going from a constituency where your favored candidate is almost certain to win to one where he or she is almost certain to lose increases the expected likelihood of supporting STV by 23%. The fact that these effects are not as strong as those predicted for the political partisan consideration likely speaks to the complexity of the wasted vote consideration: not only did voters have to know that STV would reduce wasted votes, but they also had to accurately perceive how competitive their candidate was within their local constituency. Further, the measure for this competitiveness, discussed in the appendix, was derived from the actual results of the 2005 election, making it exogenous from the survey respondents. Accordingly, the measure overlooks individuals who may have thought their candidate was competitive when he or she was not.  5.5: Conclusion  The results presented here testify to the difficulty voters faced in the B.C. referendum. Information did activate complex interests, but not to the extent of overcoming the adverse effects of incorrectly inferring party endorsements. This leaves open the question of why voters with higher levels of information are relatively less susceptible to these adverse effects. On the one hand, it should be noted that Cutler and 59  Johnston have already demonstrated the relative competence of high-information voters to rely on relevant predisposition in the B.C. referendum, with these present results perhaps being interconnected with theirs. On the other hand, information is itself an important predictor for voting behavior, positively correlating to the Yes vote. If this is the case, high-information voters may be less likely to rely on incorrectly inferred endorsements because they are more capable of relying on information itself. In either case, what is important is that the liability of incorrectly inferred partisan endorsements trumps the benefit of spreading decisions across more complex interests. In the concluding section, the implications for this research on the existing literature will be more precisely drawn.  60  C H A P T E R VI Concluding Remarks and Implications for Literature  Theories about how voters cope with low levels of information have often relied on partisan endorsements as a relatively reliable simplification that enables voters to bypass the informational requirements of democratic decision-making. While some empirical work has found support for this idea, there are a number of questions about the effectiveness of reliance on endorsements, and how susceptible voters are to particular circumstances that affect the usefulness of such simplifications. While the present research study does not claim to tie off all the loose ends surrounding this debate, it has provided some important insights into how voters use endorsements in unique circumstances, and what the role of interests is in supplementing the partisan cue. The case of the B.C. referendum was particularly unique because of the absence of clear partisan endorsements. Each of the three main political parties declined to adopt an official position of support or opposition to the proposed reform, resulting in a separation of the election and referendum campaigns. The competition for attention of these two campaigns, among other factors, lead to an absence of partisan information about STV and a general low level of information about STV. This had important implications for how voters made decisions. As was shown in the fifth chapter, a significant portion of voters mistook partisan signals, incorrectly inferring a position of support or opposition for at least one of the two largest parties. For these voters, feelings towards the party in question were used in different ways compared to those voters who did not get the party position wrong. Increasing information lessened the adverse effects of these mistakes, though not necessarily because information enabled  61  the reliance on more complex considerations based on interests. Voters with higher levels of information do demonstrate a capacity to spread decisions across complex interests, but not to the point of clearly overcoming the effects of getting party positions wrong. Insofar as the referendum provided no partisan endorsements, despite political parties having historical or interest-based reasons for supporting or opposing change, the study represents what Lau and Redlawsk would call non-stereotypical circumstances. When such is the case, relying on simplifications that are premised on stereotyping, such as was likely the case with the inferred positions for political parties in B.C., voters of all information classes are susceptible to predictable mistakes. While the data supports Sniderman et al's contention that voters spread decisions across more or fewer considerations, there are two caveats. First, low-information voters' ability to do so is premised on particular circumstances. For Snidermand et al, use of the 'likeability heuristic' requires that voters have a perception about the position of a particular political party as well as feelings about that particular political party. The present findings confirm that, but refute that application of the likeability heuristic leads to efficient decisions; rather, application in a setting where the perceived positions are incorrect leads to inefficient decisions. Second, the findings question the capacity of high-information voters to spread decisions across more numerous considerations. In the B.C. referendum, this was the case for high-information voters who did not get the party position wrong. Such voters demonstrated an ability to utilize interests such as wasted votes and partisan sentiments to achieve consistency. Yet, the impact of these interests on decision-making was diminished, if not eliminated, when high-information voters got the party position  62  incorrect. Thus, while spreading may occur, it is not a means to over come nonstereotypical conditions. The research also questions Lupia and McCubbins' suggestion that acquiring information once a partisan endorsement is activated is irrational. In the B.C. case, voters were better off by ignoring partisan endorsements and acquiring additional information to enable reliance on complex interests. Voters who inferred an endorsement instead, were left incapable of utilizing information to aid decisions. For these voters, acquiring additional information was useful in lessening the impacts of a mistaken endorsement, but not the extent of enabling reliance on complex interests, as measured here. Finally, the research questions the desirability of non-partisan debate for complex questions. On the one hand, poor decision-making was the result of low-information, with the less informed voters relatively incapable of using complex interests to determine their vote. Yet, the problems were intensified by an absence of clear party positions. By remaining out of the debate parties left voters with few resources to determine partisan endorsements. Discussed above, this lead to incorrect inferences about positions that invalidated the role of information in helping to activate complex interests. Thus, while low-information was part of the problem, increasing information alone would not have provided a complete solution; rather, greater clarity on the partisan positions was necessary. This could be accomplished within the context of party neutrality, though it is perhaps most feasible where parties play a more active role in the referendum debate.  63  BIBLIOGRAPHY Barrels, Larry M . "Uniformed Votes: Information Effects in Presidential Elections." American Journal of Political Science. 40 (February 1996): 194-230. Blake, Donald et al. Grassroots Politicians: Party Activists in British Columbia Vancouver: U B C Press, 1991. Bohn, Glenn. "Greens won't endorse, but like electoral reform." Vancouver Sun. 3 May 2005: A 5 . British Columbia Citizens' Assembly on Electoral Reform. Making Every Vote Count: the Case for Electoral Reform in British Columbia: Technical Report. Victoria: Queen's Printer, 2004. — "Presentation Summary." last viewed August 2006: http://www.citizensassembly.bc.ca/resources/victoria presentations/Carr.pdf — "Presentation Summary." last viewed August 2006, http://www.citizensassembly.bc.ca/resources/kamloops presentations/James.pdf Campbell, Angus et al. The American Voter. New York: John Wiley & Sons, 1960. Colley, Ted. "Viewpoint: STV is for losers, so I'm glad the vote failed." Now. Surrey: 25 May 2005: 8. Converse, Philip E. "Assessing the Capacity of Mass Electorates." Annual Review of Political Science. 3 (June 2000): 331-353. -—"The Nature of Mass Belief Systems." Ideology and Its Discontents. Ed. David E. Apter. New York: Free Press, 1964: 207. Crawford, Tiffany. "Citizens' Assembly ends with praise." Times Colonist. 28 November 2004: A.5. Cutler, Fred and Richard Johnston. "The B.C. Citizens' Assembly as Agenda Setter: Shaking Up Voter Choice." Forthcoming. Culber, Lori. "Coalition seeks electoral reform in B.C." The Vancouver Sun 21 October 1997: B I . Delli Carpini, Michael X . and Scott Keeter. What Americans Know about Politics and Why it Matters. New Haven: Yale University Press, 1996. Deveau, Scott. "Green's Reimer not big on Citizens' choice." Vancouver Courier. 27 Oct. 2004: 25. 64  Downs, Anthony. Economic Theory of Democracy. New York: Harper and Row Publishing, 1957. Elections B C . "2005 General Election Statement of Votes: Summary of Results by Electoral District." Last viewed August 2006: <http://w ww.elections.bc.ca/elections/sov05/resultsbyed.pdf> f  Electoral Change Coalition of British Columbia. Last viewed August 2006: <http://www.mala.bc.ca/~westj/ECCO/> Green Party of British Columbia. "Making Every Vote Count." Last viewed August 2006: <http://www.greenparty.bc.ca/frames/frame503.html> Harnett, Cindy E. "Leader urges Greens to keep the faith." Times Colonist Victoria, 21 November 2004: BI Holeman, Sean. "Are Sour Grapes Green?," Public Eye Online. 24 October 2004. Last viewed August 2006: http://wAvw.publiceveonline.com/archives/000347.html Johnston, Richard et al. The Challenge of Direct Democracy: the 1992 Canadian Referendum. Montreal and Kingston: McGill-Queen's University Press, 1996. King, Gary et al. Designing Social Inquiry: Scientific Inference in Qualitative Research. New Jersey: Princeton University Press, 1994. King, Garry, et al. "Making the Most of Statistical Analyses: Improving Interpretation and Presentation." American Journal of Political Science 44 (April, 2000): 341-355. Kuklinski, James H . and Paul J. Quirk. "Reconsidering the Rational Public: Cognition, Heuristics, and Mass Opinion." Elements of Reason: Cognition, Choice, and the Bounds of Rationality. Ed. Arthur Lupia et al. Cambridge: Cambridge University Press, 2000. Lau, Richard R. and David P. Redlawsk. "Advantages and Disadvantages of Cognitive Heuristics in Political Decision Making." American Journal of Political Science. 45 (October 2001): 951-971. Layne, Les. "Carr finally gets her wish, but she's still not happy." Times Colonist. 26 October 2004: A10 Leduc, Lawrence. "Opinion Change and Voting Behavior in Referendums." European Journal of Political Research. 41(2002): 711-732. Lupia, Arthur and Richard Johnston. "Are Voters to Blame? Voter Competence and Elite Maneuvers in Public Referendums." Referendum Democracy: Citizens. Elites, and  65  Deliberation in Referendum Campaigns. Ed. Matthew Mendelsohn and Andrew Parkin. Toronto: MacMillan/St. Martin's Press, 2001, 191-210. Lupia, Arthur. "Shortcuts Versus Encyclopedias: Information and Voting Behavior in California Insurance Reform Elections." The American Political Science Review. 88, (March 1994): 63-76. Mair, Rafe. Rafe: A Memoir. Madeira Park: Harbour Publishing, 2004. Milke, Mike. Barbarians in the Garden City: The B C N D P in Power. Victoria: Thomas and Black Publishers, 2001. Miller, Warren E. and J. Merrill Shanks. The New American Voter. Cambridge: Harvard University Press, 1996. Milner, Henry, "First Past the Post? Progress Report on Electoral Reform Initiatives in Canadian Provinces." Policy Matters. 5 (September 2004). Mitchell, David J. W.A.C. Bennett and the Rise of British Columbia. Vancouver: Douglas & Mitchell, 1994. Morley, J. Terence et al. The Reins of Power: Governing British Columbia Vancouver: Douglas & Mclntyre, 1983. New Democratic Party of British Columbia. The Constitution of the New Democratic Party of British Columbia 2005. Last viewed August 2006: http://bc.ndp.ca/upload/20060517145555 constitution2005.pdf Palmer, Vaughn. "Other Liberals not as high as Campbell on STV." The Vancouver Sun 11 May 2005: A 3 . — "May 17 vote on electoral reform not necessarily the last word, James says," Vancouver Sun 6 April 2005: A.3 Rayner, William. British Columbia's Premiers in Profile: the Good, the Bad, and the Transient Surrey: Heritage House, 2000. Schuman, Howard and Stanley Presser. Questions and Answers in Attitude Surveys: Experiments on Question Form, Wording, and Context New York: Academic Press, 1981. Seidle, F. Leslie, "Electoral System Reform in Canada: Objectives, Advocacy and Implications for Governance." Canadian Policy and Research Networks. (October 2002): 9-28. Sniderman, Paul M . et al. Reasoning and Choice: Explorations in Political Psychology. New York: Cambridge University Press, 1991. 66  Sniderman, Paul M . "Taking Sides: A Fixed Choice Theory of Political Reasoning." Elements of Reason: Cognition, Choice, and the Bounds of Rationality. Ed Arthur Lupia et al. New York: Cambridge University Press, 2000. Sobel, Joel. " A Theory of Credibility." Review of Economic Studies. 52 (Issue 4 1984): 557-573. Spector, Norman. "Lies and truths on the campaign trail." Times Colonist. 29 April 2005: A.14. Tieleman, Bill. "Greens, Liberals Drove Assembly to STV." Georgia Strait 9 December 2004 Wallas, Graham. Human Nature in Politics. New Brunswick: Transaction Books, 1981. Ward, Doug. "Ex-Premiers Barrett, Bennett say no to STV." The Vancouver Sua 6 May 2005: B I .  67  A P P E N D I X : Data, Measurements and Select Methodological Concerns  B.C. Electoral Reform Referendum Study: As described by two of the study's research  directors, "B.C. Electoral Reform Referendum Study was directed by Andre Blais, R. Kenneth Carty, Fred Cutler, Patrick Fournier, and Richard Johnston. Interviews lasted 20 minutes on average and all content referred to the referendum. The question was embedded in Computer Assisted Telephone Interviewing (CATI) software originating with the University of California Survey methods (CSM) group. The sample was released dynamically: as a weekly rolling cross-section from 17 January to 30 April and then a daily rolling cross-section (Johnston and Brady 2001) from 1 May to 16 May. 1586 responses were obtained from January through April, 1057 in M a y . "  70  Yes Vote: The yes vote appears as the dependent variable in all statistical models in the fifth chapter. The measure is a dichotomous variable recording whether or not respondents said they would vote 'yes' in the referendum, with respondents who said yes receiving a score of 1 and respondents who said no receiving a score of 0, and is included in the B.C. Referendum Study data set. As the measure is dichotomous, all statistical models in the fifth chapter use Logit models, rather than OLS regression. OLS regression models, when run, produced largely similar results.  See: Cutler and Johnston. 68  Party Position: Party position is included in the B.C. Referendum Study data set, though some minor recoding was done to make the measure interpretable as was done here. The measure is derived from the question asked to respondents about whether or not each of the party's and their leaders had taken a position on STV. Respondents who said that they had were then asked if they knew if the position was one of support or opposition.  Knowledge about S T V : A scale for knowledge about STV is comprised using several factual knowledge questions about the proposed electoral system. Scores on this scale range from 0 to 1. As one might expect, given the nature of the information campaign for STV, there is significant pooling of respondents at the low end of the knowledge scale. This is particularly dramatic when comparing the number of persons who expressed a position on the referendum, the primary dependent variable, noted above. Of respondents that declined to express a position, and thus are excluded from all models, 77.6% scored 0 on the STV knowledge scale, compared to 36.9% of respondents who did express a position. These respondents also knew less about politics in general, with 32.3% scoring 0 on a similar scale of general political knowledge compared to 25.9% of those who expressed a position. Accordingly, the model systematically excludes persons with less information. It may seem that this should cause some alarm given that hypotheses are explicitly concerned with the role of information. Yet, such systematic exclusion amounts to an artificial constraint on the variation of an independent variable, which by nature does not bias inference, though it may circumvent later attempts to generalize the  69  findings.  71  On a related note, the models that consider the political partisan and wasted  vote considerations include interaction terms between both of the heuristics and knowledge. Therefore, conclusions will only be drawn for particular categories of respondents with high, medium or low levels of information, as opposed to generalizing across all information levels. In models that distinguish low, medium and high-information, low-information is measured as those respondents who scored 0, medium is measured as those respondents who scored 0.5, and high is measured as those respondents who scored 1. The exception is table 5.1 and 5.2, where, in an effort to include all respondents, including those whose scores fell between these scores, the sample was divided roughly into thirds. Again, 0 was included in the low category, 0.5 in the medium category, and 1 in the high category. As the later analysis of different categories of information represents estimated coefficients for different information levels, the loss of respondents between categories is not of particular concern.  Partisan: For this, respondents' stated feelings about each of the three major parties are examined, with a respondent's party preference being equated to the party that received the highest score. It may seem strange to look at this measure to determine party preferences as opposed to any measure in the data that actually records respondents' vote intentions. However, by looking at the feeling thermometer scores, a measure can account for strength of party support, central to the political partisan consideration.  Gary King et al, Designing Social Inquiry: Scientific Inference in Qualitative Research. (New Jersey: Princeton University Press, 1994): 137. 7 1  70  Wasted Votes: The measure of wasted votes speaks to the basic question about individual candidate competitiveness: what chances did a respondent's favored party have of losing that respondent's constituency? This is different from measures used to rank constituencies across parties on how competitive they are. Instead, this measure must look at how competitive each candidate is within each constituency. A simple measure is proposed: Equation A . l : comp = lk  vaax{p ,p ,...)-p MJ  MJ  IJ  Where comp/kis the competitiveness of the /' party in constituency k, max(p ij, p 2j,...) is i+  i+  the vote percentage for the party that received the highest number of votes in constituency k (other than party i), and p,* is the vote percentage of / party in constituency k. Put simply, the formula subtracts a party's vote percentage from the vote percentage of the winner if it lost, or the closest competition if it won. In theory the measure ranges from 100, where the party in question received no votes and the winner received all votes, to -100, where the opposite occurred. In the present data set, the measure ranges from 56.7 (Green supporters in Vancouver-Quilchena) to -50.1 (Liberal supporters in West Vancouver-Capilano). Respondents are assigned a preferred party based on their score for the above discussed measure of 'partisan,' and measures for the vote percentages for the candidates are based on the actual results from the 2005 provincial election. As a result, the measure is exogenous from the respondents in the survey. In some respects this is a virtue, as it eliminates concerns that respondents made up the competitiveness in some way that might threaten the models. Alternatively, it places considerable demands on respondents, as they would have to accurately understand how their candidates are expected to perform in an election held in the future. Most voters 71  would likely have a difficult time recalling how well their preferred party did months after an election; knowing how well their candidates will do months before an election is thus expected to be a difficult task.  Partisan*Knowledge and Wasted*Knowledge: The interactive effects seen in tables 5.8 and 5.9 are literally, as the label implies, the product the variable measuring 'partisan' or 'wasted votes' and knowledge for each respondent. The interaction effects enable analysis of the effect of either 'partisan' or 'wasted vote' variables at different levels of information. This facilitates the hypothesis that such interests should have unique effects for voters with different levels of information, and prevents having to run separate models for each information class.  72  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0092711/manifest

Comment

Related Items