Download Implicit attitudes and inconsistent issue voting: the example

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Opinion poll wikipedia , lookup

Political spectrum wikipedia , lookup

Transcript
Implicit attitudes and inconsistent issue voting:
the example of the radical right vote*
David Johann (University of Vienna)
Sylvia Kritzinger (University of Vienna)
Colin Tucker Smith (University of Florida)
Be aware: 1st DRAFT
Please do not cite without author’s permission
Abstract
…
Keywords
Correct voting, Consistent voting, Explicit versus Implicit Attitudes, SC-IAT, Austria
*This work is supported by the Austrian National Election Study (AUTNES), a National Research
Network (NFN) sponsored by the Austrian Science Fund (FWF) [S10902-G11]. We thank Johann
Gründl and Lena Raffetseder for their helpful support.
1
Introduction
In the area of issue voting it is often assumed that voters should choose the party which is closest to
them in terms of issue preferences. Hence, voters are supposed to act rationally choosing the party
with the highest utilities. It follows that the issue distance between a party’s policy position and the
one of the voter acts as predictor of electoral preferences (e.g. Downs, 1957; Grofman and Merrill,
1999). If voters’ choose the party that corresponds best with their attitudes and preferences this is
sometimes labelled as “correct voting” (e.g. Lau and Redlawsk, 1997; Lau et al., 2008; Rosema and
de Vries, 2011; Kraft, 2012; Johann and Glantschnigg, 2013) and sometimes as “consistent voting”
(e.g. Baum and Jamison, 2006; Schmitt-Beck and Kraft, 2014).1 Consistent voting is widely
considered a normatively desirable behaviour (e.g. Lau et al., 2008; Johann and Glantschnigg, 2013;
Schmitt-Beck and Kraft, 2014). For example, Lau et al. argue: “[V]otes freely given are meaningless
unless they accurately reflect a citizen’s true preferences” (2008: 396).
Scholarly work has however shown that not all citizens vote consistently with their policy
preferences. In contrast, a considerable amount of voters rather vote for parties that do not best
represent their policy preferences (e.g. Rosema and de Vries, 2011; Kraft, 2012; Wagner et al.,
2012; Lau et al., 2013; Johann and Glantschnigg, 2013). A large number of studies have therefore
tried to identify factors that help (or hinder) the voters to choose the party that best represents
their policy preferences. Factors researchers analysed are on the one hand, voters’ cognitive
capacity, voters’ motivation and voter heuristics (e.g. Lau et al., 2008; Kraft, 2012; Popa, 2012;
Johann and Glantschnigg, 2013), and on the other hand, information effects such as political
conversation, campaign contacts and news media coverage (Baum and Jamison, 2006; Johann and
Glantschnigg, 2013; Schmitt-Beck and Kraft, 2014). What all these studies have in common is that
they first and foremost examine consistent voting as a result of an individual’s deliberative/explicit
cognitive components that underlie its decision-making processes. In other words, they only focus
on voters’ capacity and motivation to detect the party which best suits their individual policy
preferences.
When analysing citizens’ political attitudes and behaviour, recent research has however also started
to consider implicit processes and attitudes as potential influencing factors (e.g. Glaser and Finn,
1
In the following we use the term consistent voting.
2
2013). Unlike explicit attitudes these implicit attitudes derive from possibly unconscious cognitive
processes of the individual: they occur without individuals being able to modify them through
processes of conscious reflexion (e.g. Plischke, 2012). The question then arises whether implicit
affective processes influence voters’ party choice the same way as the aforementioned explicit
attitudes. As the direction and strength of an implicit attitude is relatively difficult for an individual
to consciously control, one might assume the effects of implicit attitudes to differ from those of
explicit attitudes. Particularly, in regards to inconsistent voting implicit attitudes might provide
additional explanatory power.
Hence, in this paper we aim to take implicit attitudes in addition to explicit attitudes into account
when exploring consistent voting. In particular, we are interested in analysing potential differences
across types of attitudes on consistent voting. We assume that due to the lower control
respondents can exert on implicit attitudes, they have unique effects on consistent voting.
Moreover, we expect that implicit attitudes affect inconsistent voting much stronger than explicit
attitudes. As such, we strive to provide an additional explanation why voters do not vote
consistently with their policy preferences.
We use data from a national election study, namely the Austrian national election study (AUTNES).
In an online panel survey capturing campaign effects, we conducted an experiment measuring
implicit attitude: the Single Category Implicit Association Test (SC-IAT). The experiment is designed
to capture implicit attitudes towards one of the political parties running in the campaign: the
radical-right Austrian Freedom Party, FPÖ. Hence, we study the question of (in)consistent issue
voting and implicit attitudes using the example of right-wing voting. In addition to the measure
capturing implicit attitudes, AUTNES also included several questions on explicit attitudes (e.g. party
identification) as well as issue preferences.
In the remainder of the paper, we first consider consistent and inconsistent voting before
describing implicit attitudes and our hypotheses in more detail. After presenting our data and
methods, we show evidence that implicit attitudes overall exert an unique effect on vote choice,
but most importantly, they are an important factor in explaining why individuals choose a party (in
our case a radical-right party) that does not represent their party preferences best. Implicit
attitudes influence an individual’s behaviour which – from a cost-benefit perspective – conflict with
3
her policy preferences. We conclude by pointing out new avenues for experimental research on
citizens’ behaviour.
Vote Choice Models on Consistent Voting and Explicit and Implicit Attitudes
Based on the assumption that voters act rationally, citizens’ voting decisions are considered to be
based on their issue stances (e.g. Downs, 1957; Grofmann and Merrill, 1999). Citizens should be
able to select between political actors and make a choice that is consistent with their own views,
attitudes and preferences, otherwise they may fail to choose actors that represent their interests
best (e.g. Lau and Redlawsk, 1997). This type of behaviour is commonly known as consistent voting
(e.g. Baum and Jamison, 2006; Schmitt-Beck and Kraft, 2014).
However, citizens often choose a party that does not represent their policy preferences (e.g. Lau et
al., 2013; Johann and Glantschnigg, 2013), sometimes even without realizing it (e.g. Lenz, 2012).
First, indicators such as the individuals’ education degree, their level of political knowledge and
interest in politics are considered important assets which allow them to detect the party which best
suits their individual policy preferences (e.g. Lau et al., 2008; Kraft, 2012; Popa, 2012). At the same
time, empirical research has shown that voters are often quite uninformed and unknowledgeable
about parties’ policy positions. In these circumstances, voters in general use heuristics (substitutes
for information and knowledge) to find their ideal party (and to vote for it). Party identification is
often considered as such a characteristic (e.g. Popkin 1991; Popa, 2012). The party simply supplies
cues by which the individual evaluates elements of politics, and hence, policy orientations
(preferences) are considered effects of party identification (Rohrschneider and Whitefield, 2012;
Smith et al., 2012).
From this perspective consistent voting is the outcome of citizens’ deliberate responses which are
controlled and reflective. This is a common approach in survey research interested in opinions and
behaviours of individuals. But what if consistent voting is driven by parameters other than those
explicit and deliberate ones?
Overall, research in the area of social psychology assumes that human social cognition follows a
dual process: on the one hand, there are explicit attitudes that determine the behaviour of humans,
on the other hand, humans possess also implicit attitudes. The latter humans cannot control to the
4
same extent as the explicit ones: As they are not (or less) aware of them (Bargh, 1994), humans
have greater difficulties accessing them that easily – if at all (e.g. Banaji, Lemm and Carpenter,
2001).2 The exposure to a target automatically releases a reaction, which can be considered an
implicit attitude towards the target (e.g. Lodge and Taber, 2005).3 In contrast, explicit attitudes
stem from a deliberative process of the information on the target and are put consciously in
contrast with previous beliefs and attitudes.
While recent research has thus acknowledged that implicit processes impacts citizens’ behaviours in
addition to citizens’ explicit cognitive processes, research in political science has widely neglected
the effect of spontaneous (affective) reactions on citizens’ behaviour in general and on vote choice
model in particular. Recent research in social psychology has however impressively shown that
these implicit attitudes can predict political behaviour (e.g. Jost et al., 2009; e.g. Arcuri et al., 2008),
and so, the relationship between political outcomes and implicit processes has become a research
focus in this area (Ksiazkiewicz and Hedrick, 2013).
Based on this research, in this paper we merge the different research traditions from the area of
political science and social psychology, and examine both explicit and implicit attitudes in explaining
consistent voting behaviour. The overall question we follow in this paper is the extent to which both
explicit and implicit attitudes impact citizens’ consistent voting behaviour. While we already know
how explicit attitudes affect consistent voting patterns (see before), we still lack knowledge on the
impact of implicit attitudes, and in particular, whether they can exert a unique effect. As implicit
attitudes cannot be consciously modified (Plischke, 2012), they may provide additional explanation
on citizens’ voting patterns – in particular, on those which do not reflect on their policy preferences,
namely inconsistent voting. Furthermore, implicit associations may set the respondents on the
wrong track because implicit associations can bias the interpretation of information and/or lead to
selective exposure to information in a way that is consistent with the implicit attitudes (Gawronski
et al., 2015) but inconsistent with the explicit attitudes. Hence, the more fine-grained research
question of this paper is whether and how the two types of attitudes differ in their effect on
consistent voting patterns.
2
Some research even suggests that explicit and implicit attitudes stem from different neural pathways (Stanley,
Phelps, and Banaji, 2008)
3
For a thorough review on this literature see Ksiazkiewicz and Hedrick (2013).
5
We study the following hypotheses. We expect that both types of attitudes have an effect on
citizens’ voting patterns. So far research has assumed that all factors influencing consistent voting
do conversely also provide explanation for inconsistent voting.4 However, one might also assume
that different mechanisms explain consistent and inconsistent voting patterns. Therefore, we study
the effect of both explicit and implicit attitudes separately for consistent and inconsistent voting.
Starting with explicit attitudes we follow Campbell et al. (1954) who state that party identification
acts as an explicit connection to a party. Party identification helps to reduce the policy distances
between voters and party as the latter ones “[…] formulate issues based on the interests of its social
base” (Rohrschneider and Whitefield, 2012: 1322). As such, party identification can be considered
an important heuristic to predict vote choice (e.g. Bartels 2000).
H1a: The stronger a voter’s party identification the more likely she votes consistently with
her party preferences.
H1b: PID has no effect on the likelihood to vote inconsistently.
When exploring implicit attitudes we overall do also expect a positive association between attitude
and behaviour. But with regards to inconsistent voting we expect that implicit attitudes release a
different behaviour. Hence, we state the following:
H2: The more positive a voter’s implicit association with a specific party, the more likely she
votes for the party – independent of whether she should vote for the party in terms of
attitudes congruence.
As in socially sensitive areas implicit attitudes tend to predict behaviour better than explicit
attitudes (Greenwald et al., 2009; Ksiazkiewicz and Hedrick, 2013),5 in order to test our hypotheses
we focus on right-wing voting. Voting for radical right parties is in general considered a socially
undesirable behaviour (Aichholzer et al., 2014) where implicit attitudes might provide additional
leverage to understand citizens’ behaviour.
4
In the sense, that voters formulating positive explicit/implicit attitudes towards a party are more likely to vote
consistently with their policy preferences, while voters formulating negative explicit/implicit attitudes towards a
party are less likely to vote consistently with their policy preferences.
5
In socially less sensitive areas implicit measures might be less important in explaining behaviour. There, the
explicit measures are capturing the individual’s behaviour to a large extent (e.g. Roccato and Zogmaister, 2010).
6
Data and Methods
Data
To analyze our research question we use the “AUTNES TV-Debates Panel Study 2013” (Kritzinger et
al., 2014a, 2014b), a four-wave online panel study conducted during (waves 1 to 3) and after (wave
4) the Austrian national election campaign 2013.6 The AUTNES TV-Debates Panel Study 2013 is
based on a quota sample which represents some of the main characteristics of the Austrian
population (age, gender, region).7 Only respondents with access to the internet were surveyed.
N=1414 respondents (of N=3084 respondents who took part in wave 1) participated in all four
waves (see Kritzinger et al., 2014b).
Dependent variable
Our dependent variable indicates (1) whether or not the respondents voted for the FPÖ and, in case
they voted for the FPÖ, (2) whether or not their vote choice was “attitude consistent”. “Attitude
consistent” means that the voters voted for the party that is closest to them in terms of policy
preferences, “attitude inconsistent” in turn means that the voters voted for a party that is not
closest to them in terms of policy preferences. Whether or not a respondents’ vote choice
corresponds with their policy preferences is evaluated by employing the left-right proximity. The
respondents’ left-right position is measured using the classic 11-point-scale.8 In order to measure
the parties’ left-right positions we draw on the AUTNES candidate survey (Müller et al., 2015) and
use the mean rating of the candidates of the respective parties. We consider the parties in
parliament (in addition to the FPÖ, the Social Democratic Party of Austria (SPÖ), the Austrian
People's Party (ÖVP), The Greens, The New Austria (NEOS) and the Team Stronach) as well as the
Alliance for the Future of Austria (BZÖ) that failed to reach the four-per cent threshold to enter
parliament.
6
Field time wave 1: August 14-26, 2013; Field time wave 2: September 9-12, 2013 (Split 1), and September 17-19,
2013 (Split 2); Field time wave 3: September 26-29, 2013; Field time wave 4: September 30-October 6, 2013 (see
Kritzinger et al., 2014a; Kritzinger et al., 2014b).
7
The sample is composed of 47.6% men and 52.4% women (population: 48.8% men, 51.2% women). Young
citizens (20 to 29-years-olds: 20.3% [population: 14.9%]; 30 to 39-years-olds: 18.4% [population: 15.1%]) are
overrepresented, older citizens (70+: 3.3% [population: 15.2%]) are underrepresented (see Kritzinger et al., 2014a,
2014b).
8
“In politics people often talk about ‘left’ and ‘right’. Where would you place yourself on a scale from 0 to 10,
where 0 means ‘left’ and 10 means ‘right’?”
7
Table 1 indicates how the dependent variable is distributed. 25% of the respondents declared to
have voted for the FPÖ,9 75% declared to not have voted for the FPÖ. Approximately half of the
respondents who voted for the FPÖ casted an inconsistent vote, meaning they chose the party not
closest to them in terms of left-right proximity.
Table 1: Dependent variable
Did not vote for the FPÖ
Did vote for the FPÖ & vote choice “inconsistent” (FPÖ not closest party)
Did vote for the FPÖ & vote choice “consistent” (FPÖ closest party)
Total
in %
75.00
(501)
12.57
(84)
12.43
(83)
100.00
(668)
Main co-variates
Explicit attitudes towards the FPÖ: Party identification (PID)
In order to measure the explicit attitudes towards the FPÖ we draw on four questions that indicate
whether and to what extent the respondents feel close to a particular political party (for more
information on the question wording, see Appendix B). Combining these four questions we build a
4-point-scale ranging from 0=“I do not feel close to the FPÖ” to 1=“I feel very close to the FPÖ”
(Mean=0.09, SD=0.23).
Implicit attitudes towards the FPÖ: SC-IAT
Unlike explicit attitudes, implicit attitudes cannot be measured by directly addressing these
attitudes (e.g. Gawronski et al., 2015; Plischke, 2012). Hence, indirect measures such as the
“Implicit Association Test” (IAT) (Greenwald et al., 1998) or the “Affect Misattribution Procedure”
(AMP) (Payne et al., 2005) have been widely used in the literature to capture implicit attitudes.
In order to measure implicit attitudes towards the FPÖ we use a modification of the classical IAT,
namely the “Single Category Implicit Association Test” (SC-IAT) (Karpinski & Steinman, 2006; for
more information on the AUTNES’s SC-IAT, see Kritzinger et al., 2014b, 2014c). The SC-IAT follows
9
Actual vote share of the FPÖ was 20.5 per cent.
8
the same procedure as the original IAT, but only one target category (in this case the FPÖ) and two
attribute categories are needed instead of two contrasting target categories and two attribute
categories as in a classical IAT design (Karpinski and Steinman, 2006; Kritzinger et al., 2014b). The
design of the AUTNES’s SC-IAT is based on implicit association tests that are provided by Project
Implicit10. Respondents in the AUTNES’s SC-IAT were redirected to a separate webpage after
completing the regular questionnaire of wave 3. In total N=1258 respondents completed the SC-IAT
(see Kritzinger et al., 2014a).
The SC-IAT scores are based on reaction times (see e.g. Karpinski and Steinman, 2006; Plischke,
2012): The respondents were requested to use the keys “E” and “I” of their keyboard in order to
categorize specific words or images that were displayed on their screen as fast as they could while
making as few mistakes as possible (see Kritzinger et al., 2014b, 2014c).
The AUTNES-SC-IAT consisted of three blocks (see Table 1). In the first block the respondents were
confronted with negatively (Pain, Torture, Fear, Hate, Murder) and positively (Love, Luck, Joy, Fun,
Peace) connoted words. These words had to be allocated to the attribute categories “Bad” (E-key)
and “Good” (I-key). In blocks 2 and 3 the respondents were additionally confronted with stimuli of
the FPÖ (FPÖ, Die FPÖ, Die Blauen, Die Freiheitlichen). In Block 2 the positively and negatively
connoted words as well as the stimuli of the FPÖ had to be allocated to the categories “Bad” (E-key)
and “Good or FPÖ” (I-key), in Block 3 the positively and negatively connoted words as well as the
stimuli of the FPÖ had to be assigned to the categories “Bad or FPÖ” (E-key) and “Good” (I-key)
(Block 3) (see Table 1 and Kritzinger et al., 2014c).
Table 2: Setup AUTNES SC-IAT
Block
Category labels for left
Category labels for
Aim
response key
right response key
(key „E“)
(key „I“)
1
Bad
vs.
Good
Training
2
Bad
vs.
Good or FPÖ
Measurement
3
Bad or FPÖ
vs.
Good
Measurement
The following stimuli were used: Good: Love, Luck, Joy, Fun, Peace; Bad: Pain, Torture, Fear,
stimuli: FPÖ, Die FPÖ, Die Blauen, Die Freiheitlichen
10
Trials
15
40
40
Hate, Murder; FPÖ-
https://implicit.harvard.edu/implicit/austria/
9
The first block only serves for training purposes (e.g. Plischke, 2012). In order to assess the implicit
attitudes towards the FPÖ, the Blocks 2 and 3 are needed. Respondents who complete the SC-IAT
according to the introduction unconsciously use their spontaneous affect towards the stimuli as
aids in order to cope with the task (e.g. Plischke, 2012). Hence, respondents who completed Block 2
faster than Block 3 are considered to have relatively positive associations with the FPÖ; respondents
who completed Block 3 faster than Block 2 are considered to have relatively negative associations
with the FPÖ. The SC-IAT scores were computed by (a) subtracting the reaction times of the third
block from the reaction times of the second block and by (b) dividing the resulting difference value
by the intra-individual standard deviation of all reaction times. This procedure takes into account
that the general reaction-time differs between individuals (see e.g. Plischke, 2012). The magnitude
of the relative difference between the reaction times of Block 2 and Block 3 indicates the strength
of the preferences (e.g., Ksiazkiewicz and Hedrick, 2013).
It is common practice (see e.g. Friese et al., 2012; Greenwald et al., 1998; Plischke, 2012) that
respondents, who completed at least 10% of their trials in less than 300 ms or more than 3000 ms,
respectively, and/or who completed the SC-IAT with more than 25% errors (= incorrect allocation),
are excluded from data analyses. We followed this convention in our study. Finally, the SC-IAT
scores were recoded to range from 0 (= most negative implicit association with the FPÖ) to 1 (=
most positive implicit association with the FPÖ) (Mean=0.61, SD=0.13).
Importantly, the implicit attitudes towards the FPÖ correlate moderately with the explicit attitudes
towards the FPÖ (r= 0.40, p<0.001). While this indicates that both implicit and explicit attitudes
point to the same target, the moderate correlations also illustrate that that the two attitudes are
largely distinct.
Control variables
As control variables we consider several variables which former research has shown to impact
consistent voting decisions: union membership and the degree of religiosity as indicators for
political heuristics that go beyond party identification, interest in politics and attentiveness to the
campaign as indicators for political motivation, education as indicator for political expertise, as well
as media consumption (newspapers, TV, internet). We additionally consider some control variables
that are known to have an effect on the probability to vote for the FPÖ (e.g., Plasser and Ulram,
10
2000; Johann et al., 2014): sex, age and the sympathy towards the party leader of the FPÖ (in
election of 2013 it was Heinz-Christian Strache). Except for age, all control variables were rescaled
to range from 0 to 1 (for more information on question wording, coding and the distribution of the
control variables, see Appendix A & B).
Models
We test our hypotheses by estimating seven multinomial logit (MNL) models with respondents who
did not vote for the FPÖ as baseline category. Our first model (Model 1) only contains the control
variables. The models 2 to 4 include one (Models 2 and 3) or both (Model 4) attitudes toward the
FPÖ measures (explicit and/or implicit measure, respectively). The models 5 to 7 contain the control
variables as well as one (Models 5 and 6) or both (Model 7) attitudes toward the FPÖ measures. In
addition to the logit coefficients we also present marginal effects conditional on voters’ explicit and
implicit attitudes towards the FPÖ. In calculating the marginal effects, we use the “observed-value
approach” (Hanmer and Kalkan, 2013).
Results
In Table 3 we present the results of our MNL models with respondents who did not vote for the FPÖ
as a reference category. Starting with our baseline model (Model 1) we can observe that the
sympathy for the party leader of the FPÖ, Heinz-Christian Strache, has a statistically significant
effect (p<.001) – independently of whether the vote choice for the FPÖ is issue consistent or not.
Beyond that, we can only observe a statistically significant negative effect of the degree of
religiosity on issue consistent vote decisions (p<0.05) as previous research has already indicated.
Other than that, we find no statistically significant effects. McFadden’s R² is 0.39.
Turning to our main independent variables we find that explicit as well as implicit attitudes towards
the FPÖ prove to be important factors in explaining vote decisions. The logit coefficients are
statistically significant (p<.001) and positive – independently of whether the vote choice for the FPÖ
is issue consistent or not and independently of whether we utilize control variables or not.
Moreover, all fit indices we examine (Log likelihood, McFadden’s R², AIC, BIC) indicate that explicit
as well as implicit attitudes towards the FPÖ substantially contribute to the explanation of our
11
dependent variable. This is also true if we include both explicit and implicit attitudes measures
together in our models (see Models 4 and 7).
Focusing only on logit coefficients when interpreting MNL results might be misleading because the
direction and strength of the effects depend on the reference category that is chosen (Kühnel and
Krebs, 2010). We therefore also present marginal effects conditional on voters’ explicit and implicit
attitudes towards the FPÖ. The resulting marginal effects are presented in Figure 1. Starting with
the marginal effects of the explicit attitudes towards the FPÖ we can observe that closeness to the
FPÖ in terms of party identification does not have a significant effect on the likelihood to cast an
issue inconsistent vote, but it increases the likelihood to cast an issue consistent vote for the FPÖ.
We thus could confirm H1a and H1b: while consistent voting is related to strong party
identification, inconsistent voting is not related to weak party identification. In turn, implicit
attitudes towards the FPÖ positively affect both the likelihood to cast an issue consistent and an
issue inconsistent vote for the FPÖ. The change in the predicted probabilities triggered by the
implicit attitudes towards the FPÖ amounts to more than 30 per cent. Hypothesis 2 proved to be
correct.
12
Table 3: MNL-Models (Reference category: Did not vote for the FPÖ)
(1)
Did vote for the FPÖ & vote choice “inconsistent”
Sympathy Heinz-Christian Strache
Interest in politics
Attentiveness to current campaign
Education (Matura)
Sex (Men)
Age
Degree of religiosity
Union membership
Information on political events: in newspapers
Information on political events: on TV
Information on political events: online
(2)
6.99***
(0.86)
Implicit Attitudes towards the FPÖ (SC-IAT)
Did vote for the FPÖ & vote choice “consistent”
Sympathy Heinz-Christian Strache
Interest in politics
Attentiveness to campaign
Education (Matura)
Sex (Men)
Age
Degree of religiosity
Union membership
Information on political events: in newspapers
Information on political events: on TV
Information on political events: online
-2.74***
(0.71)
-2.27***
(0.14)
8.93***
(0.88)
Implicit Attitudes towards the FPÖ (SC-IAT)
N
Mean Variance Inflation Factor
Log likelihood
McFadden’s R²
AIC
BIC
10.32***
(1.23)
-8.40***
(0.84)
6.43***
(0.91)
9.67***
(1.36)
-8.41***
(0.93)
6.37***
(0.59)
-0.35
(0.84)
1.31#
(0.75)
-0.56#
(0.33)
0.09
(0.35)
-0.00
(0.01)
-1.24*
(0.55)
-0.51
(0.39)
-0.99#
(0.59)
-0.39
(0.65)
0.53
(0.54)
Explicit attitudes towards the FPÖ (PID FPÖ)
Constant
Model
(4)
5.60***
(0.52)
-1.45#
(0.81)
0.95
(0.70)
-0.45
(0.32)
-0.54#
(0.32)
0.00
(0.01)
-0.27
(0.49)
0.09
(0.35)
-0.92
(0.57)
-0.31
(0.64)
0.48
(0.52)
Explicit attitudes towards the FPÖ (PID FPÖ)
Constant
(3)
-4.05***
(0.80)
668
1.33
-300.12
0.39
648.25
756.35
-2.93***
(0.19)
668
--362.74
0.26
733.49
751.50
13.62***
(1.37)
-10.81***
(0.97)
668
--394.35
0.20
796.71
814.72
8.23***
(0.93)
11.77***
(1.63)
-10.55***
(1.15)
668
1.19
-317.14
0.35
646.28
673.30
(5)
(6)
(7)
4.94***
(0.55)
-1.79*
(0.84)
1.63*
(0.75)
-0.46
(0.33)
-0.66*
(0.33)
0.01
(0.01)
-0.14
(0.50)
-0.05
(0.37)
-1.00#
(0.59)
-0.62
(0.65)
0.54
(0.55)
4.12***
(0.98)
5.22***
(0.56)
-1.33
(0.85)
0.88
(0.74)
-0.40
(0.33)
-0.90**
(0.34)
-0.00
(0.01)
-0.35
(0.51)
0.05
(0.36)
-0.96
(0.60)
-0.49
(0.67)
0.50
(0.55)
4.57***
(0.59)
-1.36
(0.89)
1.39#
(0.81)
-0.45
(0.34)
-1.08**
(0.36)
0.00
(0.01)
-0.22
(0.52)
-0.07
(0.39)
-1.04#
(0.62)
-0.73
(0.69)
0.60
(0.57)
4.20***
(1.05)
9.03***
(1.75)
-8.41***
(1.36)
-2.92***
(0.73)
4.64***
(0.65)
-1.18
(0.94)
2.45**
(0.86)
-0.48
(0.37)
-0.17
(0.38)
0.01
(0.01)
-1.10#
(0.61)
-0.79#
(0.45)
-1.12#
(0.66)
-0.74
(0.73)
0.68
(0.62)
6.12***
(1.00)
-4.10***
(0.87)
668
1.41
-270.35
0.45
592.69
709.80
8.74***
(1.66)
-7.95***
(1.26)
5.81***
(0.63)
-0.32
(0.90)
1.15
(0.80)
-0.51
(0.35)
-0.37
(0.37)
-0.00
(0.01)
-1.26*
(0.57)
-0.50
(0.42)
-1.08#
(0.64)
-0.70
(0.70)
0.70
(0.58)
10.81***
(1.76)
-10.34***
(1.38)
668
1.37
-274.34
0.44
600.69
717.80
4.18***
(0.69)
-0.84
(0.99)
2.22*
(0.93)
-0.46
(0.39)
-0.64
(0.41)
0.00
(0.01)
-1.17#
(0.63)
-0.79#
(0.47)
-1.17#
(0.69)
-0.97
(0.77)
0.80
(0.64)
6.17***
(1.08)
10.28***
(1.91)
-10.26***
(1.51)
668
1.44
-249.88
0.49
555.76
681.88
Note: Presented are unstandardized logit-coefficients. Standard errors in parentheses.
# p<.10, * p<.05, ** p<.01, *** p<.001.
13
Figure 1: Marginal effects conditional on explicit and implicit attitudes towards the FPÖ
Note: In calculating the marginal effects, we use the “observed-value approach” (Hanmer and Kalkan, 2013).
14
Conclusion
The literature on vote choice models in general and on consistent voting in particular has hitherto
mainly focused on individuals’ explicit cognitive components to examine their behaviour. Consistent
voting patterns of citizens were considered as a function of citizens’ deliberative decision-making
processes on which party fits their policy preferences best. Lately, in addition to explicit attitudes,
implicit attitudes have also gained importance in explaining citizens’ behaviour and beliefs. Unlike
explicit attitudes, they are released by rather unconscious processes which cannot be
modified/manipulated by the humans.
In this paper we analysed both types of attitudes assuming that both exert an influence on citizens’
voting behaviour. But while we expected a straightforward impact of explicit attitudes (e.g. stronger
party identification is connected to more consistent voting), we assumed slightly different patterns
for implicit attitudes. Unlike deliberative processes, these automatically generated affective
attitudes act independently on voters’ behaviour and impact voting patterns differently.
Our results do indeed show that both implicit and explicit attitudes impact consistent voting
patterns, but the effect differs. Explicit attitudes behave as expected: the stronger the party
identifications the more likely it is that voters vote consistently with their party preferences, hence
indicating that party identification works as a good heuristic to reflect policy interests. However,
and most interestingly, we do not see any linear relationship: inconsistent voting cannot be
explained by lower explicit attitudes. In fact, we do not see any relationship at all. It seems that the
inconsistent voting is not the outcome of a negative deliberative function. Hence, unlike assumed in
previous research, inconsistent voting does not represent the other end of the consistent voting
dimension.
When taking implicit attitudes into account the pictures slightly changes. As expected, positive
implicit attitudes also increase the likelihood to vote consistently. Unlike the explicit attitudes,
implicit attitudes do however also impact inconsistent voting. The effect of spontaneous (affective)
reactions, resulting from uncontrolled processes, is positively related to inconsistent voting: the
more positive a voter’s implicit attitudes towards the radical-right party, the more likely she will
vote for it, even if the party does not represent her policy preferences best. So, inconsistent voting
seems to be rather a function of implicit rather than explicit attitudes.
15
Dividing between consistent and inconsistent voting patterns and looking at implicit and explicit
attitudes, in this paper we have unfolded some interesting mechanisms. Consistent and
inconsistent voting do not follow the same dimension as previously assumed. While explicit
attitudes are relevant for consistent voting patterns, implicit attitudes are an important factor
explaining why individuals choose a party (in our case a radical-right party) that does not represent
their policy preferences best. Implicit attitudes influence an individual’s behaviour which from a
cost-benefit perspective conflict with her preferences/interests (e.g. Glaser and Finn 2013). They
simply follow the party they positively associate with. It hints to an overall, unconscious and
spontaneous positive association with the FPÖ, where policy preferences are not of relevance.
Which other conclusions can we draw from these first findings? Given the decline in partisanship
(explicit attitudes) and the increase of undecided voters implicit attitudes might become more
important predictors in vote choice models as it is currently the case.
Overall, we can conclude that it is important to combine individuals’ explicit and implicit attitudes
to be better able to model vote choice preferences of voters.
16
References
Aichholzer, J., Kritzinger, S., Wagner, M., Zeglovits, E. (2014). How has radical right support transformed
established political conflicts? The case of Austria. West European Politics 37(1), 113–137.
Arcuri, L., Castelli, L., Galdi, S., Zogmaister, C., Amadori, A. (2008). Predicting the Vote: Implicit Attitudes
as Predictors of the Future Behavior of Decided and Undecided Voters. Political Psychology 29(3), 369387.
Bargh, J.A. (1994). The Four Horsemen of Automaticity: Awareness, Efficiency, Intention, and Control in
Social Cognition. In: Wyer Jr., R.S., Skrull, T.K. (Eds.) Handbook of Social Cognition. Hillsdale, NJ: Erlbaum:
pp. 1-40.
Bartels, L. (2000). Partisanship and Voting Behavior, 1952-1996. American Journal of Political Science
44(1), 35-50.
Baum, M.A., Jamison, A.S. (2006). The Oprah Effect: How Soft News Helps Inattentive Citizens Vote
Consistently, The Journal of Politics 68(4), 946–959.
Campbell, A., Gurin, G., Miller, W.E. (1954). Voter Decides. Evanston, IL: Row, Peterson.
Downs, A. (1957). An Economic Theory of Democracy. New York: Harper and Row.
Friese, M., Smith, C.T., Plischke, T., Bluemke, M., Nosek, B.A. (2012). Do Implicit Attitudes Predict Actual
Voting Behavior Particularly for Undecided Voters? PLoS ONE 7(8): e44130.
doi:10.1371/journal.pone.0044130.
Gawronski, B., Galdi, S., Arcuri, L. (2015). What Can Political Psychology Learn from Implicit Measures?
Empirical Evidence and New Directions. Political Psychology 36(1), 1-17.
Glaser, J., Finn, C. (2013). How and Why Implicit Attitudes Should Affect Voting. Political Science &
Politics 46, 537-544.
Greenwald, A.G., McGhee, D.E., Schwartz, J.L.K. (1998). Measuring Individual Differences in Implicit
Cognition: The Implicit Association Test. Journal of Personality and Social Psychology 74(6), 1464-1480.
Greenwald, A.G., Smith, C.T., Sriram, N., Bar-Anan, Y., Nosek, B.A. (2009). Implicit Race Attitudes
Predicted Vote in the 2008 U.S. Presidential Election. Analyses of Social Issues and Public Policy 9(1),
241-253.
Merrill, S. III, Grofman, B. (1999). A Unified Theory of Voting. Cambridge: Cambridge University Press.
Hanmer, M.J., Kalkan, K.O. (2013). Behind the Curve: Clarifying the Best Approach to Calculating
Predicted Probabilities and Marginal Effects from Limited Dependent Variable Models. American Journal
of Political Science 57(1), 263–277.
Johann, D., Glantschnigg, C. (2013). “Correct Voting” bei der österreichischen Nationalratswahl 2008,
Österreichische Zeitschrift für Politikwissenschaft 42(4), 373-390.
Johann, D., Glantschnigg, C., Glinitzer, K., Kritzinger, S., Wagner, M. (2014). Das Wahlverhalten bei der
Nationalratswahl 2013. In: Kritzinger, S., Müller, W.C., Schönbach, K. (Eds.), Die Nationalratswahl 2013.
Wie Parteien, Medien und Wählerschaft zusammenwirken. Wien: Böhlau, pp. 191-214.
Jost, J.T., Rudman, L.A., Blair, I.V., Carney, D.R., Dasgupta, N., Glaser, J., Hardin, C.D. (2009). The
Existence of Implicit Bias is Beyond Reasonable Doubt: A Refutation of Ideological and Methodological
Objections and Executive Summary of Ten Studies that No Manager Should Ignore. Research in
Organizational Behavior 29, 39-69.
17
Karpinski, A., Steinman, R.B. (2006). The Single-Category Implicit Association Test as a measure of
implicit social cognition. Journal of Personality and Social Psychology 91, 16-32.
Kraft, P. (2012). Correct Voting in Deutschland. Eine Analyse der Qualität individueller
Wahlentscheidungen bei der Bundestagswahl 2009. In: Arbeitspapiere – Mannheimer Zentrum für
Europäische Sozialforschung No. 148.
Kritzinger, S., Johann, D., Glantschnigg, C., Aichholzer, J., Glinitzer, K., Thomas, K., Wagner, M., Zeglovits,
E. (2014a). AUTNES TV Debate Panel Survey 2013. GESIS Data Archive, Cologne. ZA5858 Data File
Version 1.0.0. doi:10.4232/1.11951
Kritzinger, S., Johann, D., Glantschnigg, C., Aichholzer, J., Glinitzer, K., Thomas, K., Wagner, M., Zeglovits,
E. (2014b). AUTNES TV-Debates Panel Study 2013 – Documentation. Vienna: University of Vienna.
Ksiazkiewicz, A., Hedrick, J. (2013). An Introduction to Implicit Attitudes in Political Science Research.
Political Science & Politics 46, 525-531.
Kühnel, S.M., Krebs, D. (2010). Multinomiale und ordinale Regression. In: Wolf, C., Best, H. (Eds.):
Handbuch der sozialwissenschaftlichen Datenanalyse. Wiesbaden: VS Verlag, pp. 855-886.
Lau, R.R., Andersen, D.J., Redlawsk, D.P. (2008). An Exploration of Correct Voting in Recent U.S.
Presidential Elections. American Journal of Political Science 52(2), 395–411.
Lau, R.R., Patel, P., Fahmy, D.F., Kaufman, R.R. (2014). Correct Voting Across Thirty Three Democracies: A
Preliminary Analysis. British Journal of Political Science 44, 239-259.
Lau, R.R., Redlawsk, D.P. (1997). Voting Correctly. The American Political Science Review 91(3), 585-598.
Lenz, G. (2012). Follow the Leader? How Voters Respond to Politicians’ Policies and Performances.
Chicago, IL: University of Chicago Press.
Lodge, M., C.S. Taber (2005). Automaticity of affect for political leaders, groups and issues: An
experimental test of the hot cognition hypothesis. Political Psychology 26, 455-482.
Müller, W.C., Eder, N., Jenny, M. (2015). AUTNES Candidate Survey 2013. Vienna: University of Vienna.
Payne, B.K., Cheng, C.M., Govorun, O., Stewart, B.D. (2005). An Inkblot for Attitudes: Affect
Misattribution as Implicit Measurement. Journal of Personality and Social Psychology 89(3), 277-293.
Plasser, F., Ulram, P. A. (2000). Rechtspopulistische Resonanzen: Die Wählerschaft der FPÖ. In: Plasser,
F., Ulram. P.A., Sommer, F. (Eds.), Das österreichische Wahlverhalten. Wien: Signum Verlag, pp. 225-241.
Plischke, T. (2012). Reaktionszeiten als Indikatoren für politische Einstellungen. Der Implizite
Assoziationstest (IAT). Methoden – Daten – Analysen 6(2), 73-98.
Popa, S. (2012). “Attitude congruent” electoral decisions. A cross-country analysis of the quality of
electoral decisions. Unpublished Paper.
Popkin, S.L. (1991). The reasoning voter: Communication and persuasion in presidential campaigns.
Chicago, IL: University of Chicago Press.
Roccato, M., Zogmaister, C. (2010). Can we improve electoral forecasts using the IAT? A field research.
Political Psychology 31, 249-274.
Rohrschneider, R., Whitefield, S. (2012). Institutional Context and Representational Strain in Party–Voter
Agreement in Western and Eastern Europe. West European Politics 35(6), 1320-1340.
18
Rosema, M., de Vries, C.E. (2011). Assessing the Quality of European Democracy Are Voters Voting
Correctly? In: Rosema, M., Denters, B., Aarts, K. (Eds.): How Democracy Works: Political Representation
and Policy Congruence in Modern Societies: Essays in Honour of Jacques Thomassen. Amsterdam, pp.
199-219.
Schmitt-Beck, R., Kraft, P. (2014). Political Information Flows and Constistent Voting: Personal
Conversations, Mass Media, Party Campaigns, and the Quality of Voting Decisions at the 2009 German
Federal Election. In: Weßels, B., Rattinger, H., Roßteutscher, S., Schmitt-Beck, R. (Eds.): Voters on the
Move or on the Run? Oxford: Oxford University Press, pp. 193-216.
Smith, C.T., Ratliff, K.A., Nosek, B.A. (2012). Rapid assimilation: Automatically integrating new
information with existing beliefs. Social Cognition 30, 199-219.
Stanley, D., Phelps, E., Banaji, M. (2008). The Neural Basis of Implicit Attitudes. Current Directions in
Psychological Science 17(2), 154-170.
Wagner, M., Johann, D., Kritzinger, S. (2012). Voting at 16: Turnout and the quality of vote choice.
Electoral Studies 31(2), 372-383.
19
Appendix A – Descriptive statistics of independent variables
Variable
Implicit Attitudes towards the FPÖ (SC-IAT)
Explicit attitudes towards the FPÖ (PID FPÖ)
Sympathy Heinz-Christian Strache
Degree of religiosity
Union membership
Interest in politics
Attentiveness to campaign
Information on political events: in newspapers
Information on political events: on TV
Information on political events: online
Education (Matura)
Sex (Men)
Age
N
668
668
668
668
668
668
668
668
668
668
668
668
668
Min.
0
0
0
0
0
0
0
0
0
0
0
0
16
Max.
1
1
1
1
1
1
1
1
1
1
1
1
87
Mean
0.61
0.09
0.31
0.29
0.24
0.65
0.63
0.73
0.74
0.63
0.54
0.54
45.60
SD
0.13
0.23
0.36
0.31
-0.27
0.28
0.31
0.30
0.35
--15.05
20
Appendix B – Question wording and coding
Variable
Implicit Attitudes towards the FPÖ (SC-IAT)
Question wording
[see Appendix C]
Explicit attitudes towards the FPÖ (PID FPÖ)
In Austria many people feel close to a
political party although they
sometimes vote for another party.
What about you? Generally speaking,
do you feel close to a particular
party?
[If “no”:] Do you generally think of
yourself as a little closer to one of the
parties than the others?
Which party is that?
Do you feel very close, fairly close or
not very close to this party?
How much do you like the following
politicians?
Sympathy Heinz-Christian Strache
Degree of religiosity
Union membership
Interest in politics
Attentiveness to campaign
Information on political events: in newspapers
Information on political events: on TV
Information on political events: online
Education (Matura)
Sex (Men)
Age
Would you say that you are not at all
religious, a little religious, somewhat
religious, or very religious?
Are you a member of a trade union?
Generally speaking, are you very,
fairly, a little or not at all interested
in politics?
How closely are you following the
current election campaign? Very
closely, fairly closely, not very closely,
or not at all closely?
How often do you inform yourself
about the political events in Austria
through newspapers?
How often do you inform yourself
about the political events in Austria
through television?
How often do you inform yourself
about the political events in Austria
through the internet?
What is the highest level of
education that you have achieved?
Are you male or female?
When were you born?
Coding
Metric variable, rescaled to range
from 0=“relatively negative implicit
associations with FPÖ” to
1=“relatively positive implicit
associations with FPÖ”
4-point-scale from 0=“I do not feel
close to the FPÖ” to 1=“I feel very
close to the FPÖ”
11-point-scale, rescaled to range
from 0=“strongly dislike” to
1=“strongly like”
4-point-scale, rescaled to range from
0=“not at all religious” to 1=“very
religious”
0=“no”, 1=“yes”
4-point-scale, rescaled to range from
0=“not at all interested” to 1=“very
interested”
4-point-scale, rescaled to range from
0=“not at all closely” to 1=“very
closely”
5-point-scale, rescaled to range from
0=“never” to 1=“almost every day”
5-point-scale, rescaled to range from
0=“never” to 1=“almost every day”
5-point-scale, rescaled to range from
0=“never” to 1=“almost every day”
1=degree that allows high school
entrance, 0=otherwise
0=“female”, 1=“male”
Age in years (year of interview – year
of birth)
21
Appendix C – AUTNES-SC-IAT (see Kritzinger et al., 2014c)
22
23
24
25