Content uploaded by Carol Galais
Author content
All content in this area was uploaded by Carol Galais on Oct 06, 2014
Content may be subject to copyright.
http://ips.sagepub.com/
International Political Science Review
http://ips.sagepub.com/content/early/2014/09/25/0192512114550372
The online version of this article can be found at:
DOI: 10.1177/0192512114550372
published online 26 September 2014International Political Science Review
Carol Galais and André Blais
Beyond rationalization: Voting out of duty or expressing duty after voting?
Published by:
http://www.sagepublications.com
On behalf of:
International Political Science Association (IPSA)
can be found at:International Political Science ReviewAdditional services and information for
http://ips.sagepub.com/cgi/alertsEmail Alerts:
http://ips.sagepub.com/subscriptionsSubscriptions:
http://www.sagepub.com/journalsReprints.navReprints:
http://www.sagepub.com/journalsPermissions.navPermissions:
What is This?
- Sep 26, 2014OnlineFirst Version of Record >>
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
International Political Science Review
1 –17
© The Author(s) 2014
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0192512114550372
ips.sagepub.com
Beyond rationalization: Voting
out of duty or expressing duty
after voting?
Carol Galais and André Blais
Université de Montréal, Canada
Abstract
It is a standard practice to include a Duty term in explanatory models of turnout. Yet the relationship
between duty and voting is not that clear. Does duty really trigger voting or is it the reverse? To address
this question, we present cross-lagged panel estimations of the impact of duty on turnout and of turnout on
duty with two different datasets: a two-wave panel Canadian survey conducted in 2008 and 2009 and a four-
wave Spanish panel conducted between 2010 and 2012. We find evidence that sense of civic duty is a true
motivation that affects people’s propensity to vote, even though duty may be reinforced by the act of voting.
Keywords
Civic duty, turnout, vote, causality, endogeneity
Introduction
From a purely utilitarian perspective, voting does not appear to be a ‘rational’ choice in a large
electorate election. Given the extremely low probability that one’s decision will be pivotal (Mueller,
2003; Owen and Grofman, 1984), the costs of voting (the time required not only to go to the polling
station but also to obtain information in order to decide which party/candidate to support) are
bound to outweigh the expected benefits. Yet most people vote, which is known as the paradox of
voting (Fiorina, 1989; Grofman, 1993).
Why do so many people vote? One possible reason is that they feel it is a citizen’s duty to vote
in a democracy. People vote not because they calculate that the benefits outweigh the costs but
because they consider that this is the ‘right’, ‘ethical’ thing to do. They believe that they have a
moral obligation to vote. This belief that voting is a civic duty is known as the ‘D term’.
Even if it is commonly assumed that the feeling that voting is a duty must be taken into account
in the turnout decision, we know little about the exact nature of the link between this belief and
Corresponding author:
Carol Galais, Postdoctoral Fellow, Canada Research, Département de science politique, Université de Montréal,
C.P.6128, Succ. Centre-ville, Montreal, Canada, H3C 3J7.
Email: carolgalais@gmail.com
550372IPS0010.1177/0192512114550372International Political Science ReviewGalais and Blais
research-article2014
Article
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
2 International Political Science Review
voting. People may vote because they are driven by an inner sense of civic duty; however, they can
also develop (or strengthen) their sense of civic duty after taking part in an election. Furthermore,
if people say they feel dutiful because they recall having voted, and if they report low levels of duty
because they abstained, then the D term would be a mere a posteriori rationalization of the act of
voting (Barry, 1970; Matsusaka and Palda, 1999). Thus the following questions: Does the belief
that voting is a civic duty really trigger electoral participation, is it the reverse or does the causality
go both ways?
For the purpose of answering these questions, we review the theoretical debate regarding the
nature of the relationship between civic duty and electoral behavior. Next, we argue that the best
way to disentangle the link between attitudes and behavior is a longitudinal approach such as the
one we use in this study. We then present our data and the estimation method chosen to disentangle
the direction of causality. We use two different datasets: a two-wave panel survey conducted in
2008 and 2009 in the Canadian provinces of British Columbia and Quebec, and a four-wave
Spanish panel conducted between 2010 and 2012. We model the relationship between voting and
civic duty by means of cross-lagged structural equations. We find evidence of both processes,
although the effect that goes from duty to voting is stronger than the other way around. We con-
clude that sense of civic duty is not purely a post-hoc rationalization of the act of voting; for some
citizens it is a true motivation that affects their propensity to vote, even though in some cases citi-
zens may align their views about voting with their past behavior or may experience a stronger
sense of civic duty after voting.
Civic duty and voting: Ethical behavior or rationalization?
Downs (1957) argued that citizens vote only if the expected costs (C) do not exceed the expected
benefits (B). The latter depend on the probability of one’s vote to be pivotal (P), which is extremely
low. Therefore, a rational citizen will soon find that PB < C, and will abstain as a consequence.
However, many citizens keep voting, which has produced what has become known as the voter’s
paradox. Many scholars have tried to solve this paradox by suggesting that the costs of voting are
negligible, that the benefits can be huge or that the P term needs to be replaced by a function includ-
ing strategic considerations or conditional expected utility. Finally, a number of authors have made
the case for a normative element, the ‘D term’ (Dowding, 2005).
The first reference to a normative foundation for the act of voting is found in Campbell et al.’s
seminal work (1954), where it is suggested that sense of civic obligation leads to a high likelihood
of showing up at the polling station on election day. Later on, Campbell et al. (1960: 105–106)
point out again that turnout is much higher among those with a strong sense of duty than among
those with none. In the same vein, Riker and Ordeshook (1968) show that duty (the ‘D term’) has
a strong impact on the propensity to vote. Dennis relates civic duty to diffuse support for the
regime, defining it as the citizen’s feeling of obligation ‘to contribute his own resources of time and
effort even when particular elections are anticipated to be unfavorable or trivial to his own inter-
ests’ (1970: 63). He finds high public endorsement of the duty to vote, consistent with the findings
for other western countries, and a strong correlation between duty and turnout.
More recently, Verba et al. (1995: 115) report that civic gratifications, among them civic duty,
are the most widespread motivation for voting. Blais (2000: 112) concludes that duty is the over-
riding motivation for about half of those who vote. Clarke et al. (2004: 259) find that the variable
with the largest effect on turnout is what they call ‘system benefits’, which in fact corresponds to
civic duty. Dalton finds a positive effect of citizens’ duty on the propensity to vote, although not on
other forms of participation, such as volunteerism, petitions or boycott (Dalton, 2008). Hence, the
inclusion of an indicator of civic duty in explanatory models of turnout is a widespread practice.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
Galais and Blais 3
Finally, recent experimental research shows that prompting citizens to think about voting as a duty
indeed boosts turnout (Gerber et al., 2008).
The nature of the ‘D term’ is, at best, unclear. Some citizens may only feel they have a duty to
vote if they think they are going to be decisive, whilst others claim that it is their duty to cast a vote
because they want to contribute to a common good (Mueller, 2003). From the most critical
approach, duty is just an artifact that makes voting an irrational matter of ‘taste’, making the con-
tribution of rational choice theory rather meaningless (Barry, 1970). For some authors, duty is an
expression of party affiliation (Fiorina, 1976) and/or some election-specific value (Aldrich, 1993).
It may also have a patriotic or altruistic connotation (Usher, 2011). From a more restricted perspec-
tive, it is an expression of social identity for those citizens who believe that political outcomes
(such as a high turnout rate) will positively affect members of their group (Fowler and Kam, 2007).
It has been equated to expressive rationality (Engelen, 2006) and to intrinsic motivations to vote
(Jones and Hudson, 2000). Its normative component has been related to system benefits and to
personal convictions about what a good citizen should do.
The goal of this paper is not to disentangle the true nature or causes of the Duty term, but rather
to examine its relationship with voting. Among the scholars suggesting a normative foundation for
the act of voting, most consider that when people express their belief that there is a duty to vote,
this reflects adherence to a social norm (Blais, 2000; Coleman, 1990; Knack and Kropf, 1998;
Mueller, 1989; Uhlaner, 1986). The assumption, which we adhere to for the purpose of this paper,
is that those who subscribe to the norm will want to behave in a way that is congruent with the
norm; hence, they will feel compelled to vote.
The causal link between the civic duty to vote and turnout is based on the theoretical assumption
that attitudes precede and cause political behavior (Marsh, 1971: 453). Most political communica-
tion and political behavior researchers work with this assumption (Ajzen and Fishbein, 1980). Some
evidence that duty is formed early and prior to voting is offered by Wolfinger and Rosenstone
(1980), who found that a higher level of education leads to higher levels of civic duty. Personality
also emerges as an antecedent of civic duty, with conscientiousness, agreeableness, openness and
extraversion positively affecting the sense that voting is a duty (Weinschenk, 2014); this suggests
that duty comes before voting in the causal chain. Campbell (2006) argues that feelings of duty
towards voting are formed in childhood and within politically homogeneous communities able to
enforce social norms about voting. From this point of view, the civic duty to vote is not different
from most psychological orientations that develop as a consequence of early socialization processes,
which make them stable over time (Easton et al., 1969). To the best of our knowledge, however, the
assumptions that duty precedes turnout and that it is stable over time have never been tested.
Despite numerous examples conceiving civic duty as an antecedent of voting, a number of dif-
ficulties beset our understanding of the causal links between these two phenomena. Since civic
duty is to be expected from a ‘good citizen’, respondents may report a sense of duty driven by a
desire for social respectability. Even worse, sense of citizen duty may be a mere a posteriori ration-
alization of the act of voting (Matsusaka and Palda, 1999). That is, a respondent may say that vot-
ing is a duty because she voted in the previous election and that it is not a duty because she
abstained. According to Dowding, ‘habitual voters justify their voting in terms of civic duty since
they cannot rationalise it any other way’ (2005: 456). Similarly, Harder points out: ‘the act of vot-
ing could increase feelings of civic duty (…); when people go to the polls they may later rationalize
that it must have been because their vote was important’ (2008: 5).
Rationalization is an internal process that can be defined as ‘an active self-justifying intensifica-
tion of belief’ (Batson, 1975: 176) or as bringing one’s attitudes in line with one’s behavioral inten-
tions (Finkel and Muller, 1998: 40). In other words, rationalizing may mean changing attitudes to
align them with actual behavior, or generating them when they do not exist. In the first case,
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
4 International Political Science Review
respondents report ‘consistent’ attitudes after recalling behavior because ‘inconsistencies among
beliefs and attitudes are noxious to people, and they are inclined to eliminate such inconsistencies
by changing, adding, or deleting the beliefs or attitudes responsible for the inconsistency ‘(Rahn et
al., 1994: 586). This explanation is based on cognitive dissonance avoidance, which drives indi-
viduals to strive for internal consistency.
Alternatively, and according to Bem’s self-perception theory (1967), when individuals do not have
previous attitudes, they can produce a consistent answer when asked by observing their own behavior
and inferring what attitudes must have caused it. People would report high levels of duty to vote if
they recall having voted because this answer is in line with their past behavior. Since cause does not
precede consequence, any causal inference about the impact of this attitude would be erroneous.
From another point of view, finding that the relationship between reported civic duty and
reported electoral behavior goes from the latter to the former could reveal the presence of learning
and reinforcement processes. As described by several works in political behavior, this is the case
of trust (Delli Carpini et al., 2004), knowledge (Fishkin, 1995) or internal political efficacy (Clarke
and Acock, 1989; Finkel, 1985; Leighley, 1991). To the best of our knowledge, no research has
examined this reinforcement and learning process for voting and civic duty. There is, however, a
literature on learning models of voting that describes the probability of engaging in an act in the
future as a function of the positive or negative feedback received for this action in the past (e.g.
voting for the winner increases the chances of voting in the next election, see Bendor et al., 2003;
Kanazawa, 1998). It makes sense, for instance, to think of an individual with a moderate level of
civic duty who, after voting, feels proud and experiences a stronger sense of civic duty.
This possible reciprocal relationship between turnout and the civic duty to vote means that the
direction of causality between the attitude and the associated behavior is unclear (Raney and
Berdahl, 2009). Our goal is to contribute to disentangling this intriguing relationship, and to ascer-
tain to what extent this belief precedes voting or whether it is the reverse, that is, voting causes
civic duty.
Data and methods
The main challenge of this research is to disentangle the direction of causality between an attitude
and a behavior. Causal statements imply change in variables along a time axis, so in order to speak
of causality ‘there is a time ordering between causes and effects. The cause must precede the effect
in time’ (Blossfeld and Rohwer, 1997: 366). The requirement that the cause must precede the con-
sequence in time can only be fulfilled, in a non-experimental design, with the use of panel data
(Kenny, 1975).
For this purpose, we rely on two panel surveys from Canada and Spain. These are two very dif-
ferent countries in terms of the longevity of their democracy, since Canada is a well-established
democracy while Spain is still young, its Constitution having been adopted in 1978. Both countries
are also different with respect to their electoral system (plurality in Canada, proportionality in
Spain), and party system (with higher fragmentation in Spain, where 13 parties obtained seats in
the national assembly in 2011, while only 5 entered the Canadian House of Commons in 2008).
Turnout rates in the national elections held between 2006 and 2011 were higher in Spain. At any
rate, if the same patterns are found in both contexts, this would strengthen the robustness and exter-
nal validity of our findings. We will first test our hypotheses with the Canadian data, and then
replicate the analysis on the Spanish database.
Our first data source is an internet panel survey conducted by YouGov Polimetrix in the
Canadian provinces of British Columbia and Quebec. The first wave of fieldwork was carried out
one week before the Canadian federal election that was held in October 2008. The second wave
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
Galais and Blais 5
took place just before the provincial elections held in Quebec in December 2008 and in British
Columbia in May 2009. Hence, a national general election was held between the two waves. The
sampling frame was designed to match the demographic profile of both provinces. We have
repeated measures (waves 1 and 2) of duty and turnout for 1268 Quebeckers and 873 British
Columbians.
The source for the Spanish data is also an internet panel survey that includes citizens up to 45 years
old, as it was designed to detect and track attitudinal change, which is less likely in adulthood.1 The
study includes 2100 individuals in the first wave in November 2010, plus 600 freshly recruited indi-
viduals in the second wave, conducted in May 2011. The fieldwork for the third wave took place in
November 2011, just one week before the national election held that year. It retained 1514 individuals
from the original sample and 465 of the refreshment. The last wave (May 2012) retains 1322 indi-
viduals from the original sample and 395 from the refreshment pool. In order to better tap abstention,
we kept in our database only those citizens who were eligible to vote in the two elections held within
each country during the time span covered by the panel surveys. Hence, in our Canadian survey we
only kept individuals who were 20 years old or more in 2008 (and thus had the right to vote in the
2006 election). For Spain, we dropped respondents under 21 years old in 2011 (those younger than
18 years old in 2008). The final total N is 2141 Canadians and 2569 Spaniards.
The first relevant variable is turnout in the national elections. For Canada, this refers to the
federal elections held in 2006 and 2008; in Spain, the two elections at stake were held in 2008 and
2011. The Canadian respondents were asked in the first wave (October 2008) about their electoral
behavior in the previous federal election, which took place in 2006. They were asked in the second
wave whether they voted in the 2008 federal election, held between the two waves. Similarly, the
third and fourth waves of the Spanish survey took place just before and after a general election (20
November 2011). This means that we have three measures of vote recall (waves 1, 2 and 3) that
refer to a previous election (held in 2008) and one measure referring to the 2011 elections, a time
structure very similar to the Canadian data. For both countries, turnout has been coded as a dichot-
omous variable: those who reported having voted are assigned the value 1 and those who did not
– including those that said ‘they could not’ – the value 0.
Regarding the measure of the civic duty to vote, we have borrowed Blais and Achen’s (2010)
wording. In order to minimize the social desirability problem, Blais and Achen proposed a question
wording that offers a ‘positive’ option (‘voting is a choice’) to those who do not feel a duty to show
up at the polling station. Both questionnaires included the same formulation of the duty to vote
question with a slight difference: the question is framed in general terms for Spain, whereas in the
Canadian survey it refers to different types of elections, for different levels of government: federal,
provincial and local elections. As the risk of measurement error increases when attitudes are sus-
ceptible to social desirability biases (Liska, 1974), we are fortunate that our data include several
indicators for civic duty, so we can obtain coefficients unbiased by random measurement error.
This allows us to capture the latent construct of ‘civic duty to vote’. In order to take full advantage
of the multiple indicators, our estimations include a measurement model using Confirmatory
Factor Analysis (CFA).
CFA is a multivariate statistical technique used to test the relationship between a latent or unob-
served factor and a series of observed variables or indicators (Brown, 2006). It is used to test
whether the indicators of such latent construct are consistent with theoretical expectations, that is,
whether the measurement model fits the data. The magnitude and significance of the factor load-
ings confirm or disconfirm such expectations, and several measurement model fit measures, such
as the Root Mean Squared Error of Approximation (RMSEA) or the Comparative Fit Index (CFI),
indicate to what extent the co-variation matrix produced by the model accurately reproduces that
of the data.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
6 International Political Science Review
We have different measurement models for the Duty to vote in Canada and Spain. In the
Canadian case, the latent construct of ‘duty’ was estimated using the answers regarding duty at
different levels of government, which corrects for measurement error potentially biasing our
estimates for the relationship between duty and turnout.2 In order to account for contextual
effects -such as the presence of an election- affecting latent constructs, and also to take acquies-
cence bias into account (Saris and Aalberts, 2003), we have correlated the error terms of the duty
indicators between waves.
The Spanish questionnaire included only a single indicator of duty per wave.3 Hence, we have
estimated the latent construct ‘previous duty’ using the questions tapping the civic duty to vote
included in waves 1–3. In this way we correct for measurement errors that may affect the unbiased-
ness of our estimates.4 Similarly, we have also estimated the latent construct ‘previous turnout
2008’ taking into account the questions on voting behavior in the 2008 general election, included
in waves 1–3. In this fashion, we are able to overcome electoral behavior recall inconsistencies due
to forgetfulness or cognitive bias, which are likely to increase as time and panel waves go by (van
Elsas et al., 2013).
A last challenge is that our research question involves more than one endogenous variable:
turnout and duty to vote. Single-equation models ignore the possibility of reciprocal causal rela-
tions among variables. This is why some scholars have used Structural Equation Modeling – SEM
(Markus and Converse, 1979; Page and Jones, 1979). This approach has also been adopted for the
sake of parsimony, since SEM allows for the estimation of two or more dependent variables, each
measured by different indicators, with a minimum number of parameters to be estimated and
reported.
We have the limitation that our panel data for Canada only cover two sampling moments. The
literature recommends at least three time points to have enough statistical power to account for
attitudinal change (Venter et al., 2002). We choose a research design appropriate for a two-wave
panel, which is a cross-lagged panel model. This is a method that tests spuriousness by comparing
cross-lagged correlations and regression coefficients (Burkholder and Harlow, 2003). It allows us
to estimate the strength of the causal effect of X on Y and the reverse. Both variables are regressed
at the same time on both their own lagged score and the lagged score of the other variable measured
in the past (t–1), producing auto-regressive and cross-lagged regression coefficients. While the first
coefficient gives us a clue about the stability of the phenomenon at stake, the cross-lagged regres-
sion parameters tell us how much variation in a phenomenon measured in t–1 predicts variation in
the other variable between t–1 and t1, that is, between panel waves.
The possible outputs when estimating these models are several: if none of the cross-lagged coef-
ficients are statistically significant, we can discard any causal relationship between X and Y. If all
of the cross-lagged coefficients are significant, this points to reciprocal effects; if only one cross-
lagged coefficient is statistically significant, this points to a unidirectional relationship. Control
variables are not essential as these models do not focus on the prediction or explanation of a phe-
nomenon, but on the relationship between two variables whose causal link is unclear. Nevertheless,
the inclusion of a lagged dependent variable accounts to some extent for the effects of unobserved
time-constant variables (Berrington et al., 2006). Correlating the error terms of the endogenous
variables also takes into account that other factors may be at work. Cross-lagged panel models
have been used in the political behavior literature for examining questions of reciprocal causality
(Campbell et al., 1960; Campbell and Kenny, 1999; Finkel, 1995; Hooghe and Quintelier, 2013;
Marsh and Yeung, 1997), or more specifically, to disentangle the relationship between an attitude
and a behavior with short panel data (Lenz, 2009).5
The first model to be estimated is presented in Figure 1. In the upper part of the figure the meas-
urement model for civic duty is displayed. Besides duty indicators, the model involves four
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
Galais and Blais 7
variables, two attitudes and two behaviors, measured at two points in time represented by the left
(wave 1) and the right (wave 2) sides of the graph. The two variables measured in October 2008
(time 1) are exogenous and thus correlated as is the standard in SEM. The two variables measured
in time 2 are endogenous, and both receive the effect of duty and reported turnout measured at time
1. The error terms of these variables are correlated since other variables (including contextual
covariates) may be affecting both of them.
The cross-lagged auto-regressive model has been replicated with the Spanish data, extracting the
information necessary to estimate the two latent variables from the first three waves of the panel
survey. Only the communalities of the indicators are therefore taken into account, which means that
we get rid of the bias associated with random measurement error for both constructs. Hence, the
only coefficients that will tap the stability of both phenomena will be those linking the latent con-
structs with the indicators of duty and turnout measured in the last wave of the panel. The diagonal
arrows specify the two coefficients of greatest interest in this model (see Figure 2). The one going
from ‘previous duty’ to turnout in the 2011 elections points to a causal effect from the attitude to
subsequent behavior. The one linking turnout in the 2008 general election to duty measured in wave
4 taps a reverse causality phenomenon in which the individual produces an answer to be in line with
past behavior, or experiences a genuine change in her beliefs about voting, after having participated
(or abstained). Note again that an election took place between waves 3 and 4, and therefore it makes
sense to assume that previous duty is affecting behavior in that particular election, and not before.
Finally, this model estimates the effects of rationalization or learning and duty-driven turnout con-
trolling for the lags of duty and turnout. The two models displayed in Figures 1 and 2 are to be
estimated using maximum likelihood, which makes uses of all available data points in the presence
of missing data (Little and Rubin, 1987).
A last test of causality will be conducted using these very same data and a completely different
methodological approach. The effects of previous duty on turnout in the last election of each country
will be estimated only among those young enough to have participated in these elections for the first
Figure 1. Cross-lagged estimation model for duty and turnout: Canada.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
8 International Political Science Review
time, which are Canadians under 20 years old in 2008 and Spaniards under 21 years old in 2011 (32
and 83 individuals, respectively, who were excluded from the previous analyses). These subsamples
of younger citizens had their first opportunity to participate in an election between panel waves;
hence their reported duty in t–1 (wave 1 for Canada, wave 3 in the case of Spain) necessarily precedes
their reported vote in t1, the survey immediately following their first general election.
In the next section we present the empirical evidence. We show the results of estimating the
cross-lagged models displayed in Figures 1 and 2. Finally, we will test a simple logistic regression
that estimates the predictive power of duty on the electoral behavior of our youngest respondents.
Results
The crucial piece of evidence for testing the direction of causality is provided by estimating the
cross-lagged longitudinal models.6 Note that some of the equations involved in these analyses are
equivalent to a Granger causality test.7 As noted, these estimations have the virtue of considering
several dependent variables at the same time, controlling for lags and measurement error when
more than an indicator is available for a particular construct. The results for the first estimation
with the Canadian data are displayed in Figure 3 and Table 1, including factor loadings that stem
from the measurement model (CFA) and regression coefficients that read as classic ordinary least
squares (OLS) regression coefficients.8
The high factor loadings for the two latent ‘duty’ constructs suggest that civic duty is accurately
measured. However, the regression weights are the most revealing estimates. All of them are
Figure 2. Cross-lagged estimation model for duty and turnout: Spain.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
Galais and Blais 9
significant and positive and since these are standardized coefficients, they can be compared and read
in terms of how many standard deviations the dependent variable will change with an increase of one
standard deviation in the independent variables. Stability is high, as the effect of the lags on duty and
Figure 3. Results of the cross-lagged estimation: Standardized coefficients, Canada.
Table 1. The cross-lagged longitudinal model: Unstandardized coefficients, Canada.
Independent variable Dependent variable Estimate S.E. P
Turnout federal elections
2006 (wave 1)
→Duty (wave 2) .6 .075 .000
Duty ( wave 1) →Duty (wave 2) .55 .020 .000
Duty (wave 1) →Duty federal elections (wave 1) 1.0 –
Duty (wave 1) →Duty provincial elections (wave
1)
1.06 .013 .000
Duty (wave 1) →Duty local elections (wave 1) .74 .017 .000
Duty (wave 2) →Duty federal elections (wave 2) 1.0 –
Duty (wave 2) →Duty provincial elections (wave
2)
1.06 .016 .000
Duty (wave 2) →Duty local elections (wave 2) .74 .018 .000
Duty (wave 1) →Turnout federal elections 2008
(wave 2)
.05 .005 .000
Turnout federal elections
2006 (wave 1)
→Turnout federal elections 2008
(wave 2)
.52 .021 .000
Sample size: 2141 Number of distinct sample moments: 44 CFI: .999
Number of distinct parameters to be estimated: 31 RMSEA: .026
Degrees of freedom (36 – 22): 13 Chi-square = 31,31
P-value = .003
CFI: Comparative Fit Index; RMSEA: Root Mean Squared Error of Approximation.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
10 International Political Science Review
Figure 4. Results of the cross-lagged estimation: standardized coefficients, Spain. Gray, italic parameters
indicate non-significant estimates.
Table 2. The cross-lagged longitudinal model estimates: Unstandardized coefficients, Spain.
Independent variable Dependent variable Estimate S.E. P
Previous turnout 2008 →Duty (t4) –.03 .078 .000
Previous turnout 2008 →Turnout 2008 (wave 1) 1.0 –
Previous turnout 2008 →Turnout 2008 (wave 2) 1.03 .025 .000
Previous turnout 2008 →Turnout 2008 (wave 3) .98 .026 .000
Previous Duty →Duty wave 1 1.0 –
Previous Duty →Duty wave 2 1.05 .033 .000
Previous Duty →Duty wave 3 1.13 .034 .000
Previous Duty →Duty wave 4 1.13 .039 .000
Previous turnout 2008 →Turnout 2011 ( wave 4) .51 .031 .000
Previous Duty →Turnout 2011 ( wave 4) .08 .012 .000
Sample size: 2569 Number of distinct sample moments: 44 CFI: .985
Number of distinct parameters to be
estimated: 28
RMSEA: .05
Degrees of freedom (44 – 38): 16 Chi-square = 118.4
P-value=.000
CFI: Comparative Fit Index; RMSEA: Root Mean Squared Error of Approximation.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
Galais and Blais 11
turnout measured in wave 2 shows. Nevertheless, what stand out are the cross-lagged coefficients.
The rationalization hypothesis finds empirical support in the standardized coefficient of .15, but there
is also clear support for the hypothesis that duty causes turnout (.19). Since both coefficients are sig-
nificant, we cannot reject our two hypotheses, but we can conclude that the causal effect from civic
duty to vote is at least as strong – if not stronger – as the rationalization effect. Table 1 displays the
unstandardized coefficients for the same model. All the coefficients for the measurement model are
significant. The most interesting coefficients are the first and the next to last, estimating a rationaliza-
tion and a duty causal effect, respectively. Both of them are significant and positive.
The last relevant piece of information in this table refers to the fit of the model. A significant chi-
square is not generally good news for the model fit but it is also known that when structural equation
models are run with a sample over 400 observations, the chi-square is almost always statistically
significant. Moreover, the stronger the correlations between the variables included in the model, the
higher the chances that this statistic is significant (Kenny and McCoach, 2003). We should thus look
for alternative measures of fit. One of the most common alternatives to chi-square is the RMSEA,
which ideally should be lower than 0.05 to be considered a ‘good’ model, or, at least, lower than 0.08
to be considered acceptable (MacCallum et al., 1996). In our case, the RMSEA is 0.026, which
indicates that the model fits accurately the data. The same conclusion can be drawn from the value
of the CFI, which is close to 1 and therefore indicates a very good fit of the model.9
Figure 4 and Table 2 display the standardized and unstandardized estimates for the cross-lagged
model with the Spanish data. All the regression coefficients are significant except for one: the one
from previous turnout to duty measured at time 4. This suggests that there is no rationalization in the
last two waves of the panel.10 It is also interesting to note that duty to vote is remarkably stable. Last,
but not least, we find a positive, significant effect of previous duty to vote on subsequent electoral
behavior. This supports the claim that the act of voting is a product of a pre-existing sense of duty.
We found a similar pattern in Canada, which strengthens the robustness of our results. Finally, and
just as for Canada, the model fit statistics point to a good reproduction of the co-variation matrix.
RMSEA is 0.05 and CFI is close enough to 1 to indicate a good model fit.
A last robustness check on the predictive power of duty, performed only among those who had
their first chance to vote in the Canadian 2008 election and the Spanish 2011 election, is displayed in
Table 3. The first model for each country estimates the effect of lagged duty to vote on the reported
electoral behavior, that is, whether or not they voted in the general election. What we see is a positive
and significant effect of previously reported duty on their first reported electoral behavior. Unlike
previous estimations, we have included here sex and education as controls, and we see that the effect
of previous duty stands when controlling for these factors. This indicates that a pre-existing sense that
voting is or is not a duty shapes individuals’ decision to vote or abstain in their first election. More
precisely, the predicted probability of voting for a young Canadian with a low level of duty (value 1,
hence dutiful but not very much) keeping the values of the other covariates as they are – is 54%. If
we move to the maximum level of duty, this probability increases to 70%. In Spain, the predicted
probability of having voted in the 2011 election (wave 4) for a youngster with a low level of duty in
the previous wave – and at given values for gender and education – is 57 %. If he or she very strongly
feels a duty to vote, then the likelihood of voting increases to 82%.
Conclusion and discussion
A large body of literature has shown that benefits and costs cannot fully account for the decision to
vote or abstain, and that some form of expressive benefits or moral obligations should be added to
the equation. Thus, most models include a ‘Duty’ term that improves their explanatory power. Yet
there is still a lot to learn about this ‘D’ term. Citizens can develop or reinforce their sense of civic
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
12 International Political Science Review
duty after an election. Furthermore, civic duty may be a mere rationalization of past voting behav-
ior, and then duty would not have any real causal effect on turnout. Against this critical view, a
more classical, ‘cultural’ explanation stands out, characterized by the belief that attitudes are the
product of socialization. According to this perspective, the civic duty norm is internalized (or not)
at some point in early stages of life and translates into predispositions for or against voting in elec-
tions. Our study puts these perspectives to empirical test.
The only way to address the issue of endogeneity, outside experiments, is to use longitudinal
data. This allows us to put some order in the sequence of events, guaranteeing that the cause pre-
cedes the consequence. For this purpose, we have used a two-wave panel survey conducted in 2008
and 2009 in two Canadian provinces and a four-wave survey conducted between 2010 and 2012 in
Spain. The two surveys are different in terms of their time structure (two versus four waves) and
also because the Canadian survey established a difference between the levels of government that
was absent in the Spanish case, and of course the two surveys were conducted in two countries that
differ considerably in terms of political culture and institutions. Despite all these differences, the
findings are similar in the two countries. The distribution of duty is similar and also appears to be
similarly stable over time.
Two cross-lagged panel models estimated the effects of the latent construct ‘civic duty’ –thus,
controlling for measurement error – on electoral behavior in posterior elections. The estimates of
the structural equation models are similar in the two countries. Hence, there is strong evidence of
a causal effect from civic duty to subsequent turnout in both Canada and Spain. Our logit estima-
tions restricted to the youngest citizens for whom the 2008 Canadian and 2011 Spanish election
was their first opportunity to cast a vote point in the same direction. Duty to vote measured a wave
before their first election that has a positive and significant effect on the decision to vote or abstain.
Yet our findings differ in one respect. The results for Canada are more ambiguous, revealing that
the act of voting in turn increases the propensity to construe voting as a duty. In other words, there
is evidence of some rationalization or learning process in Canada but not in Spain. It is possible
that measurement errors are more efficiently corrected by tapping civic duty at different points in
time (as was the case in Spain) than by asking about sense of duty for different types of elections.
However, these results also suggest that further research should examine the contextual factors that
encourage or discourage rationalization.
Table 3. Logistic estimation of turnout among those who were eligible to vote for the first time.
Canada Spain
Duty t–1 1.3** 1.3** .67*** .66**
(.61) (.57) (.21) (.23)
Sex (male) .09 –.75
(.85) (.52)
Education .34 .21
(.35) (.21)
Constant −2.9 −4.9 −.43 −1.0
(1.6) (2.7) (.38) (1.02)
Pseudo R2.18 .21 .1 .13
N32 32 83 83
***p < .001; **p < .05. Standard errors in parentheses.
Duty t−1 (values from 0, choice, to 3, duty very strongly) is measured in wave 3 for Spain and wave 1 for Canada (only
considering federal elections). The dependent variable is turnout in the last election, which is the 2008 election in
Canada and the 2011 election in Spain. The analyses are restricted to the third and fourth waves for the Spanish case.
Education is a 10-category variable from ‘no schooling’ to ‘complete MA or PhD’.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
Galais and Blais 13
Whether there is rationalization or not may well depend on where sense of duty comes from. It
is thus crucial to understand the reasons that make people believe that it is their duty to vote, and
especially to unravel the cognitive and emotional bases of that belief. Do people feel that they
ought to vote because they recognize that democracy works better if most people vote or is duty an
emotional reaction that leads to pride (when voting) or guilt (when abstaining)? Clearly our study
cannot address these fundamental questions.
Whatever the case, a clear lesson we draw is about the necessity of using panel data to
unravel the ‘true’ impact of civic duty on turnout. Because of the presence of the public norm
that it is every citizen’s duty to vote in democratic elections, we cannot rule out the possibility
that people pay lip service to the norm even if they have not internalized it, especially if they
have voted.
In short, our findings clearly show that civic duty is not mere rationalization. In both Canada
and Spain, those who feel that they have a moral obligation to vote are subsequently prone to act
consistently with their ethical views. We can conclude therefore that there is evidence about the
effects of sense of duty on the decision to vote in elections. The bottom line is that how one con-
strues voting does matter. The D term should be included in explanatory models of the decision to
vote or not to vote in elections. The next logical step is to examine the sources of duty. Future
research should investigate how and why some citizens believe that voting is a civic duty while
others think that it is a matter of personal choice.
Funding
The Spanish survey (study CIS 2855) was sponsored and funded by the Centro de Investigaciones Sociológicas
and the Universitat Autónoma de Barcelona (UAB, P.I. Eva Anduiza), and has benefited from a grant of the
Spanish Ministry of Science and Innovation (CSO2010-18534).
Notes
1. The sample was built from invited survey firm (Millward Brown) panelists. A comparison of this panel
survey with a representative survey reveals that in wave 1 there are 14 percentage points more respond-
ents with a college degree. This may lead to some overestimation of the overall levels of duty to vote, but
we do not think this compromises our estimates.
2. Individuals were asked the aforementioned Blais and Achen question on duty three times, with respect to
the federal, provincial and local elections. For every government level for which they answered ‘duty’,
individuals were then asked ‘How strongly do you feel personally that voting is a duty?’. We then coded
duty to vote for every government level so as choice = 0, duty not very strongly = 1, duty somewhat
strongly = 2, duty very strongly = 3.
3. In this case, the Achen and Blais indicator on duty to vote did not refer to an election in particular, but
to elections in general. As for Canada, the indicators take the values 0–3, where 0 equals ‘choice’ and 3
equals ‘duty, very strongly’.
4. Other alternatives for latent variables measured using a single indicator include fixing each error vari-
ance to one and specifying those errors to be uncorrelated across waves (Wiley and Wiley 1970, 1974),
or allowing correlated errors. We discarded these options after realizing that (a) they yield similar results
to ours despite their greater complexity and (b) they yield poor model fit measures. Other solutions, such
as latent growth linear and curvilinear models, were tested with similar outputs. Yet the results were
inadmissible due to negative variances.
5. Gastil and Xenos (2010) also deal with reciprocal causality between political engagement and civic atti-
tudes using a two- wave panel and SEM, but they do not control for lagged attitudes or behaviors, and
their study is confined to one particular setting (King County).
6. All descriptive statistics are available on request. Sense of civic duty in Canada is relatively strong for
both federal and provincial elections (about 36–38% of respondents expressing strong duty) and it is
weaker with respect to local elections. Less than 30% of Spaniards feel a very strong duty to vote. Both
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
14 International Political Science Review
countries exhibit a general pattern of stability, although Canada seems more stable and the proximity of
an election seems to slightly arouse civic duty in Spain.
7. A set of ‘variables, X, is said to Granger-cause another set, y, if adding past values of x in a regression
equation for predicting y, which already includes all past values of y as regressors, improves the predic-
tive power of the equation in the sense that it reduces the mean squared forecast error’ Buiter, 1984: 161).
8. In Figure 4, the factor loadings are located above the arrows that go from the latent ‘duty’ construct to
each of the duty indicators, while in Table 3 they correspond to the rows where arrows go from duty to
its indicators. The first regression coefficient is the one linking Duty (wave 1) to Duty (wave 2), that is,
the result of regressing Duty 2 on Duty 1, controlling for all the other relationships included in the model.
There are three more regression coefficients resulting from regressing Turnout 2008 on Turnout 2005,
Duty (wave 2) on Turnout 2006 and Turnout 2008 on Duty (wave 1).
9. An additional check of the robustness of this model – not shown here – is a multi-group analysis for
Quebec and British Columbia. The results – available on request – showed almost identical results for
both samples. In both cases our two competing hypotheses find empirical support, and the fit model
indicators were equally satisfactory.
10. Several alternative models were estimated, including some without latent constructs. None of them
yielded significant estimates for rationalization between waves 3 and 4, which confirms the robustness
of our results.
References
Ajzen, Icek and Martin Fishbein (1980) Understanding Attitudes and Predicting Social Behavior. Englewood
Cliffs, NJ: Prentice-Hall, Inc.
Aldrich, John H (1993) Rational choice and turnout. American Journal of Political Science 37(1): 246–278.
Barry, Brian (1970) Sociologists, Economists, and Democracy. London: Collier-Macmillan.
Batson, C Daniel (1975) Rational processing or rationalization? The effect of disconfirming information on a
stated religious belief. Journal of Personality and Social Psychology 32(1): 176–184.
Bem, Daryl J (1967). Self-perception: An alternative interpretation of cognitive dissonance phenom-
ena. Psychological review 74(3), 183.
Bendor, Jonathan, Daniel Diermeier and Michael Ting (2003) A behavioral model of turnout. American
Political Science Review 97(2): 261–280.
Berrington, Ann, Peter Smith and Patrick Sturgis (2006) An overview of methods for the analysis of panel
data. Discussion Paper. N/A. Mimeo. Available at: http://eprints.ncrm.ac.uk/415/.
Blais, André (2000) To Vote or Not to Vote? The Merits and Limits of Rational Choice. Pittsburgh, PA:
University of Pittsburgh Press.
Blais, André and Christopher H Achen (2010) Taking civic duty seriously: political theory and voter turnout.
Working paper. Available at: http://www.princeton.edu/csdp/events/Achen 031110/Achen031110.pdf.
Blossfeld, Hans-Peter and Götz Rohwer (1997) Causal inference, time and observation plans in the social
sciences. Quality and Quantity 31(4): 361–384.
Brown, Timothy A (2006) Confirmatory Factor Analysis for Applied Research. New York: Guilford.
Buiter, Willem H (1984) Granger-causality and policy effectiveness. Economica 51(202): 151–162.
Burkholder, Gary J and Lisa L Harlow (2003) An illustration of a longitudinal cross-lagged design for larger
structural equation models. Structural Equation Modeling 10(3): 465–486.
Campbell, Angus, Gerald Gurin and Warren Miller (1954) The voter decides. Evanston, IL: Row, Peterson.
Campbell, Angus, Philip E Converse, Warren E Miller, et al. (1960) The American Voter. New York: Wiley.
Campbell, David E (2006) Why We Vote: How Schools and Communities Shape our Civic Life. Princeton, NJ:
Princeton University Press.
Campbell, Donald Thomas and David A Kenny (1999) A Primer on Regression Artefacts. New York: Gilford.
Clarke, Harold D and Alan C Acock (1989) National elections and political attitudes: The case of political
efficacy. British Journal of Political Science 19(4): 551–562.
Clarke, Harold D, David Sanders, Marianne C Stewart, et al. (2004) Political Choice in Britain. Oxford:
Oxford University Press.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
Galais and Blais 15
Coleman, James S (1990) Foundations of Social Theory. Cambridge, MA: Harvard University Press.
Dalton, Russell J (2008) The Good Citizen: How a Younger Generation is Reshaping American Politics.
Washington, DC: CQ Press.
Delli Carpini, Michael X, Fay Lomax Cook and Lawrence R Jacobs (2004) Public deliberation, discursive
participation, and citizen engagement: A review of the empirical literature. Annual Review of Political
Science 7: 315–344.
Dennis, Jack (1970) Support for the institution of elections by the mass public. The American Political
Science Review 64(3): 819–835.
Dowding, Keith (2005) Is it rational to vote? Five types of answer and a suggestion. The British Journal of
Politics and International Relations 7(3): 442–459.
Downs, Anthony (1957) An Economic Theory of Democracy. New York: Harper Collins Publishers.
Easton, David, Jack Dennis and Sylvia Easton (1969) Children in the Political System: Origins of Political
Legitimacy. New York: McGraw-Hill.
Engelen, Bart (2006) Solving the paradox. The expressive rationality of the decision to vote. Rationality and
Society 18(4): 419–441.
Finkel, Steven E (1985) Reciprocal effects of participation and political efficacy: A panel analysis. American
Journal of Political Science 29(4): 891–913.
Finkel, Steven E (1995) Causal Analysis with Panel Data (No. 105). Sage.
Finkel, Steven E and Edward N Muller (1998) Rational choice and the dynamics of collective political
action: Evaluating alternative models with panel data. American Political Science Review 92(1):
37–49.
Finkel, Steven E, Edward N Muller and Karl-Dieter Opp (1989) Personal influence, collective rationality, and
mass political action. The American Political Science Review 83(3): 885–903.
Fiorina, Morris P (1976) The voting decision: Instrumental and expressive aspects. Journal of Politics 38(2):
390–413.
Fiorina, Morris P (1989) Retrospective Voting in American National Elections. New Haven, CT: Yale
University Press.
Fishkin, James S (1995) The Voice of the People: Public Opinion and Democracy. New Haven, CT: Yale
University Press.
Fowler, James H and Cindy D Kam (2007) Beyond the self: Social identity, altruism, and political participa-
tion. Journal of Politics 69(3): 813–827.
Gastil, John and Michael Xenos (2010) Of attitudes and engagement: Clarifying the reciprocal relationship
between civic attitudes and political participation. Journal of Communication 60(2): 318–343.
Gerber, Alan S, Donald P Green and Christopher W Larimer (2008) Social pressure and voter turn-
out: Evidence from a large-scale field experiment. American Political Science Review 102(01):
33–48.
Grofman, Bernard (1993) Is turnout the paradox that ate rational choice theory? In: Bernard Grofman (ed.)
Information, Participation, and Choice: An Economic Theory of Democracy in Perspective. Ann Arbor,
MI: University of Michigan Press.
Harder, Joshua K (2008) Why do people vote? The case for political efficacy. Stanford University work-
ing papers. Mimeo. Available at: http://cspc.nonprofitsoapbox.com/storage/documents/Fellows2008/
Harder.pdf.
Hooghe, Marc and Ellen Quintelier (2013) Do all associations lead to lower levels of ethnocentrism? A two-
year longitudinal test of the selection and adaptation model. Political Behavior 35(2): 289–309.
Jones, Philip and John Hudson (2000) Civic duty and expressive voting: Is virtue its own reward? Kyklos 53(1):
3–16.
Kanazawa, Satoshi (1998) A possible solution to the paradox of voter turnout. The Journal of Politics 60(04):
974–995.
Kenny, David A (1975) Cross-lagged panel correlation: A test for spuriousness. Psychological Bulletin 82(6):
887–903.
Kenny, David A and D Betsy McCoach (2003) Effect of the number of variables on measures of fit in struc-
tural equation modeling. Structural Equation Modeling 10(3): 333–351.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
16 International Political Science Review
Knack, Stephen and Martha E Kropf (1998) For shame! The effect of community cooperative context on the
probability of voting. Political Psychology 19(3): 585–599.
Leighley, Jan (1991) Participation as a stimulus of political conceptualization. Journal of Politics 53 (1):
198–211.
Lenz, Gabriel S (2009) Learning and opinion change, not priming: Reconsidering the priming hypothesis.
American Journal of Political Science 53(4): 821–837.
Liska, Allen E (1974). Emergent issues in the attitude-behavior consistency controversy. American
Sociological Review 39(2): 261–272.
Little, Roderick JA and Donald B Rubin (1987) Statistical Analysis with Missing Data. New York: John
Wiley.
MacCallum, Robert C, Michael W Browne and Hazuki M Sugawara (1996) Power analysis and determination
of sample size for covariance structure modeling. Psychological methods 1(2): 130.
Markus, Gregory B and Philip E Converse (1979) A dynamic simultaneous equation model of electoral
choice. American Political Science Review 73(4): 1055–1070.
Marsh, David (1971) Political socialization: The implicit assumptions questioned. British Journal of Political
Science 1(4): 453–466.
Marsh, Herbert W and Alexander Seeshing Yeung (1997) Causal effects of academic self-concept on aca-
demic achievement – structural equation models of longitudinal data. Journal of Educational Psychology
89(1): 41–54.
Matsusaka, John G and Filip Palda (1999) Voter turnout: How much can we explain? Public Choice 98(3–4):
431–446.
Mueller, Dennis C (1989) Public Choice II. Cambridge: Cambridge University Press.
Mueller, Dennis C (2003) Public Choice III. Cambridge: Cambridge University Press.
Owen, Guillermo and Bernard Grofman (1984) To vote or not to vote: The paradox of nonvoting. Public
Choice 42(3): 311–325.
Page, Benjamin I and Calvin C Jones (1979) Reciprocal effects of policy preferences, party loyalties and the
vote. American Political Science Review 73(4): 1071–1089.
Rahn, Wendy M, Jon A Krosnick and Marijke Breuning (1994) Rationalization and derivation processes in
survey studies of political candidate evaluation. American Journal of Political Science 38(3): 582–600.
Raney, Tracey and Loleen Berdahl (2009) Birds of a feather? Citizenship norms, group identity, and political
participation in western Canada. Canadian Journal of Political Science 42(01): 187–209.
Riker, William. H and Peter C Ordeshook (1968) A theory of the calculus of voting. The American Political
Science Review 62(1): 25–42.
Saris, Willem E and Chris Aalberts (2003) Different explanations for correlated disturbance terms in MTMM
studies. Structural Equation Modeling 10(2): 193–213.
Uhlaner, Carole Jean (1986) Political participation, rational actors, and rationality. Political Psychology 7(3):
551–573.
Usher, Dan (2011) What exactly is a duty to vote? Queen’s Economics Department Working Paper, Nº 1266,
10–2011. Available at: http://qed.econ.queensu.ca/working_papers/papers/qed_wp_1266.pdf.
Van Elsas, Erika J, Rozemarijn Lubbe, Tom WG van der Meer, et al. (2013) Vote recall: A panel study on the
mechanisms that explain vote recall inconsistency. International Journal of Public Opinion Research.
Epub ahead of print October 2013. DOI:10.1093/ijpor/edt031.
Venter, Anre, Scott E Maxwell and Erika Bolig (2002) Power in randomized group comparisons: The value
of adding a single intermediate time point to a traditional pretest–posttest design. Psychological Methods
7(2): 194–209.
Verba, Sidney, Kay Lehman Schlozman and Henry E Brady (1995). Voice and Equality: Civic Voluntarism
in American Politics. Cambridge, MA: Harvard University Press.
Weinschenk, Aaron C (2014) Personality Traits and the Sense of Civic Duty. American Politics Research 42(1):
90–113.
Wiley, David E and James A Wiley (1970) The estimation of measurement error in panel data. American
Sociological Review 35(1): 112–117.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
Galais and Blais 17
Wiley, James A and Mary G Wiley (1974) A note on correlated errors in repeated measurements. Sociological
Methods & Research 3(2): 172–188.
Wolfinger, Raymond E and Steven Rosenstone (1980). Who Votes? New Haven, CT: Yale University Press.
Author biographies
Carol Galais is currently a Postdoctoral Fellow at the Canada Research Chair in Electoral Studies, in the
Political Science Department, Université de Montréal. Her research interests are political socialization, civic
duty and political engagement.
André Blais holds the Canada Research Chair in Electoral Studies and is a professor in the Department of
Political Science at the Université de Montréal. He is the principal co-investigator of the Making Electoral
Democracy Work Project and the chair of the Planning Committee of the Comparative Study of Electoral
Systems (CSES). He is past president of the Canadian Political Science Association. His research interests are
elections, electoral systems, turnout, public opinion and methodology.
at UNIVERSITE DE MONTREAL on September 30, 2014ips.sagepub.comDownloaded from
