UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Examining the effect of partisanship on the 2005 and 2009 BC-STV referenda Wong, Danica Michelle Waih-Mahn 2010

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

Item Metadata

Download

Media
24-UBC_2010_fall_wong_danica.PDF [ 1.76MB ]
Metadata
JSON: 24-1.0071436.json
JSON-LD: 24-1.0071436-ld.json
RDF/XML (Pretty): 24-1.0071436-rdf.xml
RDF/JSON: 24-1.0071436-rdf.json
Turtle: 24-1.0071436-turtle.txt
N-Triples: 24-1.0071436-rdf-ntriples.txt
Original Record: 24-1.0071436-source.json
Full Text
24-1.0071436-fulltext.txt
Citation
24-1.0071436.ris

Full Text

EXAMINING THE EFFECT OF PARTISANSHIP ON THE 2005 AND 2009 BC-STy REFERENDA  by Danica Michelle Waih-Mahn Wong B.A. (Hons), The University of British Columbia, 2007  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS in The Faculty of Graduate Studies (Political Science) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2010  © Danica Michelle Waih-Mabn Wong, 2010  ABSTRACT In the past decade, British Columbians were twice given the option of replacing the province’s first-past-the-post method of electing Members of the Legislative Assembly with the single-transferable vote system designed by the Citizens’ Assembly. The central inquiry of this paper is: Did partisanship affect British Columbians’ intentions to vote for or against the 2005 and 2009 BC-STy referenda? Employing the 2005 and 2009 B.C. Referendum Studies, this paper investigates whether partisanship, combined knowledge about the Citizens’ Assembly and the BC STV proposal itself, and/or predisposition towards electoral system reform influenced individuals’ referendum vote intentions. This paper also seeks to identify the specific mechanisms via which party interests came to be, and subsequently operated, as a key factor in partisans’ decision-making processes. Regression analysis reveals that in both 2005 and 2009, individuals who planned to vote for different parties in the general election exhibited noticeably different probabilities of supporting BC-STy. On average, supporters of the Green Party of B.C. were the most likely to vote in favour of BC-STy, followed by supporters the New Democratic Party of BC voters and then supporters of the B.C. Liberal Party. Partisanship, knowledge and predisposition all played a significant role in bringing about these different vote intentions. Partisanship affected voters increasingly over time via two processes. When party leader cues were given, partisans responded accordingly. Absent such cues, however, voters were nevertheless able to determine on their own, likely through mental calculation, how best to vote given their party preferences. Both the progression of the referendum campaigns and increased knowledge helped individuals become more accurate and more confident at matching their partisan interests to their vote intention. Predisposition affected voters steadily over time, though different levels of predisposition had a greater effect on BC-STy vote intentions in the two referendum years.  11  TABLE OF CONTENTS Abstract  .  ii  Table of Contents  iii  List of Tables  iv  List of Figures  v  Introduction  1  Hypotheses  3  How Partisanship Likely Existed at the Time of the Referenda  3  How Partisanship May Have Influenced BC-STy Vote Intentions  4  How Partisanship May Have Affected the Two Separate Referenda  7  Measuring BC-STy Vote Intention  10  Step 1: Examining the Effect of Partisanship Alone  12  Step 2: Investigating Other Possible Influences  17  2-A: Controlling for Knowledge  17  2-B: Interacting Partisanship and Knowledge  22  2-C: Interacting Partisanship and Predisposition  27  Step 3: Examining Partisanship and Predisposition Over Time  35  3-A: Interacting Partisanship and Survey Day  36  3-B: Interacting Predisposition and Survey Day  39  3-C: Interacting Partisanship and Survey Day, Controlling for Predisposition  43  Conclusion  50  References  52  Appendices  54  Appendix A Appendix B Appendix C Appendix D  —  —  —  —  Survey Information  55  Survey Questions Used  56  Logit Regression Tables Analyzing 2005 Survey Data  57  Logit Regression Tables Analyzing 2009 Survey Data  64  111  LIST OF TABLES Table 1  Breakdown of BC-STV Vote Intentions  10  Table 2  Breakdown of General Election Vote Intentions  13  Table 3  The Effect of Partisanship Alone  14  Table 4  Breakdown of Combined Knowledge  18  Table 5  Partisan Means of Combined Knowledge  18  Table 6a  The Effect of Partisanship with Knowledge Controlled at .5  19  Table 6b  The Effect of Partisanship with Knowledge Controlled at 1  20  Table 7  The Effect of Interacted Partisanship and Knowledge  24  Table 8  Partisan Means of Predisposition  28  Table 9  The Effect of Interacting Partisanship and Predisposition  30  iv  LIST OF FIGURES Figure 1 a Partisanship Interacted with Knowledge in 2005  23  Figure lb Partisanship Interacted with Knowledge in 2009  24  Figure 2a Partisanship Interacted with Predisposition in 2005  29  Figure 2b Partisanship Interacted with Predisposition in 2009  30  Figure 3a Partisanship Interacted with Predisposition in 2005 (enlarged)  31  Figure 3b Partisanship Interacted with Predisposition in 2009 (enlarged)  32  Figure 4a Breakdown of Predisposition in 2005  33  Figure 4b Breakdown of Predisposition in 2009  33  Figure 5a Partisanship Interacted with Survey Day in 2005 (120 days)  36  Figure Sb Partisanship Interacted with Survey Day in 2005 (25 days)  37  Figure Sc Partisanship Interacted with Survey Day in 2009  37  Figure 6a Breakdown of Predisposition Categories in 2005  40  Figure 6b Breakdown of Predisposition Categories in 2009  40  Figure 7a Predisposition Categories Interacted with Survey Day  41  in 2005 (120 days) Figure 7b Predisposition Categories Interacted with Survey Day in 2005 (25 days) ....41 Figure 7c Predisposition Categories Interacted with Survey Day in 2009  42  Figure 8a Partisan Category Predisposition Means Over Time in 2005 (120 days)  43  Figure 8b Partisan Category Predisposition Means Over Time in 2005 (25 days)  44  Figure 8c Partisan Category Predisposition Means Over Time in 2009  44  Figure 9a Partisanship Interacted with Survey Day with Predisposition Controlled in 2005 (120 days)  45  Figure 9b Partisanship Interacted with Survey Day with Predisposition Controlled in 2005 (25 days)  46  Figure 9c Partisanship Interacted with Survey Day with Predisposition Controlled in 2009  46  v  INTRODUCTION In the past decade, British Columbians were twice granted the rare opportunity to change the way they elect Members of the Legislative Assembly. These opportunities occurred in the form of special referenda, run in conjunction with the 2005 and 2009 provincial elections.  Voters were given the choice between adopting the single-  transferable vote (STV) method suggested by the province’s Citizens’ Assembly, or retaining the existing first-past-the-post (FPTP) system. Although both ballots ultimately failed, the processes have drawn significant attention among academics and pundits. A great deal of research has been done on the surprising near-success of the 2005 round and on the unique features of the Citizens’ Assembly, for example.  However, despite  journalist-reported evidence that supporters of different parties exhibited different levels of support for BC-STy, not much scholarly attention has been paid to the role partisanship may have played in either referendum.’  The following paper seeks to  remedy this oversight. The primary inquiry throughout this paper is therefore: Did partisanship affect British Columbians’ intentions to vote for or against the 2005 and 2009 BC-STy referenda? To answer this question, I will analyze survey data gathered by the 2005 and 2009 B.C. Referendum Studies, employing summary statistics and using logit 2 regression. If partisanship did affect British Columbians’ BC-STy vote intentions, we should find evidence of significant differences between the average yes-vote probabilities exhibited by supporters of the various parties. Simply finding such differences, however, is not conclusive evidence that partisanship did in fact affect British Columbians’ vote intentions. Such variations in support may merely have been a byproduct of some other, Neal Hall, “New voting system headed for referendum loss,” The Vancouver Sun, May 9, 2009. Data from the B.C. Referendum Study 2005 and B.C. Referendum Study 2009 were provided by the Institute for Social Research, York University. The 2005 BC Referendum Study survey was funded by the Social Sciences and Humanities Research Council of Canada (SSHRC), the Centre for the Study of Democratic Institutions, the Canada Research Chair in Electoral Democracy, and the Fonds and was completed for Fred Cutler (University of British Columbia), André Blais (Université de Montreal), Patrick Fournier (Université de Montreal), R. K. Carty (University of British Columbia) and Richard Johnston (University of British Columbia). These individuals are not responsible for the analyses and interpretations presented here. Additional information about the studies and dataset questions can be found in Appendix A and Appendix B. 2  I  more important and underlying, influence(s) at work. To complete our analysis, we must therefore also investigate the more obvious of these possible alternative influences. These are namely individuals’ levels of combined knowledge about the Citizens’ Assembly and the BC-STy proposal itself, and voters’ intrinsic predispositions towards electoral system reform. It is only after conducting these additional queries that we may approach some understanding of whether, and how, partisanship played a role in the outcomes of the 2005 and 2009 BC-STV referenda.  2  HYPOTHESES To investigate the influence of partisanship on the 2005 and 2009 BC-STy referenda, we must begin by outlining several of the theoretical presumptions that underpin the paper’s main inquiry. First, we must establish why we should expect British Columbians to have exhibited strong partisan ties during the time of the referenda. Next, we must explain why we ought to suppose that these party attachments had any effect on citizens’ BC-STy vote intentions. Finally, given that two separate referenda took place, we must spell out a few separate expectations about the effect of partisanship for each year.  HOW PARTISANSHIP LIKELY EXISTED AT THE TIME OF THE REFERENDA The notion of partisanship has been central to the study of political behaviour for many decades.  First discussed in relation to voting behaviour in American federal  politics, the idea that individuals form durable attachments to political parties has since been applied in other countries, including Canada.  The debate over whether or not  Canadian voters exhibit strong ties to particular political parties has been intense. Many scholars argue that Canadians exhibit only weak or merely provisional ties to the national political parties. 3 Particularly in recent years however, the argument that partisanship indeed influences Canadian politics has been increasingly vindicated. In a study using Canadian Election Survey data, for example, Belanger and Stephenson found that even Canadians who self-identify as having only weak ties to a party were nevertheless highly likely to support that party in elections. 4 They also discovered that since 2004, some of the federal parties have started to develop stronger and more durable attachments with 5 partisans. E.g., Harold D. Clarke, Allan Kornberg and Thomas J. Scotto, Making Political Choices: Canada and the United States (Toronto: University of Toronto Press, 2009). Eric Belanger and Laura B. Stephenson, “Parties and Partisans: The Influence of Ideology and Brokerage on the Durability of Partisanship in Canada,” in Voting Behaviour in Canada, eds. Cameron D. Anderson and Laura B. Stephenson, 107-138 (Vancouver: University of British Columbia Press, 2010). Ibid.  ‘  3  HOW PARTISANSHIP MAY HAVE INFLUENCED BC-STy VOTE INTENTIONS  Although scholars dispute the exact maimer via which individuals form partisan ties, “all agree that party loyalties condition political opinions and behaviour [sic].” 6 In particular, voters often rely on their partisan attachments to perform decision-making short cuts.  Such “cognitive heuristic” processes occur when people have scarce  information and little incentive or desire to search out the additional data necessary for performing a more complex, rational decision-making procedure. 7 In such situations, parties will often emit cues, usually through their leaders via the media, regarding how they think voters should react to new and evolving political situations. Individuals who feel ties towards a particular party may pick up on, or even search out, and then follow that party’s cues. 8 The most often-cited example of such behaviour occurs during elections, when partisans rely on information about candidates’ party affiliations when they cast their ballot to elect local representativ 9 es.  However, research has shown that voters also  search out and follow partisan cues when deciding how they ought to feel about policy questions, especially when these issues are novel.’ 0 Furthermore, empirical testing has proven that voters are particularly likely to employ heuristic processes when faced with complex choice situations.” As few political situations are more novel, multi-faceted or complex than a citizen referendum, it therefore stands to reason that individuals might use their partisan ties to determine how to respond such proposals. The results of Borges 6  Jennifer Merolla, Laura Stephenson and Elizabeth Zechmeister, “Have Cue, Will Travel? Political Parties as Heuristics in Three Countries” Paper presented at the annual meeting ofthe American Political Science Association, Marriott Wardman Park, Omni Shoreham, Washington Hilton, Washington, DC, September 1, 2005. http://www.allacademic.com/metaJp4 1490 index.html/. E.g. Richard R. Lau and David P. Redlawsk, “Advantages and Disadvantages of Cognitive Heuristics in Political Decision Making,” American Journal ofPolitical Science 45 (October 2001): 953-971. http://web.ebscohost.com]. 8 E.g. Peverill Squire and Smith, Eric R. A. N, “The effect of partisan information on voters in nonpartisan elections,” Journal ofPolitics 50, no. 1 (February 1988):169—79. http://web.ebscohost.com!. Lau and Redlawsk, 953. ‘° E.g. Cindy D. Kam, “Wbo Toes the Party Line? Cues, Values, and Individual Differences,” Political Behavior 27, no. 2 (June 2005): 163-182. http://web.ebscohost.coml. ‘ Lau and Redlawsk, 959-961. 4  and Clarke’s 2008 internet survey experiment seem to support such reasoning. These authors found that during polity shaping referenda at least, “leader and partisan cues typically are available, and voters’ use of them can help explain public opinion 2 dynamics.” Despite this abundance of literature on how partisanship often influences voters via party cues and on how this can occur during referenda, we cannot invoke the usual “party cue followed by partisan response” model in the context of the 2005 and 2009 BC STV referenda. Neither in 2005, nor in 2009 did the provincial parties offer their voters cues on how they ought to react to the Citizens’ Assembly’s proposal. The two dominant parties, the B.C. Liberal Party and the New Democratic Party of B.C., never publicly stated whether they were for or against BC-STy.  In huge contrast to their normal  behaviour, these parties with seats in the legislature remained silent, at least on the referendum issue, throughout both campaign 3 periods.’ Only fairly late during the 2009 provincial election campaign did any party leader, Jane Sherk of Green Party of B.C., publicly endorse the yes-vote in a maimer that was fairly prominent in the media.’ 4 Given this general lack of direction from party leaders, if partisanship had any effect on the BC-STy vote, this influence would have had to occur more subtly and on a more individual basis. Partisanship may have iniluenced voters in this fashion once people came to realize that not all the parties stood to gain, nor equally gain, legislature seats or political influence over public policy should BC-STy be adopted. Indeed, the more voters learned about the referendum, the more they should have become aware that some parties stood to benefit greatly by switching to BC-STy, while others were much more likely to maintain or increase their success by perpetuating first-past-the-post. Although the consequences of moving from a first-past-the-post to a single transferable vote method of counting votes are fundamentally uncertain, there are several generally held predictions. For instance, it is commonly thought that STV treats parties whose support is geographically dispersed, parties representing “fringe issues,” and small 12  Walter Borges and Harold D Clarke, “Cues in Context: Analyzing the Heuristics of Referendum Voting with an Internet Survey Experiment,” Journal ofElections, Public Opinion & Parties 18, no. 4 (November 2008): 433-448. http://www.informaworId.com!. 13 Ostensibly the parties remained silent in order to allow individuals to make up their own minds about the proposal. Certain commentators, however, argue that this silence was actually part of an organized strategy to minimize public awareness of the referenda, and thus maintain the status quo. 14 Hall.  5  parties more favourably than does FPTP. On the flip side, geographically concentrated parties and parties that have previously formed governments thanks to the artificial majorities permitted by FPTP often see their electoral prospects reduced under STy. Given these predictions, individuals’ party preferences may have influenced their BC-STy vote intentions via a mental utility maximization process, through which individuals attempted to maximize the gain from their votes in future general elections. In plainer English, if they cared at all about the future electoral prospects of their party, partisans may have attempted to determine which electoral system stood a better chance of benefiting their party. Having or acquiring at least some basic knowledge about the predicted consequences of BC-STy and the nature of their preferred party would obviously have been a prerequisite to individuals performing this process.’ 5  Then  gathering what they knew about their party on the one hand and what they learned about the consequences of BC-STy on the other, these individuals could have mentally weighted the predicted consequences of BC-STy andlor of first-past-the-post for their party. The more knowledgeable the individual, the more accurately he/she ought to have performed this mental calculation and the more confident he/she should have been about their consequent vote intention. After determining which of the two systems would better serve their party, it would be a simple matter for individuals to then link that choice to the correct “yes” or “no” response given the syntax of the referendum question. If partisanship did, in fact, influence BC-STy vote intentions in the manner just hypothesized, we can also propose the following, albeit basic, party-specific predictions regarding how the different supporters ought to have behaved:  People who preferred the Green Party of B.C. should have been extremely favourable towards BC-STy. The Green Party of B.C. was, and continues to be, a smaller party whose supporters are geographically dispersed. In the 2001 and 2005 elections, the party won approximately 10% of the popular vote, but was awarded no seats in the legislature under first-past-the-post.  15  Given this intrinsic link between the hypothesis and the effect of knowledge, this paper will have to investigate both the possibility that vote intentions were entirely based on knowledge and the possibility that knowledge worked hand in hand with partisanship in a manner that supports the hypothesis.  6  •  Supporters of the B.C. Liberal Party ought to have been very unlikely to support the referenda. The B.C. Liberal Party was the dominant provincial party in 2001 and 2005, due in large part to the artificial majority granted it in 2001 by firstpast-the-post counting.  •  More difficult to pin down, supporters of the New Democratic Party of B.C. should most likely have demonstrated a level of support for BC-STy between that of the Greens and Liberals. The second dominant party in the province, New Democratic Party of B.C. formed government in the 1990s. It won by an artificial majority in 1996, but was then promptly and soundly defeated by a Liberal artificial majority in 2001. Thus, though the party had known both success and failures under the existing electoral system, its resounding defeat in 2001 may have made some New Democrats more willing to adopt a system based on STy. To recap, even though the parties remained largely silent throughout the  campaigns, partisanship may nevertheless have affected British Columbians’ BC-STy vote intentions through an internal thought process aimed at maximizing the future electoral success of their preferred parties.  As individuals’ knowledge about the  consequences of BC-STy increased, supporters of different parties ought to have come to increasingly divergent and confident decisions on how to cast their ballots.  HOW PARTISANSHIP MAY HAVE AFFECTED THE TWO SEPARATE REFERENDA Since the BC-STy referendum took place in both 2005 and 2009, and since survey data was collected in both campaign periods, we have the opportunity to analyze two full sets of respondent data related to BC-STy vote intentions. That said, until we conclude our analysis, it is unclear whether having two sets of data to analyze will increase the certainty with which we are able to draw conclusions, or whether studying the two referenda simultaneously will simply raise additional questions that we are unable to answer. To be sure, we cannot treat the 2005 and 2009 referenda as separate,  7  isolated events as the 2009 referendum only occurred due to the near-success of the 2005 vote. Nor, since so many years passed between the two ballots, can we assume that the two events featured identical choices for citizens as feelings towards BC-STy in 2009 probably picked up where the 2005 ballots left off. Rather, we must develop both joint and year-specific hypotheses regarding the expected effect of partisanship on BC-STV vote intentions. If partisanship affected vote intentions through the individual thought process just outlined, it makes sense to hypothesize that partisanship mattered in both 2005 and 2009. If anything, the influence of partisanship via this process ought to have been equal, if not greater, in 2009 because those who learned to perform the mental calculation in 2005 should have easily recalled or repeated the process in 2009. Looking to the unique features and circumstances of the two campaigns brings to light other interesting notions regarding when partisanship ought to have been more influential. The hypotheses based on these factors are somewhat contradictory, however. Partisanship may have mattered more in 2005 because voters’ memories of the “unfair” artificial majority results of the 2001 election were quite fresh.’ 6  New  Democrats, who were so blatantly disadvantaged by first-past-the-post in that election, ought to have reacted by strongly supporting BC-STy in 2005, and Liberals vice versa. This knee-jerk reaction may have subsided by 2009, however, after the 2005 first-pastthe-post general election produced a more equitable division of seats vis-à-vis the popular 17 vote. If voters, in particular Liberals, recalled all the way back to the 1996 election, however, the story may have been more complicated. 8’ With longer memories, Liberals and New Democrats may have acted similarly to one another in 2005. They would have done so not because partisanship did not matter to them, but rather because they saw their long term partisan interests as dominant parties under first-past-the-post as more similar than dissimilar. 16  In 2001, the B.C. Liberal Party won 77 of the 79 seats in the Legislature with only 56.72% of the popular vote. The New Democratic Party of B.C. won only 2 seats, despite securing one-fifth of the popular vote 17 In 2005, the B.C. Liberal Party won 46 of the 79 seats in the legislature with 45.80% of the popular vote, while the New Democratic Party of B.C. won 33 seats with 41.52% of the popular vote. 18 In 1996, the New Democratic Party of B.C. won a 6-seat artificial majority over the B.C. Liberal Party, despite the fact the parties’ shares of the popular vote were 39.45% and 4 1.82% respectively.  8  Alternatively, the effect of partisanship may have been greater in 2009 for at least  two separate reasons.  First, in 2009 the provincial government provided a total of  $500,000 to fund recognized yes- and no-vote campaigns run by citizen proponent groups. While advocating their positions, these groups raised public awareness about the consequences of moving from a first-past-the-post system to the BC-STy method of voting. This more prominent availability of information in 2009 may have enabled more voters to make the partisan-based mental calculation to determine their vote intention, spreading the effect of partisanship more widely throughout the population than in 2005. Also in 2009, the provincial government determined and publicized how the electoral ridings would change under BC-Sly. Given the nature of single-transferable vote, the projected constituencies were much larger than the existing ridings. In some cases, this led to the amalgamation of areas that previously supported different parties. People residing in these areas, particularly if they did not understand how votes are counted under BC-STy, may have felt that their ability to elect their usual party might be threatened by this amalgamation, inducing them to vote no. To summarize, if partisanship indeed affected British Columbians’ BC-STy vote intentions through an internal calculation process aimed at maximizing the value of their future general election votes, partisanship should have mattered during both the 2005 and the 2009 referenda. The influence of partisanship may have had a greater effect in one of these years, but it is difficult to predict which year that might be.  9  VOTING FOR BC-STV The dependent variable throughout this paper is “vote.” Coded from 0 to 1, “vote” measures the probability of a respondent voting in favour of either the 2005 or the 2009 BC-STy referendum.’ 9 Respondents were read the year’s respective referendum question, then asked: “Do you think you will vote yes or no?” in 2005 or “Do you think you will vote for the existing system or for the proposed BC-STy system?” in 2009. Only respondents who said they were at least somewhat likely to vote in the referendum were asked about their BC-STy vote intention. Respondents who replied “don’t know” or “refused” are not included in the analysis. In 2005 and 2009, respondents’ BC-STy vote intentions were as follows: Table I  —  Breakdown of BC-SW Vote Intentions  BC-SW Vote Intention Yes—In Favour No—Against Total  2005 #of %Of Respondents Respondents 911 65.3 484 34.7 1,395 100.0  2009 #of Respondents 365 348  %of Respondents 51.2 48.8  713  100.0  Both the 2005 and the 2009 survey over-predicted the success of the BC-STy proposal, which failed in both years. The 2005 survey nevertheless managed to correctly predict that yeses would outnumber nos.  However, the 2009 survey predicted that  popular support would come in at just over fifty percent, when in fact the final tally was only 3 8.8%. This disparity between the survey statistics and the actual referenda results, especially in 2009, is likely the result of several factors, including survey features, such as question design ° and sample size, 2 21 and changing Elections BC ballot rules. 22 19  Given the dummy, probabilistic nature of the “vote” dependent variable, the method of analysis throughout this paper is logit regression. All the regressions, tables and line graphs in this paper were calculated and/or drawn using Stata 11. The graphs in Steps 1 and 2 were drawn using the “line” method. The graphs in Steps 3-A and 3-2 use the “lowess” method. The graphs in Step 3-C use the “lflt” method. The comparative probability figures were all calculated using the Clarify first differences method. The bar graphs were created using Microsoft Excel 2000. 20 As has already been mentioned, only respondents who identified themselves as being at least somewhat likely to vote in the referendum were asked about their BC-STy vote intention. Thus people who identified themselves as being “slightly unlikely” and “very unlikely” to vote in the referendum were not  10  asked about their BC-STy vote intention. Assuming that the “slightly unlikely” respondents were less politically engaged and knowledgeable, and thus on average less likely to support the BC-STy proposal, it is possible that including those respondents in the BC-STy vote intention variable would have produced predictions closer to the actual results. 21 The 2005 survey interviewed a total of 2643 respondents, while the 2009 survey interviewed less than half that number. Not surprisingly, therefore, there is BC-STV yes-vote intention information for twice as many respondents in 2005 as compared to 2009. As is always the case in statistics, smaller sample size makes population inferences less statistically certain. 22 In 2005, voters were permitted to decline the referendum ballot and cast only a general election ballot. For the 2009 referendum, the rules were changed such that all voters had to accept and cast both the ballots. It was therefore easy in 2005 for voters who did not know or care much about BC-STy (identified in the surveys as “somewhat unlikely” and “very unlikely” to vote in the referendum) to not take part in the BC STV voting. In 2009, however, it is likely that a substantial portion of these same voters ended up voting “no,” not realizing that they had the option of simply turning in a blank or spoiled referendum ballot instead.  11  STEP 1: EXAMINING THE EFFECT OF PARTISANSHIP ALONE As previously noted, if partisanship did in fact affect British Columbians’ vote intentions towards BC-STy, we should find significant differences between the average yes-vote probabilities exhibited by supporters of the various parties.  The first step  towards determining whether partisanship influenced respondents’ votes is therefore to regress BC-STy vote intention by partisanship alone. The key independent variables throughout this paper are a set of dummy variables that represent vote intention in the 2005 and 2009 British Columbia general elections. Respondents were asked: “Will you vote for the Liberal Party, the NDP, the Green Party, or another party?” Binary variables were created for the following response options: Green Party of B.C. (Green), New Democratic Party of B.C. (NDP), B.C. Liberal Party (Liberal), Conservative Party of B.C. (Conservative) and “don’t know” (Undecided), though the Conservative response option was only offered in 2005.23 Only respondents who said they were at least somewhat likely to vote in the general election were asked their party vote intention. People who responded “other,” “refused” or “will not vote” are not included in the analysis. 24 Unfortunately, vote intention is not a perfect measure of partisanship. Not every person who votes for a candidate in the general election has firm ties to that candidate’s party. This is especially the case if a person votes strategically because his/her preferred party has a very low probability of winning under the first-past-the-post system. Nevertheless, vote intention serves as an adequate proxy for partisanship, particularly in regards to this paper’s survey data. Both the 2005 and the 2009 survey were conducted during the weeks preceding voting day.  Since respondents had the  option of stating that they were “undecided,” those who chose to voice a party preference —  especially if they were interviewed early in the survey periods  —  most likely indeed felt  23  As the Conservative response option was not offered in 2009 and since so few respondents self-identified into this category in 2005, it will be necessary to oftentimes ignore this category during the analyses that follow. 24 The Undecided category in both years thus encompasses respondents who planned to cast a ballot, but had not yet firmly decided on their party preference and/or voting strategy at the time of their interview.  12  loyalty to the party they named. Thus, more so than if the data had stemmed from a post election survey, we can classif’ Undecideds as non-partisans on the one hand, and those who voiced a party preference as partisans on the other.  Analysis also shows that strategic voting was not widespread in either 2005 or 2009.25 So the vast majority of  individuals who planned to vote for a particular party, planned to do so because they preferred that party above all the others. 26 In 2005 and 2009, vote intentions for upcoming general election were as follows: Table 2— Breakdown of General Election Vote Intentions General Election Vote Intention Undecided Green NDP Liberal Conservative Total  2005 #of Respondents 895 140 510 706 32 2,283  %of Respondents 39.2 6.1 22.3 30.9 1.4 100.0  2009 #of %of Respondents Respondents 274 32.2 66 7.7 225 26.4 287 33.7  n/a  n/a  852  100.0  Over 60% of respondents in both 2005 and 2009 demonstrated a preference towards one of the political parties. 27 This ratio bodes well for our ability to analyze whether or not partisanship affected BC-STy vote intentions.  The figure at least  confirms that a large number of voters in both years harboured the partisan attachment necessary for basing a party-influenced BC-STy decision  —  if they so chose.  Ignoring the Undecideds, the survey statistics for party vote intention were generally consistent with the actual popular vote results. In both 2005 and 2009, the Liberal Party and NDP won the most ballots by far, with support for other parties 25  Although the 2005 and 2009 elections occurred under the first-past-the-post system, the survey data suggests (analysis not shown) that strategic voting was not a significant phenomenon in either election. If strategic voting did take place, it should be most apparent within the NDP category, as Greens sometimes vote for the NDP as their candidates are more likely to be elected than Green candidates under first-pastthe-post. However, in 2005 and 2009 the number of respondents who voted for the NDP while actually preferring the Green party was only 55 and 28 respectively. 26 In 2005 on a 100-point scale, Liberal voters preferred their party to the NDP by 42 points on average. NDP voters preferred their party to the Liberal party by an average of 41 points. In 2009, the preference differences were 46 and 35 points respectively. 27 In fact, if the survey period in 2005 had been shorter, it is possible that the 2005 ratio of partisans to nonpartisans might have been even greater. The survey period in 2005 spanned from the beginning of January until voting day, and approximately one-third of those who identified themselves as Undecided did so in January.  13  remaining below 10% of the popular vote. As predicted by both surveys, the Liberal Party of B.C. won the popular vote in both years. However, the final vote-share gap between the Liberals and New Democrats was not as wide as predicted by the surveys. 28 The regression model for calculating the effect of partisanship on BC-SW vote intention is as follows:  BC-SW vote  =  f3  +  (13 1 undecidedvote)  (13 5 conservativevote)  +  +  (3 2 greenvote)  +  (f3 3 ndpvote)  +  (13 4 liberalvote)  +  u  Table 3 summarizes the average differences in BC-STy yes-vote probability between the different partisan categories. 29  Table 3  —  The Effect of Partisanship Alone Voting In Favour of BC-SW  Undecided Green NOP Liberal Conservative Comparison Category s Baseline Probability  base group .005 (0.052) -.543 (0.035) .158* (0.034) -.174 (0.110)  (0.112)  .723  .727  .563  N 1227 Pseudo-R 1 .0169 (Levels of Significance: *<5%)  ]  2009  2005 -.005 (0.053) base group -.059 (0.054) .165* (0.054) -.176 (0.117)  F  f  .159* (0.034) .164* (0.055) .104* (0.035) base group  -.012  base group .151 (0.077) .090 (0.053) .256* (0.050)  -14.9 (0.078) base group -.062 (0.078) 408* (0.073)  (0.049) base group  n/a  n/a  n/a  .545  .697  .288  .261 * (0.048) .409* (0.072) 347*  1227  1227  604  604  604  .0169  .0169  .0748  .0748  .0748  Reading down any of the vertical columns in Table 3, it is clear that respondents who planned to vote for different political parties in the general election did, in fact, demonstrate different levels of support for the BC-STy referendum. This suggests that 28  This suggests that a large portion of the Undecided respondents in both 2005 and 2009 ended up voting for NDP candidates. However, the survey data does not include the information necessary to test this hypothesis. 29 The information in Table 3 should be read vertically, as each column displays the probability differences of a logit regression that drops a different comparison category. Thus, the “Comparison Category’s Baseline Probability” line toward the bottom of the table gives the baseline yes-vote probability of supporting BC-STy for a different partisan category in each column.  14  partisan ties may indeed have affected how people voted in both the 2005 and 2009 BC STV referenda. 2009.  That said, partisanship had a greater influence on vote intentions in  In 2005, the party categories’ yes-vote probabilities ranged within 17.5  percentage-points of each other, beginning at roughly 55% yes-vote likelihood. In 2009, the range of partisan differences was more than double the 2005 figure. There was a 40 percentage-point gap between the most and least supportive partisan categories, beginning at roughly 30% yes-vote likelihood. 30 Interestingly, in both 2005 and 2009, the pattern of partisan yes-vote probabilities shadowed somewhat the parties’ placements along the left-right political spectrum. More importantly, however, the resulting pattern echoed the three party-specific hypotheses. Support for BC-STV fell moving from the far-left Green supporters, to the middle-left NDP voters, to the middle-right Liberal supporters, to the far-right Conservative voters. Thus in 2005, for example, the average Green voter demonstrated the highest probability of voting in favour of the referendum at 72.7%.  In comparison, NDP, Liberal and  Conservative voters were respectively approximately 6, 16.5 and 17.5 percentage-points less likely to vote yes. That said, the left-right pattern blurs somewhat once we take into account the size and standard errors of some of the likelihood differences in 2005 and 2009.’ Indeed, in both surveys, only the Liberal yes-vote differences were statistically significant at the 5% point. Given these qualifications, we should more cautiously simply conclude that in both 2005 and 2009, there was a clear divide in yes-vote probabilities between two groups of partisan categories. In 2005, Liberals and Conservatives were less likely over 10 percentage-points  —  —  by  to support the referendum than were Greens and New  Democrats. In 2009, Liberals were extremely less likely points  —  —  by at least 35 percentage-  to vote yes than were Greens and New Democrats.  That the range of partisan yes-vote probability differences begin at a much lower percentage in 2009 than in 2005 is very much in keeping with the fmal popular vote results of the two referenda. The same can be said of the fact that all the partisan base group baseline probabilities, some by more than 10 percentagepoints, are lower in 2009 than in 2005. 31 In both years, the 6 percentage-point difference between the Green and NDP categories was not very large given the yes-vote variable’s scale, nor was the difference statistically significant below the 5% level. In 2005, the difference between the Liberal and Conservative categories was minuscule, at only 1.2 percentage-points.  15  Undecideds behaved quite differently in the two referendum years, in keeping with the different outcomes of the two referenda. In 2005, the category demonstrated an average 72.3% likelihood of voting in favour of BC-STy. In 2009, however, average Undecided support dropped to only 54.5%. Thus in 2005, Undecideds were among the most supportive respondents, while in 2009 they were the second least supportive. In short, when vote intention is regressed on partisanship alone, we find ample evidence that partisanship may indeed have influenced how British Columbians voted in the BC-STy referenda. People who supported different parties demonstrated different probabilities of voting yes.  More importantly, the resulting pattern of differences  matches the party-specific hypotheses with support declining from Greens, to New Democrats, to Liberals. The survey data also suggests that partisanship had a greater effect in 2009, since the range of partisan probability differences in the later referendum was more than double the range exhibited in 2005.  16  STEP 2: INVESTIGATING OTHER POSSIBLE INFLUENCES Although the probability differences found in the previous section suggest that partisanship did indeed influence people’s BC-STy vote intentions, we must investigate the possibility that these differences are simply indicating that some other mechanism or influence was at work. The partisan-based yes-vote differences may merely have been the byproduct of some other influence, which was the sole influence on British Columbians’ referendum vote intentions. Alternatively, and more likely, the differences may be the result of partisanship working in concert with other processes.  2-A: CONTROLLING FOR KNOWLEDGE  In “Deliberation, information, and trust: the BC Citizens’ Assembly as agenda setter,” the authors identify a strong link between information and voting for the BC-STy referendum in 2005.32 In short, voters became, on average, more likely to favour the proposal as their knowledge about the Citizens’ Assembly andlor the single-transferable vote alternative increased. It is therefore possible that the partisan yes-vote differences found in Step 1 were merely caused by the different partisans exhibiting different levels of knowledge. If this was the case, including a control variable for knowledge in the regression equation should dissipate, if not erase, the partisan-based yes-vote differences identified in Step 1. To control for effect of respondent knowledge, I created an additive index to amalgamate two survey questions. The first question measured respondent knowledge about the Citizens’ Assembly. In 2005, interviewees were asked: “How much would you say you know about the BC Citizens’ Assembly on Electoral Reform?” In 2009, the equivalent query was: “How much would you say you know about the BC Citizens’ Assembly on Electoral Reform which was in operation during 2004?”  The second  question measured respondent knowledge about the BC-STy proposal itself, by asking: 32  Fred Cutler, Richard Johnston, R. Kenneth Carty, André Blais and Patrick Fournier, “Deliberation, information, and trust: the BC Citizens’ Assembly as agenda setter,” in Designing Deliberative Democracy, eds Mark E. Warren and Hilary Pearse, 166-191 (Cambridge: Cambridge University Press, 2008).  17  “Would you say you know a lot, some, not very much, or nothing about this proposal?” 33 The result is a new variable called “knowledge,” which measures a respondent’s combined knowledge about the Citizens’ Assembly and the BC-STy proposal.  The  variable runs from 0 to 1, moving from no knowledge to a lot of knowledge. In 2005 and 2009, respondents’ combined knowledge levels were as follows: Table 4— Breakdown of Combined Knowledge Level of Combined Knowledge 0 no knowledge .1667 .3333 .5 .6667 .8333 I a lot of knowledge Total —  —  2005 # of Respondents 954 476 441 254 291 56 60 2,532  % of Respondents 37.7 18.8 17.4 10.0 11.5 2.2 2.4 100.0  2009 # of Respondents 286 193 222 126 141 33 28 1,029  % of Respondents 27.8 18.8 21.6 12.2 13.7 3.2 2.7 100.0  In 2005 and 2009, the mean levels of combined knowledge within each partisan category were as follows:  Table 5— Partisan Means of Combined Knowledge Mean Level of Combined Knowledge Undecided Green NDP Liberal Conservative Total  2005 .194 .244 .309 .295 .220 .258  2009 .264 .250 .352 .318  n/a .310  The survey respondents, both as a whole and within their party categories, were very uninformed. Indeed, 38% and 28% of respondents in 2005 and 2009 respectively exhibited no knowledge whatsoever about the Citizens’ Assembly or the BC-STy proposal. In both years, approximately three-quarters of respondents fell below .5 on the The question was prefaced by the following: “The Citizens’ Assembly has proposed a change to the way we elect the B.C. Legislature in Victoria. The system they proposed is called BC-STy.” Thus the words “this proposal” were roughly equivalent to: “the Citizens’ Assembly proposed BC-STV change to the way we elect the B.C. Legislature in Victoria.”  18  knowledge scale. NDP voters exhibited the highest average knowledge. However, none of the categories’ means differed significantly from their year’s overall average. Respondents were slightly more knowledgeable in 2009 than in 2005, both overall and within the partisan categories. However these inter-year differences were negligible given the scale of the “knowledge” variable. The regression model for calculating the effect of partisanship on BC-STV vote intention while controlling for knowledge is as follows: BC-SW vote  =  13o  +  (undecidedvote) 1 13  (conservativevote) 5 3  +  +  (greenvote) 2 3  (knowledge) 6 3  +  +  (ndpvote) 3 f3  +  (liberalvote) 4 13  +  u  Table 6a and 6b summarize the average differences in BC-STy yes-vote probability between the different partisan categories while holding combined knowledge constant at a medium level and the highest possible level respectively. 34 Table Ga  —  The Effect of Partisanship with Knowledge Controlled at .5  Undecided Green NDP Liberal Conservative knowledge  Voting In Favour of BC-STy holding knowledge fixed at 0.5 2005 2009 .169* .274* base base group group (0.032) (0.050) .176* .408* .003 .139 (0.047) (0.048) (0.073) (0.050) 345* .077* .093* .070 (0.034) (0.035) (0.053) (0.047) -.170 * base -.272 * base group (0.031) group (0.049) -.146 .025 nla n/a (0.106) (0.113) 337* .250* .207* .193* (0.036) (0.048) (0.070) (0.069)  Comparison Category’s .780 Probability N 1180 Pseudo-Ri .0449 (Levels of Significance: *=<5%)  .611  .585  1180 .0449  600 1 600 .082J .0828  .314  In both Table 6a and 6b, the “knowledge” coefficient represents the average change in yes-vote likelihood that results from moving from 0 to 1 on that variable or, in other words, from no combined knowledge to a lot of combined knowledge. However, the rest of the coefficients (baseline probabilities and partisan differences) have been calculated holding “knowledge” fixed at the level specified in the table.  19  Table 6b  —  The Effect of Partisanship with Knowledge Controlled at I  Undecided Green NDP Liberal Conservative knowledge  Voting In Favour of BC-SW holding knowledge fixed at I 2005 2009 .114* base base .271 * group group (0.024) (0.049) .116* .381* .002 .115 (0.028) (0.033) (0.061) (0.062) .049* .064* .332* .060 (0.021) (0.026) (0.046) (0.047) 113 * base -.270 * base group (0.024) group (0.048) -.109 .009 n/a n/a (0.081) (0.087) 337* .250* .207* .193* (0.036) (0.048) (0.070) (0.069) -.  Comparison .880 Category’s Probability N 1180 PseudoRz .0449 (Levels of Significance: *<5%)  .767  .685  .419  1180 .0449  600 .0828  600 .0828  As expected given the conclusions reached by the authors of “Deliberation, information, and trust: the BC Citizens’ Assembly as agenda setter,” combined knowledge about the Citizens’ Assembly and the BC-STy proposal did, in fact, affect individuals’ likelihoods of supporting the 2005 and 2009 referenda.  Combined  knowledge was strongly, positively correlated with BC-STV vote intention and this effect was statistically significant at well below the 1% point. 35 knowledge was greater in 2005 than in 2009.  That said, the effect of  In 2005, moving from no combined  knowledge to a lot of combined knowledge made respondents 25 to 34 percentage-points more likely to support BC-STy. 36 In 2009, this move led to only an average 19 to 21 percentage-point increase in support.  Additionally, each of the 2005 baseline  probabilities increased by at least 15 percentage-points when combined knowledge is held fixed at the highest value, whereas the 2009 baseline probabilities only increased by at least 10 percentage-points. Although combined knowledge had a strong positive effect on individuals’ yesvote intentions, including the “knowledge” control variable nevertheless does not erase See the logit regression tables in Appendix C and Appendix D. As can be seen by looking horizontally along the “knowledge” lines of tables 6a and 6b, the exact percentage-point increase depends on the partisan category dropped from the regression equation. 36  20  the partisan-based yes-vote differences in either 2005 or 2009. Rather, the differences identified in Step 1 persisted, though they did so more stubbornly in 2009 when the influence of partisanship was greater.  With knowledge held constant at the halfway  mark, the range yes-vote differences in both years remained almost identical to those found in Step  The 2009 full 40 percentage-point range endured even when  knowledge is held fixed at the highest possible value. 38 The pattern of partisan yes-vote differences identified in Step 1 also remained largely intact in both years, despite the introduction of the “knowledge” control variable. Liberal voters remained much less likely to support BC-STy than supporters of either the NDP or Green parties. In 2009, the pattern of partisan differences remained identical to the findings in Step 1 when knowledge is held fixed at either .5 or i. In 2005, if anything, including combined knowledge into the regression equation clarified the pattern of partisan differences. ° When knowledge is held fixed at the midpoint, for 4 example, the New Democrats’ level of BC-STy support stood firmly between that of the Greens and the Liberals and Conservatives. Thus certain levels of knowledge in 2005 distinguish three, rather than just two, groups of significant differences. In sum, combined knowledge had a strong positive effect on BC-STy vote intention both in 2005 and 2009. Respondents with higher levels of knowledge were more inclined to support the referendum, though this effect was more powerful in 2005. Including knowledge into the regression equation did not, however, erase in either year the effect of partisanship on respondents’ BC-STy yes-vote likelihoods. At the various levels of knowledge, the partisan-based yes-vote differences identified in Step 1 persisted for the most part. Similarly, these differences continued to conform to the party-specific expectations put forth in the Hypotheses section. Given these findings, we must conclude  The ranges were approximately 17 and 40 percentage-points respectively. These began at a higher starting value, however, as discussed in the previous paragraph. 38 At full knowledge in 2005 however, when the effect of knowledge was stronger, the partisan yes-vote differences began to disappear as the range of these differences narrowed to within 10 percentage-points of each other. Thus as before, Liberals in 2009 were extremely less likely by roughly 35 percentage-points to vote yes than were Greens and New Democrats, the latter two of whom were within 6 percentage-points of each other. 40 Including combined knowledge into the regression equation also reverses the Liberal and Conservative categories’ positions in the 2005 difference pattern as can be seen in both tables 6a and 6b. This should be ignored, however, given the magnitude and statistical significance of this difference. —  —  21  that both partisanship and knowledge influenced British Columbians’ vote intentions towards the BC-STy referenda. differed in 2005 and 2009.  However, the relative strength of these influences  In 2005, the effect of knowledge began to mitigate the  influence of partisanship at high levels of combined knowledge. In 2009, the effect of partisanship dominated at all levels of knowledge.  2-B: INTERACTING PARTISANSHIP AND KNOWLEDGE Since introducing a control variable for the effect of combined knowledge did not erase the partisan yes-vote differences found in Step 1, we must also consider the possibility that the relationship between knowledge, partisanship and vote intention was far more complex than suggested by the regression equation tested in Step 2-A. Instead of having a separate linear effect on each partisan category, for example, knowledge may have affected each category differently. If knowledge did indeed affect the various partisan categories differently, interacting knowledge and partisanship will give us a better idea of the extent to which the effect of partisanship truly stemmed from individuals calculating the advantages and disadvantages of BC-STV in relation to their preferred party. If such mental calculations were indeed the primary mechanism via which partisanship came to matter among voters, we should find that moving towards higher levels of knowledge, supporters of the different parties came to progressively more divergent decisions on how to cast their ballots in keeping with the party-specific hypotheses. ’ 4  Thus whereas the regression  equations in Steps 1 and 2-A were merely designed to discover whether or not yes-vote differences existed between different partisans, Step 2-B should also uncover whether and how acquiring knowledge caused these differences to arise.  41  As previously expressed in the Hypotheses section, this expectation is based on knowledge being one of the two prerequisites to partisans being able to mentally tabulate whether their party’s future electoral success would be better served by single-transferable vote or by first-past-the-post. The more knowledgeable a person became regarding the consequences of the different electoral systems, the more accurately he/she should have concluded how to vote in keeping with his/her party preference. Since the parties’ future electoral chances were most likely different should BC-STy be adopted, the supporters of the different parties ought to have come to increasingly divergent conclusions on how to cast their ballots at higher levels of knowledge.  22  To interact the partisanship categories with combined knowledge about the Citizens’ Assembly and the BC-STy proposal, five new variables were created. Each multiplies one of the partisanship categories with the “knowledge” variable. The regression model for calculating the effect on BC-STy vote intention of partisanship interacted with knowledge is as follows: BC-SW vote  =  13o  +  (undecidedvote) 1 13  (conservativevote) 5 f3  +  (greenvoteXknowledge) 8 I3 knowledge)  Figure Ia  —  +  +  (greenvote) 2 f3  (knowledge) 6 f3 +  +  (ndpvote) 3  +  (Iiberalvote) 4 13  (undecidedvoteXknowledge) 7 3  +  3g(ndpvoteXknowledge)  (conservativevoteXknowledge) 1 3 1  +  +  + +  (IiberalvoteX 10 13  u  Partisanship Interacted with Knowledge in 2005  >  0  0 ci)  Cl)  ci)  >-  0 (N  0 .2 .4 .6 .8 1 Level of Combined Knowledge about the Citizens Assembly and BC-SW Undecided NDP Conservative  Green Liberal  23  Figure lb  —  Partisanship Interacted with Knowledge in 2009  >.  .0 0  0  ‘a  cD.  a)  F.  0 .2 .4 .6 .8 1 Level of Combined Knowledge about the Citizens Assembly and BC-SW Undecided NDP  Green Liberal  Table 7 summarizes the average changes in BC-STy yes-vote likelihood within each partisan category from moving between various levels of combined knowledge, with knowledge permitted to affect each category differently. 42 Table 7— The Effect of Interacted Partisanship and Knowledge Voting In Favour of BC-SW moving between various levels of knowledge 2005  2009  0—.5 .250* (0.054) .326* (0.109)  .25-.75 .182* Undecided (0.029) .196* Green (0.060) .204* .176* NDP (0.055) (0.420) .094 .091 * Liberal (0.040) (0.039) -.196 -.191 Conservative (0.216) (0.203) N 1 1,180 1 1,180 Pseudo-R .0545 .0545 (Levels of Significance: *<5%) *  .5-.l .118* (0.012) .100* (0.028) .140* (0.026) .086 (0.035) -.142 (0.148) 1,18q[ .0545 *  F  [  0—.5  .25-.75  .092 (0.070) .496* (0.135) .093 (0.071) (0.056)  .088 (0.067) .277* (0.072) .086 (0.064) .042 (0.060)  .5-i .082 (0.060) .108* (0.046) .077 (0.054) .045 (0.063)  n/a  n/a  n/a  .037  600 1 600 600 jO O .0930 .0930 93  42  Table 7 contains no cross-party comparisons. All BC-STy yes-vote likelihood differences are calculated by comparing same-party supporters at different levels of knowledge.  24  From Figures 1 a and lb it is clear that increased combined knowledge did indeed affect the partisan categories’ vote intentions differently from one another. Furthermore, the variation in these effects occurred in a manner that echoes the expected outcome if partisanship indeed influenced voters through individual mental calculations. Achieving higher levels of knowledge generally increased respondents’ likelihoods of supporting the referendum. 43 These likelihood increases did not always occur with the same strength across the party categories, however. Rather, in both years, higher knowledge induced smaller increases in BC-STy support among Liberals than among Greens, or even among New Democrats. In 2005, moving between knowledge extremes increased the average Liberal’s yes-vote probability by 18 percentage-points, while increasing the average Green’s likelihood by over 42 percentage-p 44 oints. In 2009, the corresponding probability increases from moving between knowledge extremes were even more disparate at 8.2 and 60.4 percentage-points for Liberals and Greens 45 respectively. The different effects of knowledge on partisanship did share one common feature, nevertheless: a diminishing rate of yes-vote probability change. In other words, moving from low- to mid-level knowledge about the Citizens’ Assembly and BC-STy led to a greater increase in support for BC-STy than did moving from mid- to high-level knowledge. NDP voters in 2005, for example, gained 20.4 percentage-points on average when going from 0 to .5 on the knowledge scale, but only 14 percentage-points when moving from .5 to 1. In 2009, the corresponding Green increases in support dropped from a 49.6 to only a 10.8 percentage-point bump. The only exception to this common trend was the Liberal category in 2009. Liberals experienced a slight, but statistically  The sole exception to this was the Conservative category, within which increased knowledge led to decreased support for BC-STy. However, this category can and should be ignored for statistical reasons. When partisanship and knowledge are interacted, the Conservative category is extremely unlikely to represent the true story of Conservative supporters. The 2005 survey did not include any Conservatives with knowledge above .65. Furthermore, the survey likely included several statistical outliers in the Conservative category, as can be seen by the fact that Conservative support for BC-STy was markedly higher than any other partisan group at zero knowledge. The yes-vote probability increases discussed in this paragraph are all calculated using the “0 .5” column and “.5 1” column figures in Table 7. Between these two 2009 extremes, however, the NDP and Undecided categories displayed almost identical slopes. Referring back to the previous paragraph, this means that the increased range of probability differences in 2009 was mostly due to the Greens moving away from the other three parties. In contrast, in 2005, all of the party categories moved away from one another as knowledge increased. —  —  25  insignificant, growth in yes-vote probability increases when moving up the knowledge scale. More noticeable than this common feature, is that the range of partisan yes-vote differences grew with increasing knowledge. Diverse yes-vote likelihoods existed when respondents had no knowledge, though some of the 2005 differences were quite closely grouped.  At full knowledge, however, the average partisan differences were much  46 In 2005, even discounting Conservatives, the range of partisan differences more larger. than tripled, to approximately 30 percentage-points. 47 In 2009, the range doubled, from 30 to just over 60 percentage-points, despite already being extremely large at no 48 knowledge. In short, interacting partisanship and knowledge illustrates that knowledge did indeed affect the partisan categories differently, though this occurred to a greater extent in 2005.  All respondents, with the exception of Conservatives, exhibited a greater  inclination to support BC-STy as their level of combined knowledge rose. However, Greens and New Democrats were more strongly affected by increased knowledge than 49 As a result, voters exhibited different yes-vote probabilities in keeping with Liberals. the party-specific hypotheses after achieving just the barest amount of combined knowledge.  The range of these differences then widened considerably with further  increases in knowledge, though usefulness of additional knowledge tapered off once  46  The obvious exception to this is when two partisan categories’ prediction lines crossed, which caused the  two categories to exhibit the same yes-vote likelihood at a very specific level of knowledge. However this occurred rarely, particularly if one discounts the Conservative line in 2005 since it is based on so few data points. In Figure Ia at no knowledge, discounting the Conservative category, Undecideds were the most likely and Liberals were the least likely to vote yes. Using Clari1j, the range of this difference can be calculated as 6.0 + 1.4 = 7.4, since Undecideds were 6 percentage-points more likely and Liberals were 1.4 percentage-points less likely to support BC-STy than New Democrats, who were the base category in this regression. At a lot of knowledge, Greens were the most likely and Liberals were the least likely to vote yes. Using the same Clarify method, this range can be calculated as 10.7 + 18.2 = 28.9. 48 In Figure lb at no knowledge, New Democrats were the most likely and Liberals were the least likely to vote yes. Using Clarifj, the range of this difference can be calculated as 0 + 30.2 = 30.2, since Liberals were 30.2 percentage-points less likely to vote yes than New Democrats, who were the base category in this regression. Similarly, at a lot of knowledge, Greens were the most likely and Liberals were the least likely to vote yes. Using the same Clarif,’ method, this range can be calculated as 24.0 + 37.8 = 61.8. Thus for Liberals, increased knowledge about the consequences of BC-STV led to a yes rather than a no vote, albeit at a much lower rate of change than for New Democrats or Greens. It seems, therefore, that knowledge had crosscutting effects on Liberals, making them more positive towards BC-STy even though particularly as compared to the results of the 2001 election the Liberal Party stood a fair chance of losing seats should the referendum pass. —  —  26  respondents became fairly knowledgeable.  All together, these findings confirm the  notion that partisanship and knowledge worked hand in hand to influence British Columbians’ BC-STy vote intentions. Moreover, the findings echo what ought to have occurred if partisanship did indeed influence voters the hypothesized individual mental calculations: partisan yes-vote differences existed in keeping with the party-specific hypotheses as in Steps 1 and 2-A, but the range of these differences grew as respondents became more knowledgeable about the referendum.  Step 2-B thus provides the first  evidence suggesting that not only did partisanship affect BC-STy vote intentions, but also that it may have done so via the specific mental process hypothesized in this paper.  2-C: INTERACTING PARTISANSHIP AND PREDISPOSITION  Although we now have a fairly clear picture of how partisanship and knowledge worked together to influence BC-STy voting, we would be negligent if we did not also investigate the possibility that the partisan yes-vote differences were merely the byproduct of respondents’ inherent predispositions towards electoral system reform. Logically, people predisposed against electoral system reform ought to have voted against BC-STy, and vice versa. If the partisan categories harboured different intrinsic feelings about electoral reform, including these predispositions in the regression equation should erase, or at least mitigate, the partisan-based yes-vote differences. To measure respondent predisposition towards electoral system reform, I created an additive index to amalgamate five survey questions.  These questions measured  respondent feeling towards different electoral system options and their associated trade offs. The questions were as follows: •  “In your opinion, which is better: governments made up of 2 or 3 parties because they are forced to compromise, or one party governments so they can get things done?”  •  “Under our present system, a party can win a majority of seats without winning a majority of votes. Do you find this acceptable, unacceptable, or do you not have an opinion on this?”  27  •  “Under the proposed BC-STV system it will be hard to follow how the ballots are counted. Are you very concerned, somewhat concerned, not very concerned, or not at all concerned about this?”  •  “Some people say that with the BC-STy voting system the parties in power will change too often. How concerned are you about this? Are you very concerned, somewhat concerned, not very concerned, or not at all concerned about this?”  •  “Some people like a two-party system because you know who to blame, other people like a system with more than two parties because that way the voter has more choices. Which is more important to you, knowing who to blame or having more choices?”  The result is a new variable called “predisposition,” which measures a respondent’s intrinsic feeling towards electoral system reform. The variable runs from 0 to 1, moving from adamantly anti-reform to extremely pro-reform. In 2005 and 2009, the mean levels of predisposition towards electoral system reform within each partisan category were as follows: Table 8  —  Partisan Means of Predisposition  Undecided Green NDP Liberal Conservative Total  Mean Level of Predisposition Towards Reform 2005 2009 .700 .660 .783 .727 .725 .722 .586 .487 .643 n/a .676 .623  The survey samples from both years were, on average, mildly pro-reform. Liberal voters were always the least predisposed to favour electoral reform, while Green supporters were always the most supportive. None of the parties strayed more than ± .15 from their year’s general mean. Nevertheless, given the scale of the “predisposition” variable, the differences between some of the category means, particularly in 2009, must be considered quite large.  28  Both overall and within the partisan categories, respondents were slightly more favourable towards reform in 2005 than in 2009. However these inter-year differences were very small given the scale of the “predisposition” variable. To interact the partisanship categories with predisposition, five additional variables were created. Each multiplies one of the partisanship dummy variables with the “predisposition” variable. The regression model for calculating the effect on BC-STy vote intention of partisanship interacted with predisposition towards electoral system reform is as follows: BC-STy vote  Figure 2a  —  =  (3 + 13 (greenvote) + 3 2 (ndpvote) + f3 (Hberalvote) + 4 (undecidedvote) + 13 1 (conservativevote) + 6 5 f3 (predisposition) + f3 (undecidedvoteXpredisposition) + 7 (greenvoteXpredisposition) + 3g(ndpvoteXpredisposition) + 13 8 o(IiberalvoteX 1 predisposifion) + 1 + ( 1 conservativevoteXpredisposition) u i 3  Partisanship Interacted with Predisposition in 2005  >‘  co 0  a  0  .2 .4 .6 .8 Level of Predisposition Towards Electoral System Reform Undecided NDP Conservative  1  Green Liberal  29  Figure 2b  —  Partisanship Interacted with Predisposition in 2009  >..  -Q 0a10  >-.  0 0 I  0  I  -  .2 .4 .6 .8 Level of Predisposition Towards Electoral System Reform Undecided NDP  1  Green —— Liberal  Table 9 summarizes the average differences in BC-STy yes-vote probability between the partisan categories at different levels of the “predisposition” variable, with predisposition permitted to affect each category differently. ° 5  Table 9  —  The Effect of Interacting Partisanship and Predisposition Voting In Favour of BC-SW at various levels of predisposition 2005 2009  Undecided Green Liberal Conservative  0 .127 (0.097) .192 (0.245) -.024 (0.048) -.034 (0.127)  Comparison Category’s .074 (NDP’s)_Probability N 864 2 Pseudo-R .1694 (Levels of Significance: *<5%)  1 ]  .5 .137 (0.073) .115 (0.157) .000 (0.066) -.107 (0.156)  1 -.008 (0.039) -.026 (0.063) .025 (0.035) .039 (0.123)  0  .5  1  -.036 (0.065) .011 (0.143) -.064 (0.057)  -.047 (0.102) .036 (0.170) .167* (0.091)  .028 (0.054) .046 (0.074) .046 (0.054)  n/a  n/a  n/a  .452  90.5  .074  .413  .884  864 .1694  864 .1694  447 .3097  447 .3097  447 .3097  1  ]  50  All BC-STy yes-vote likelihood differences in Table 9 are calculated in relation to the NDP category, whose baseline probability is provided in the “Comparison Category’s (NDP’s) Probability” line towards the bottom of the table.  30  As is obvious from Figures 2a and 2b, predisposition towards electoral system reform had a strong, statistically significant, effect on respondent support for BC-STV. As expected, in both 2005 and 2009, people on the anti-reform end of the scale were very likely to vote no, while people on the pro-reform side were extremely likely to vote yes. Also in both years, people who felt neutrally towards reform exhibited middling yes-vote likeithoods. Indeed, interacting predisposition and partisanship in the regression equation erased some of the previously found partisan vote intention differences. In 2005 for example, discounting the Conservative category, the partisan yes-vote probabilities at the extreme pro-reform end fell within just 4 percentage-points of each other. In 2009, the range of differences at the two extremes was only 5 and 6 percentage-points respectively. All together, this suggests that people’s predispositions definitely played a role in determining how individuals voted in the 2005 and 2009 referenda. Despite this strong influence of predisposition however, partisan-based yes-vote probability differences persisted in both 2005 and 2009 within certain segments of the predisposition scale. Figures 3a and 3b offer an enlarged view of these predisposition levels between which partisanship continued to strongly influence BC-STy vote intentions.  Figure 3a  —  Partisanship Interacted with Predisposition in 2005 (enlarged) CD U)  w  0  .2 .3 .4 Level of Predisposition Towards Electoral System Reform Undecided NDP  .5  Green Liberal  31  Figure 3b  —  Partisanship Interacted with Predisposition in 2009 (enlarged)  Co  a)  -D 0  0  a)  .3  .4 .5 .6 Level of Predisposition Towards Eledoral System Reform Undecided NDP  .7  Green Liberal  In 2005, large partisan differences remained, in spite of the strong influence of predisposition, between approximately 0 and .5 on the predisposition scale.  These  differences ranged between 20 percentage-points at the widest gap because Undecideds and Greens remained approximately 25% likely to vote in favour of BC-STy even when they were adamantly anti-reform. The vote intentions exhibited by Liberals and New Democrats, on the other hand, fell much more in line with their predispositions. In 2009, enduring partisan differences occurred around the predisposition mid point, between approximately .3 and .7 on the predisposition scale. Despite the strong effect of predisposition, between those levels of predisposition, partisan yes-vote probabilities ranged within 20 percentage-points.  Three.-quarters of this range was  entirely due to just the Liberals, however, diverging from the expected influence of predisposition.  Neutral Liberals at .5 on the predisposition scale, for example, only  displayed a 25% probability of supporting BC-STV, whereas Undecided, Green and New Democrat neutrals exhibited likelihoods around 40%. Despite the previous two paragraphs, it is difficult to judge the influence of partisanship’s true strength between some levels of predispositions. The amount of data upon which some of the findings discussed in the previous two paragraphs are based on is not always as great as is usually preferred in statistical analysis.  Figures 4a and 4b  32  illustrate how the various partisans were roughly situated along the predisposition scale in 2005 and 2009. The graphs thus illustrate the number of data points upon which the regression graphs in Step 2-C were predicted.  Figure 4a  —  Breakdown of Predisposition in 2005 350 300 250 C  C)  •0 0 0. U)  200 150 100 50 0 Predispositions Undecided (Anti-Reform  Figure 4b  —  =  • Green 0-35  U NDP  Neutral  =  Liberal U Conservative .35-65  Pro-Reform  =  .7-1)  Breakdown of Predisposition in 2009 120 100 U) C C)  C 0 0. U)  80 60  a) 0  40  4l  20 0 Anti-Reform  Neutral  Pro-Reform  Predispositions Undecided (Anti-Reform  =  0-35  • Green Neutral  =  U NDP .35-65  • Libera[] Pro-Reform  =  .7-1)  33  The effect of partisanship in 2005 was found to be strongest among respondents who were anti reform. However, anti-reform respondents comprised less than 10% of the 2005 survey sample. Furthermore, less than a quarter of this 10% were Undecideds and only one of these respondents preferred the Green party. In 2009, partisanship was found to have a strong effect on neutral respondents. This finding, at least, was based on a reasonable amount of data.  Neutral respondents comprised 40% of the 2009 survey  sample, within which “Liberal” was the most populated category. To recapitulate, in both 2005 and 2009, partisan-based yes-vote differences disappeared somewhat when partisanship and predisposition towards electoral system reform are interacted. This leaves little doubt that individuals’ predispositions strongly affected their BC-STy vote intentions. Partisanship remained influential in both years, however, between certain levels of predisposition. Even at these points, however, just one or two parties diverged from the yes-vote probabilities expected under the influence of predisposition.  Furthermore, some of these enduring partisan-differences are not  based on as many data points as is usually preferred in statistical analysis. All together, these findings suggest that when it came to deciding their BC-STy vote, individuals were influenced by both their predispositions and their party preferences, with predispositions being the greater influence for most people. Nevertheless, there was some strategic thinking going on in voters’ minds, over and above the influence of their predispositions.  34  STEP 3: EXAMINING PARTISANSHIP AND PREDISPOSITION OVER TIME Both the 2005 and the 2009 B.C. Referendum Studies conducted respondent interviews over several weeks leading up to voting day.  As a result, we have the  information necessary to investigate how the influence of partisanship and the effect of predisposition towards electoral system reform both evolved over the referendum campaigns. In particular, analyzing these effects over time may help clarify how British Columbians’ BC-STy vote intentions were influenced by both their party preferences and their predispositions towards electoral system reform simultaneously, as the results of Step 2-C suggest. To observe how partisanship and predisposition towards electoral system reform affected BC-STV vote intentions over time, I created a variable that indicates the day of the survey round on which each respondent’s interview took place. In 2005, interviews were conducted on the 120 consecutive days between January 17 and May l7.’  Between 1 and 102 interviews were conducted each day, with an  average of 22 interviews completed per day. In 2009, interviews were only conducted during the month immediately prior to the referendum. On the 26 consecutive days between April 16 and May 11, between 16 and 80 interviews were conducted each day. 52 53 54  51  Within the new “surveyday” variable, January 17 is therefore considered survey day 1 for the 2005 data. Within the new “surveyday” variable, April 16 is considered survey day 1 for the 2009 data. To take advantage of the fact that interviews were conducted over 4 months in 2005, discussion of the 2005 referendum will include details about the entire survey period. To facilitate comparing the 2005 and 2009 periods, two graphs are presented for each of the 2005 regressions. One graph illustrates the entire 120-day survey period, while the other only shows the last 25 days to match the 2009 survey period that lasted for 26 days. To aid the reader, the 2005 graphs showing only the last 25 days of the survey period and the 2009 graphs are formatted to the same y-axis scale. “ Conservative respondents were dropped entirely from the Step 3 analyses. Because so few respondents identified as Conservatives, “surveyday” is an extremely volatile, and thus uninfonnative, variable for that partisan category. 52  35  3-A: INTERACTING PARTISANSHIP AND SURVEY DAY Since we are interested in clarifying the relationship between the effect of partisanship and the influence of predisposition on BC-STy vote intentions, we must begin by establishing how the effect of partisanship alone evolved over the course of the two survey periods.  This can be achieved by simply interacting the partisanship  categories with survey day. To interact partisanship and survey day, I created four additional variables. Each one multiplies one of the partisanship dummy variables with the “surveyday” variable. The regression model for calculating the effect on BC-STy vote intention of partisanship interacted with survey day is as follows: BC-SW vote  +  =  (undecidedvote) 1  (surVeyday) 5 f3  +  —  (greenvote) 2 3  +  (ndpvote) 3  (undecidedvoteXsurveyday) 6 3  (ndpvoteXsurveyday) 8 13  Figure 5a  +  +  +  (Nberavote) 4 j3  +  (greenvoteXsurveyday) 7 f3  +  (IiberalvoteX surveyday) 9 13  +  +  u  Partisanship Interacted with Survey Day in 2005 (120 days)  Cu .0  0• o LO  >co. Co  LO. -  0  20  I  -Th  40  60 Survey Day  100  Undecided  Green  NDP  Liberal  120  36  Figure 5b  —  Partisanship Interacted with Survey Day in 2005 (25 days)  >‘  .0 .0  2co  95  100  105 110 Survey Day Undecded NDP  Figure 5c  —  115  120  Green Liberal  Partisanship Interacted with Survey Day in 2009  15 Survey Day Undecided NDP  20  25  Green Uberal  Figure 5a clearly illustrates that the partisan categories’ average yes-vote probabilities diverged increasingly over the course of the 2005 campaign. When the BC STV referendum was first introduced to the public, Liberals, New Democrats and Greens all exhibited an approximate 60% probability of supporting the proposal. As time passed, however, partisans began to develop different attitudes towards the referendum. Liberal  37  support for BC-STy dropped, while NDP support rose steadily and Green support escalated dramatically. By voting day, and indeed beginning several weeks before that, the partisan differences aligned in accordance with the party-specific expectations as previously found in Steps 1 and 2-A. These growing likelihood differences imply that in 2005, partisanship grew more influential as voting day drew nearer. Moreover, it implies that voters were learning about the consequences of BC-STy, and then figuring out more and more decisively over time how these consequences most likely stood to affect their preferred parties. In 2009, the yes-vote differences exhibited by different partisans were much larger than in 2005, also echoing the previous findings of Steps 1 and 2-A. 55 With the exception of the NDP however, the inter-party ranges of these differences remained largely stable over time. This suggests that the effect of partisanship remained fairly stable in 2009, as compared to in 2005.  Since we only have data for the month  immediately prior to voting day in 2009, it is possible that the 2009 party-based differences were established earlier in that campaign period, picking up from the closing 2005 differences before widening to the levels seen in Figure 5c. Uncovered for the first time by this step’s time-related regression, the most noticeable trend in late 2009 was that all respondents experienced a steady decline in support for BC-STy as voting day drew nearer. As the campaign period drew to a close, something occurred to depress all voters’ support of BC-STy. The exact character and cause of this phenomenon are impossible to determine based on the data available. 56 Since the decline in BC-STy support occurred among all partisans, however, we can rule out the influence of partisanship as the explanatory factor. 57 Rather, it seems that partisan learning was complete by the last month of the 2009 campaign,  This echoes the findings from Steps 1 and 2-A, which suggested that partisanship was more influential during the second referendum. Additionally, throughout the entire period for which we have information, once again the partisan differences aligned in accordance with the left-right spectrum as previously discovered. 56 Nevertheless, a likely explanation is that declining support across the board may have had something to do with public weariness of the BC-STy issue, which by May 2009 had been on the political agenda for over 3 years. ‘ Since Green support declined as well as Liberal and NDP support, we can rule out the explanation that Liberal and NDP voters were simply learning that STV stood to hurt their respective parties.  38  In sum, interacting partisanship and survey day confirms the findings from previous steps, suggesting that partisanship had a strong effect on individuals’ BC-STy vote intentions during both referenda. The time-oriented regression equation in this step offers new insight, however, into exactly how the influence of partisanship came to be more powerful in 2009 than in 2005. As the campaign period progressed in 2005, it seems more and more voters learned with greater accuracy how they ought to vote based on their party preferences. As hypothesized, the effect of partisanship was therefore stronger in 2009, as additional partisans joined those many voters who had already figured out in 2005 how best to vote according to their party’s best interest.  3-B: INTERACTING PREDISPOSITION AND SURVEY DAY If we hope to clarify the relationship between the effect of partisanship and the influence of predisposition on BC-STy vote intentions, we must also establish how the effect of predisposition evolved independently over time.  This can be achieved by  simply interacting predisposition towards electoral system reform with survey day. To observe the relationship between predisposition and survey day, I created three new dummy variables corresponding to three behaviours of interest. Copying the coding used to create Figures 4a and 4b, the relationship between the new categories and the existing “predisposition” variable’s scale was as follows: anti-reform=0-.35, neutral=.35.65 and pro-reform=.7-1. Three additional variables were created to interact the new predisposition categories with survey day.  Each multiplies one of the predisposition  categories with the “surveyday” variable. In 2005 and 2009, respondents were distributed among the three predisposition categories as follows:  39  Figure 6a  —  Breakdown of Predisposition Categories in 2005 60% 50% 0 ‘  0  40%  0  30% 0  20% 10% 0% Anti-Reform  Neutral  Pro-Reform  Predisposition Categories  Figure 6b  —  Breakdown of Predisposition Categories in 2009 60% 50%  •  30%  & 20% 10%  0% Anti-Reform  Neutral  Pro-Reform  Predisposition Categories  In 2005 and 2009, pro-reform respondents were most numerous. The anti-reform category was always the least populated, never comprising more than 15% of the sample. The regression model for calculating the effect of predisposition towards electoral system reform interacted with survey day on BC-STV vote intention is as follows: BC-SW vote  =  +  (antireform) 1 13  +  (j3 2 neutral)  (s_antireformXsurveyday) 5 3 surveyday)  +  + +  (proreform) 3  +  (v_antireformXsurveyday) 4  (neutralXsurveyday) 6 13  3 ( 7 v proreformXsurveyday)  +  +  +  (f3 7 sproreformX  u  40  Figure 7a  —  Predisposition Categories Interacted with Survey Day in 2005 (120 days)  [. 0• I  I  I  I  I  0  20  40  60 Survey Day  80  Anti-Reform Pro-Reform  Figure 7b  —  —--—----  100  120  Neutral  Predisposition Categories Interacted with Survey Day in 2005 (25 days)  >.  0  a 11 0  >.. (I) ‘3)  I  I  95  100  I  —  I  105 110 Survey Day Anti-Reform Pro-Reform  —--—-—  115  120  Neutral  41  Figure 7c  —  Predisposition Categories Interacted with Survey Day in 2009 a). >.. -D  -co. 0  a 1) 0 >‘J-. Co  ID  10  15 Survey Day  Anti-Reform Pro-Reform  —  20 Neutral  In both 2005 and 2009, respondents in the three predisposition categories demonstrated extremely different yes-vote probabilities.  Pro-reform  individuals  exhibited likelihoods around 80 percentage-points, while anti-reformers expressed probabilities below 20 percentage-points. In keeping with the referenda results, yes-vote likelihoods were lower in 2009 than in 2005. For the most part, the levels of support generated by people’s predispositions in both years were characterized by remarkable stability over time. During the final month of the two campaigns, the predisposition categories’ probabilities fluctuated within only a few percentage-points. This was also mostly the case during the entire 120-day survey period in 2005, with the exception of the neutral category whose support for BC-STy dropped approximately 15 percentage-points. In short, interacting the predisposition categories and survey day echoes the findings from Step 2-C, suggesting that predisposition towards electoral system reform had a strong effect on voters’ BC-STy support in both 2005 and 2009.  This time-  oriented regression equation offers new insight, however, into just how stable this influence was, for the most part, over both survey periods. It seems that individuals did not require time, as they did with partisanship, to figure out how best to vote based on their predispositions.  42  3-C: INTERACTING PARTISANSHIP AND SURVEY DAY, CONTROLLING FOR PREDISPOSITION  With the independent effects over time of partisanship and predisposition towards electoral system reform firmly established by Steps 3-A and 3-B, we are now in a position to merge the two influences back together into one regression equation. The easiest way to do this is to simply add “predisposition” as a control variable to the regression from Step 3-A, which formerly just interacted partisanship and survey day. Doing so, we can control for the influence of predisposition by holding the “predisposition” variable fixed at each party category’s daily “predisposition” mean. The regression model for determining how each partisan category’s mean level of predisposition changed over time is as follows: predisposition mean  Figure 8a  —  =  -  (surveyday) + u, if [partisan category]=1 1 3  Partisan Category Predisposition Means Over Time in 2005 (120 days)  CD  a’ Cr-.. 0 Co (j)CD 0 CD  CD CD  0  20  40  60 Survey Day  Undecided NDP  80  100  120  Green Liberal  43  ___________________________  Figure 8b  —  Partisan Category Predisposition Means Over Time in 2005 (25 days)  I  95  I  100  I  105 110 Survey Day Undecided NDP  Figure 8c  —  I  I  115  120  Green Liberal  Partisan Category Predisposition Means Over Time in 2009  0)  C a> a>  0  5  10  15  Survey Day Undecided NDP  20  25  Green Liberal  In both 2005 and 2009, the party categories’ predisposition means remained remarkably steady over time. This is particularly apparent in Figure 8a, which illustrates that virtually no changes occurred to the partisan predisposition means over the 1 20-day survey period in 2005.  44  The regression model for calculating the effect of partisanship interacted with survey day, with predisposition towards reform controlled, on BC-STy vote intention is as follows: BC-SW vote  =  13o  +  (undecidedvote) 1 13  (surveyday)  +  —  (greenvote) 2 f3  +  (ndpvote) 3  (undecidedvoteXsurveyday) 6 13  (ndpvoteXsurveyday) 8 3  Figure 9a  +  +  +  +  135  (greenvoteXsurveyday) 7 f3  +  (liberalvoteX surveyday) 9 13  +  +  (liberalvote) 4 13  (predisposition) 10 13  +  u  Partisanship Interacted with Survey Day with Predisposition Controlled in 2005  (days) 0) U)  a) 0)  0  a)  >C.”, C) 0)  0  20  40  60 Survey Day  Undecided NDP  80  100  120  Green Lberai  45  Figure 9b  —  Partisanship Interacted with Survey Day with Predisposition Controlled in 2005  (25 days)  (D. U)  a) >L0.  c)  95  100  105 110 Survey Day Undecided NDP  Figure 9c  —  115  120  Green Liberal  Partisanship Interacted with Survey Day with Predisposition Controlled in 2009  U)  1: 0  5  10  15  Survey Day  Undeaded NDP  20  25  Green Liberal  Interacting partisanship and survey day while holding the “predisposition” variable fixed at each party category’s daily “predisposition” mean generally confirms the findings from Step 2-C.  By introducing the element of time into the regression  equation, however, the Step 3-C analysis reveals a more detailed story of how the  46  influences of partisanship and predisposition evolved concurrently over the course of the two survey periods. Figures 9a, 9b and 9c illustrate that individuals’ referendum vote intentions were neither solely the result of their party preferences, nor were they entirely the product of people’s predispositions towards electoral system reform. Had partisanship been the sole influence on people’s vote intentions, the graphs would have featured partisan yes-vote differences that continuously grew over the course of 2005.58 Instead, we fmd that the range of partisan differences narrowed during the last three weeks of 2005. On the other hand, had respondents been affected solely by their predispositions, the partisan yes-vote differences would have been chiefly stable over time. 59 This was clearly not the case, as can be seen in Figure 9a in particular.  We can therefore decisively conclude that  partisanship and predisposition indeed worked together to influence voter support for BC-STy in both the 2005 and 2009 referenda. Since the partisan categories’ predisposition means were relatively stable over time and since the individual effect of predisposition was also stable over time, predisposition most likely had a significant, but constant effect on voters’ BC-STy vote intentions in both years.  As the party categories’ predisposition means differed  somewhat from one another, predisposition was therefore responsible for some, but not all, of the inter-party probability differences in both 2005 and 2009. Had the different partisans all expressed the same mean level of predisposition, the range of inter-party yes-vote differences at each point in time would have been smaller, though it is impossible to say by exactly how many percentage-points. Predisposition could only have marginally been responsible, however, for the changes in each partisan category’s individual likelihood trajectory that occurred over time in 2005 and in 2009. Together, the stability of the predisposition effect and the stability of the partisan categories’ predispositions means indicate that the link between people’s predispositions and how they ought to vote based on those predispositions did not evolve over time.  Rather, voters were able to make a fairly accurate match-up  between their predispositions and the corresponding correct ballot alternative very early 58  This assumption is based on the fmdings from Step 3-A. This assumption is based on the findings from Step 3-B given the information outlined in Figures 6a, 6b and6c.  47  in 2005. They then maintained this link throughout the rest of 2005 campaign and had no difficulty resurrecting it in 2009. The effect of partisanship was also strong in both years, making party preferences responsible for the remaining portion of the inter-party yes-vote differences. parties had all stood to gain or lose equally by adopting BC-STy  —  If the  regardless of whether  that message was spread via party cues or individual mental calculations  —  the range of  inter-party yes-vote differences would have been smaller at each point in time. Since the independent effect of partisanship grew over time, as was discovered in Step 3-B, the influence of partisanship increased with time in this more complicated Step 3-C regression as well. This growing effect indicates that partisanship was primarily, and indeed almost entirely, responsible for the changes in each partisan category’s individual likelihood trajectory that occurred over time, particularly in 2005.  Since these  fluctuations led to the significant widening of the range of inter-party differences in that year, it seems that more and more partisans learned how to better vote in accordance with their party interests as the 2005 campaign period progressed. By the final month of the 2005 campaign, the effects of this learning period were well established and becoming somewhat stabilized. In 2009, as was first discovered in Step 2-A, the yes-vote differences of almost all the party categories were suppressed by some unknown dampening effect. Greens were exempted from this phenomenon, however, as a result of this Step 3-C regression equation.  This opens up the possibility that Green supporters were affected by  partisanship via the more traditional cognitive heuristics process, rather than the less orthodox one hypothesized in this paper. As was briefly mentioned in the Hypotheses section, the leader of the Green Party of B.C. publicly endorsed the yes-vote in a manner that was fairly prominent in the media. This endorsement occurred fairly late during the 2009 campaign period. The rising yes-vote likelihood among Greens in the immediate weeks preceding the 2009 voting day may therefore have been a reaction to this blatant party cue. In sum, interacting partisanship and survey day while holding the “predisposition” variable fixed at each partisan category’s daily “predisposition” mean confirms that both partisanship and predisposition towards electoral system reform strongly affected  48  individuals’ BC-STy vote intentions.  The influence of partisanship grew over time,  while the effect of predisposition remained constant.  As a result, partisanship and  predisposition shared the responsibility for creating the yes-vote probability differences between parties. However, partisanship alone was mostly responsible for the trajectory changes in yes-vote probability experienced by individual parties over time, which served to increase the range of the inter-party differences over the 2005 campaign. It seems then that, voters were able to match-up their predisposition and vote choice early in 2005, but required several months of learning to determine how best to vote according to their party preferences.  49  CONCLUSION In both 2005 and 2009, British Columbians who planned to vote for different parties in the general election exhibited noticeably different probabilities of supporting the BC-STy referendum.  In both years, Green supporters were on average the most  likely to support BC-STy, while Liberal voters were on average the least likely to favour supplanting first-past-the-post. The average likelihood of New Democrats supporting the referendum fell between the probabilities exhibited by Greens and Liberals. Regression analysis reveals that party preference and affiliation were indeed at the root of these differing levels of support. Even though the provincial parties remained largely silent on referendum-related issues, partisanship undeniably influenced BC-STV vote intentions in both 2005 and 2009.  Individuals, it therefore seems, were able to  determine all on their own how best to vote given their party preference. Knowledge, specifically about the Citizens’ Assembly and the BC-STy proposal itself, appears to have nurtured this process. Individuals became more accurate and more confident at matching their party’s best interests with the more appropriate vote choice as they became more knowledgeable about the aspects of the referendum. Time was also an essential element of this learning process, as partisans became more apt to and more adept at voting according to their partisan preferences as the campaign periods progressed. The effect of partisanship grew over time, as a result, eventually becoming much stronger in 2009 than in 2005. On the whole then, the findings from this paper’s various investigative analyses are consistent with the hypothesis that partisanship affected individuals’ BC-STy vote intention. The findings are also consistent with the notion that partisanship influenced voters via an individual, internal thought process aimed at maximizing the future electoral success of their preferred parties, as many of the expectations linked to such a process taking place indeed occurred in both referenda years. Partisanship was not the only influence behind the partisan category yes-vote likelihood differences in the 2005 and 2009 referenda, however.  The investigative  findings force us to additionally conclude that predisposition towards electoral system reform also strongly affected individuals’ BC-STy vote intentions, though this effect  50  remained largely stable over time. Pro-reform people were extremely likely to vote yes in the referendum, while anti-reform individuals were very likely to vote no. In fact, at certain levels of predisposition, voters’ intrinsic feelings about electoral system reform were far more important than their party preferences in determining their degree of support for BC-STy. In the end, it seems that partisanship, knowledge and predisposition all played a significant role in influencing how individuals planned to vote in the 2005 and 2009 BC STV referenda. This paper offers valuable insight into how these influences worked simultaneously and often hand in hand.  Even so, more scholarly study into these  extraordinary referenda is needed, as several important aspects of them remain unexplained  —  such as the reason for which BC-STy support declined globally in 2009  and the sources from which individuals gained their knowledge about referendum-issues.  51  REFERENCES Belanger, Eric, and Laura B. Stephenson. “Parties and Partisans: The Influence of Ideology and Brokerage on the Durability of Partisanship in Canada.” In Voting Behaviour in Canada, edited by Cameron D. Anderson and Laura B. Stephenson, 107-138. Vancouver: University of British Columbia Press, 2010.  Canada. British Columbia. Elections BC. Guide to Recounts. [Victoria B.C.], 2009. http://www.elections.bc.caldocs/guidebooks/879.pdf Canada. British Columbia. Elections BC. Statement of Votes: Referendum on Electoral Reform. [Victoria B.C.], 2005. http://www.elections.bc.ca/docs/rpt/SOV-2005ReferendumOnElectoralReform.pdf  Clarke, Harold D., Allan Kornberg and Thomas J. Scotto, Making Political Choices: Canada and the United States. Toronto: University of Toronto Press, 2009. Cutler, Fred, Richard Johnston, R. Kenneth Carty, André Blais and Patrick Fournier. “Deliberation, information, and trust: the BC Citizens’ Assembly as agenda setter.” In Designing Deliberative Democracy, edited by Mark E. Warren and Hilary Pearse, 166-191. Cambridge: Cambridge University Press, 2008. Hall, Neal. “New voting system headed for referendum loss.” The Vancouver Sun. May 9, 2009.  Lau, Richard R., and David P. Redlawsk. “Advantages and Disadvantages of Cognitive Heuristics in Political Decision Making.” American Journal ofPolitical Science 45 (October 2001): 95 1-971. http://web.ebscohost.com/. Kam, Cindy D. “Who Toes the Party Line? Cues, Values, and Individual Differences.” Political Behavior 27, no. 2 (June 2005): 163-182. http://web.ebscohost.coml.  52  Merolla, Jennifer, Laura Stephenson and Elizabeth Zechmeister, “Have Cue, Will Travel? Political Parties as Heuristics in Three Countries” Paper presented at the annual meeting ofthe American Political Science Association, Marriott Wardman Park, Omni Shoreham, Washington Hilton, Washington, DC, September 1, 2005. http://www.allacademic.comlmeta/p4 1 490_index.html/.  Squire, Peverill and Smith, Eric R. A. N.”The effect of partisan information on voters in nonpartisan elections.” Journal ofPolitics 50, no. 1 (February 1988):169—79. http://web.ebscohost.com/.  Stephenson, Laura B., Thomas J. Scotto and Allan Komberg. “Slip, Sliding Away or Le Plus ca Change...: Canadian and American Partisanship in Comparative Perspective.” American Review ofCanadian Studies 34, no. 2 (August 2004):283312. http://www.informaworld.com!.  Stevenson, H. Michael. “Ideology and Unstable Party Identification in Canada: Limited Rationality in a Brokerage Party System.” Canadian Journal ofPolitical Science 20, no. 4 (December 1987): 813-850. http://www.jstor.org/. Walter Borges and Harold D Clarke. “Cues in Context: Analyzing the Heuristics of Referendum Voting with an Internet Survey Experiment.” Journal ofElections, Public Opinion & Parties 18, no. 4 (November 2008): 433-448. http://www.informaworld.coml.  53  APPENDIX A -SURVEY INFORMATION The following information is mostly taken from the appendix section of “Deliberation, information, and trust: the BC Citizens’ Assembly as agenda setter” by Fred Cutler, Richard Johnston, André Blais, R. Kenneth Carty, and Patrick Foumier:  The BC Electoral Reform Referendum Stud[ies werel directed by Richard Johnston, Fred Cutler, André Blais, R. Kenneth Carty, and Patrick Fournier. Interviews lasted twenty minutes on average and almost all of the content referred to the referendum. The question was embedded in Computer Assisted Telephone Interviewing (CAT!) software originating with the University of California Survey Methods (CSM) group. Fieldwork was conducted by the Institute for Social Research at York University, under the direction of David Northrup.  The [2005] sample was released dynamically: as a  weekly rolling cross-section from January 17 to April 30 and then a daily rolling crosssection from May 1 to 16. [The 2009 sample was released similarly, but only between April l6andMay 11.1 The [2005] survey is representative of the eligible voting population, with the important exception of those who could not complete the interview in English.  One  result is that in our sample, 4.5 per cent responded “Chinese” to the question asking “to what ethnic or cultural group do you belong?”. The Census figure is 15.4 percent. We also get 1.5 percent “Indian”; the Census figure is 7.3 percent. The Census figures are overestimates of the voting population however, since a significant proportion of these groups are non-citizens. Nonetheless, these groups are under-represented in our sample. Language is obviously the main reason, but there must also be a cultural barrier to participation. We can only flag this shortcoming and be sensitive to the possibility that decision-making processes observed in the sample do not generalize perfectly to the population of voters. [The 2009 survey did not include questions about respondents’ ethnic backgrounds, but it is likely that non-English speakers were once again under represented.]  54  APPENDIX B -SURVEY QUESTIONS USED The survey questions and variables used in this paper are as follows. Questions are grouped thematically for the reader’s convenience. Only items used in this paper’s analyses are presented.  BC-STy Vote Intention •  The referendum question will be: “Should British Columbia change to the BC STV electoral system as recommended by the Citizens Assembly on Electoral Reform?” [If you vote do/Do] you think you will vote YES or NO? (2005)  •  The referendum question will be: “Which electoral system should BC use to elect members to the provincial Legislative Assembly? The existing electoral system, First-Past-the-Post or the single transferable vote electoral system (BC-STy) proposed by the Citizens’ Assembly on Electoral Reform.” Do you think you will vote for the existing system or for the proposed BC-STy system? (2009)  General Election Vote Intention •  [If you vote, do you think/Will] you will vote for the Liberal Party, the NDP, the Green Party, or another party?  Intervener Ratings •  Use a scale from ZERO to ONE HUNDRED.  ZERO means you REALLY  DISLIKE that group and ONE HUNDRED means you REALLY LIKE that group. How do you feel about the BC Liberal Party? •  Use a scale from ZERO to ONE HUNDRED.  ZERO means you REALLY  DISLiKE that group and ONE HUNDRED means you REALLY LIKE that group. How do you feel about the BC New Democratic Party? •  Use a scale from ZERO to ONE HUNDRED.  ZERO means you REALLY  DISLIKE that group and ONE HUNDRED means you REALLY LIKE that group. How do you feel about the BC Green Party?  55  Combined Knowledge •  How much would you say you know about the BC Citizens’ Assembly on Electoral Reform? Would you say a lot, some, not very much, or nothing? (2005)  •  How much would you say you know about the BC Citizens’ Assembly on Electoral Reform which was in operation during 2004? Would you say a lot, some, not very much, or nothing? (2009)  •  The Citizens’ Assembly has proposed a change to the way we elect the BC Legislature in Victoria. The system they proposed is called BC-SW. Would you say you know a lot, some, not very much, or nothing about this proposal?  Predisposition Towards Electoral System Reform •  In your opinion, which is better: governments made up of 2 or 3 parties because they are forced to compromise, or one party governments so they can get things done?  •  Under our present system, a party can win a majority of seats without winning a majority of votes. Do you find this acceptable, unacceptable, or do you not have an opinion on this?  •  Under the proposed BC-STy system it will be hard to follow how the ballots are counted. Are you very concerned, somewhat concerned, not very concerned, or not at all concerned about this?  •  Some people say that with the BC-STy voting system the parties in power will change too often. How concerned are you about this? Are you very concerned, somewhat concerned, not very concerned, or not at all concerned about this?  •  Some people like a two-party system because you know who to blame, other people like a system with more than two parties because that way the voter has more choices. Which is more important to you, knowing who to blame or having more choices?  Survey Day •  The day of the interview  •  The month of the interview  56  APPENDIX C  LOGIT REGRESSION TABLES  -  ANALYZING 2005 SURVEY DATA The logit regression tables calculated using Stata to analyze the 2005 survey data for this paper are as follows.  STEP 1: EXAMINING THE EFFECT OF PARTISANSHIP ALONE BC-SW vote  =  f3  +  f3  +  (3 1 undecidedvote)  (13 5 conservativevote)  VARIABLES undecidedvote  +  ndpvote iberavote  ,  +  REGRESSION #3  #1  #2  vote  vote  omitted  -0.0326 (0.265)  (0.151) 0.741 conservativevote (0.488) 0.964*** Constant (0.116) Observations  (13 2 greenvote)  (j3 3 ndpvote)  +  (13 4 Iiberalvote)  +  u  0.0326 (0.265) -0.261 (0.167) 0.708***  greenvote  +  omitted  #4  #5  vote  vote  vote  0.261 (0.167) 0.294 (0.267)  0.708***  0.741 (0.488) 0.773 (0.531) 0.48 (0.489) 0.0323 (0.484)  (0.151) 0.741*** (0.258) 0.447***  -0.294 (0.267) 0.741***  0.447***  (0.258) -0.773 (0.531) 0.996***  (0.154) -0.48 (0.489) 0.703***  -0.0323 (0.484) 0.255***  (0.239)  (0.12)  (0.0969)  1227  1227  1227  1227  omitted  (0.154) omitted  omitted 0.223 (0.474)  I  1227  Standard errors in parentheses p<0.01, ** p<0.05, * p<0.1  57  STEP 2: INVESTIGATING OTHER POSSIBLE INFLUENCES  2-A: CONTROLLING FOR KNOWLEDGE BC-SW vote  +  =  (undecidedvote) 1 3  (conservativevote) 5  VARIABLES undecidedvote  +  ndpvote liberalvote conservativevote knowledge Constant  Observations  (greenvote) 2 3  (knowIedge) 6 3  #2  vote  vote  omitted  -0.0375 (0.271)  0.0375 (0.271) 0.407**  omitted  (0.136)  -0.445 (0.274) 0.857*** (0.264) -0.753 (0.539) 1.481*** (0.234) 0.573** (0.25)  1180  1180  (0.175) 0.819*** (0.158) -0.716 (0.496) 1.481*** (0.234) 0.536***  +  +  (ndpvote) 3  +  (IiberaIvote) 4 3  +  u  REGRESSION #3  #1  _____________  greenvote  +  #4  #5  vote  vote  vote  O.407**  0.819***  (0.175) 0.445 (0.274)  (0.158) 0.857***  0.716 (0.496) 0.753 (0.539) 0.309 (0.498) -0.103 (0.493)  omitted 0.412** (0.161) -0.309 (0.498) 1.481***  (0.264) 0.412** (0.161) omitted 0.103 (0.493) 1.481***  omitted 1.481***  (0.234) 0.129 (0.151)  (0.234) 0.283** (0.131)  (0.234) -0.18 (0.486)  1180  1180  1180  Standard errors in parentheses p<0.01, ** p<O.05, * p<0.l  58  2-B: INTERACTING PARTISANSHIP AND KNOWLEDGE BC-STy vote  =  13  +  (undecidedvote) 1 13  (conservativevote) 5 13  +  (greenvote) 2 3  +  (knowledge) 6 13  (greenvoteXknowledge) 8 j3 knowledge)  +  +  +  +  (ndpvote) 3 13  +  (liberalvote) 4 13  (undecidedvoteXknowledge) 7 13  (ndpvoteXknowledge) 9 f3  (conservativevoteXknowledge) 11 13  +  +  + +  (IiberalvoteX 10 13  u  REGRESSION #1 VARIABLES  vote 0.245 (0.271) 0.0869 (0.4)  undecidedvote greenvote ndpvote  omitted  liberalvote  -0.0529 (0.259) 0.679 (0.783) 1.741 (0.456) 0.732 (0.7) 1.804 (1.303)  conservativevote knowledge undecidedvoteXknowledge greenvoteXknowledge ndpvoteXknowledge  omitted  IiberalvoteXknowledge conservativevoteXknowledge Constant  Observations  1.002* (0.571) 3.480* (2.075) 0.036 (0.204) 1180  Standard errors in parentheses p<0.01, p<O.O5, p<0.1 **  *  59  2-C: INTERACTING PARTISANSHIP AND PREDISPOSITION BC-STy vote  =  13o  +  (13 1 undecidedvote)  (5 conservativevote)  +  (3 2 greenvote)  +  (f3 3 ndpvote)  +  (13 4 Iiberalvote)  +  (undecidedvoteXpredisposition) + 7 (13 6 predisposition) + 3 (greenvoteXpredisposition) + 13 8 13 (ndpvoteXpredisposition) + 13 9 (IiberalvoteX 10 predisposition) + 3 (conservativevoteXpredisposition) + u 8 +  REGRESSION  #1 VARIABLES undecidedvote greenvote  vote 1.147 (0.831) 1.131 (1.715)  ndpvote  omitted  liberalvote  -0.39 (0.746) -3.485 (3.064) 4.898***  conservativevote predisposition undecidedvoteXpredisposition greenvoteXpredisposition ndpvoteXpredisposition liberalvoteXpredisposition onservativevoteXpredispositior. Constant  Observations  (0.839) -1.22 (1.171) -1.327 (2.209) omitted 0.739 (1.109) 5.854 (5.26) 2.627*** (0.599) 864  Standard errors in parentheses p<0.01, ** p<0.05, * p<0.1  60  STEP 3: EXAMINING PARTISANSHIP AND PREDISPOSITION OVER TIME 3-A: INTERACTING PARTISANSHIP AND SURVEY DAY BC-STy vote  =  +  (undecidedvote) 1 3  (surveyday) 5 3  +  +  (greenvote) 2 13  +  (ndpvote) 3  (undecidedvoteXsurveyday) 6 13  (ndpvoteXsurveyday) 8 13  +  +  (IiberaIvote) 4 3  +  (greenvoteXsurveyday) 7 13  +  I3gOiberalvoteX surveyday)  +  +  u  REGRESSION  #1 VARIABLES  vote 0.885***  undecidedvote greenvote  (0.337) -0.0526 (0.567)  ndpvote  omitted  liberalvote campday undecidedvoteXcampday greenvoteXcampday  0.00495 (0.323) 0.00382 (0.00277) -0.0081 6** (0.00401) 0.0046 (0.00656)  ndpvoteXcampday  omitted  IiberalvoteXcampday  -0.0055 (0.0037) 0.384 (0.239)  Constant  Observations  1227  Standard errors in parentheses p<0.01, p<0.05, p<O.1 **  *  61  3-B: INTERACTING PREDISPOSITION AND SURVEY DAY  BC-SW vote  =  13o  +  (antireform) 1 13  +  (neutraI) 2 3  (neutralXsurveyday) 5 13  +  +  (proreform) 3 13  (antireformXsurveyday) 4 13  +  (proreformXsurveyday) 6 13  +  +  u  REGRESSION #1 VARIABLES  vote  -2. 029***  antireform  (0.699)  neutral proreform surveyday antireformXsurveyday  omitted 0.555* (0.326) -0.00561 ** (0.00279) 0.00383 (0.00805)  neutralXsurveyday proreformXsurveyday Constant  omitted 0.0091 7** (0.0038) 0.638***  Observations  (0.239) 979  Standard errors in parentheses p<0.01, ** p<0.05, * p<0.1  62  3-C: INTERACTING PARTISANSHIP AND SURVEY DAY, CONTROLLING FOR PREDISPOSITION  BC-SW vote  =  13  +  (undecidedvote) 1 13  (surveyday)  +  +  (greenvote) 2 f3  +  (ndpvote) 3 13  (undecidedvoteXsurveyday) 6 13  (ndpvoteXsurveyday) 8 13  +  +  (Iiberalvote) 4 f3  +  135  (greenvoteXsurveyday) 7 13  +  (IiberalvoteX surveyday) 9 13  +  +  (predisposition) 10 13  +  u  REGRESSION #1 VARIABLES  vote 0.91 5** (0.452) -1.184 (0.748)  undecidedvote greenvote ndpvote liberalvote campday undecidedvoteXcampday greenvoteXcampday  omitted 0.698 (0.43) 0.00371 (0.00358) -0.00799 (0.00542) 0.0165* (0.00898)  omitted  rtdpvoteXcampday IiberalvoteXcampday pref_sum  0.00823* (0.00489) 4.946***  Constant  Dbservations  (0.443) -2. 956*** (0.435) 864  Standard errors in parentheses p<0.01, ** p<0.05, * p<0.1  63  ______  APPENDIX D  -  LOGIT REGRESSION TABLES  ANALYZING 2009 SURVEY DATA The logit regression tables calculated using Stata to analyze the 2009 survey data for this paper are as follows.  STEP 1: EXAMINING THE EFFECT OF PARTISANSHIP ALONE BC-SW vote  =  13o  +  13o  +  (undecidedvote) 1 13  (conservativevote) 5 3  +  ndpvote liberalvote Constant  Observations  +  REGRESSION #2 #3  vote  undecidedvote omitted greenvote  (greenvote) 2  (ndpvote) 3 13  +  (Iiberalvote) 4 13  +  u  #1 VARIABLES  +  vote  vote  vote  0.669*  0371*  1.093***  (0.354)  (0.223) 0.298 (0.358)  (0.214) 1.761***  0.669*  omitted (0.354) 0.371* -0.298 omitted (0.223) (0.358) 1.093*** 1.761*** 1.463*** (0.214) (0.352) (0.219) 0.857*** 0.560*** 0.189 (0.154) (0.319) (0.162) 604  I  #4  604  604  (0.352) 1.463*** (0.219) omitted 0904*** (0.148) 604  Standard errors in parentheses *** p<O.Ol, ** p<0.05, * p<0.1  64  STEP 2: INVESTIGATING OTHER POSSIBLE INFLUENCES  2-A: CONTROLLING FOR KNOWLEDGE  BC-SW vote  f3  +  (undecidedvote) 1 f3  (conservativevote) 5 13  +  +  (knowledge) 6 13  VARIABLES  vote  undecidedvote  omitted  (ndpvote) 3 13  (Iiberalvote) 4 j3  +  #4  vote  vote  0.633*  -0.305 (0.227) 0.329 (0.362)  1.135***  (0.358)  +  u  vote  0.633*  greenvote  +  +  REGRESSION #2 #3  #1  (0.217) 1 .769*** (0.356) 1 .440’ (0.221)  omitted (0.358) 0.305 -0.329 omAtted (0.227) (0.362) 1.135*** 1.769*** 1.440*** omitted (0.217) (0.356) (0.221) 0.887*** 0.887*** 0.887*** 0.887*** (0.325) (0.325) (0.325) (0.325) 1.227*** -0.0919 0.542 0.213 (0.187) (0.338) (0.205) (0.194)  ndpvote liberalvote knowledge Constant Observations  (greenvote) 2 13  I  600  I  600  600  I  600  Standard errors in parentheses p<0.01, ** p<0.05, * p<01  65  2-B: iNTERACTING PARTISANSHIP AND KNOWLEDGE BC-SW vote  =  +  (undecidedvote) 1  (conservativevote) 5 j3  +  +  (knowledge) 6 13  (greenvoteXknowledge) 8 knowledge)  +  (greenvote) 2  +  (ndpvote) 3 13  +  +  +  (Iiberalvote) 4 13  (undecidedvoteXknowledge) 7 f3  (ndpvoteXknowledge) 9 13  (conservativevoteXknowledge) 11 f3  +  +  + +  (liberalvoteX 10 13  u  REGRESSION #1 VARIABLES undecidedvote greenvote ndpvote liberalvote knowledge undecidedvoteXknowledge greenvoteXknowledge ndpvoteXknowledge liberalvoteXknowledge Constant Observations  vote -0.296 (0.37) -0.78 (0.589) omitted -1 .299*** (0.377) 0.801 (0.587) -0.0445 (0.837) 4.272** (1.83) omitted -0.373 (0.811) 0.246 (0.277) 600  Standard errors in parentheses p<0.01, ** p<O.05, * p<0.1  66  2-C: iNTERACTING PARTISANSHIP  BC-SW vote  =  +  (undecldedvote) 1 13  (conservativevote) 5 13  +  +  (greenvote) 2 13  (predisposition) 6 13  (greenvoteXpredisposition) 8 f3 predisposition)  +  AND  +  +  PREDISPOSITION  +  (ndpvote) 3 13  (Iiberalvote) 4 13  +  (undecidedvoteXpredisposition) 7 j3  (ndpvoteXpredisposition) 9 j3  (conservativevoteXpredisposition) 8 13  +  + +  (IiberalvoteX 10 f3  + U  REGRESSION #1 VARIABLES u ndecidedvote greenvote ndpvote liberalvote predisposition undecidedvoteXpredispositior. greenvoteXpredisposition ndpvoteXpredisposition liberalvoteXpred isposition Constant Observations  vote -0.824 (1.144) -0.672 (1.971) omitted 2.284** (1.098) 4.802*** (1.063) 1.198 (1.612) 1.577 (2.786) omitted 2.951* (1.637) -2. 745*** (0.782) 447  Standard errors in parentheses p<0.01, ** p<0.05, * p<0.1  67  STEP 3: EXAMINING PARTISANSHIP AND PREDISPOSITION OVER TIME 3-A: INTERACTING PARTISANSHIP AND SURVEY DAY BC-SW vote  =  (undecidedvote) + 3 1 (greenvote) + 13 2 (IiberaIvote) 4 (ridpvote) + 3 3 13o + 3 (surveyday) + 3 5 3 (undecidedvoteXsurveyday) + 13 6 (greenvoteXsurveyday) 7  (ndpvoteXsurveyday) 8 13  +  (IiberalvoteX surveyday) 9 13  +  +  +  u  REGRESSION #1 VARIABLES undecidedvote greenvote ndpvote liberalvote surveyday undecidedvoteXsurveyday greenvoteXsurveyday  vote -0.8 (0.611) -0.0331 (0.94) omitted 1 .736*** (0.575) 0.0425* (0.0224) 0.0236 (0.0333) 0.0181 (0.0506)  ndpvoteXsurveyday  omitted  IiberalvoteXsurveyday  0.0137 (0.0314) 1 .298*** (0.43)  Constant Observations  604  Standard errors in parentheses p<0.01, ** pcO.05, * p<0.1  68  3-B: INTERACTiNG PREDISPOSITION AND SURVEY DAY BC-SW vote  =  13o  +  (3 1 antireform)  +  (2 neutral)  (3 5 neutralXsurveyday)  +  +  (r3 3 proreform)  (3 4 antreformXsurveyday)  +  (13 6 proreformXsurveyday)  +  +  u  REGRESSION #1 VARIABLES  vote  2.474*  antireform  (1.49)  neutral  omitted 2.194***  proreform  (0.543) 0.00247 (0.0208) 0.0185 (0.0797)  surveyday antireformXsurveyday neutralXsurveyday  omitted  proreformXsurveyday  -0.0195 (0.0298) -0.598 (0.374)  Constant Observations  532  Standard errors in parentheses p<0.01, p<0.05, p<01 **  *  69  3-C: INTERACTING PARTISANSHIP AND SURVEY DAY, CONTROLIJNG FOR PREDISPOSITION BC-STy vote  =  +  (undecidedvote) 1 13  (surveyday)  (greenvote) 2 3  +  +  (ndpvote) 3 13  (undecidedvoteXsurveyday) 6 13  +  (ndpvoteXsurveyday) 8 13  +  +  (Iiberalvote) 4 13  +  135  (greenvoteXsurveyday) 7 13  +  (IiberalvoteX surveyday) 9 13  +  +  (predisposition) 10 13  +  u  REGRESSION #1 VARIABLES undecidedvote greenvote  vote -0.693 (0.5) -0.676 (1.145)  ndpvote  omitted  liberalvote  -0.713 (0.772) -0.04 12 (0.0307) 0. 04 16 (0.0462) 0.0681 (0.0667)  surveyday  undecidedvoteXsurveyday greenvoteXsurveyday  nd pvoteXsurveyday  omitted  IiberalvoteXsurveyday  0.0171 (0.0417) 6.218***  predisposition Constant  Observations  (0.648)  3.028*** (0.731) 447  Standard errors in parentheses p<0.01, ** p<0.05, * p<0.1  70  

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.24.1-0071436/manifest

Comment

Related Items