no job namepatrick-fournier.com › d › fournier2013.pdf · patrick fournier, centre for the...

Riding the Orange Wave: Leadership, Values,Issues, and the 2011 Canadian Election

PATRICK FOURNIER Université de MontréalFRED CUTLER University of British Columbia

STUART SOROKA McGill UniversityDIETLIND STOLLE McGill University

ÉRIC BÉLANGER McGill University

The 2011 Canadian federal election was historic. The New DemocraticParty ~NDP! formed the official opposition for the first time. The Lib-eral party suffered its worst defeat ever at the national level.1 The 2011contest produced the first majority government for the reunited right underthe banner of the Conservative party. And, in Quebec, the election resultedin a complete collapse ~in seats, if not also votes! of the main player inthat province over the last two decades, the Bloc Québécois.

What accounts for these dramatic and largely unexpected shifts inthe Canadian electoral landscape, particularly the sudden leap in popu-larity for the NDP? The aftermath of election day produced no shortageof potential explanations. Pundits, politicians, and political scientists havesuggested that fluke polls swayed public opinion, that decisive campaignevents such as the leaders’ debates shifted support dramatically, that“Layton-mania” took hold, or that results reflected a leftward shift or arightward shift or both, that is, increasing ideological polarization in Cana-dian politics. In Quebec, where the NDP saw its greatest increase in sup-port, it has also been suggested that the election outcome was the productof increasing cynicism towards major parties or that it stemmed from adecline in support for Quebec sovereignty. To a great extent, these pos-sible explanations reflect past research on the forces that affect electoraloutcomes: campaign events ~Blais et al., 1996, 1999, 2003; Geer, 1988;Johnston et al., 1992!, leadership ~Aarts et al., 2011; Bean and Mughan,

Patrick Fournier, Centre for the Study of Democratic Citizenship, Département descience politique, Université de Montréal, Pavillon Lionel-Groulx, C.P. 6128, Succ.Centre-ville, Montréal QC, H3C 3J7, [email protected]

6 6

CJP 00087 1035 12003013 1:37 PM REVISED PROOF

�

�

�

CJP00087

Canadian Journal of Political Science / Revue canadienne de science politique46:4 (December/décembre 2013) 1–35 doi:10.10170S0008423913000875

© 2013 Canadian Political Science Association ~l’Association canadienne de science politique!and0et la Société québécoise de science politique

1989; McAllister, 1996!, persuasion ~Hillygus and Shields, 2009; Zaller,1992, 1996!, and priming ~Iyengar et al., 1984, 1987; Krosnick andKinder, 1990; Krosnick and Brannon, 1993; Mendelsohn, 1996; Millerand Krosnick, 2000!.

We use the 2011 Canadian Election Study ~CES! to evaluate theempirical validity of these accounts of the “orange wave.” Our goal hereis to separate fact from fancy. We begin by reviewing the election’s out-come, exploring just how much aggregate opinion change occurred, andwhere. Then we sketch a picture of campaign dynamics and key moments,using a combination of CES data and media content analyses. This isfollowed by an individual-level analysis of CES data. Models of votechoice and turnover from 2008 to 2011 are used to uncover who exactlychanged their minds and to identify the factors and motivations that ledto the unexpected electoral results of 2011.

The 2011 Election

Table 1 presents the actual vote shares ~from Elections Canada! receivedby the main political parties during the federal elections of 2008 and2011 and the change between the two ~indicated by the symbol D!. Atthe national level ~see the last row!, two parties made gains among theelectorate: the Conservatives were up very slightly ~�1.9 percentagepoints!, while the NDP expanded its support extensively ~�12.4 points!.The other three parties’ shares of the vote declined. The Liberals lost thegreatest number of votes ~over 850,000!, and the Greens lost nearly halfof their support. Most visible in lost seats, the Bloc Québécois was aban-doned by close to 40 per cent of its former voters.

The regional story is one of relative homogeneity, at least in the restof Canada ~ROC!. The Conservatives made gains in every province exceptQuebec, though somewhat more in the East than in the West, where theywere already strong. They finished first in all provinces except Quebecand Newfoundland and Labrador, stealing the lead from the 2008 Liber-als in two ~Nova Scotia and Prince Edward Island!. The New Democratsimproved their standing in every province save Newfoundland and Lab-rador. NDP gains were rather uniform across the ROC and, what is prob-ably more significant, the party finished second in every province exceptPEI. The Liberals, in contrast, lost support everywhere. They finishedthird or worse in eight out of ten provinces ~the exceptions being PEIand Newfoundland and Labrador!. The Greens also lost ground in everysingle province.

Quebec is often distinctive in federal elections, of course, but theresults of 2011 seem particularly dramatic in this regard—indeed, themost dramatic since the Bloc arrived on the scene in 1993. Only one

6 6

CJP 00087 2035 12003013 1:37 PM REVISED PROOF PAGE:2

�

�

�

2 PATRICK FOURNIER ET AL.

party improved its position in Quebec in 2011, the NDP. Its advances inthat province were about three times as large as anywhere else ~�30.7percentage points!. And the Bloc suffered the largest fall in its history~�14.7 points!, winning only four seats, making this the first time it didnot win the majority of seats in Quebec in the last 20 years. The Conser-vatives failed to make headway in the province ~�5.2 points!, in spite ofthe fact that they did so everywhere else. The Liberals also lost the larg-est proportion of their support in that province ~down roughly 10 pointsor 40%!. The changes in vote shares from 2008 to 2011 in Quebec were,in comparison to the rest of the country, much more sweeping.

Campaign dynamics in 2011 also point to a more volatile electoralenvironment in Quebec. Figure 1 depicts the daily movements of the aver-age shares in vote intentions2 using data from the Canadian ElectionStudy.3 The rolling cross-sectional design of the CES campaign waveallows us to track public opinion with a common survey methodologythroughout the campaign. Quebec and the rest of Canada are treatedseparately.

Outside Quebec, shifts during the campaign were rather modest. TheConservative party led wire to wire. It experienced a notable but tempo-rary boost in mid-April. Apart from this bump, however, the Conserva-tive line is essentially f lat. The Liberals ran second for most of thecampaign. Their share dipped first briefly during the second week of Apriland then decidedly during the last week, when they lost about 10 per-centage points and fell below the New Democrats. The NDP’s line mir-

Abstract. The results of the 2011 Canadian federal election were surprising. What accountsfor the dramatic and largely unexpected shift in electoral fortunes? Most importantly, whataccounts for the sudden leap in popularity of the New Democratic Party, particularly in Que-bec? The aftermath of election day produced no lack of potential explanations. Pundits, politi-cians, and political scientists have suggested many. This paper examines the empirical validityof various explanations swirling about the 2011 election, especially regarding the “orange surge.”The analysis relies upon the 2011 Canadian Election Study and the content of news mediacoverage. It concludes that the most important factors behind the orange wave were the imagegap between Jack Layton and the other party leaders, as well as the proximity between theNDP’s values and issue positions and those of many Quebeckers.

Résumé. Le résultat de l’élection fédérale canadienne de 2011 a été surprenant. Commentexpliquer les renversements dramatiques et largement inattendus dans les appuis aux partis poli-tiques? Surtout, comment expliquer le bond soudain de popularité du Nouveau parti démocra-tique, particulièrement au Québec? Plusieurs explications potentielles ont été suggérées par lescommentateurs, les politiciens et les politologues au lendemain de l’élection. Cet article exam-ine la validité empirique de nombreuses explications entourant l’élection de 2011, notammentcelles portant sur la « vague orange ». L’analyse repose sur l’Étude électorale canadienne de2011 ainsi que sur le contenu de la couverture médiatique. Elle indique que les facteurs les plusimportants à l’origine de la vague orange ont été l’écart entre l’image de Jack Layton et celledes autres chefs de partis, de même que la proximité entre le NPD et plusieurs Québécois quantaux valeurs et aux enjeux.

6 6


�

�

�

TAB

LE

1V

ote

Sha

res

in20

08an

d20

11~%

!

Con

serv

ativ

eL

iber

alN

DP

Blo

cG

reen

’08

’11

D’0

8’1

1D

’08

’11

D’0

8’1

1D

’08

’11

D

BC

44.5

45.6

�1.

119

.313

.4�

5.9

26.1

32.5

�6.

49.

47.

7�

1.7

Alb

.64

.766

.8�

2.1

11.4

9.3

�2.

112

.716

.8�

4.1

8.8

5.2

�3.

6S

ask.

53.8

56.3

�2.

514

.98.

5�

6.6

25.5

32.3

�6.

85.

52.

6�

2.9

Man

.48

.953

.5�

4.6

19.1

16.6

�2.

524

.025

.8�

1.8

6.8

3.6

�3.

2O

nt.

39.2

44.4

�5.

233

.825

.3�

8.5

18.2

25.6

�7.

48.

03.

8�

4.2

Que

.21

.716

.5�

5.2

23.8

14.2

�9.

612

.242

.9�

30.7

38.1

23.4

�14

.73.

52.

1�

1.4

NB

39.4

43.8

�4.

432

.522

.6�

9.9

21.9

29.8

�7.

96.

13.

2�

2.9

NS

26.1

36.7

�10

.629

.828

.9�

0.9

28.9

30.3

�1.

48.

03.

9�

4.1

PE

I36

.241

.2�

5.0

47.7

41.0

�6.

79.

815

.4�

5.6

4.7

2.4

�2.

3N

fld

16.6

28.3

�11

.746

.837

.9�

8.9

33.7

32.6

�1.

11.

70.

9�

0.8

Can

.37

.739

.6�

1.9

26.3

18.9

�7.

418

.230

.6�

12.4

10.0

6.1

�3.

96.

83.

9�

2.9

Off

icia

lel

ecti

onre

sult

s,so

urce

:E

lect

ions

Can

ada

6 6


�

�

�


FIGURE 1Campaign Dynamics of Vote Intensions

6 6


�

�

�

Riding the Orange Wave 5

rors to a great extent the Liberal one; steady for the first half of thecampaign, with a slow rise starting at the halfway point, for about a10-point gain overall.

In Quebec, shifts in vote intentions exhibited more amplitude andstarted earlier than in the ROC. The vague orange ~orange wave! is indeedimpressive, a swing of approximately 30 percentage points in 30 days.The NDP started the race in fourth place, exactly where it left off in2008, with around 15 per cent of popular support. Nothing at the begin-ning of the campaign hinted at the historic breakthrough that was aboutto unfold. It is interesting to note that the orange surge occurred in threedistinct stages. First, very early on, the NDP jumped to 25 per cent, fromfourth to second place. NDP support then held steady for about two weeksbefore taking off again in mid-April, putting the NDP in the lead in Que-bec for the first time ever. There was a slip in the NDP’s vote share aroundApril 23, but this was a temporary setback, as the wave gained poweragain in the last week of the campaign. The Bloc Québécois’ line nearlymirrors the NDP’s, at least until the final weeks when the NDP rise seemsto be driven as much by diminishing support for the Conservatives andLiberals as by declining support for the Bloc. Ultimately, all three ofthese parties lost close to 10 percentage points in Quebec during the cam-paign, with apparently almost all of that going to the NDP.

Who changed their vote in this election? Table 2 cross-tabulates vot-ers’ choices in the two most recent federal elections, in the ROC and in

TABLE 2Sources of the 2011 Vote ~%!

2008

2011 Conserv. Liberal NDP Bloc Green Abstain Entering

ROCConserv. 86.9 9.6 8.9 11.1 18.3 20.0Liberal 3.1 55.9 6.0 20.4 13.8 12.0NDP 5.3 24.2 79.7 33.3 22.9 36.0Green 0.9 3.2 2.1 29.6 3.7 8.0Abstain 3.8 7.1 3.2 5.6 41.3 24.0Total 100 100 100 100 100 100

QuebecConserv. 59.4 11.3 4.4 4.6 0.0 15.1 5.9Liberal 1.9 43.4 4.4 3.9 8.7 7.5 0.0NDP 25.8 33.3 86.8 31.1 65.2 20.8 64.7Bloc 3.2 1.9 2.9 55.0 8.7 7.5 5.9Green 0.0 1.3 0.0 0.7 13.0 0.0 0.0Abstain 9.7 8.8 1.5 4.6 4.3 49.1 23.5Total 100 100 100 100 100 100 100

6 6


�

�

�


Quebec; it shows how the supporters of each party in 2008 voted in 2011.The cells reporting the proportion of stable voters ~who did not changetheir vote from 2008 to 2011! have a grey background; all other cellsdescribe the percentage of each party’s 2008 voters who voted differ-ently in 2011.4

Results confirm that changes in vote choice were more common inQuebec than elsewhere. While approximately 87 per cent of 2008 ROCConservative voters endorsed the party again in 2011, only six out of tenQuebec Conservative voters remained loyal to the party. Liberal votersin 2008 in Quebec were also more likely than those in the ROC to aban-don their party in 2011, mostly to the advantage of the NDP. Indeed,looking across all parties, it is clear that New Democrats were the mainbeneficiaries of preference changes, particularly in Quebec. The NDPattracted the support of 26 per cent of 2008 Conservatives, 33 per centof 2008 Liberals, 31 per cent of 2008 Bloquistes, and 65 per cent of2008 Greens in that province. In this sense, the new NDP vote in Que-bec represents a kind of rainbow coalition. The party even had successamong 2008 abstainers ~21% voted NDP!, and especially new voters ~65%voted NDP!.5 The proportion of 2011 NDP voters who had voted other-wise ~or not at all! in 2008 was 21 per cent in the ROC, while it was astunning 47 per cent in Quebec.

Clearly, there is much change in opinion to explain. Figure 1 makesclear that big changes occurred during the campaign, not before. Alongwith Table 2, it also suggests that Quebec deserves special attention. Weturn to these tasks below, giving specific consideration to the possibleexplanations set out in the introduction. We start by exploring some ofthe most common propositions relating to campaign events.

The Campaign Events

This section focuses on the dynamics of vote intentions and media con-tent, and explores the potential impact of campaign events. Vote inten-tions are illustrated in Figure 1, above. For media content, we rely ondata from a content analysis of all election-related stories in ten dailynewspapers.6 We focus on a very simple statistic: the daily proportionof “first mentions” for each of the parties0leaders. The “first mentions”—the leader or party mentioned first in an article—serve as a simple butvaluable measure of the presence and prominence of parties and leadersin campaign coverage ~Andrew et al., 2008; Daku et al., 2009!. Theyalso capture a degree of campaign framing. When a campaign is viewedmainly as a competition highlighting a single party, for instance, thatparty tends to receive first mentions even in articles that end up focus-ing on other parties. As a result, first mentions tend to be relatively

6 6


�

�

�


strong predictors of campaign-period trends in vote shares and of elec-toral outcomes more generally. They have the added advantage of beingan objective measure of campaign coverage, captured using automatedcontent analysis in Lexicoder.7

Figure 2 displays trends in party0leader first mentions over the 2011campaign. The Conservative party dominated coverage: in the ROC,roughly 60 per cent of all campaign-related stories mention either theConservatives or Harper before any other party or leader; in Quebec, theConservative share is lower but usually greater than any other party. Thatsaid, there are some important shifts and we use these below in conjunc-tion with the data on vote intentions to explore the potential impact ofcampaign events on the final outcome.

Media commentators are often tempted to see campaign events astriggers of change. Indeed, some accounts of the orange wave noted abovepoint towards such events ~such as polling results, leaders’ debates, andother events!. The evidence in Figures 1 and 2 suggests that several ofthem can be ruled out, however. In the aftermath of his resignation, GillesDuceppe alleged that a bizarre outlier poll startled the Quebec electorateand pushed many into the arms of the NDP. It is not clear to which sur-vey he was referring, but the most likely candidates are ~a! the first pollsplacing the NDP in second place in the province ~Angus Reid, Forum,and Ipsos on April 7!, or ~b! the first polls putting the NDP in front inQuebec ~CROP, Ekos, Forum and Ipsos on April 21!. In neither opinionnor media data do we see that these surveys produced marked shifts. TheApril 7 polls came in the midst of a plateau in NDP support that startedaround April 5, and the April 21 polls come in the middle ~not the begin-ning! of a NDP increase.8 The consistency exhibited by these polls andtheir compatibility with the CES estimates indicate that they were noflukes.9 Furthermore, media content shows that both sets of polls cameduring lulls, not surges, in NDP coverage.

What about the leaders’ debates? None of the media stories on thedebates suggested that they were game-changers and our survey data areconsistent with that impression. The debates of April 12 and 13 do coincidewith an important shift in party support in ROC vote intentions. Forinstance, a t-test reveals that Conservative support in the four days priorto the English debate was significantly lower than support in the four daysafter. But, as is sometimes observed ~Blais et al., 1996, 1999, 2003!, thispossible impact of debates was temporary; by April 17, vote shares in theROC were just about where they were pre-debate.10 In Quebec, on theother hand, the debates appear inconsequential: there is no significant dif-ference in vote intentions for any party during the 8- to 14-day windowspanning either side of the French debate. The fortunes of the NDP didlift off in Quebec starting on April 18, but that is five days after the Frenchdebate. Typically, when debates have had an impact, it has been instanta-

6 6


�

�

�


FIGURE 2Campaign Dynamics of Media Coverage

6 6


�

�

�


neous, especially among debate viewers ~Blais et al., 1996, 1999, 2003!.In 2011, the debates do not line up with the second part of the orange surgein Quebec, even when we examine only debate viewers.

Other campaign events did coincide with the main shifts in aggre-gate vote intentions. As Figure 1 illustrates, the first phase of the NDPsurge within Quebec begins on April 3, corresponding with Jack Lay-ton’s very successful appearance on Quebec’s main talk show Tout lemonde en parle. There is a statistically significant 8-point difference inNDP vote intentions when comparing the week before this appearance tothe following week. This is a first sign that the NDP leader played animportant role in the 2011 election results.

The second phase begins on April 18, when Bloc support starts toslip and NDP support starts to rise again. T-tests confirm that this pointis a significant hinge in the unfolding of the campaign. Just one day ear-lier, on April 17, Gilles Duceppe stood with Parti Québécois leader PaulineMarois on the stage of a PQ convention and confidently asserted that there-election of the Bloc would be followed quickly by the election of thePQ at the provincial level and a victory in a third referendum on sover-eignty. Indeed, for the remainder of the campaign the Bloc continued tofocus on sovereignty and to remind voters that the NDP is a federalistparty, thus producing an explosion in sovereignty coverage over the lasttwo weeks of the campaign. This is readily evident in Figure 3, whichshows the number of times sovereignty keywords were used in election-related articles over time.11 The Bloc message might have been effectiveamong diehard sovereignists in getting out the party’s core vote, but itmay well also have backfired among the more numerous soft nationalistswho often support the Bloc in order to defend the interests of Quebec onthe federal scene. Indeed, Figure 1 suggests that this emphasis on sover-eignty may have pushed a significant number of voters in Quebec awayfrom the Bloc Québécois, not just around April 17, but also during thethird campaign phase beginning on April 26.

Consider the following. Before April 26, the NDP gains were stalledor, worse, collapsing, and some early Bloc deserters may have been trick-ling back to camp. After this date, New Democrats reached significantlyhigher levels, while the Bloc declined a bit further. This date is brack-eted by controversial speeches from sovereignist leaders Jacques Parizeau~April 25! and Gérald Larose ~April 27!. These two speeches were partof a change in strategy aiming to rally the party’s core sovereignist sup-porters back to the Bloc ~Bélanger and Nadeau, 2011: 125!. This tacticmay have caused more harm than good. In fact, the volume of coveragededicated to the sovereignty issue in Figure 3 is linked to the evolutionof NDP support over the last portion of the Quebec campaign, not onlyvisually but also in time-series analysis. Sovereignty coverage spikes fol-lowing the PQ convention, then falls, and then increases again at the time

6 6


�

�

�


of the two speeches, exactly when the size of the NDP ranks swell andshrink.

Of course, we cannot be confident that these events had significantimpacts based on the simple examination of these graphical displays alone.But, as we shall see in individual-level data analyses below, a combina-tion of positive impressions of Layton and the exodus of nationalists fromthe Bloc to the NDP seems to have been critical.

Individual-Level Determinants of Vote Choice and Change

Tables 3 and 4 present the results of multinomial regression analyses ofthe 2011 voting decision. Since the nature of this choice is quite differ-ent in Quebec and the rest of Canada, the two regions are examined sep-arately. The dependent variable is the reported vote from the post-electionwave of the CES. To establish a baseline for comparison, the first set ofcolumns in the tables provides regression results from the 2008 election.All models contain the same independent variables, measured identi-cally. ~The description of all variables is specified in the online appen-dix.! Inspired by previous accounts of vote choice in Canada ~Blais et al.,2002; Gidengil et al., 2012; Nevitte et al., 2000!, the explanatory vari-

FIGURE 3The Evolution of Media Coverage about Quebec Sovereignty

6 6


�

�

�


TAB

LE

3T

heD

eter

min

ants

ofV

ote

Cho

ice

inth

eR

est

ofC

anad

a

Con

serv

ativ

eL

iber

alN

DP

’08

’11

Yea

rD

Cam

D’0

8’1

1Y

ear

DC

amD

’08

’11

Yea

rD

Cam

D

Dem

ogra

phic

sA

tlan

tic

�.1

4**

�.0

5*

.06*

.04

.08*

*.0

1*

Wes

t.1

0**

.11*

*�

.16*

*�

.16*

**

.06*

*.0

4*W

omen

�.0

6**

�.0

3.0

1�

.00

**.0

5**

.03

Und

er35

year

sol

d�

.03

�.0

4�

.05

.03

.09*

*.0

2O

ver

54ye

ars

old

.01

�.0

5**

.02

.07*

*�

.03

�.0

1C

atho

lic

�.0

2�

.01

*.0

4*.0

0*

�.0

2.0

0V

isib

lem

inor

ity

�.2

2**

�.2

4**

**.2

7**

.15*

***

�.0

5.0

9*

Hig

hsc

hool

drop

out

�.0

5.0

3�

.05

�.0

4.1

0**

.00

****

Uni

vers

ity

grad

uate

�.1

2**

�.1

4**

.13*

*.1

1**

�.0

1.0

3U

nion

hous

ehol

d�

.09*

*�

.11*

*�

.04

�.0

2.1

2**

.13*

*

Gen

eral

orie

nta

tion

sA

ccom

mod

ate

Que

bec

�.2

3**

�.2

2**

.17*

*.1

7**

*.0

6**

.05

Con

tine

ntal

ism

.30*

*.2

6**

�.1

6**

�.1

5**

�.1

3**

�.1

2**

Mar

ket

libe

rali

sm.2

5**

.26*

*�

.07*

*.0

2**

�.1

9**

�.2

8**

Mor

altr

adit

iona

lism

.22*

*.2

2**

�.1

0**

�.0

7**

�.1

2**

�.1

5**

Poli

tica

ldi

saff

ecti

on�

.22*

*�

.25*

*.0

6*.1

3**

*.1

7**

.12*

**

Reg

iona

lal

iena

tion

�.0

7**

�.0

4*.0

5**

.02

.03*

.01

Par

tisa

nsh

ipC

onse

rvat

ive

.36*

*.3

3**

**�

.22*

*�

.17*

***

�.1

4**

�.1

6**

Lib

eral

�.2

3**

�.2

4**

.31*

*.2

9**

�.0

9**

�.0

5**

ND

P�

.26*

*�

.41*

***

�.1

3**

�.1

5**

.38*

*.5

6**

**

6 6


�

�

�


Eco

n.p

erce

ptio

ns

Pers

onal

retr

ospe

ct.

.01

.02

**�

.01

.01

**.0

0�

.03*

**

Nat

iona

lre

tros

pect

..0

2*.0

2*�

.02*

.01

*�

.00

�.0

3**

Issu

epo

siti

ons

Pers

onal

taxe

s�

.04*

*�

.03*

*.0

4**

.01

*.0

0.0

2C

orpo

rate

taxe

s�

.01

�.0

5**

**�

.01

.00

.02

.05*

*H

ealt

hsp

endi

ng.0

2�

.01

*�

.06*

*.0

1**

**.0

4*�

.01

**D

efen

cesp

endi

ng.0

4**

.03*

***

�.0

4**

�.0

2�

.00

�.0

1E

nvir

onm

ent

spen

d.�

.04*

*�

.04*

*.0

6**

�.0

2**

�.0

2.0

6**

***

Mor

eim

mig

rati

on�

.01

�.0

1.0

1.0

1�

.01

.00

Lea

der

eval

uat

ion

sS

teph

enH

arpe

r.2

6**

.19*

***

�.1

6**

�.0

7**

**�

.10*

*�

.12*

*S

.D

ion0

M.

Igna

tief

f�

.09*

*�

.09*

*.1

3**

.18*

**

**�

.04*

*�

.09*

*Ja

ckL

ayto

n�

.15*

*�

.10*

**

�.0

1�

.09*

***

.16*

*.1

9**

Year

2008

2011

Num

ber

ofca

ses

1835

1725

Pse

udo

R-s

quar

ed.5

6.5

5

Not

e:In

the

firs

ttw

oco

lum

ns,

entr

ies

are

mar

gina

lef

fect

sof

mul

tino

mia

llo

git

esti

mat

ions

,an

dth

ede

pend

ent

vari

able

isth

epo

st-e

lect

ion

repo

rted

vote

.M

argi

nal

effe

cts

repr

esen

tth

eav

erag

ech

ange

inth

epr

obab

ilit

yof

the

depe

nden

tva

riab

lefo

ral

lon

e-po

int

mov

emen

tsof

the

inde

pend

ent

vari

able

.Con

trar

yto

MN

Lco

effi

cien

ts,

they

dono

tim

ply

are

fere

nce

cate

gory

.The

thir

dco

lum

nsh

ows

the

sign

ific

ance

ofth

edi

ffer

ence

betw

een

the

firs

ttw

oco

lum

ns~y

ear

D!.

The

four

thco

lum

nsh

ows

the

sign

ific

ance

ofth

edi

ffer

ence

betw

een

the

vari

able

’sim

pact

onvo

tein

tent

ions

duri

ngth

efi

rst

two

wee

ksof

the

2011

cam

paig

nan

dit

sim

pact

duri

ngth

ela

sttw

ow

eeks

~cam

paig

nD

!.S

tati

stic

alsi

gnif

ican

ce:

*,.1

0;**

,.0

5

6 6


�

�

�


TAB

LE

4T

heD

eter

min

ants

ofV

ote

Cho

ice

inQ

uebe

c

Blo

cC

onse

rvat

ive

Lib

eral

ND

P

’08

’11

YD

CD

’08

’11

YD

CD

’08

’11

YD

CD

’08

’11

YD

CD

Dem

ogra

phic

sW

omen

.04

.01

�.1

0**

�.0

2.0

4.0

3**

.02

�.0

2U

nder

35ye

ars

old

.03

�.0

4�

.04

.04

�.0

6�

.04

.07*

.03

*O

ver

54ye

ars

old

�.0

3.0

3.0

4.0

0.0

6.0

3�

.07*

*�

.06

Non

-fra

ncop

hone

�.4

3**

�.2

5**

�.0

1.1

3**

**.3

5***

.16*

*.0

9*�

.04

**H

igh

scho

oldr

opou

t�

.09

�.0

3.0

8.1

2**

.03

�.0

0�

.03

�.0

9U

nive

rsit

ygr

adua

te.0

2.0

9**

�.0

5�

.03

.06

.02

�.0

2�

.08*

Uni

onho

useh

old

.03

.05

**�

.01

�.0

1�

.03

�.0

9**

*.0

1.0

4**

Gen

eral

orie

nta

tion

sQ

uebe

cso

vere

ignt

y.2

5**

.23*

*�

.09*

*�

.09*

*�

.15*

*�

.06*

**

�.0

1�

.08*

***

Con

tine

ntal

ism

�.0

3�

.15*

**

.11*

*.1

6**

.01

�.0

4�

.09*

*.0

3*

Mar

ket

libe

rali

sm�

.14*

*.0

2**

*.1

8**

.12*

*�

.08

�.0

1.0

4�

.13*

*M

oral

trad

itio

nali

sm�

.20*

*.0

3**

.13*

*.1

1**

.07

.03

*.0

1�

.17*

*Po

liti

cal

disa

ffec

tion

.12*

*�

.03

*�

.14*

*�

.10*

*.0

1�

.06

.02

.18*

*R

egio

nal

alie

nati

on.0

9**

�.0

1**

.00

�.0

5*�

.02

�.0

2�

.07*

*.0

8**

**

Par

tisa

nsh

ipB

loc

.22*

*.1

7**

�.0

7�

.11*

***

�.1

1**

�.0

2*

�.0

3�

.04

Con

serv

ativ

e�

.22*

*�

.16*

***

.36*

*.4

1**

�.2

2**

�.1

4**

.08

�.1

1**

Lib

eral

�.2

5**

�.0

8***

**�

.10*

*�

.11*

*.3

3**

.23*

*.0

2�

.04

ND

P�

.27*

*�

.13*

*�

.22*

*�

.11*

***

�.1

3**

�.0

9**

**.6

2**

.33*

***

6 6


�

�

�


Eco

n.p

erce

ptio

ns

Pers

onal

retr

ospe

ct.

�.0

0.0

5**

.01

�.0

1.0

2�

.01

�.0

2�

.04

Nat

iona

lre

tros

pect

..0

0.0

0*

.06*

*�

.02

**�

.05*

*.0

0*

�.0

1.0

2

Issu

epo

siti

ons

Pers

onal

taxe

s.0

1.0

1�

.03

�.0

1.0

4.0

3�

.01

�.0

3C

orpo

rate

taxe

s�

.02

�.0

4.0

2�

.02

�.0

2�

.01

.02

.07*

*H

ealt

hsp

endi

ng.0

5�

.01

**�

.03

�.0

5***

�.0

5�

.04*

.03

.09*

*D

efen

cesp

endi

ng�

.06*

*�

.02

.04*

.04*

*.0

4*�

.01

�.0

1�

.01

Env

iron

men

tsp

end.

.08*

*�

.03

**�

.06*

*�

.07*

*.0

1.0

1�

.02

.09*

***

Mor

eim

mig

rati

on.0

3.0

2�

.03

�.0

7**

�.0

2.0

5**

**.0

1�

.01

Lea

der

eval

uat

ion

sG

ille

sD

ucep

pe.2

0**

.16*

*�

.12*

*�

.06*

*�

.06*

�.0

3�

.03

�.0

7S

teph

enH

arpe

r�

.11*

*�

.00

*.2

9**

.20*

**

�.1

1**

�.0

5**

�.0

6**

�.1

5**

S.

Dio

n0M

.Ig

nati

eff

�.0

5.0

1*

�.0

9**

�.0

5**

.20*

*.1

1**

�.0

7**

�.0

6*

Jack

Lay

ton

�.1

1**

�.1

7**

�.0

0�

.11*

***

**�

.06

�.0

5*.1

6**

.34*

*

Year

2008

2011

Num

ber

ofca

ses

580

645

Pse

udo

R-s

quar

ed.6

1.4

7

Not

e:In

the

firs

ttw

oco

lum

ns,

entr

ies

are

mar

gina

lef

fect

sof

mul

tino

mia

llo

git

esti

mat

ions

,an

dth

ede

pend

ent

vari

able

isth

epo

st-e

lect

ion

repo

rted

vote

.M

argi

nal

effe

cts

repr

esen

tth

eav

erag

ech

ange

inth

epr

obab

ilit

yof

the

depe

nden

tva

riab

lefo

ral

lon

e-po

int

mov

emen

tsof

the

inde

pend

ent

vari

able

.C

ontr

ary

toM

NL

coef

fici

ents

,th

eydo

not

impl

ya

refe

renc

eca

tego

ry.T

heth

ird

colu

mn

show

sth

esi

gnif

ican

ceof

the

diff

eren

cebe

twee

nth

efi

rst

two

colu

mns

~yea

rD

!.T

hefo

urth

colu

mn

show

sth

esi

gnif

ican

ceof

the

diff

eren

cebe

twee

nth

eva

riab

le’s

impa

cton

vote

inte

ntio

nsdu

ring

the

firs

ttw

ow

eeks

ofth

e20

11ca

mpa

ign

and

its

impa

ctdu

ring

the

last

two

wee

ks~c

ampa

ign

D!.

Sta

tist

ical

sign

ific

ance

:*,

.10;

**,

.05

6 6


�

�

�


ables cover a comprehensive mixture of short-term and long-term fac-tors: demographics, general orientations, party identification, economicperceptions, issue positions, and leader evaluations.12

In line with the tradition of block-recursive models of voting behav-iour ~Bélanger and Nadeau, 2009; Blais et al., 2002; Gidengil et al., 2012;Miller and Shanks, 1996; Nevitte et al., 2000!, we estimate all models ina step-wise fashion, starting with demographics and adding each sub-sequent block ~separated by headlines in the tables! in turn. All estima-tions rely on multinomial logit, appropriate for a categorical variable suchas vote choice. Not all effects of each model are presented here, how-ever. We only report the effects of a block when that block first enters,controlling for the anterior blocks, but excluding the posterior blocks. Sothe effects of demographic variables come from a model which includesdemographics only; the effects of general orientation variables originatefrom a model that contains general orientations as well as the precedingdemographic block; partisanship’s effects take into account controls forgeneral orientations and demographics; and so on. We do this based onthe assumption that the ordering of blocks is linked to causal precedenceand the stability of variables ~see especially Miller and Shanks, 1996!.The full results are available in the online appendix.

Note that our intention here is not to dissect in detail the effects ofeach factor on vote choice in 2008 and 2011. Future work may well doso; there is much worthy of further analysis in these tables. But our cur-rent aim is to identify the factors that can help explain why the outcomesof these two elections were so different, especially in Quebec. The analy-ses presented below are just those that are most useful for this specificpurpose.

In order to highlight what we see as most critical to our storyline,we make several concessions. First, we present marginal effects ratherthan raw multinomial logistic coefficients. Marginal effects are more intu-itive but also less comparable from one model to the next. ~Issues relatedto marginal effects are discussed in Hanmer and Kalkan, 2013.! Resultsbelow were confirmed using both raw coefficients and other transforma-tions of these coefficients as well. Second, as noted above, we do notpresent all the results from each of the fourteen step-wise multinomialestimations on which our story relies; rather, we present only the mostimportant effects. This means that we do not show instances in which avariable that is significant at an earlier stage in the estimation becomesinsignificant at some later stage.13 We do not regard this as problematicfor our purposes since the central premise of the block-recursive approachis that the full estimate of a given variable’s impact comes before caus-ally posterior variables are added ~which may well subsume that variable’simpact!. But we readily admit that shifts in the significance of variableswill be of real interest for those with objectives different than ours. Impor-

6 6


�

�

�


tantly, then, we have included all the estimations in their entirety in theonline appendix. That appendix also includes the syntax necessary to rep-licate these results from the publicly available 2011 CES file. We expectthat other researchers will be interested in focusing on some of the resultsthere. Here, we focus just on the rather broad picture of change.

Towards that end, we search below for vote determinants where therewas either ~a! a change in the distribution of opinion between 2008 and2011, or ~b! a change in the impact of a variable on vote choice between2008 and 2011. Our assumption is as follows. In order for a variable toaccount for the differing election results, it needs to exhibit either a move-ment in its distribution or a modification of its effect; otherwise, thatfactor cannot explain the massive shift in party fortunes. For instance,evaluations of Jack Layton can only be responsible for the orange surgeif Layton is much more popular in 2011, or if his ratings have muchmore weight on the 2011 voting calculus.

Evidence concerning the change in the effects of the variables ispresented in Tables 3 and 4. The marginal effects reported for the 2008and 2011 elections are relatively straightforward, but we add two tests ofthe significance of differences in effects. The first test focuses on dis-crepancies between the two elections: the significance of the differencesbetween the 2008 and 2011 marginal effects is shown in the third col-umn for each party ~year D!. These estimates are based on a model thatpools the 2008 and 2011 data and includes interactive terms for all 2011items. The second test focuses on divergences across the 2011 cam-paign: capturing the significance of the differences in marginal effectsmeasured at the start and at the end of that campaign ~the columns labelledcampaign D!.14 Because NDP support lifted off very early in the Quebeccampaign and continued to move until very late, we cannot simply splitthe campaign halfway through. In a compromise between the desire tocapture periods before and after the orange surge and the need to have asufficient number of cases, we compare the first two weeks of the cam-paign to the last two weeks of the campaign ~the eight days in the middleare excluded!. Again, this model relies on interactive terms to uncoverthe significance of differences in effects.

Evidence regarding changes in the distribution of each independentvariable is displayed in Table 5. It includes the mean for 2008 and 2011,the significance of the differences in means across the two elections ~yearD!, and the significance of the differences between the means during thecampaign’s beginning and those of the campaign’s end ~campaign D!.15

Finally, where appropriate, we also discuss below the results of regres-sion analyses that model change in vote choice between 2008 and 2011.These specifications compare the supporters of a party that remained loyalto those who left, or alternatively the new supporters of a party to theolder ones. Since they rely on few cases, especially in Quebec, their sta-

6 6


�

�

�


tistical power is limited.16 Nevertheless, these analyses occasionally pin-point individuals who were significantly more likely to desert a particularpolitical party or to rally to another.

Demographics

We capture basic demographics using a series of binary variables ~seeappendix!. Results in Table 3 suggest that the demographic profile ofConservative and Liberal supporters in the ROC did not change muchbetween 2008 and 2011.17 In both elections, the Conservatives had moresuccess among westerners, whites, the less educated and non-unionizedhouseholds. One notable difference between the two elections was the

TABLE 5The Distribution of Variables

ROC Quebec

2008 2011 Year DCampaign

D 2008 2011 Year DCampaign

D

General orientations (-1-1)Accommodate Quebec �.02 .07 **Quebec sovereignty �.13 �.19 *Continentalism .09 .19 ** .05 .13 **Market liberalism �.06 �.05 �.10 �.09Moral traditionalism �.26 �.40 ** �.26 �.39 **Political disaffection .11 .12 .15 .23 **Regional alienation .25 .19 ** * .14 .34 ** **

Partisanship (0/1)Bloc .27 .22 **Conservative .27 .30 ** .08 .08Liberal .20 .21 .15 .14NDP .10 .10 .05 .08 **

Econ. perceptions (-1-1)Personal retrospective �.04 �.03 �.00 �.09 ** **National retrospective �.38 .20 ** �.27 �.02 **

Issue positions (-1-1)Personal taxes �.29 �.33 ** ** �.38 �.40Corporate taxes .26 .26 ** .43 .45Health spending .75 .73 ** .76 .77 *Defence spending .13 .00 ** �.39 �.37Environment spend. .57 .51 ** .71 .60 **More immigration �.09 �.13 ** �.12 �.23 **

Leader evaluations (-1-1)Gilles Duceppe .08 �.01 **Stephen Harper .04 �.02 ** �.09 �.23 ** **S. Dion0M. Ignatieff �.20 �.26 ** �.13 �.16Jack Layton .01 .02 ** .09 .19 ** **

Note: Entries are means.

6 6


�

�

�


weakening of the Conservative gender gap. In 2008, Harper’s party wasless successful among women than men. In 2011, the difference wassmaller and no longer significant. The same pattern occurred with Atlan-tic residents. As for the Liberals, during both 2008 and 2011, they didbetter among easterners, visible minorities and the more educated. WhileAtlantic residents and the elderly were slightly more inclined to vote Lib-eral, the effects were not consistently significant across the two elec-tions. Interestingly, the propensity for Catholics to vote Liberal, alreadyweak in 2008, vanished completely in 2011, marking the first federalelection since 1965 where this relationship is not observed ~Blais, 2005!.

The profile of ROC NDP voters changed a great deal. In 2008, NDPsupporters tended to be from Atlantic Canada and the West, female, youn-ger, less educated and union members. In 2011, the New Democrats madegains in central Canada, among the middle-aged and male voters, as wellas those with pre-university education, resulting in a more equally dis-tributed popularity across various portions of the electorate. Only twosignificant differences persisted; NDP support was higher in the Westthan elsewhere, and higher among union households as well. It is alsoworth mentioning that the NDP obtained the support of a greater propor-tion of visible minorities in 2011, a group that generally favours the Lib-erals ~Blais, 2005!.18

In Quebec ~Table 4!, vote choice was less determined by demograph-ics, and the relationships did not change much between 2008 and 2011despite wholesale changes in voting preferences. During both elections,the Bloc attracted fewer of the non-francophone voters. In 2011, the partymanaged to hold on to more university graduates. As in the ROC, theConservative gender gap of 2008 disappeared in 2011. But during thelast election, two significant relationships emerged: non-francophones andvoters without a high school diploma were more likely to vote Conser-vative. The Liberal base was among non-francophones, but the party alsohad success in 2011 with citizens from non-union households. NDP sup-port was higher among younger voters and non-francophones in 2008,but these relationships became insignificant three years later, as the NDPdiversified the profile of its clientele. Only the more educated were mark-edly less prone to vote NDP in 2011.

Overall, there is not much material in the demographic results toexplain the divergent outcomes of the two most recent federal electionsin both Quebec and the rest of Canada. To make further headway in thisdirection, we need to examine people’s attitudes.

General orientations

Next, we turn to general orientations—or values—with respect to polit-ical, economic, and social life ~controlling for demographics!. We rely

6 6


�

�

�


here on a battery of general orientations used in previous CES analyses~Gidengil et al., 2012; see appendix!.

Outside Quebec, it is difficult to find differences between the twoelections in the attitudinal correlates of support for each political party~see Table 3!. The drivers of Conservative appeal were exactly the samein 2008 and 2011: the party attracted more votes among those with neg-ative views about Quebec’s demands, those who favour closer ties to theUnites States, those who prefer market liberalism, those with traditionalattitudes about women and homosexuals, the politically satisfied, andthose who think their province is treated fairly.

On all but one of the general orientations, the Conservative votersare on one side and the Liberal and NDP voters are together on the otherside. In other words, these orientations, with the exception of market lib-eralism in 2011, do not distinguish Liberals from New Democrats. Inboth electoral contests, NDP and Liberal supporters tended to approvethe accommodation of Quebec, to oppose closer links with Americansand moral traditionalism, and to be politically disaffected.

One difference is noteworthy. Citizens who prefer state interventionto unfettered capitalism split their votes in 2008 between Liberals andNew Democrats, all else being equal, whereas in 2011 this value orien-tation pushed voters toward the NDP. Vote change models also show thatnew NDP voters in 2011 were more likely to be market liberals. Theshift was consequential for the outcome. Simulations reveal that the NDPreaped approximately one percentage point in ROC aggregate vote sharesin 2011 as a result of the change in the impact of this attitude. Had mar-ket liberalism mattered in 2011 to the same extent it did in 2008, theNDP would have ended up with one point less in the ballot boxes, to thebenefit of the Liberals.19 Since little about the NDP platform or its leaderchanged, there are two possible and complementary explanations. Votersperceived the Liberals under Michael Ignatieff as further right than underStéphane Dion, or the ROC voters who moved to the NDP for other rea-sons were more left-leaning than those who stayed put. We expect bothare true to some extent. This NDP gain on market liberalism, however, isbalanced by a loss on political disaffection. In 2008, cynics tended tovote for New Democrats, while three years later they divided their votesbetween Liberals and the NDP. This swing in the impact of disaffectionalso resulted in a one-point gap, but to the advantage of the Liberal party.Bottom line: the dynamics in the effects of general orientations between2008 and 2011 are not responsible for the NDP’s growth in the rest ofCanada.

Table 5 shows that the distribution of ROC opinion moved signifi-cantly between the two elections in the case of four general orientations:accommodating Quebec, continentalism, moral traditionalism and regionalalienation. Two of these movements could have aided the Conservatives;

6 6


�

�

�


people became more favourable to close ties with the US and less criti-cal of their province’s treatment in the federation. Two changes mighthave helped the parties on the left; people became more open regardingQuebec and more liberal in social terms. Simulations that predict the 2011electoral outcome had the attitudes not changed confirm the direction ofthese expectations. Overall, though, the simulations suggest that the NDPonly garnered about one percentage point as a result of the shift in thedistribution of opinions, particularly due to the rise in opposition to moraltraditionalism, a political terrain which the party occupied very effec-tively in 2011.

In Quebec, the partisan landscape of general orientations changeddramatically over the last two elections ~see Table 4!. In 2008, the NDPclientele, small as it was ~12%!, was distinctive on two fronts: more dis-tant towards the US and less alienated regionally. Three years later, theNDP attracted more opponents to market liberalism and to moral conser-vatism, along with the politically disaffected and the alienated. In someways, the NDP stole the Bloc’s not fully sovereignist voters. In 2008,state interventionists, social liberals, cynics and the regionally alienatedwere relatively sympathetic to the BQ. In 2011, those voters inclinedtoward the NDP.20 The only group that the Bloc retained in both elec-tions was the one clearly in favour of sovereignty, and even there theBloc lost ground ~see below!. One new correlate of Bloc support alsosurfaced in 2011: negative views of continentalism. Did this transforma-tion of the BQ and NDP vote structures happen during the inter-electionperiod or during the 2011 campaign? Despite the fact that few differ-ences in the fourth column of Table 4 are significant, the changes in rela-tionships mostly took place during the course of the latest campaign. Thedeterminants of Bloc and NDP vote intentions during the first portion ofthe 2011 campaign resemble to a great extent those of the 2008 votechoice. The relationships evolved during the campaign and ultimatelyreached the values of the reported effects on the 2011 vote choice. Thelack of significant campaign differences in the table is due to the smallsample size of each group in the Quebec campaign split.21

The attitudinal profiles of Conservative and Liberal voters were, incomparison with those of the NDP and Bloc, more stable over time. Inboth elections, Harper’s party had greater success among federalists, con-tinentalists, market liberals, moral traditionalists, non-cynics and non-alienated ~only in 2011!. Each time, the Liberal voters’ only distinctivecharacteristic was being opposed to sovereignty.

Did the Bloc fail in 2011 because Quebeckers turned against thesovereignty agenda? Support for sovereignty was significantly but onlyslightly less popular in 2011 ~40%! than in 2008 ~42%!. Though manyelectors ditched the Bloc in 2011, they nevertheless continued to endorsesovereignty. In fact, over a third of 2011 NDP supporters would like to

6 6


�

�

�


see Quebec become a country ~38%!.22 Yet the key voters were in themiddle. The NDP, like the Action Démocratique du Québec ~ADQ! hadbeen in 2007 ~Bélanger and Nadeau, 2009!, was more successful amongmoderates on both sides of the national issue. It attracted more peoplesomewhat favourable to sovereignty than very favourable, and more peo-ple somewhat opposed to sovereignty than very opposed. Still, all fourgroups supported the NDP more in 2011 than in 2008. If the orange surgedid not embody a rejection of Quebec sovereignty, it nonetheless signalsthe weakening of its power as the structuring dimension in federal elec-toral politics in Quebec. The fact that so many sovereignists were willingto vote for the federalist NDP indicates that the national question was nolonger their overriding motivation.

Was the orange wave a form of protest? The level of political disaf-fection did not change outside Quebec between 2008 and 2011 ~�.01!,while it did rise significantly in Quebec during that period ~�.08 on ascale ranging from �1 to �1! ~see Table 5!. Although the increase incynicism is not huge, it is there, and it occurred only in Quebec. Evenamong survey panelists interviewed in both years, the results are strik-ingly similar. The rise in political discontent in Quebec that we observein the CES data corresponds well with what has been described by manyas a climate of growing distrust in politics following two years of con-tinuous public allegations of corruption in that province ~see, for exam-ple, Bellavance, 2011; Noël, 2011!. That climate may have laid the tablefor significant voter volatility in Quebec prior to the 2011 federal elec-tion. Politically distrustful voters are generally attracted to non-traditionalparty alternatives ~see Bélanger, 2007: 97–101!. Was it a factor behindthe NDP wave in Quebec? Mean cynicism scores increased among allparty groups in Quebec between 2008 and 2011, the largest shift how-ever in cynicism ~�.14! is found among NDP voters ~climbing from .12to .27!. In addition, those Quebecers who defected from their 2008 partychoice were more distrustful than those who remained loyal to their party~�.06!. As well, new NDP voters in 2011 were slightly more cynicalthan NDP voters who had already supported this party three years earlier~�.05!. Simulations, however, indicate that the increase in political dis-affection does not explain the orange surge; it only brought the NDPapproximately one percentage point. The transformation of cynicism’s rela-tionships to vote choice is not much more decisive; the fact that the NDPstole this topic from the Bloc accounts for a two-point swing. Therefore,we cannot uncover definitive evidence that cynicism was the dominantmotivation behind vote switching in favour of the New Democrats. Allwe can say is that Quebec voters in 2011 were embedded in an overallatmosphere of political alienation that favoured discontinuity.

Disaffection in Quebec may also have had something to do witha rejection of the other parties. Apart from the NDP, all parties’ ther-

6 6


�

�

�


mometer scores declined from 2008 to 2011 in the province. The Bloc’sthermometer score decreased by 8 points on a 101-point scale, the Con-servatives lost 5 points, and the Liberals fell by 4 points.23 Were the NDPranks a refuge for people disappointed by these political parties? Actu-ally, all voters saw these parties in a less favourable light. The Conserva-tive supporters in 2011 gave lower ratings to the Liberals and the Bloccompared to those of 2008. The evaluations of the Conservatives and theBQ dropped in the eyes of Liberal voters. And the Bloquistes adjusteddownward their scores of Conservatives and Liberals. The key point is thatNDP voters were not more likely to have diminished thermometer scoresof the opposing parties. Also, those who switched to the NDP between2008 and 2011 did not hold more negative assessments of other partiesthan all other voters. So, again, there are signs of growing political dis-enchantment surrounding the 2011 election in Quebec, but this contextonly loosened the partisan commitments of citizens, it did not necessar-ily lead them massively to the New Democrats.

Although no single general orientation appeared to have been a magicbullet, the orientations overall do constitute a large part of the story forthe orange surge in Quebec. In 2008, the NDP captured only a few valuedimensions among Quebeckers: opposition to closer ties with the US andregional satisfaction. In 2011, they managed to occupy a much widerspan of the value spectrum in the province: mild opposition to sover-eignty, market interventionism, social liberalism, political disaffection andregional alienation. Much of this territory was previously held by theBloc. Citizens modified their opinions about some of these general ori-entations, but the dynamics in the attitudinal profiles resulted in a muchgreater pay-off electorally for New Democrats. Simulations show thisclearly. If the relationships between general orientations and vote choicehad not changed between 2008 and 2011, the NDP would have receivedas much as nine percentage points less in Quebec vote shares ~close to athird of the orange wave!, while attitude change only brought the partyabout three points.24 The success of the NDP in 2011 did indeed par-tially hinge on the party’s capacity to rally many Quebec voters sharingthese ideas. Numerous people became convinced that the NDP could bea good ~or better! vehicle for these ideas.

Partisanship

The next block of variables to enter models in Tables 3 and 4 measurespartisanship through a set of simple binary variables capturing strong orfairly strong identification with each of the major parties during the cam-paign wave, with non-identifiers as the residual category ~see appendix!.Results suggest a good deal of stability in the size of the effects acrosselections, except in the case of the NDP. In the rest of Canada, NDP

6 6


�

�

�


partisans were more likely to vote NDP in 2011 than in 2008. In Que-bec, the effect declined in strength. The different patterns reflect a com-bination of the number of partisans, the pattern of their loyalty, and theirproportion among a party’s voters. The percentage of NDP partisans inthe ROC stayed around 10 per cent, but they were more inclined to voteNDP in 2011 ~86%! than in 2008 ~80%!. In Quebec, while most NDPpartisans cast a ballot for that party in both elections, the NDP’s recentsuccess extended way beyond its partisan core. While 35 per cent of NDPvoters in 2008 were NDP partisans, this proportion fell to 17 per cent in2011. The party managed to capture many Liberal partisans ~31%!, Blocpartisans ~30%!, and Conservative partisans ~12%!.

Our focus so far has been on explaining the NDP surge, and wehave emphasized the factors leading voters to desert the other parties.But now we must also emphasize that we detect very little evidence tosuggest that the short-term shift to the NDP solidified quickly duringand after the campaign in terms of partisanship. Figure 4 shows the dis-tribution of post-election partisanship from 2004 to 2011 in Canada as awhole, using only those who say their attachment is “fairly” or “very”strong.25 The Liberals have been in steady decline since 2006, the Con-servatives on the ascent, and 2011 simply continued these trends. In thecase of the Bloc, the shock of 2011 is indeed observed in partisan affil-

FIGURE 4Party Identification over Time

6 6


�

�

�


iation, with the BQ partisans dropping from 29 to 14 per cent of Quebeck-ers between 2008 and 2011.26 The party that experienced the largest voteswings in 2011, the NDP, moved up in party identifications from 9 to 14per cent of Canadians and from 5 to 14 per cent in Quebec. Consideringthat New Democrats more than tripled their vote share in the provinceand ended up with 43 per cent of votes cast during the 2011 election,their number of identifiers rose somewhat modestly. These results dem-onstrate that party identification is not a simple reflection of vote choice~see also Gidengil et al., 2012; Green et al., 2002; Green and Yoon, 2002!.The Liberals still have as many partisans as the NDP. The BQ hangs onto a base from which to rebound. And the NDP has not yet built a solidcore in Quebec.

How much stability in party identification was there from the cam-paign to the post-election in 2011? Measuring the stability of partisan-ship is fraught with measurement error ~Green and Palmquist, 1990!.Undoubtedly, some survey respondents take the question to be about theirmost recent feelings rather than a long-term attachment. Nonetheless, wecan observe patterns across parties in 2011 that tell us a good deal abouthow a party might fare going into the following election. Looking at loy-alty of partisanship from the campaign-wave interview to the post-electioninterview, we observe that the Conservatives held on to the largest pro-portion of their identifiers ~90%!. The NDP is next with 87 per cent par-tisan stability, while the Liberals and Bloc are down at 80 per cent and77 per cent respectively. On the one hand, the differences indicate thatthe 2011 election shook the faith of a fair number of Liberal and Blocpartisans. On the other hand, the high proportions highlight the stayingpower of party identification. The Bloc saw more partisan evaporationbetween the campaign and the post-election because their partisans weremuch more likely to abandon their identification after supporting a dif-ferent party in the voting booth in 2011. The BQ retained the attach-ments of only 30 per cent of its identifiers that voted for another party.The Conservatives, Liberals and NDP retained the same proportion oftheir partisans who strayed in the voting booth ~60%!. Choosing anotherparty shook the foundation of Bloc partisanship far more than it did forany of the other parties.

Economy and issues

The next two blocks are short-term factors: judgments about the recentevolution of the economy and views on public policy issues. Economicperceptions were infrequently and then only weakly related to vote choice.In the ROC ~Table 3!, people with positive evaluations of the Canadianeconomy were slightly more prone to vote for the incumbent Conserva-tives in both elections. Those with negative economic views leaned toward

6 6


�

�

�


the Liberals in 2008, but toward the NDP in 2011. Since the relation-ships are tiny in both cases, however, this swing was not beneficial forLayton’s party ~nil impact on vote shares!. Moreover, perceptions of thenational economy improved significantly between the two elections. Andthis attitude change helped the Conservatives marginally to the detri-ment of New Democrats ~one percentage point!. In Quebec ~Table 4!,economic perceptions were not at all linked to NDP support in eitheryear. Consequently, the two economic variables simply cannot be the mainexplanations for the upswing in NDP fortunes from 2008 to 2011.

Overall, issue positions were minor determinants of vote choiceacross the two regions in both elections. They were generally outshoneby general orientations, party identification and leader evaluations ~whichwe examine in the next section!. But changes both in the effects and thedistributions of issues were observed between 2008 and 2011 in bothregions. Thus, we have to consider the role of issues in the 2011 surpriseoutcome. In the rest of Canada, positions on four issues predicted Lib-eral support in 2008—personal taxes, spending on health, defence andthe environment—while none did so in 2011. People who demanded extraspending on the environment tended to back the NDP during the lastelection. Also, one issue materialized as a relevant factor in 2011 ~cor-porate taxes, notably correlated to a NDP vote!, as another issue disap-peared from the scene ~spending on health!. According to simulations,these modifications of the vote structure account for a climb of two per-centage points in NDP 2011 ROC shares. Additionally, significant shiftsin the mean issue scores between the two elections are evident in Table 5.ROC respondents became less favourable to increases in personal taxes,spending on defence, spending on the environment, and immigration. Onlythe second of these changes could have helped the NDP. In actuality,simulations show that the party would have done equally well had therebeen no movement in the distribution of the four issues.

While issues did not make a difference outside Quebec, they turnout to be critical elements of the orange surge story in that province. Incomparison to 2008, three issues emerged as significantly associated withvoting for the New Democrats in Quebec in 2011 ~Table 4!. People whofavoured higher corporate taxes, more spending on the environment andmore spending on health care were more likely to cast an NDP ballot.These gains were mostly made at the expense of the Bloc which, threeyears earlier, was attracting supporters of higher environmental and healthexpenditures ~though the latter effect barely misses being significant!.Another way to put this is that the NDP carved off much of the ideolog-ical left of the Bloc, as we saw above for the market and social liberal-ism orientations. Because the distribution of views on the three topicswas quite skewed in Quebec, the issues were very profitable for the NDP.Simulations report that if these three factors had mattered in 2011 the

6 6


�

�

�


same way they did in 2008, the New Democrats would have obtainednine percentage points less in Quebec vote shares ~about a third of theorange wave!. The opinions of Quebeckers regarding two issues evolvedduring this period ~environment and immigration!, but simulations revealthis was of no consequence for the 2011 electoral outcome.

Leadership

The final block is composed of leader evaluations on a 101-point ther-mometer scale recoded to run from �1 to �1 ~see appendix!. Unsurpris-ingly, these variables behave a lot like partisanship does: the more positiveone’s assessment of a party’s leader, the greater the propensity to votefor that party. Inversely, an individual is less likely to mark a ballot infavour of that party when he or she likes the leader of another party. Theimpact of leader evaluations is substantial and rather consistent acrosselections, even for the Liberals who changed leaders. Nevertheless changesover time in both the opinions about the leaders and the effects of theseopinions on voting preferences did occur and we need to consider whetherthe changes were driving forces behind the orange wave.

Figure 5 displays the evolution of the mean ratings of party leadersduring the campaign in both regions. It is clear that Jack Layton wasperceived positively in 2011. In fact, he was the lonely leader on thepositive side of the scale. He just edged out Stephen Harper in the ROC,but he floated high above other leaders in Quebec. Was Layton morepopular in 2011 than in 2008? Outside Quebec, Layton’s overall meanratings in both years were essentially identical, though Harper’s and theLiberal leaders’ scores declined slightly but significantly ~Table 5!. InQuebec, all leaders took a tumble except Layton, who was perceived morefavourably. Furthermore, Layton’s evaluations improved significantly overthe course of the campaign in both regions, rising nearly 10 points betweenthe first two weeks and the last two weeks. In the ROC, these campaignpersonal popularity gains coincide with the upward movement of NDPvote intentions ~starting around April 17!. In Quebec, the first ~April 3!and second ~April 17! steps of the NDP climb took place when the eval-uations of Layton became more positive. Visually, the party’s perfor-mance therefore seems to improve at the same time the NDP leader wasdoing better. Are the movements in leader popularity large enough toexplain the dramatic surge of NDP support? Simulations that comparepredicted vote choice using the leaders’ ratings at the beginning and theend of the 2011 campaign ~holding everything else constant! suggest thatthe NDP harvested two percentage points in the ROC and six points inQuebec as a result of the growing gap in popularity between Jack Laytonand his opponents. This is another important part of the story of the 2011election.

6 6


�

�

�


FIGURE 5Campaign Dynamics of Leader Evaluations

6 6


�

�

�


There are other signs that leadership played a prominent role in the2011 voting calculus as well, particularly in Quebec. Asking respondentsabout the main reason they voted the way they did produces a wide rangeof responses, of course ~see Table 6!, but a disproportionate number ofNDP voters in Quebec name Jack Layton as the top reason ~36%!. Thisproportion is larger than for any other party, in either region. And whilescholars are rightfully hesitant to rely on respondents’ own analyses oftheir behaviour, on this point the regression models in Tables 3 and 4support voters’ assessments. The impact of Layton’s ratings on 2011 NDPsupport in Quebec was the largest observed during the two electionsamong any party. It should be noted that this shift in the weight of lead-ership did not occur during the interlude period between the two elec-tions. Regression models conducted on campaign vote intentions indicatethat the effect of Layton’s evaluations on NDP voting in Quebec increasedduring the 2011 campaign, not before. Some movement along these linesis also present in the ROC, but on a much smaller scale. According tosimulations, because of the change in the effects of Layton’s ratings, theNDP ended up with an extra two points in the ROC and three points inQuebec. Of course, we cannot argue with certainty that people chose tomodify the weights they allocated to leadership considerations. They couldalso have switched preferences for other reasons and this would haveinfluenced the relationships of leader evaluations.

Further analysis aimed at distinguishing voters who switched to theNDP from those who stayed with their 2008 party confirm that favour-able impressions of Jack Layton in 2011 powerfully drove 2008 support-

TABLE 6Most Important Reason for Voting ~%!

Conservative Liberal NDP Bloc

Rest of CanadaLiked the leader 23.5 5.4 24.0 —Liked the local candidate 17.9 36.4 13.4 —Liked the policies 25.6 21.5 33.0 —Did not like the other parties 26.9 26.9 23.3 —Party was doing well in the polls 2.2 0.5 2.6 —Other 3.9 9.2 3.7 —

QuebecLiked the leader 10.2 14.0 35.5 11.6Liked the local candidate 24.5 23.0 4.4 18.0Liked the policies 37.4 30.0 24.6 39.0Did not like the other parties 14.3 17.0 19.3 14.0Party was doing well in the polls 4.8 6.0 3.4 2.9Other 8.8 10.0 12.8 14.5

6 6


�

�

�


ers of all other parties, as well as non-voters, to the NDP. Negative feelingsabout the other leaders were also relevant, but these impacts were party-specific; negative feelings about the leader of the party one voted for in2008, but not the other leaders, pushed voters away from their previouschoice and toward the NDP. This is clearest in the case of Gilles Duceppe.The 2008 Bloc voters with negative evaluations of Duceppe in 2011 werea great deal more likely to jump to the NDP. For example, a former Blocvoter whose view of Layton went up 20 points and whose view of Duceppewent down 20 points was about 20 percentage points more likely to switchto the NDP than a 2008 Bloc voter whose impressions of the leaderswere unchanged.

Conclusion

Why did the NDP make a significant breakthrough ~at least in Quebec!during the 2011 federal election? Our analysis of data from the 2011Canadian Election Study as well as news media content reveals whichexplanations withstand empirical scrutiny. Neither fluke polls, leaders’debates, nor a decline in support for Quebec sovereignty were the driv-ing forces behind the orange wave. Instead, the surge was the result of aset of forces driving supporters of all the other political parties towardthe NDP: a combination of Jack Layton’s leadership and the discoveryby many voters of the NDP’s proximity on some values and issues.

First, multiple sources of evidence indicate that positive ratings ofJack Layton were vital for the NDP victory in Quebec. The leader ofthe New Democrats gained popularity both between the two most recentelections and over the course of the 2011 campaign. At the aggregatelevel, NDP support rose in tandem with an expanding gap between Lay-ton and other leaders during the campaign. Key events in Quebec, suchas Layton’s appearance on the province’s main talk show early in thecampaign and a change in the Bloc Québécois’ campaign strategy, likelycontributed. At the individual level, evaluations of Layton dominatedother variables in comprehensive vote choice models. These evaluationseven gathered strength in explaining vote choice over the campaign. Like-wise, a substantial number of NDP voters in Quebec cited leadershipqualities as the most relevant motivation for their final choice. Andvoters who switched to the NDP in 2011 had particular leadershipassessments.

Second, the NDP managed to occupy some valuable ideological ter-ritory and convince Quebeckers of their credibility in the centre-left ofthe spectrum. Cutting across the main cleavage in Quebec politics, theparty attracted both soft sovereignists and soft federalists. It was alsoable to credibly embody values traditionally owned by the Bloc in Que-

6 6


�

�

�


bec: state interventionism, moral liberalism, political disaffection andregional alienation. Their territory was broader than this, however, as theNDP was also successful in conquering centrist and right-of-centre Lib-erals and Conservatives on the dimension of market liberalism.

Relatedly, questions on budgetary policy reveal three domains thatwere cornerstones of the orange surge. Voting for the New Democrats inQuebec in 2011 was related to support for increased spending on healthcare, more spending on the environment, and higher corporate taxes.Again, many of these gains were made at the expense of the Bloc Québé-cois, which had attracted during the previous election numerous individ-uals sharing these positions.

Interestingly, the more modest advances of the NDP outside Quebecwere also connected to this trio of factors: values, issues and leadership.The ROC electorate moved slightly to the left on the topic of social lib-eralism. The party took hold of the clientele for two relevant issue posi-tions: more spending on the environment and higher corporate taxes.Opinions of Jack Layton became more favourable relative to other lead-ers and also more pertinent for vote choice.

Our conclusions do not stem from a simple analysis of the corre-lates of vote choice in 2011. Rather, they come from models and simu-lations designed to identify differences over time; specifically differencesin the distribution of opinion and differences in the effects of variables,either across two elections ~2008 and 2011! or during the 2011 cam-paign. We believe this is the most appropriate strategy to account forchanges in the electoral fortunes of political parties.27

Earthquakes of the magnitude that shook the foundations of the Cana-dian party system in 2011 do not occur frequently. One could argue thatthe leadership and platform of the NDP have changed little in the lastfew elections, yet only in 2011 was there a major breakthrough. Why didthe orange surge take place then? Our results highlight which types ofindividuals were more likely to change their mind and vote differently,but they do not provide evidence of causality nor of the fundamentalreasons why voters switched preferences. Only speculation is possible.Perhaps part of the story lies with growing political disaffection and cyn-icism in Quebec following years of allegations of political corruption atthe municipal and provincial levels. These sentiments did not directlylead citizens toward the New Democrats, but they created space for weak-ened partisan ties and thus electoral volatility. In so doing, they may haveindirectly contributed to the success of the NDP in the province. Thecoincidence of a few campaigning coups with a sequence of polling suc-cesses could have played a role as well, by creating momentum and rais-ing the electoral expectations for the party. A strategic blunder probablyalso mattered. The decision of Bloc leaders partway through the cam-paign to focus on Quebec sovereignty rather than their usual mantra—

6 6


�

�

�


the defence of Quebec’s interests—seems to have amplified the NDP’ssuccess. Finally, we suspect that the Bloc’s success over the previous twodecades indirectly and partially paved the road for New Democrats inthe province. The BQ demonstrated time and time again that Quebeckersare not required to vote for a government on the federal scene. In a sense,the NDP faced a lower bar in Quebec than elsewhere; it did not have torepresent a credible government alternative, it only needed to convincevoters that the party could champion Quebec’s interests in Ottawa.

What do these results imply for the future fortunes of the NDP, espe-cially in Quebec? Of course, one of the main draws of the party in 2011,Jack Layton, has passed away, and his loss may be devastating for theparty. The selection of Thomas Mulcair as NDP leader could lessen theblow, at least in Quebec where he is very popular among citizens of allpolitical stripes. It remains true, however, that New Democrats are facedwith holding together a Quebec rainbow coalition composed notably ofsovereignists and federalists, along with both left- and right-leaning vot-ers. As other parties who suddenly find themselves supported by con-trasting clienteles often discover, the NDP will be forced into a delicatebalancing act. Pleasing all sides of the “question nationale” is invariablytroublesome, particularly when the party has to deal with sustaining itsappeal in English Canada as well. And since the NDP’s support has notyet crystallized as genuine partisan attachment in Quebec, their task willbe challenging indeed.

The significant shift of Quebec’s voting preferences in the 2011 elec-tion may thus mark a paradigmatic change in the Canadian electoral land-scape, or it may reflect a perfect but fleeting electoral storm. This articlehas attempted to identify the combination of both long- and short-termfactors behind this unexpected outcome. As for the durability of the result-ing change, however, only time will tell. The NDP surge during, notbefore, the 2011 campaign, suggests that we will have to wait for thenext general election to answer this question

Supplementary Material

To view supplementary material for this article, please visit: http:00dx.doi.org0S0008423913000875

Notes

1 The Liberal party had always finished either first or second in their existence before2011.

2 Data in Figures 1 and 5 are smoothed using five-day weighted averages. For the ROC,we use the following weights: t�5, t-1�4, t-2�3, t-3�2, t-4�1. For Quebec, wesmooth somewhat more due to a smaller sample size: t�5, t-1�4, t-2�4, t-3�3,

6 6


�

�

�


t-4�3. Media data in Figures 2 and 3 also rely on five-day weighted averages, asfollows: t�5, t-1�4, t-2�3, t-3�2, t-4�1. Note that smoothing gives more weightto current and most recent days, so as to properly capture the timing of shifts. In allcases, weighting was based on the need to both take advantage of cases across mul-tiple days ~to get more accurate estimates!, and to smooth data series to make chartsmore legible.

3 The 2011 CES was fielded by the Institute for Social Research at York University,under the supervision of David Northrup. The campaign response rate was 41 percent. The RDD sample is only weighted based on provincial population estimatesand the number of household members. Questionnaires, data and technical documen-tation can be obtained from ces-eec.org.

4 The 2008 reported vote was captured during the 2011 campaign wave, while the 2011reported vote was measured during the 2011 post-election wave. Despite the fact thatthe first question refers to an event which took place over two-and-a-half years ear-lier, the distribution of responses resembles the official results of that electoral con-test. The following numbers can be compared to those in the final row of Table 1 for2008: 41 per cent for the Conservatives, 30 per cent for the Liberals, 14 per cent forthe NDP, 11 per cent for the Bloc, and 5 per cent for the Greens.

5 Among those who did not vote in 2008 but did so in 2011, 41 per cent chose the NDP.6 Calgary Herald, Montreal Gazette, National Post, Regina Leader-Post, Globe and Mail,

Toronto Star, Vancouver Sun, Winnipeg Free Press, Le Devoir, and La Presse.7 Lexicoder is a Java-based content analysis utility freely available for academic use at

lexicoder.com.8 We were very careful to use an appropriate smoothing strategy before making this

determination. We verified that our rolling-window smoother produced a very simi-lar tracking of vote intentions as other smoothing methods. The most technically sound,a state-space ~Kalman! filter, with subsequent smoothing, as recommended by Greenand Yoon ~2002!, produces a tracking almost identical to our simpler version depictedin Figure 1.

9 It is possible that reaching first place in Quebec vote intention polls helped the NDPmake even more gains among the electorate, for instance, by affecting expectationsabout the final electoral outcome. But we cannot examine this issue; a separate ded-icated study would be required.

10 The NDP vote share in the ROC seems to be moving upwards as of April 17, but itseems unlikely that this can be credited to the April 12 debate, since the NDP voteshare was totally flat through the debate period.

11 The keywords used here are referendum, référendum, sovereignty, séparation, sou-veraineté, souverain, and souverainiste, along with the names of the three party lead-ers and activists most active in talking about these issues in the news: Pauline Marois,Jacques Parizeau and Gérald Larose. Both the Quebec newspapers and the entire setshow exactly the same story.

12 The cases with missing data were excluded listwise. To minimize the number of exclu-sions, we avoided variables with too many missing cases ~such as those originatingfrom the mailback and web waves, along with income, which always elicits a greatnumber of refusals!.

13 For instance, in the 2011 Quebec NDP final model containing all blocks, voting forthat party remains significantly related to university education, Quebec sovereignty,political disaffection, NDP partisanship, health spending, environment spending,Harper ratings and Layton evaluations, but other variables are washed out by theadditional controls.

14 To avoid overloading the table, the campaign marginal effects are not reported; onlythe significance in gaps between campaign marginal effects is presented. The fullresults are available from the authors upon request.

6 6


�

�

�


15 The campaign means are not reported but are available from the authors upon request.16 The situation is even worse in the case of the panel component of the survey. For this

reason, we cannot build vote change models based solely on panel respondents.17 Note that we do not necessarily attribute causal status to these demographic variables.18 This resulted in a decline of the corresponding effect in the 2011 Liberal column, but

it remains substantial and significant.19 The simulations compare the actual probability of voting NDP in 2011 with the prob-

ability where the impact of market liberalism in 2011 is the same as the impact in2008.

20 Note that the NDP also had success among centrist and right-of-centre Liberals andConservatives.

21 Recall that here we compare respondents interviewed during the first two weeks ofthe campaign to those surveyed during the last two weeks.

22 In 2008, the proportion was 29 per cent.23 The NDP gained four points.24 The two most important orientations were moral traditionalism and regional alienation.25 The following numbers differ slightly from those reported in Table 5 because the

latter come from the campaign wave.26 This amounts to a drop from 7 to 4 per cent nationally.27 Such a strategy was previously employed to explain differing levels of public support

across three referendums on electoral reform ~Fournier et al., 2011!.

References

Aarts, Kees, André Blais and Hermann Schmitt. 2011. Political Leaders and DemocraticElections. Oxford: Oxford University Press.

Andrew, Blake, Lori Young and Stuart Soroka. 2008. “Back to the Future: Press Coverageof the 2008 Canadian Election Campaign Strikes both Familiar and Unfamiliar Tones.”Policy Options ~November!: 79–84.

Bean, Clive and Anthony Mughan. 1989. “Leadership Effects in Parliamentary Electionsin Australia and Britain.” American Political Science Review 83: 1165–79.

Bélanger, Éric. 2007. “Third Party Success in Canada.” In Canadian Parties in Transi-tion, ed. Alain-G. Gagnon and A. Brian Tanguay. Peterborough: Broadview Press.

Bélanger, Éric and Richard Nadeau. 2009. Le comportement électoral des Québécois. Mon-tréal: Les Presses de l’Université de Montréal.

Bélanger, Éric, and Richard Nadeau. 2011. “The Bloc Québécois: Capsized by the OrangeWave.” In The Canadian Federal Election of 2011, ed. Jon H. Pammett and Christo-pher Dornan. Toronto: Dundurn Press.

Bellavance, Joël-Denis. 2011. “Méfiance extrême envers les élus.” La Presse, October 19,A6.

Blais, André. 2005. “Accounting for the Electoral Success of the Liberal Party in Can-ada.” Canadian Journal of Political Science 38: 821–40.

Blais, André and Martin Boyer. 1996. “Assessing the Impact of Televised Debates: TheCase of the 1988 Canadian Election.” British Journal of Political Science 26: 143–64.

Blais, André, Elisabeth Gidengil, Richard Nadeau and Neil Nevitte. 2002. Anatomy of aLiberal Victory: Making Sense of the Vote in the 2000 Canadian Election. Peterbor-ough: Broadview Press.

Blais, André, Elisabeth Gidengil, Richard Nadeau and Neil Nevitte. 2003. “CampaignDynamics in the 2000 Canadian Election: How the Leader Debates Salvaged the Con-servative Party.” PS: Political Science & Politics 36: 45–50.

Blais, André, Richard Nadeau, Elisabeth Gidengil and Neil Nevitte. 1999. “CampaignDynamics in the 1997 Canadian Election.” Canadian Public Policy 25: 197–205.

6 6


�

�

�


Daku, Mark, Adam Mahon, Stuart Soroka and Lori Young. 2009. “Media Content andElection Campaigns: 2008 in Comparative Perspective.” Paper presented at the annualmeeting of the Canadian Political Science Association, Ottawa.

Fournier, Patrick, Henk van der Kolk, André Blais, R. Kenneth Carty and Jonathan Rose.2011. When Citizens Decide: Lessons from Citizen Assemblies on Electoral Reform.Oxford: Oxford University Press.

Geer, John. 1988. “The Effects of Presidential Debates on the Electorate’s Preferences forCanadidates.” American Politics Research 16: 486–501.

Gidengil, Elisabeth, Neil Neville, André Blais, Joanna Everitt and Patrick Fournier. 2012.Dominance and Decline: Making Sense of Recent Canadian Elections. Toronto: Uni-versity of Toronto Press.

Green, Donald P. and Brad Palmquist. 1990. “Of Artifacts and Partisan Stability.” Ameri-can Journal of Political Science 34: 872–902.

Green, Donald P. and David H. Yoon. 2002. “Reconciling Individual and Aggregate Evi-dence Concerning Partisan Stability: Applying Time-Series Models to Panel SurveyData.” Political Analysis 10: 1–24.

Green, Donald P., Brad Palmquist and Eric Schickler. 2002. Partisan Hearts and Minds.New Haven: Yale University Press.

Hanmer, Michael J. and Kerem Ozan Kalkan. 2013. “Behind the Curve: Clarifying theBest Approach to Calculating Predicted Probabilities and Marginal Effects from Lim-ited Dependent Variable Models.” American Journal of Political Science 57: 263–77.

Hillygus, D. Sunshine and Todd G. Shields. 2009. The Persuadable Voter: Wedge Issues inPresidential Campaigns. Princeton: Princeton University Press.

Iyengar, Shanto and Donald R. Kinder. 1987. News That Matters. Chicago: University ofChicago Press.

Iyengar, Shanto, Donald R. Kinder, Mark D. Peters and Jon A. Krosnick. 1984. “The Eve-ning News and Presidential Evaluations.” Journal of Personality and Social Psychol-ogy 46: 778–87.

Johnston, Richard, André Blais, Henry E. Brady and Jean Crête. 1992. Letting the PeopleDecide: Dynamics of a Canadian Election. Montreal: McGill-Queen’s University Press.

Krosnick, Jon A. and Laura A. Brannon. 1993. “The Impact of the Gulf War on the Ingre-dients of Presidential Evaluations: Multidimensional Effects of Political Involve-ment.” American Political Science Review 87: 963–75.

Krosnick, Jon A. and Donald R. Kinder. 1990. “Altering the Foundations of Support forthe President through Priming.” American Political Science Review 84: 497–512.

McAllister, Ian. 1996. “Leaders.” In Comparing Democracies: Elections and Voting inGlobal Perspective, ed. L. LeDuc, R. Niemi and P. Norris. Thousand Oaks CA: SagePublications.

Mendelsohn, Matthew. 1996. “The Media and Interpersonal Communications: The Prim-ing of Issues, Leaders, and Party Identification.” Journal of Politics 58: 112–25.

Miller, Joanne M. and Jon A. Krosnick. 2000. “News Media Impact on the Ingredients ofPresidential Evaluations: Politically Knowledgeable Citizens are Guided by a TrustedSource.” American Journal of Political Science 44: 301–15.

Miller, Warren E. and J. Merrill Shanks. 1996. The New American Voter. Cambridge MA:Harvard University Press.

Nevitte, Neil, André Blais, Elisabeth Gidengil and Richard Nadeau. 2000. Unsteady State:The 1997 Canadian Federal Election. Toronto: Oxford University Press.

Noël, Alain. 2011. “Zone inondable.” Policy Options ~August!: 72.Zaller, John R. 1992. The Nature and Origins of Mass Opinion. Cambridge: Cambridge

University Press.Zaller, John R. 1996. “The Myth of Massive Media Impact Revived: New Support for a

Discredited Idea.” In Political Persuasion and Attitude Change, ed. D.C. Mutz, P.M.Sniderman and R.A. Brody. Ann Arbor MI: University of Michigan Press.

6 6


�

�

�


no job namepatrick-fournier.com › d › fournier2013.pdf · patrick fournier, centre for the...

Documents