The embarrassment of riches? A meta-analysis of
individual-level research on voter turnout
Kaat Smets a,*,1
, Carolien van Ham b,1
a
Centre for the Study of Political Change (CIRCaP), University of Siena, Via Mattioli 10, 53100 Siena, Italy
b
Centre for the Study of Democracy, University of Twente, The Netherlands
a r t i c l e i n f o
Article history:
Received 19 July 2012
Received in revised form 13 December 2012
Accepted 13 December 2012
Keywords:
Individual level turnout
Review
Meta-analysis
a b s t r a c t
Voter turnout has puzzled political scientists ever since Anthony Downs postulated the
paradox of voting. Despite decades of research aiming to understand what drives citizens
to the polls, the jury is still out on what the foundations of micro-level turnout are. This
paper aims to provide a modest yet important contribution by taking a step back and
summarizing where we stand and what we know. To this end, we review 90 empirical
studies of individual level voter turnout in national elections published in ten top-journals
during the past decade (2000–2010). Through a meta-analysis of the results reported in
these studies, this paper identifies those factors that are consistently linked to individual
level turnout.
Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction
Perceived as fundamental for the functioning and
legitimacy of representative democracy, the question why
citizens participate in elections has received unabated
attention in empirical research. Many different hypotheses
have been proposed to explain voter turnout at the individual
level: from the more conventional rational choice,
sociological, and psychological explanations, to more
‘exotic’ explanations like rainfall or genetic variation. As
almost every possible factor seems to have been explored,
slowly but surely, it has become difficult to see the wood for
the trees.
From a rational choice perspective, the decision to vote is
conceptualized as the result of a personal cost-benefit calculation
in which the expected benefits of voting should
outweigh its costs (Downs, 1957). ‘Extended’ rational choice
models posit that in addition to cost-benefit considerations,
a sense of civic duty drives citizens to the polls (Riker and
Ordeshook, 1968; Blais, 2000). Alternatively, voting is seen
as an act involving the consideration not only of personal
benefits but also those of others (Fowler, 2006). The resource
model of turnout, on the other hand, hypothesizes that
turnout is driven by resources and expects turnout to be
higher for citizens with a higher economic status, more
skills, and more knowledge (Verba and Nie, 1972).
Theories of mobilization view voting essentially as social
behavior guided by norms and sanctions, and argue that
citizens go to the polls just because their family and peers
do so, or even simply because they are asked to vote by
campaigners (Arceneaux and Nickerson, 2009; Gerber and
Green, 2000). Sociological explanations of turnout have
regained prominence recently with research demonstrating
that turnout is subject to (parental) socialization,
learning and habit-formation (Plutzer, 2002; Gerber et al.,
2003). Yet another strand of research are psychological
models of turnout that stress the role of attitudes and
psychological predispositions such as political interest,
partisanship, and political efficacy in explaining voter
turnout. Lastly, the political institutional model sees the
decision to turn out as a by-product of the political and
institutional context in which citizens live.
* Corresponding author.
E-mail addresses: smets@unisi.it (K. Smets), c.t.vanham@utwente.nl
(C. van Ham).
1
Authors names are listed alphabetically, both authors contributed
equally to this project.
Contents lists available at SciVerse ScienceDirect
Electoral Studies
journal homepage: www.elsevier.com/locate/electstud
0261-3794/$ – see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.electstud.2012.12.006
Electoral Studies 32 (2013) 344–359
Clearly, the jury is still out on what the foundations of
micro-level turnout are (Arceneaux and Nickerson, 2009;
Blais, 2006). The fact that so many different theoretical
explanations exist and have found empirical support
points to the possibility that multiple causal mechanisms
explain turnout and that different causal mechanisms may
be prominent for different voters or in different contexts
(Gallego, 2010; Arceneaux and Nickerson, 2009). At the
same time, the wealth of individual-level explanations is
also the consequence of much existing research focusing
on demonstrating the validity of one particular theory or
variable, instead of assessing the relative strength of
competing theoretical frameworks in explaining turnout
(notwithstanding exceptions like Plutzer, 2002 and
Fieldhouse et al., 2007). There seems to be a lack of consensus
within the research community on a ‘core model’
of turnout (Geys, 2006). Of course, lack of consensus is
generally a sign of health for any scientific community,
however if it leads to under-specified models the resulting
omitted variable bias might lead to spurious and simply
wrong findings.
The aim of this paper is to provide a modest yet
important contribution to the current situation by taking
a step back and summarizing where we stand and what we
know. To this end, this paper reviews empirical studies of
individual level voter turnout in national elections, published
in ten top-journals in political science and political
behavior during the past decade (2000–2010). To illustrate
the non-parsimonious nature of research on individual
level turnout: the 90 studies reviewed in this paper included
over 170 different independent variables, none of
which were included in all studies.
Through a meta-analysis of the results of these studies
we aim to shed light on the factors that are consistently
linked to individual level turnout. In doing so, our research
seeks to complement the meta-analysis of aggregate level
turnout by Geys (2006). Our paper follows a similar set-up:
in section two we discuss our sample selection and coding
procedure, as well as the methods used to review the
studies. Next we consider the ways in which the dependent
variable, individual level turnout, has been measured in the
studies included in our sample. In section four we present
our empirical results, and section five presents our
conclusions.
2. Data and methods
The sheer amount of studies on individual level turnout
renders a review of all available research results impossible.
Hence, in this research project we have restricted our analyses
in a number of ways. First of all, we have chosen to
consider only peer-reviewed journal articles. We realize
that our sample selection suffers from the ‘file drawer
problem’ as research findings that are insignificant are less
likely to get published. However, while acknowledging the
selection bias this might generate in our sample, we think
the problem is less severe in the meta-analyses carried out
here. The file drawer problem is likely to affect results for
explanatory variables of interest to researchers, not for
control variables. The meta-analyses presented here cover
a wide variety of theoretical approaches and explanatory
variables, that should mitigate selection bias in favor of
certain explanatory variables.2
Secondly, we include only national parliamentary or
presidential elections, excluding local elections, as well as
European elections in the case of Europe, and primaries and
‘second-order’ elections in the United States (i.e. Senate and
Congress elections). Pooling studies that analyze turnout in
different types of elections might lead to mixed findings as
certain independent variables can affect turnout differently
in first order elections than in second order elections.
Hence, we prefer to limit our study to a sample of comparable
elections, knowing that our findings will be
restrained to that sub-set of elections, but having more
confidence in our findings.
The third restriction of our sample follows the same line
of argument and is to exclude studies on turnout in new
democracies. The literature on individual level turnout in
new democracies has only recently started to emerge.
While certainly finding similarities with established democracies,
turnout in new democracies seems to be affected
by a number of important factors that are not found to
affect electoral participation in established democracies
(Blais and Dobrzynska, 1998; Pérez-Liñán, 2001; Norris,
2002).
Finally, for more pragmatic reasons we focus on studies
published in 10 top-journals in political science (see
Table 1). Our goal was to include both general political
science journals as well as journals specializing in political
behavior, and to strike a balance between European and
American journals. We have limited the time-period to
studies published between 2000 and 2010, based on the
idea that the near past is of most interest to both scholars
and policy makers. This sample selection results in a total of
90 articles analyzing individual level turnout in national
elections in established democracies.3
Table 1 shows the
distribution of articles over journals.
The selection of journals and especially the time-period
covered is important as it might generate sample bias that
could result in our review covering certain theoretical approaches
more extensively than others. For example, work
on the impact of mobilization on turnout has received quite
some attention in recent years, leading to more studies on
variables associated with that theoretical framework, and
more robust findings for those variables.
To check the representativeness of our sample in terms
of journal selection, we bench-marked our selection of
journals against the journals cited by the Social Science
Citation Index as having published most articles on turnout
in the period between 2000 and 2010. Almost 70% of
articles on turnout were published in political science
journals, compared to about 10% for economics journals
2
Moreover, robustness checks including only control variables demonstrate
that our results remain virtually the same, even if variables of
interest are excluded (cf. online Appendix C).
3
Articles that analyzed a composite index of various forms of political
participation including turnout were excluded as the effects of independent
variables on turnout cannot be isolated. Articles based on data
from laboratory experiments were also not included in our analyses.
We did, however, include articles using data generated from field
experiments.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359 345
and about 5% for sociology journals, justifying our focus on
political science journals. Moreover, within the political
science journals, we included 8 of the 15 journals that
published most articles on turnout, and added 2 additional
European journals to have a better spread over articles
using American and European data.4
Based on these statistics,
we feel confident our sample is representative of
studies of individual level turnout in national elections of
established democracies published between 2000 and
2010.
In addition to journal selection, as a way of gauging the
‘representativeness’ of our sample in terms of theoretical
approaches, we have categorized explanatory factors of
turnout in six broad theoretical models that we feel reflect
the main theoretical approaches in the literature (see
Section 4). Subsequently we coded each study according to
the model or models it reflects. As Table 2 demonstrates,
resource and mobilization studies are somewhat more
common in our sample, while socialization and rational
choice studies are less common. This may be a consequence
of the time-period assessed in this study. Note
that some studies test multiple theoretical models, hence
the total number of studies in the table exceeds the original
number 90.
After a double-blinded selection of studies based on
careful reading of the abstracts of all papers published in
the ten journals specified above, the sample was coded
following a precise coding procedure. A codebook has been
developed that specifies characteristics of the data, the
dependent variable used, the independent variables modeled,
statistical techniques used, as well as the study results.
Inter-coder reliability was enhanced by test-coding
a substantial sub-set of the data at the start of the
research project and reconciling differences, as well as an
assessment by both authors when questions in coding
arose. Coding decisions were recorded for each study in
separate log-files. Finally, all final codings were double
checked by one of the authors.5
2.1. Meta-analysis
Meta-analysis is often defined as an analysis of analyses
(Glass, 1976, 3). Instead of reviewing studies on a given
topic in a descriptive way, the aim of meta-analysis is to
analyze test results from previous studies through quantitative
methods and to summarize the findings.
In this paper we use a combination of the vote-counting
procedure and the combined-tests technique (see Imbeau
et al., 2001; Geys, 2006). In the vote-counting procedure,
each test of a hypothesis is considered a ‘success’ when
a coefficient is statistically significant and has the hypothesized
direction. On the other hand, the hypothesis
test is considered a ‘failure’ when it is found not to be
significant and an ‘anomaly’ when the coefficient is statistically
significant but is in the opposite direction than
expected. We have used the two-tailed p < 0.05 level as
the cut-off point for significant effects. Considering all
tests together for each independent variable, the modal
category gives an estimate of the most common relationship
between the independent variable and turnout, and
dividing the number of ‘successes’ by the total number of
tests provides a measure of the success rate (see Equation
(1)). The higher the success rate, the more confident we
are that an independent variable has the hypothesized
effect on individual level voter turnout, both in terms of
direction and significance.
success rate ¼ ðsuccesses=number of testsÞÃ100 (1)
Because some studies include more tests than others
(e.g. the same hypothesis can be assessed in multiple
models within in a given study/article), looking at the
separate test results may lead to biased results when the
distribution of tests over studies is highly skewed. Moreover,
various test results from a single study are not independent
from one another as they often use the same
data (Wolf, 1986, 14). To resolve this problem we calculate
the success rate not only per test but also per study,
implicitly assigning a weight to each test result that is the
inverse of the number of tests performed in the study. A
hypothesis is considered a ‘success’, ‘failure’, or ‘anomaly’
based on which of these three categories the majority of
Table 1
Distribution studies among journals.
Journal Studies coded
Journal of Politics 22 (24.4%)
Political Behavior 18 (20.0%)
American Journal of Political Science 14 (15.6%)
British Journal of Political Science 8 (8.9%)
Electoral Studies 8 (8.9%)
Journal of Elections, Public Opinion and Parties 6 (6.7%)
European Journal of Political Research 5 (5.6%)
Political Analysis 5 (5.6%)
American Political Science Review 3 (3.3%)
Acta Politica 1 (1.1%)
Total 90
Table 2
Distribution of the main theoretical models.
Theoretical model Studies Tests
Resource model 35 (31.5%) 133 (28.4%)
Mobilization model 29 (26.1% 137 (29.3%)
Socialization model 5 (4.5%) 29 (6.2%)
Rational choice model 11 (9.9%) 54 (11.5%)
Psychological model 19 (17.1%) 71 (15.2%)
Institutional model 12 (10.8%) 44 (9.4%)
Total 111 468
4
Note that the fact that the majority of studies were carried out in the
United States might also affect the ‘representativeness’ of our sample. To
be precise, 68% of the studies in our sample were based on data from the
US, the remaining 32% are based mostly on cross-national survey data
from West European democracies. Space does not permit thorough
analysis of the potential differences in findings between the US and
Europe in this paper, however, this is a topic that will be taken up in
future research.
5
The dataset, codebook, and coding instructions are available upon
request from the authors.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359346
tests within the study fall. Like for separate tests, the
modal category is considered to give the best approximation
of the true direction of the relationship between
the dependent and independent variable.
The vote-counting procedure is purely based on direction
and statistical significance. It does not allow to take
into account the size of effects (Glass, 1976, 5; Lau et al.,
2007, 1179). Combined test techniques allow to summarize
the test statistics provided by different studies. However,
since the studies in our sample use different statistical
techniques and therefore provide different test statistics,
such a comparison is out of order. Instead we use proxies of
effect sizes based on whether the effects were ‘successes’,
‘failures’, or ‘anomalies’.
Combining the vote-count procedure and the
combined-tests technique in this way allows us to calculate
a proxy of the average effect size. To this end, one first
calculates the approximate effect size r for each individual
test by assigning successes (significant and in hypothesized
direction) a weight of 1, failures (not statistically significant)
a weight of 0 and anomalies (significant but not in
hypothesized direction) a weight of À1. A proxy of the effect
size at the level of tests can then be calculated with the
formula in Equation (2):
r ¼ ðsuccesses À anomaliesÞ=number of tests (2)
The average effect size of a given independent variable
across all studies (rav) is subsequently given by the mean
effect size (see Equation (3)). This metric, that has a theoretical
lower bound of À1 and an upper bound of þ1, behaves
like a correlation coefficient and gives the number of
standard deviation units with which individual level
turnout is affected if the independent variable changes by
one standard deviation. By calculating a confidence interval
around this statistic we can judge whether or not there is
a statistically significant effect on the dependent variable
(testing the null hypothesis that the mean effect across all
studies is zero).
rav ¼
X
ri=number of studies (3)
As an example, consider the effect of age on turnout,
which we hypothesize to be positive. Imagine study one
includes four tests of this hypothesis. In two of those the
effect for age is positive and significant, in one test the
effect is not significant, and in the last test the effect is
negative and significant. The modal category of this study
is ‘success’. The first two tests will receive an effect size
score of 1, the third test a 0, and the fourth test a À1. At the
level of tests, the success rate is (2/4)*100 ¼ 50%, the proxy
of the effect size r is (2–1)/4 ¼ 0.25 and a t-test will
demonstrate that this effect is not significantly different
from 0. Now, say in hypothetical studies two, three and
four the modal category is also ‘success’ and the effect size
r equals 0.75, while in study five the modal category is
‘failure’ and the effect size r is 0. At the level of studies
then, the success rate is (4/5)*100 ¼ 80%. The average effect
size rav is the sum of r divided by the number of studies
((0.25 þ 0.75 þ 0.75 þ 0.75 þ 0)/5 ¼ 0.5) and a t-test of this
effect will demonstrate that it is significantly different
from 0 (p < 0.05, two-tailed).6
3. Description dependent variable
Crucial in any attempt to review the extensive literature
on individual level turnout is to distinguish the different
ways in which the dependent variable is measured. At the
individual level, turnout is most commonly measured by
post-election surveys that ask respondents whether they
voted in the past election. Such self-reported turnout is
affected by problems such as recall bias and social desirability,
and hence reported turnout tends to have an upward
bias when compared to data on actual turnout. For
example, Karp and Brockington (2005, 825) estimate the
difference between actual turnout and reported turnout in
the American National Election Studies to be around 20
percentage points during the 1990s.
For this reason, scholars generally prefer to use validated
turnout data, based on official voter records. However,
since such data are often only released after some
time and official voter records are not kept in all countries,
validated turnout data is more difficult to obtain. A third
type of measurement that is used in studies of individual
level turnout is based on surveys that are held at some
point before elections, such as general social surveys, and
that ask respondents about their turnout intention, either
in the upcoming election or “if elections were to be held
tomorrow”. Turnout intention is likely to be even more
prone to social desirability bias and is therefore not used
often.
In terms of validity then, validated turnout data is the
most robust, but at the same time the most difficult data to
obtain, while reported turnout data (and turnout intention
data) are more prone to bias, but generally easier to obtain.
As Table 3 demonstrates, about 82% of the studies included
in this paper measure turnout as reported turnout, 11% of
the studies use validated turnout, and 7% use turnout
intention.7
4. Analysis and findings
We present our findings dividing all independent variables
into six main theoretical models of individual level
voter turnout: the resource model, the mobilization model,
the socialization model, the rational choice model, the
psychological model, and the political–institutional model.
Note that our classification of studies in these six broad
theoretical models is purely meant as a practical way to
organize our results. There are multiple ways to group
variables into theoretical models and different scholars are
likely to have different preferences. Also, as will become
clear below, the models chosen for the purpose of this
6
Note that in case of a tie, the modal study effect is always coded
conservatively. For example, if a study includes 4 tests, 2 of which are
a success and 2 are a failure, the modal effect of the study is coded as
failure.
7
Note that while the total number of studies included in our review is
90, five studies use two dependent variables, for example analyzing
models with reported turnout and validated turnout.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359 347
paper are not necessarily mutually exclusive as certain
variables can be argued to influence voter turnout through
multiple theoretical pathways.
Since our aim is to compare effects over different studies,
we keep the direction of the hypotheses constant, disregarding
the hypotheses proposed in each particular study.
For example, in some studies authors hypothesize men to
turn out at higher rates than women, while in other studies
women are expected to turn out at higher rates. In order to
compare results for these different studies, we code results
following a single hypothesis for all studies, in this case that
men turn out more than women. For each variable the hypothesized
direction of the effect is denoted by the ‘þ’ or ‘À’
sign after the variable name in the results tables.
We focus exclusively on main effects, disregarding
interaction terms.8
Moreover, we do not report variables
that were included in only one or two studies, since this
would not allow us to carry out t-tests. Online Appendix A
presents the results for the variables that were only
included in one or two studies. Variables are reported in
descending order based on the frequency with which they
were included in the studies. Overall, we found that of the
176 different independent variables included in the 90
studies reviewed, only eight (less than 5%) were included in
more than 25% of the studies: age, gender, income, education,
race, marital status, political interest and party
identification. Even the two most common independent
variables – age and education – were included in only 72%
and 74% of studies respectively. Generally, the motivation
for the inclusion of control variables in the articles
reviewed was very brief (i.e. “we include the ‘usual suspects’
as control variables”) or absent.
4.1. The resource model
The basic idea behind the resource model is that political
participation is an act driven by resources, particularly time,
money, and skills (Verba and Nie, 1972; Verba et al., 1995).
Those with jobs, a high income, and a high socio-economic
status are more likely to have a wider range of resources
and are, thus, more likely to vote (Brady et al., 1995, 273).
Education also contributes to resources, acting as a social
sorting mechanism and serving as a proxy for social class
and skills. Higher resource citizens also have larger networks
and higher stakes in elections, characteristics that
may in turn may act as motivators and lead to targeted
mobilization efforts from political parties. The cluster of
resource variables is the largest of all discussed in this paper,
comprising 22 explanatory variables, many of which are
demographic indicators. Results are summarized in Table 4.
4.1.1. Education
Education is considered one of the strongest predictors
of voter turnout and has been used as an independent
variable in 67 out of 90 studies (see e.g. Hillygus, 2005b;
Campbell, 2009; Gallego, 2010). Notwithstanding the rise
of educational levels in advanced western democracies,
turnout levels have not risen: a fact that continues to
puzzle scholars of political behavior (Burden, 2009, 540).
Our meta-analysis shows that education is indeed positively
related to individual level turnout as most studies
fall into the success category (success rate is 67–72%). The
average effect size (rav) is statistically significant both at the
level of tests and studies. A standard deviation change in
educational level increases turnout with roughly 0.72
standard deviation units. However, living in an area with
relatively higher levels of education does not seem to significantly
affect turnout (see results for contextual
education).
4.1.2. Age and age squared
Age is among the three most common independent
variables in research on individual voter turnout (65 out of
90 studies). Young adults are notorious abstainers. Turnout
is, however, hypothesized to rise with the transition into
adulthood (Lane, 1959, 218, Strate et al., 1989, 444,
Jankowski and Strate, 1995, 91) and decline again when
citizens at an older age start to withdraw from social life
(Cutler and Bengtson, 1974, 163).
This suggests the relationship between age and turnout
is curvilinear rather than linear, which is why some studies
also include age squared. Our findings show that most tests
and studies are successful, indicating support for a positive
effect of age and a negative effect of age squared on turnout
(the success rate being around 75% for the studies).
4.1.3. Gender
Because of their different role in society (e.g. being the
breadwinner, historically having the right to vote) men
have long been considered to have more resources and are,
therefore, thought to turn out more than women. Recent
research, however, suggests that the gender gap in turnout
has gradually disappeared (see Inglehart and Norris, 2003;
Childs, 2004).
Our meta-analysis indicates that gender in most instances
is no longer a statistically significant predictor of turnout in
national elections. The success rate of gender is very low
(around 3–5%), as in most tests and studies the variable does
not reach statistical significance. The average effect size is
statistically significant, but in comparison to other variables
relatively close to zero. Moreover, its negative sign indicates
that when gender is found to be significant it is usually
women that turn out at higher rates, not men.
Table 3
Measurement dependent variable.
Dependent variable Studies Tests
Reported turnout 78 (82.1%) 324 (80.0%)
Validated turnout 10 (10.5%) 54 (13.3%)
Turnout intention 7 (7.4%) 27 (6.7%)
Total 95 405
8
As an anonymous reviewer pointed out, main effects of interaction
terms are affected by the inclusion of the interaction effect and could
hence distort our results. However, secondary analysis of our results
excluding all variables that were main effects of interaction terms gave
near-identical results to the ones reported here. As additional robustness
checks we also tested whether our results hold when including only
control variables and when weighting the effect by the number of variables
included in each test. The results did not change substantively, as
reported in online Appendix C.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359348
Table 4
Resource model – results meta-analysis.
Variable Success (1) Failure (0) Anomaly (À1) Modal category Success rate Effect size (rav) p-Value
Education (D)
Tests (239) 173 66 0 Success 72.38 0.72 ***
Studies (67) 45 22 0 Success 67.16 0.71 ***
Age (D)
Tests (233) 167 60 6 Success 71.67 0.69 ***
Studies (65) 49 15 1 Success 75.38 0.74 ***
Age squared (L)
Tests (78) 44 32 2 Success 56.41 0.54 ***
Studies (17) 13 4 0 Success 76.47 0.78 ***
Gender (male) (D)
Tests (225) 11 160 54 Failure 4.89 À0.19 ***
Studies (61) 2 45 14 Failure 3.28 À0.20 **
Race (black, Latino, non-white) (L)
Tests (156) 38 109 9 Failure 24.36 0.19 ***
Studies (43) 10 30 3 Failure 23.26 0.21 *
Income (D)
Tests (151) 76 75 0 Success 50.33 0.50 ***
Studies (40) 21 19 0 Success 52.50 0.57 ***
Marital status (married) (D)
Tests (90) 41 47 2 Failure 45.56 0.43 ***
Studies (30) 17 12 1 Success 56.67 0.53 ***
Residential mobility (L)
Tests (54) 28 23 3 Success 51.85 0.46 ***
Studies (18) 10 7 1 Success 55.56 0.46 **
Region (south/periphery) (L)
Tests (52) 28 24 0 Success 53.85 0.54 ***
Studies (18) 10 8 0 Success 55.56 0.55 ***
Occupational status (employed) (D)
Tests (50) 17 33 0 Failure 34.00 0.34 ***
Studies (18) 5 13 0 Failure 27.78 0.32 **
Home ownership (D)
Tests (58) 30 27 1 Success 51.72 0.50 ***
Studies (16) 8 8 0 Failure 50.00 0.55 ***
Citizenship (nationalized/born in country) (D)
Tests (41) 13 22 6 Failure 31.71 0.17 þ
Studies (13) 5 6 2 Failure 38.46 0.26 n.s.
Occupational type (white collar) (D)
Tests (22) 8 12 2 Failure 36.36 0.27 *
Studies (9) 4 5 0 Failure 44.44 0.35 þ
Socio-economic status/class (D)
Tests (21) 14 7 0 Success 66.67 0.67 ***
Studies (8) 4 4 0 Failure 50.00 0.50 *
Residential location (rural) (D)
Tests (18) 0 17 1 Failure 0.00 À0.06 n.s.
Studies (7) 0 7 0 Failure 0.00 À0.02 n.s.
Children (D)
Tests (18) 2 11 5 Failure 11.11 À0.17 n.s.
Studies (6) 1 3 2 Failure 16.67 À0.10 n.s.
Occupational type: students (L)
Tests (15) 0 8 7 Failure 0.00 À0.47 **
Studies (6) 0 3 3 Failure 0.00 À0.43 þ
Contextual race (L)
Tests (37) 5 24 8 Failure 13.51 À0.08 n.s.
Studies (5) 1 4 0 Failure 20.00 0.05 n.s.
Contextual education (D)
Tests (23) 9 13 1 Failure 39.13 0.35 **
Studies (4) 1 3 0 Failure 25.00 0.33 n.s.
Contextual citizenship (L)
Tests (22) 6 16 0 Failure 27.27 0.27 *
Studies (3) 1 2 0 Failure 33.33 0.43 n.s.
Generation (60s/70s vs. pre-WW II) (L)
Tests (14) 12 2 0 Success 85.71 0.86 ***
Studies (3) 3 0 0 Success 100.00 0.88 **
Contextual income (D)
Tests (9) 2 6 1 Failure 22.22 0.11 n.s.
Studies (3) 1 2 0 Failure 33.33 0.28 n.s.
Note: T-test with two-tailed significance levels. þp < 0.01, *p < 0.05, **p < 0.01, ***p < 0.001.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359 349
4.1.4. Race and citizenship
Race and citizenship are included in models of voter
turnout based on the idea that ethnic minorities often
have fewer resources and skills. We therefore assess the
hypotheses that turnout among ethnic minorities is
lower than among ethnic majorities, and that those
nationalized or born in their country of residence participate
more.
Race is included in about half of the studies, while citizenship
is considered in 13 out of 90 studies. The modal
category for both variables is ‘failure’ which implies that
a non-significant effect was found in most of the tests and
studies. The success rate of the race variable lies around
23%, while citizenship does a little better at 32–38%. The
average effect size for citizenship does not reach statistical
significance at the p < 0.05 level though. Breaking down
results by various ethnic and racial groups does not change
the overall picture (see online Appendix B). Living in an
area with relatively high proportions of inhabitants from
ethnic minorities or naturalized citizens also does not significantly
affect turnout, as the variables contextual race
and contextual citizenship show.
4.1.5. Income, occupational status, occupational type and social
status
Following the resource model, we expect income to
have a positive impact on turnout, middle-class citizens to
turn out at higher levels than lower class citizens, those
with white collar jobs to participate more than citizens
with blue collar jobs, and students to vote less.
As indicated by the modal category, income appears to
have a positive effect on turnout, even though the number
of successes and failures are almost even. The average
success rate for income lies around 50% with an average
effect size of 0.50 and 0.57 for tests and studies respectively.
At both levels rav is highly significant. Living in more
affluent areas however does not seem to have a significant
impact on turnout (see contextual income). Those from
higher social classes indeed systematically turn out at
higher rates, even though at the study level the modal
category is a tie between success and failure. The average
effect size is significant both for tests and studies though,
and varies between 0.50 and 0.67.
Lastly, occupational status and type did not appear to
have statistically significant effects on turnout in most
studies. White collar workers were not found to turn out at
higher rates, nor were students found to turn out less. In
fact, judging by the zero successes and the negative sign of
the average effect size for students (rav ¼ À0.43), students
were rather found to turn out at higher rates. While students
may not have high paid jobs, they do often come from
higher socio-economic backgrounds and moreover are
potentially socialized into voting by a politically more
stimulating environment than their non school-going peers
(see e.g. Highton and Wolfinger, 2001; Tenn, 2007).
4.1.6. Marital status and children
Marriage not only has practical consequences such as
increased residential stability, married citizens may also be
motivated by a politically active spouse (Stoker and
Jennings, 1995, 422). Married couples are, moreover,
thought to be more likely to conform to the idea of ‘good
citizenship’ and consider political engagement a civic duty
(Denver, 2008). In a similar vein Lane (1959, 218) points
out that having children increases the awareness of social
needs and the responsibility to perform as a good role
model. Solt (2008), however, argues that while married
couples may be more likely to remind each other to vote,
they are nonetheless less likely to have free time and
spend this scarce time to engage in politics. The arrival of
children likewise distracts parents from participating in
politics.
While marital status is included in one-third of the
studies, the impact of having children is much less frequently
researched. At the level of studies the positive effect
of marital status on turnout is confirmed, though at the
level of tests the modal effect is insignificant. The average
effect size for marriage is nonetheless statistically significant
for both tests and studies. The effect of having children
appears to be insignificant in most studies.
4.1.7. Home ownership and residential mobility
Citizens that own a property are usually more grounded
in a community than those that rent, thus strengthening
community ties (Lane, 1959; Highton and Wolfinger, 2001).
People that have been living in their community for a longer
time are moreover better informed about (local) political
affairs.
Our meta-analysis shows that home ownership and
residential mobility largely influence voter turnout as
expected. Residential mobility leads to lower levels of
turnout in most tests and studies, while the modal category
for home ownership is success at the test level and a tie
between failure and success at the study level. The average
effect size lies around 0.50 and is highly significant for both
variables.
4.1.8. Urbanization and region
Citizens in rural areas historically turn out at higher
levels as rural societies tend to have higher levels of associational
life. The results of the meta-analysis in Table 4
show that this hypothesis may be outdated as almost all
tests and studies find insignificant effects. Region is often
included as a control variable in countries where there are
stark differences in turnout levels for different parts of the
country, such as lower turnout rates in the South of the
United States. These variables are indeed often found to
pick up on regional differences.
4.2. The mobilization model
Turning to the next group of variables, the mobilization
model of voter turnout centers around the idea that citizens
are mobilized to participate in politics by parties,
candidates, interest groups and new social movements
(Rosenstone and Hansen, 1993). Such social networks
reduce the costs of political participation by providing information
about parties, candidates and the electoral process.
Associational life, moreover, emphasizes values that
are thought to mobilize citizens. This section covers a total
of 11 mobilization variables. The results of the metaanalysis
are shown in Table 5.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359350
4.2.1. Attendance of religious services, religion, organizational
membership and social capital
Attendance of religious services is a measure of the
mobilization characteristics of religion (see Solt 2008 and
Claassen and Povtak 2010). People can however also be
members of other organizations such as political parties,
social associations, sports clubs, etc. In all of these cases
turnout is expected to increase with higher involvement in
associational life. The impact of associational membership
on voter turnout is not solely linked to the mobilization
model but also to the socialization and resource model as
organizational membership is also thought to promote
civic commitment and skills that stimulate political participation
(Verba et al., 1995).
The results of the meta-analysis indicate that while
organizational membership is positively related to turnout,
having a religious denomination (as opposed to not being
religious) in most instances does not significantly affect
turnout. Attendance of religious services is found to affect
individual turnout roughly half of the time, which is
reflected in the success rate that lies between 50 and 57%.
General measures of social capital were included in only
three studies and were most often found not to have
a significant effect on turnout.
4.2.2. Union membership and union density
Unions mobilize their members to participate in politics
and reduce class bias by enhancing participation of those
with fewer resources. Moreover, a strong presence of
unions at the aggregate level (i.e. union density), may lead
political parties to adopt policy positions that represent
union members in an effort to win their votes (Leighley and
Nagler, 2007, 432).
As indicated by the modal category both union membership
and union density do not have a statistically significant
effect on individual level turnout in most tests and
studies. The average effect size for union membership
ranges between 0.29 and 0.48 and is statistically different
from zero. The effect size for union density is insignificant.
4.2.3. Partisan and non-partisan mobilization
Voter mobilization efforts such as Get Out The Vote
(GOTV) phone calls, canvassing and personal contacts
boost turnout as they reduce information costs. However,
the impact of partisan and non-partisan mobilization is
thought to be mediated by a more general propensity to
vote and is therefore not expected to affect all voters in
a similar way (see e.g. Karp et al., 2008; Arceneaux et al.,
2006; Dale and Strauss 2009).
Both partisan and non-partisan mobilization efforts are
indeed found to positively affect individual turnout in national
elections in most instances (see modal category and
the significant t-tests). The success rate is higher for partisan
(70%) than for non-partisan mobilization (56%). The
average effect size is likewise higher for mobilization efforts
by parties.
Table 5
Mobilization model – results meta-analysis.
Variable Success (1) Failure (0) Anomaly (À1) Modal category Success rate Effect size (rav) p-Value
Attendance of religious services (D)
Tests (40) 20 20 0 Failure 50.00 0.50 ***
Studies (14) 8 6 0 Success 57.14 0.59 ***
Union membership (D)
Tests (48) 14 34 0 Failure 29.17 0.29 ***
Studies (13) 6 7 0 Failure 46.15 0.48 **
Mobilization (partisan) (D)
Tests (27) 19 8 0 Success 70.37 0.70 ***
Studies (10) 7 3 0 Success 70.00 0.83 ***
Media exposure (D)
Tests (20) 14 6 0 Success 70.00 0.70 ***
Studies (10) 6 4 0 Success 60.00 0.57 **
Mobilization (non–partisan GOTV) (D)
Tests (32) 18 14 0 Success 56.25 0.56 ***
Studies (9) 5 4 0 Success 55.56 0.67 ***
Political advertising exposure (D)
Tests (19) 4 14 1 Failure 21.05 0.16 n.s.
Studies (7) 1 6 0 Failure 14.29 0.05 n.s.
Religious denomination (D)
Tests (17) 4 12 1 Failure 23.53 0.18 n.s.
Studies (6) 1 5 0 Failure 16.67 0.14 n.s.
Organizational membership (D)
Tests (8) 7 1 0 Success 87.50 0.88 ***
Studies (5) 4 1 0 Success 80.00 0.80 *
Total political advertisements (D)
Tests (6) 0 6 0 Failure 0.00 0.00
Studies (4) 0 4 0 Failure 0.00 0.00
Social capital (D)
Tests (8) 4 4 0 Failure 50.00 0.50 *
Studies (3) 1 2 0 Failure 33.33 0.22 n.s.
Union density (D)
Tests (3) 1 2 0 Failure 33.33 0.33 n.s.
Studies (3) 1 2 0 Failure 33.33 0.33 n.s.
Note: T-test with two-tailed significance levels. þp < 0.01, *p < 0.05, **p < 0.01, ***p < 0.001.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359 351
4.2.4. Media exposure and campaign advertisements
Exposure to (political) news in the media arguably leads
to higher levels of political information among citizens.
Prior (2005, 577) warns, however, that increasing media
choice does not per se lead to higher levels of turnout. As
the number of media outlets increases, the likelihood that
a person will encounter political news by chance diminishes
significantly. Campaign advertisements are a way
to get potential voters to focus on issues of interest to them
(Solt, 2008). Nonetheless, although political campaigns are
intended to get out votes, negative campaigns can also have
a demobilizing effect (see Goldstein and Freedman, 2002;
Anduiza, 2005; Stevens, 2008, 2009).
The results of our meta-analysis show that reading
newspapers, watching the news, listening to the radio, etc.
indeed has a positive effect on turnout (success rate 60–
70%). The average effect size rav reaches statistical significance
both at the level of tests and studies. Campaign advertisements
were generally not found to have
a statistically significant effect.9
4.3. The socialization model
The impressionable or formative years between childhood
and adulthood are generally considered a key period
during which citizens form the basis of political attitudes
and behaviors (Plutzer, 2002). The political learning curve
is mediated through various socializing agents such as
family, peers, school, mass media, and even the political
context. In this third result section we discuss 3 variables
related to the socialization model. The results are summarized
in Table 6.
4.3.1. Parental influences during adolescence
The process of political learning from parents and other
family members is tapped by various indicators. Only two
measures were included in three or more studies (see
online Appendix A for variables that are included in less
than 3 studies). Both parental educational levels and
parental socio-economic status are expected to have
a positive impact on children’s turnout levels in later life
(Sandell and Plutzer, 2005; Sandell Pacheco, 2008).
Parental income and social class seem more successful
in explaining turnout than parental educational level, for
which the modal category is failure. The positive effect of
parental income and social class is confirmed at the level of
tests and at a tie between success and failure at the level of
studies. Average effect sizes are however highly significant
at both the test and study level (rav ¼ 0.73).
4.3.2. Political discussion
Social exchange theories of political participation
emphasize how talking to friends, family members or
neighbors may persuade people to participate in politics
(see Cutts and Fieldhouse, 2009). Not only does political
discussion potentially lead to higher levels of interest and
political knowledge, emphasizing social norms (such as
turning out in elections) may also induce norm-conforming
behavior (Gerber and Rogers, 2009). The results of the
meta-analysis for political discussion are however inconclusive.
In only half of the tests and studies the positive
impact of discussing politics is confirmed, while in the
other half no significant effect was found.
4.4. The rational choice model
The rational choice model emphasizes that there is
a cost-benefit calculus of voting whereby benefits should
outweigh costs in order for a person to turn out to vote
(Downs, 1957; Riker and Ordeshook, 1968). In this section
we consider 10 variables related to the rational choice
model. The results of the meta-analysis are summarized in
Table 7.
4.4.1. Past turnout, new voter, propensity to vote, and costs of
voting
The large impact of past turnout on current turnout
decisions observed in the literature is thought to be
mediated through various mechanisms (see Cutts et al.,
2009 for an overview). First, turnout is caused by a set of
factors such as political interest or partisanship that are
relatively stable over the life span. Moreover, voting might
be self-reinforcing as it increases positive attitudes toward
voting and alters one’s self-image to the extent that voting
contributes to that image. Third, once voters have been to
the polls they face lower information barriers and can make
use of their hands-on experience and knowledge. Because
they are inexperienced, new voters are thought to turn out
less.
Past turnout and general propensity to vote are both
consistently linked with higher turnout levels. The success
rate is close to 100% and the effect size close to 1. The results
for new voters seem to indicate that those entering the
electorate are not significantly less likely to turn out, as in
most studies this variables does not reach statistical significance.
Lastly, higher costs of voting are found to
decrease the likelihood of turnout even if the effect size is
only significant at the test level.
4.4.2. Cares who wins, personal benefits and civic duty
The higher the stakes in the elections, the more inclined
citizens will be to turn out to vote. Caring about the outcome
of the election and perceived personal benefits may
both increase turnout. In fact, both variables in most instances
fall into the modal category success. The success
rate for caring about the outcome of the election is high
between 75 and 89%. The average effect size is found to be
significantly different from zero and ranges between 0.81
and 0.89. While the chances that a single person will influence
the outcome of the election are infinitely small,
a sense of civic duty may convince citizens to cast a vote
nonetheless. This hypothesis is confirmed in the vast majority
of studies, and the success rate and average effect size
are accordingly very high.
9
Note that in some cases, as for example ‘total political advertisements’
an effect size of 0.00 (or 1.00) is reported without a p-value, this
means that there was no variation in the effect sizes for that variable and
hence no p-value could be calculated. This often happens when a variable
was included in only a few tests, and those results should hence generally
be taken with a grain of salt.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359352
4.4.3. Evaluation economy, parties and candidates
Citizens may consider managing the national economy
as part of the government’s responsibilities. Experiencing
economic strain may therefore lead citizens to blame the
government for their situation and mobilize them to the
polling booth to vote the government out of office (Lipset,
1969, 187; Schlozman and Verba, 1979, 12–19). The alternative
hypothesis stipulates that economic suffering withholds
people from participating in politics. Someone who
has just lost his job is more likely to be pre-occupied with
personal economic well-being than with remote concerns
like politics (Rosenstone,1982). In this instance, the costs of
voting outweigh the benefits. Similar dual mechanisms are
at work when considering the evaluation of parties and
candidates. High approval rates foster a positive atmosphere
that may encourage and stimulate citizens to turn
out to vote. Unpopular parties and candidates, on the other
hand, may also stimulate turnout levels as citizens seek
change.
Evaluation of the national economic situation as well as
the evaluation of one’s own economic situation fail to
predict turnout in most of the studies and tests for which
these variables were included. Positive evaluations of
parties and candidates do seem to boost turnout, though
the results are not conclusive. The success rate is high
(67–82%), however the average effect size is insignificant at
the study level.
4.5. The psychological model
The fifth model of voter turnout discussed in this paper
focuses on psychological determinants of voter turnout.
Table 6
Socialization model – results meta-analysis.
Variable Success (1) Failure (0) Anomaly (À1) Modal category Success rate Effect size (rav) p-Value
Parental social class/income (D)
Tests (22) 16 6 0 Success 72.73 0.73 ***
Studies (4) 2 2 0 Failure 50.00 0.73 *
Political discussion (D)
Tests (10) 5 5 0 Failure 50.00 0.50 *
Studies (4) 2 2 0 Failure 50.00 0.58 n.s.
Parental educational level (D)
Tests (16) 7 9 0 Failure 43.75 0.44 **
Studies (3) 1 2 0 Failure 33.33 0.37 n.s.
Note: T-test with two-tailed significance levels. þp < 0.01, *p < 0.05, **p < 0.01, ***p < 0.001.
Table 7
Rational choice model – results meta-analysis.
Variable Success (1) Failure (0) Anomaly (À1) Modal category Success rate Effect size (rav) p-Value
Vote in previous election (D)
Tests (77) 75 2 0 Success 97.40 0.97 ***
Studies (12) 12 0 0 Success 100.00 0.98 ***
Cares who wins (D)
Tests (18) 16 2 0 Success 88.89 0.89 ***
Studies (8) 6 2 0 Success 75.00 0.81 ***
Evaluation national economic situation (L)
Tests (15) 4 11 0 Failure 26.67 0.27 *
Studies (7) 2 5 0 Failure 28.57 0.25 n.s.
Evaluation candidates/parties (D)
Tests (22) 18 3 1 Success 81.82 0.77 ***
Studies (6) 4 1 1 Success 66.67 0.48 n.s.
Civic duty (D)
Tests (17) 16 1 0 Success 94.12 0.94 ***
Studies (6) 6 0 0 Success 100.00 0.94 ***
Personal benefits of voting (D)
Tests (13) 10 3 0 Success 76.92 0.77 ***
Studies (4) 2 2 0 Failure 50.00 0.58 n.s.
Costs of voting (L)
Tests (9) 6 3 0 Success 66.67 0.67 **
Studies (4) 2 2 0 Failure 50.00 0.58 n.s.
New voter (L)
Tests (12) 1 7 4 Failure 8.33 À0.25 n.s.
Studies (3) 0 2 1 Failure 0.00 À0.14 n.s.
Propensity to vote (D)
Tests (6) 6 0 0 Success 100.00 1.00
Studies (3) 3 0 0 Success 100.00 1.00
Evaluation own economic situation (L)
Tests (5) 1 1 3 Anomaly 20.00 À0.40 n.s.
Studies (3) 1 1 1 Failure 33.33 0.00 n.s.
Note: T-test with two-tailed significance levels. þp < 0.01, *p < 0.05, **p < 0.01, ***p < 0.001.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359 353
Here explanatory factors range from more cognitive
characteristics such as political interest, political knowledge,
or cognitive ability to personal preferences associated
with expressive voting such as party identification
and ideology. Cognitive characteristics are expected to
function as resources, lowering the costs of voting and
increasing turnout, while ideological preferences are
expected to increase the intrinsic benefits from the act of
voting. Moreover, citizens that are politically interested
and involved are expected to have higher levels of confidence
in the influence they have on the political system
(i.e. trust in institutions, perceptions of external and internal
efficacy). Personality characteristics are also
expected to explain the degree to which people engage in
altruistic behavior – such as voting – or perceive voting as
a civic duty. Clearly, the psychological model represents
a wide variety of approaches to explaining voter turnout.
We consider 14 variables related to the psychological
model. The results of our meta-analysis are summarized in
Table 8.
4.5.1. Party identification, political interest, political knowledge
and cognitive ability
While it is seemingly evident that levels of political
involvement are positively related to voter turnout, the
proximity of the concepts of political involvement and
political participation is often considered problematic as
the decision to vote is very similar to the decision to acquire
political information and knowledge (Rubenson et al.,
2004; Denny and Doyle, 2008). Cognitive abilities are correlated
with educational levels and it has been suggested
that the large impact of education on turnout found in
many studies may be overestimated due to a lack of control
for measures of cognitive ability (see Denny and Doyle,
2008 for an overview).
Our meta-analysis shows that on the whole party
identification, political interest and political knowledge are
indeed positively related to turnout. The success rate lies
between 72 and 85% for party identification and political
interest. The average effect size rav lies between .72 and .85
and is statistically significant in all instances for these
Table 8
Psychological model – results meta-analysis.
Variable Success (1) Failure (0) Anomaly (À1) Modal category Success rate Effect size (rav) p-Value
Party identification (D)
Tests (147) 109 36 2 Success 74.15 0.73 ***
Studies (43) 31 12 0 Success 72.09 0.72 ***
Political interest (D)
Tests (91) 73 18 0 Success 80.22 0.80 ***
Studies (27) 23 4 0 Success 85.19 0.85 ***
Political efficacy (D)
Tests (48) 29 18 1 Success 60.42 0.58 ***
Studies (15) 7 8 0 Failure 46.67 0.47 **
Political knowledge (D)
Tests (21) 20 1 0 Success 95.24 0.95 ***
Studies (10) 10 0 0 Success 100.00 0.98 ***
Trust in institutions (D)
Tests (20) 6 14 0 Failure 30.00 0.30 *
Studies (9) 3 6 0 Failure 33.33 0.37 *
Cognitive ability (D)
Tests (20) 8 12 0 Failure 40.00 0.40 **
Studies (7) 2 5 0 Failure 28.57 0.33 n.s.
Ideological self-placement (right/conservative) (D)
Tests (10) 3 7 0 Failure 30.00 0.30 þ
Studies (5) 1 4 0 Failure 20.00 0.12 n.s.
Satisfaction with democracy (D)
Tests (9) 3 6 0 Failure 33.33 0.33 þ
Studies (5) 2 3 0 Failure 40.00 0.40 n.s.
Alienation/political cynicism (L)
Tests (18) 4 12 2 Failure 22.22 0.11 n.s.
Studies (4) 1 3 0 Failure 25.00 0.13 n.s.
Trust in others (D)
Tests (9) 0 9 0 Failure 0.00 0.00
Studies (4) 0 4 0 Failure 0.00 0.00
Ambivalence (L)
Tests (9) 9 0 0 Success 100.00 1.00
Studies (4) 4 0 0 Success 100.00 1.00
Ethnic identification (L)
Tests (5) 0 5 0 Failure 0.00 0.00
Studies (4) 0 4 0 Failure 0.00 0.00
Mental health (D)
Tests (9) 2 7 0 Failure 22.22 0.22 n.s.
Studies (3) 0 3 0 Failure 0.00 0.13 n.s.
Personality (hardworking) (D)
Tests (9) 8 1 0 Success 88.89 0.89 ***
Studies (3) 2 1 0 Success 66.67 0.67 n.s.
Note: T-test with two-tailed significance levels. þp < 0.01, *p < 0.05, **p < 0.01, ***p < 0.001.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359354
variables. The results for political knowledge also seem to
indicate a significant and positive effect on turnout, though
these results are based on fewer studies. The results for
cognitive ability are less conclusive: the majority of tests
and studies found this variable to be insignificant.
4.5.2. Efficacy, trust in institutions and others, satisfaction with
democracy
Internal efficacy usually refers to the degree to which
people think they can influence the government or policy
outcomes, while external efficacy indicates the extent to
which citizens perceive a government to be responsive to
their interests. Both forms of efficacy are expected to
increase turnout. The empirical findings are mixed however:
while the majority of tests appears to confirm a positive
and significant effect of efficacy, at the level of studies
the modal category is failure. The average effect size is
positive and significant however.
People that have more confidence in the political system
and in others will more likely have a positive outlook on the
workings of the electoral process (Bélanger and Nadeau,
2005). Likewise satisfaction with democracy is hypothesized
to increase turnout. However, the results of the metaanalysis
seem to disconfirm these hypotheses as most
studies find these variables not to be statistically significant.
4.5.3. Ideological self-placement
With respect to ideological self-placement a common
hypothesis is that right-wing or conservative voters tend to
perceive voting as a civic duty more often than left-wing or
liberal voters. Testing this hypothesis the results of the
meta-analysis show, however, that ideological selfplacement
in most instances does not have a statistically
significant effect on voter turnout.
4.5.4. Alienation and ambivalence
Alienation from the political system and ambivalence
toward parties and candidates are usually not seen as signs
of democratic health (see Adams et al., 2006). While alienation
is most often found not to have an effect on turnout,
ambivalence indeed has a negative effect on individual
level turnout in national elections.
4.5.5. Personality, mental health
Citizens that are hardworking and mentally fit will either
want to be more involved in politics or are more capable to
become involved (Denny and Doyle, 2008). While mental
health is not found to affect turnout significantly, having
a hardworking personality appears to boost turnout. The
success rate ranges between 67 and 89%, however the
average effect size is significant only at the study level.
4.6. The political–institutional model
The notion that the decision to participate in politics is
a by-product of the political system in which people live, is
one that is prevalent in research on voter turnout. Especially
in cross-national research of voting behavior, the
political–institutional context has proven important to
explain differences in levels of voter turnout. In this last
result section we review the influence of 7 political–
institutional variables on turnout. Results of the metaanalysis
are presented in Table 9.
4.6.1. Closeness of elections, concurrent elections
High stake elections tend to attract more voters than
elections where the outcome is a foregone conclusion as
the chances of influencing the outcome increases in close
elections. However, Franklin (2004) points out that measures
of competitiveness such as closeness of the race,
margin of victory, and party polarization affect younger
voters disproportionally as they are still developing turnout
habits.10
Included in 20 out of 90 studies, we find that
competitiveness of elections at the national level does not
have a statistically significant impact on turnout in most
tests and studies. As a consequence the success rate is fairly
low (29–35%). Margin of the victory at the district level also
does not seem to significantly influence turnout.
Concurrent elections have been proposed to increase
voter turnout (due to increased party mobilization, campaigning,
and heightened media attention) as well as to
lower voter turnout (due to voter fatigue). In the metaanalyses
we followed the latter hypothesis, however the
results do not seem to confirm it. Most tests and studies
find that holding concurrent elections in the same year as
parliamentary or presidential elections does not influence
turnout.
4.6.2. Voter facilitation rules and compulsory voting
Legal characteristics of elections are considered to influence
turnout in various ways (see Blais and Dobrzynska,
1998 and Geys, 2006 for overviews). Compulsory voting, for
example, is expected to boost turnout as abstention leads to
punishment which consequently increases the costs of
non-voting. Vote facilitating rules, on the other hand, can
be seen as institutional measures to motivate and mobilize
people. Examples are holiday or weekend voting; postal,
proxy, advance, or e-voting; the placement of special polling
booths (for example in and around shopping centers);
and spreading elections over a couple of days. All these
provisions are aimed at lowering the costs of voting.
All tests and studies for which the variable is included
confirm the mobilizing effect of compulsory voting. The
results are less conclusive regarding the voter facilitation
rules with about half of the tests and studies falling into the
category ‘success’ and the other half in the category
‘failure’. However, the average effect size is positive and
significant, ranging between 0.52 and 0.60.
4.6.3. Electoral system and effective number of parties
Since the translation of votes into seats is much less
precise in majority electoral systems, the number of wasted
votes is higher than in proportional systems (Geys, 2006).
This decreases the probability of a voter influencing the
outcome of elections. Majority electoral systems, on the
10
Note that, as mentioned in the introduction, different causal mechanisms
might affect voter turnout for different groups of voters or in
different contexts. For example, closeness might have an effect on turnout
only among young voters, and hence its effect in studies including voters
of all ages, as most studies reviewed here, might appear to be weaker.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359 355
other hand, are easier to understand for voters. Also, proportional
systems often lead to coalition governments,
which decreases the chance of a voter influencing the
outcome of the elections (Blais, 2006,118; Geys, 2006, 649–
650). Hence the effect of electoral systems on turnout could
run in both ways. However, the four studies that took into
account the electoral system did not find significant effects
of this variable on individual level voter turnout.
Proportional systems produce more parties than majority
systems. The more parties there are, the higher the
number of options a voter will have, and the more likely it is
that a voter will find a party he or she can identify with,
increasing turnout. Also, if more parties compete in elections
more parties will seek to mobilize citizens to turn out
and vote. On the other hand, fractionalization leads to
complexity and increased information costs. The results
reflect these competing hypotheses as studies have both
confirmed and disconfirmed the hypothesis that more
parties lead to lower turnout levels.
5. Conclusion and discussion
In a research context where almost every possible indicator
of voter turnout has been explored it has become
difficult to get a good grip on those factors that matter most
for electoral participation. In this paper we have taken
a step back to assess where we stand, reviewing 90 articles
on individual level turnout published in 10 top-journals of
political science between 2000 and 2010. While initially we
felt the title of our paper to be a slight exaggeration, by the
end of the project we have come to consider ‘the embarrassment
of riches’ a rather accurate depiction of the current
state of voter turnout research.
In 90 articles we found over 170 independent variables
used to explain voter turnout, none of which were included
in all studies. Only 8 of these independent variables were
included in more than 25% of the studies we reviewed:
education, age, gender, race, income, marital status, party
identification and political interest. Even the two most
common independent variables – age and education – were
included in only 72% and 74% of studies respectively. Not
only does this imply that there is no consensus on a ‘core
model’ of voter turnout, it also implies that authors rarely
include the same control variables in their models – despite
often referring to these as ‘the usual suspects’. This possibly
leads to underspecified models and spurious inferences.
In this review we aimed to shed light on those factors
that are consistently linked to individual level turnout. The
variables that we found to have a consistent effect on
turnout (i.e. both at the level of tests and studies the modal
category is ‘success’ and the average effect size is significantly
different from zero) in 10% or more of studies are:
age and age squared, education, residential mobility, region,
media exposure, mobilization (partisan and nonpartisan),
vote in previous election, party identification,
political interest, and political knowledge. Variables that
we consistently found to have no effect on turnout (i.e. both
at the level of tests and studies the modal category is
‘failure’ and/or the effect size is insignificant) in 10% or
more of studies are: gender, race, occupational status and
type, citizenship, union membership, trust in institutions,
and the closeness of elections.
Based on the results presented in this paper, we would
like to make two concluding arguments. First of all, the
current state of turnout research seems to be one where
models are often underspecified theoretically and empirically.
While the theoretical argumentation for the variable
of interest is mostly well developed, often too little attention
is paid to other factors that evidently influence turnout
and that may confound the impact of the variable of interest
on turnout.
Secondly, when including “usual suspects” as controls,
our meta-analysis suggests that scholars should at least
control for the variables listed above, unless of course there
Table 9
Political–institutional model – results meta-analysis.
Variable Success (1) Failure (0) Anomaly (À1) Modal category Success rate Effect size (rav) p-Value
Closeness of election at national level (D)
Tests (51) 15 36 0 Failure 29.41 0.29 ***
Studies (20) 7 13 0 Failure 35.00 0.36 **
Concurrent second order election (L)
Tests (31) 1 28 2 Failure 3.23 À0.03 n.s.
Studies (6) 1 4 1 Failure 16.67 À0.02 n.s.
Voter facilitation rules (D)
Tests (21) 11 10 0 Success 52.38 0.52 ***
Studies (5) 2 3 0 Failure 40.00 0.60 **
Compulsory voting (D)
Tests (12) 12 0 0 Success 100.00 1.00
Studies (5) 5 0 0 Success 100.00 1.00
Electoral system (FPTP/plurality) (L)
Tests (10) 1 9 0 Failure 10.00 0.10 n.s.
Studies (4) 1 3 0 Failure 25.00 0.25 n.s.
Effective number of electoral parties (L)
Tests (5) 2 1 2 Failure 40.00 0.00 n.s.
Studies (4) 2 1 1 Success 50.00 0.25 n.s.
Closeness of election district level (D)
Tests (13) 0 13 0 Failure 0.00 0.00
Studies (3) 0 3 0 Failure 0.00 0.00
Note: T-test with two-tailed significance levels. þp < 0.01, *p < 0.05, **p < 0.01, ***p < 0.001.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359356
are good theoretical reasons not to include those variables
(for example because variables are path-dependent and the
variable of interest is causally prior to the control variable,
as could be the case when investigating the impact of such
variables as parental education or income on turnout c.f.
Verba et al., 1995).
Note that we do not claim that the variables found to be
relevant in this review constitute the only factors to be
included in an eventual core model of turnout. This is
because there are a number of important caveats with
respect to the results found in this paper, apart from the
obvious limitations such as the restricted time period and
exclusive focus on national elections. First of all, while we
believe our sample is representative of research conducted
in the past decade, our results should be considered truly
robust only for those variables that were included in
a substantial number of studies. We found several other
factors that appeared to have a strong relation to turnout,
however since these variables were only included in
a limited number of studies our meta-analysis results are
less robust. Further testing of the impact of these factors on
turnout to corroborate their importance would be
extremely useful. Secondly, since different studies use
widely varying model specifications, the present metaanalysis
cannot be used to draw definite conclusions
about the relative strength of variables.
Thirdly, it is important to note that we analyze effects
for models that are usually based on a random sample of
the entire population, and this means that variables that
have a clear and significant effect on only a particular group
or part of the population might not turn up as important
variables in these analyses. For example, the results of
closeness of the elections are mixed, whereas aggregate
level studies find quite consistent evidence that closeness
explains turnout (Geys, 2006). This might be because, as
Franklin (2004) has found, closeness affects mainly young
voters. Likewise, language proficiency or citizenship status
may be important explanatory factors in studies of turnout
among immigrants or ethnic minorities. Hence, this review
does not provide insight in possible conditional effects of
independent variables on turnout.
Fourthly, many variables explaining turnout are
interconnected, be-it because of path-dependency or
because they measure closely related concepts (as is the
case with for example education, political knowledge and
political interest), and hence our findings for each independent
variable are highly dependent on whether its
covariates or more proximal causal factors are included in
the model as controls or not. We explored this issue more
in-depth in the robustness checks reported in online
Appendix C.
Concluding, while we hope our review of the literature
is useful to the research community, we want to warn
against taking the easy route of discarding all variables
for which we did not find significant results. Rather, we
hope that our meta-analysis will encourage future
research to a.) further develop a ‘core model’ of turnout,
b.) improve our understanding of conditional effects on
turnout, and c.) carry out more extensive meta-analysis
reviews to get a clearer view of the effects of less often
studied variables.
Acknowledgments
We are grateful for financial support from the European
Union Democracy Observatory (EUDO), to Trajche Panov for
research assistance, and to Alexander Trechsel, Mark
Franklin, and Benny Geys for advice in earlier stages of this
project. Preliminary versions of this paper were presented
at EPOP, Colchester, 10–12 September 2010 and at the
ELECDEM Closing Conference, Florence, 28–30 June 2012.
We thank Kees Aarts, Bernard Grofman, Gabriel Katz, and
panel participants in these conferences for their comments.
We also thank two anonymous reviewers and the journal
editor for their thoughtful and helpful comments. The usual
disclaimer applies.
Appendices. Supplementary data
Supplementary data related to this article can be found
at http://dx.doi.org/10.1016/j.electstud.2012.12.006.
References
Articles marked with an * are part of the meta-analysis.
* Adams, J., Dow, J., Merrill III, S., 2006. The political consequences of
alienation-based and indifference-based voter abstention: applications
to presidential elections. Political Behaviour 28, 65–86.
* Anduiza, E., 2005. Campaign effects in the Spanish election of 2000.
Journal of Elections, Public Opinion and Parties 15, 215–236.
Arceneaux, K., Gerber, A.S., Green, D.P., 2006. Comparing experimental
and matching methods using a large-scale voter mobilization
experiment. Political Analysis 14, 37–62.
Arceneaux, K., Nickerson, D.W., 2009. Who is mobilized to vote? a reanalysis
of 11 field experiments. American Journal of Political Science
53, 1–16.
* Bélanger, E., Nadeau, R., 2005. Political trust and the vote in multiparty
elections: the Canadian case. European Journal of Political Research
44, 121–146.
Blais, A., 2000. To Vote or Not to Vote? The Merits and Limits of Rational
Choice Theory. The University of Pittsburgh Press, Pittsburgh.
Blais, A., 2006. What affects voter turnout? Annual Review of Political
Science 9, 111–125.
Blais, A., Dobrzynska, A., 1998. Turnout in electoral democracies. European
Journal of Political Research 33, 239–261.
Brady, H.E., Verba, S., Schlozman, K., 1995. Beyond ses: a resource model
of political participation. American Political Science Review 89,
271–294.
* Burden, B.C., 2009. The dynamic effects of education on voter turnout.
Electoral Studies 28, 540–549.
* Campbell, D.E., 2009. Civic engagement and education: an empirical
test of the sorting model. American Journal of Political Science 53,
771–786.
Childs, S., 2004. A british gender gap? Gender and political participation.
The Political Quarterly 75, 422–424.
* Claassen, R.L., Povtak, A., 2010. The christian right thesis: explaining
longitudinal change in participation among evangelical christians.
The Journal of Politics 72, 2–15.
Cutler, N.E., Bengtson, V.L., 1974. Age and political alienation: maturation,
generation and period effects. Annals of the American Academy of
Political and Social Science 415, 160–175.
* Cutts, D., Fieldhouse, E., 2009. What small spatial scales are relevant as
electoral contexts for individual voters? The importance of the
household on turnout at the 2001 general election. American Journal
of Political Science 53, 726–739.
Cutts, D., Fieldhouse, E., John, P., 2009. Is voting habit forming? The
longitudinal impact of a GOTV campaign in the UK. Journal of Elections,
Public Opinion and Parties 19, 251–263.
Dale, A., Strauss, A., 2009. Don’t forget to vote: text message reminders as
a mobilization tool. American Journal of Political Science 53, 787–804.
* Denny, K., Doyle, O., 2008. Political interest, cognitive ability and personality:
determinants of voter turnout in Britain. British Journal of
Political Science 38, 291–310.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359 357
Denver, D., 2008. Another reason to support marriage? Turnout and the
decline of marriage in Britain. British Journal of Politics and International
Relations 10, 666–680.
Downs, A., 1957. An Economic Theory of Democracy. HarperCollins, New
York.
* Fieldhouse, E., Tranmer, M., Russell, A., 2007. Something about young
people or something about elections? Electoral participation of young
people in Europe: evidence from a multilevel analysis of the European
social survey. European Journal of Political Research 46, 797–822.
Fowler, J.H., 2006. Altruism and turnout. The Journal of Politics 68,
674–683.
Franklin, M.N., 2004. Voter Turnout and the Dynamics of Electoral
Competition in Established Democracies Since 1945. Cambridge
University Press, Cambridge, MA.
* Gallego, A., 2010. Understanding unequal turnout: education and voting
in comparative perspective. Electoral Studies 29, 239–248.
Gerber, A.S., Green, D.P., 2000. The effects of canvassing, telephone calls,
and direct mail on voter turnout: a field experiment. American Political
Science Review 94, 653–663.
Gerber, A.S., Green, D.P., Shachar, R., 2003. Voting may be habit-forming:
evidence from a randomized field experiment. American Journal of
Political Science 47, 540–550.
Gerber, A.S., Rogers, T., 2009. Descriptive social norms and motivation to
vote: everybody’s voting and so should you. The Journal of Politics 71,
178–191.
Geys, B., 2006. Explaining voter turnout: a review of aggregate-level
research. Electoral Studies 25, 637–663.
Glass, G.V., 1976. Primary, secondary, and meta-analysis of research.
Educational Researcher 5, 3–8.
* Goldstein, K., Freedman, P., 2002. Campaign advertising and voter
turnout: new evidence for a stimulation effect. The Journal of Politics
64, 721–740.
* Highton, B., Wolfinger, R.E., 2001. The first seven years of the political life
cycle. American Journal of Political Science 45, 202–209.
* Hillygus, D.S., 2005b. The missing link: exploring the relationship between
higher education and political engagement. Political Behaviour
27, 25–47.
Imbeau, L.M., Pétry, F., Lamari, M., 2001. Left-right party ideology and
government policies: a meta-analysis. European Journal of Political
Research 40, 1–29.
Inglehart, R., Norris, P., 2003. Rising Tide: Gender Equality and
Cultural Change Around the World. Cambridge University Press,
Cambridge, MA.
Jankowski, T.B., Strate, J.M., 1995. Modes of participation over the adult
life span. Political Behaviour 17, 89–106.
* Karp, J.A., Banducci, S.A., Bowler, S., 2008. Getting out the vote: Party
mobilization in a comparative perspective. British Journal of Political
Science 38, 91–112.
Karp, J.A., Brockington, D., 2005. Social desirability and response validity:
a comparative analysis of overreporting voter turnout in five countries.
The Journal of Politics 67, 825–840.
Lane, R.E., 1959. Political Life. Why People Get Involved in Politics. The
Free Press of Glencoe, Toronto.
Lau, R.R., Sigelman, L., Rovner, I.B., 2007. The effects of negative political
campaigns: a meta-analytic reassessment. The Journal of Politics 69,
1176–1209.
* Leighley, J.E., Nagler, J., 2007. Unions, voter turnout, and class bias in the
U.S. electorate, 1964–2004. The Journal of Politics 69, 430–441.
Lipset, S.M., 1969. Political Man. Heinemann, London.
Norris, P., 2002. Democratic Phoenix. Reinventing Political Activism.
Cambridge University Press, Cambrdige, MA.
Pérez-Liñán, A., 2001. Neoinstitutional accounts of voter turnout: moving
beyond industrial democracies. Electoral Studies 20, 281–297.
* Plutzer, E., 2002. Becoming a habitual voter: inertia, resources, and
growth in young adulthood. American Political Science Review 96,
41–56.
* Prior, M., 2005. News vs. entertainment: how increasing media choice
widens gaps in political knowledge and turnout. American Journal of
Political Science 49, 577–592.
Riker, W.H., Ordeshook, P.C., 1968. A theory of the calculus of voting.
American Political Science Review 62, 25–42.
Rosenstone, S.J., 1982. Economic adversity and voter turnout. American
Journal of Political Science 26, 25–46.
Rosenstone, S.J., Hansen, J.M., 1993. Mobilization, Participation and Democracy
in America. MacMillan, New York, NJ.
* Rubenson, D., Blais, A., Fournier, P., Gidengil, E., Nevitte, N., 2004. Accounting
for the age gap in turnout. Acta Politica 39, 407–421.
* Sandell, J., Plutzer, E., 2005. Families, divorce and voter turnout in the
US. Political Behaviour 27, 133–162.
* Sandell Pacheco, J., 2008. Political socialization in context: the effect of
political competition on youth voter turnout. Political Behaviour 30,
415–436.
Schlozman, K.L., Verba, S., 1979. Injury to Insult. Unemployment, Class,
and Political Response. Harvard University Press, Cambridge, MA.
* Solt, F., 2008. Economic inequality and democratic political engagement.
American Journal of Political Science 52, 48–60.
Stevens, D., 2008. The relationship between negative political advertising
and public mood: effects and consequences. Journal of Elections,
Public Opinion and Parties 18, 153–177.
* Stevens, D., 2009. Elements of negativity: volume and proportion in
exposure to negative advertising. Political Behaviour 31, 429–454.
Stoker, L., Jennings, M.K., 1995. Life-cycle transitions and political participation:
the case of marriage. American Political Science Review 89,
421–433.
Strate, J.M., Parrish, C.J., Elder, C.D., Ford, C., 1989. Life span civic development
and voting participation. American Political Science Review
83, 443–464.
* Tenn, S., 2007. The effect of education on voter turnout. Political Analysis
15, 446–464.
Verba, S., Nie, N.H., 1972. Participation in America. Political Democracy
and Social Equality. Harper & Row, New York.
Verba, S., Schlozman, K., Brady, H.E., 1995. Voice and Equality. Civic
Voluntarism in American Politics. Harvard University Press, Cambridge,
MA.
Wolf, F.M., 1986. Meta-analysis. Quantitative Methods for Research Synthesis.
In: Quantitative Applications in the Social Sciences 07-059.
Sage Publications, Beverly Hills, CA.
Further reading
Articles marked with an * are part of the meta-analysis.
* Aarts, K., Semetko, H.A., 2003. The divided electorate: media use and
political involvement. The Journal of Politics 65, 759–784.
* Adams, J., Merrill III, S., 2003. Voter turnout and candidate strategies in
American elections. The Journal of Politics 65, 161–189.
* Alex-Assensoh, Y., Assensoh, A.B., 2001. Inner-city contexts, church
attendance, and African American political participation. The Journal
of Politics 63, 886–901.
* Alvarez, R.M., Hopkins, A., Sinclair, B., 2010. Mobilizing Pasadena democrats:
measuring the effects of partisan campaign contacts. The
Journal of Politics 72, 31–44.
* Anduiza, E., 2002. Individual characteristics, institutional incentives and
electoral abstention in western Europe. European Journal of Political
Research 41, 643–673.
* Banducci, S.A., Donovan, T., Karp, J.A., 2004. Minority representation,
empowerment, and participation. The Journal of Politics 66,
534–556.
* Bass, L.E., Casper, L.M., 2001. Impacting the political landscape: who
registers and votes among naturalized Americans? Political Behaviour
23, 103–130.
* Blais, A., Gidengil, E., Nevitte, N., Nadeau, R., 2004. Where does
turnout decline come from? European Journal of Political Research
43, 221–236.
* Bromley, C., Curtice, J., 2003. The lost voters of Scotland: devolution
disillusioned or westminster weary? British Elections & Parties Review
13, 66–85.
* Busch, M.L., Reinhardt, E., 2000. Geography, international trade, and
political mobilization in U.S. industries. American Journal of Political
Science 44, 703–719.
* Busch, M.L., Reinhardt, E., 2005. Industrial location and voter participation
in Europe. British Journal of Political Science 35, 713–730.
* Cho, W.K.T., Gimpel, J.G., Dyck, J.J., 2006. Residential concentration,
political socialization, and voter turnout. The Journal of Politics 68,
156–167.
* Chong, D., Rogers, R., 2005. Racial solidarity and political participation.
Political Behaviour 27, 347–374.
* Clarke, H., Sanders, D., Stewart, M., Whiteley, P., 2002. Downs, stokes and
modified rational choice: modelling turnout in 2001. Journal of
Elections, Public Opinion and Parties 12, 28–47.
* Clarke, H., Sanders, D., Stewart, M., Whiteley, P., 2006. Taking the bloom
off new labour’s rose: Party choice and voter turnout in Britain, 2005.
Journal of Elections, Public Opinion and Parties 16, 3–36.
* Clinton, J.D., Lapinski, J.S., 2004. ”Targeted” advertising and voter turnout:
an experimental study of the 2000 presidential election. The
Journal of Politics 66, 69–96.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359358
* DeFrancesco Soto, V.M., Merolla, J.L., 2006. Vota por tu futuro: Partisan
mobilization of Latino voters in the 2000 presidential election. Political
Behaviour 28, 285–304.
* Denny, K., Doyle, O., 2009. Does voting history matter? Analysing persistence
in turnout. American Journal of Political Science 53, 17–35.
* Fischer, S.D., Lessard-Philips, L., Hobolt, S.B., Curtice, J., 2008. Disengaging
voters: do plurality systems discourage the less knowledgeable
from voting? Electoral Studies 27, 89–104.
* Fowler, J.H., Dawes, C.T., 2008. Two genes predict voter turnout. The
Journal of Politics 70, 579–594.
* Freedman, P., Franz, M., Goldstein, K., 2004. Campaign advertising and
democratic citizenship. American Journal of Political Science 48,
723–741.
* Gastil, J., Deess, E.P., Weiser, P., 2002. Civic awakening in the jury room:
a test of the connection between jury deliberation and political participation.
The Journal of Politics 64, 585–595.
* Gastil, J., Deess, E.P., Weiser, P., Meade, J., 2008. Jury service and electoral
participation: a test of the participation hypothesis. The Journal of
Politics 70, 351–367.
* Gimpel, J.G., Dyck, J.J., Shaw, D.R., 2004. Registrants, voters, and
turnout. Variability across neighborhoods. Political Behaviour 26,
343–375.
* Goldstein, K.M., Ridout, T.N., 2002. The politics of participation: mobilization
and turnout over time. Political Behaviour 24, 3–29.
* Green, D.P., Shachar, R., 2000. Habit formation and political behaviour:
evidence of consuetude in voter turnout. British Journal of Political
Science 30, 561–573.
* Green, D.P., Vavreck, L., 2008. Analysis of cluster-randomized experiments:
a comparison of alternative estimation approaches. Political
Analysis 16, 138–152.
* Griffin, J.D., Keane, M., 2006. Descriptive representation and the composition
of African American turnout. American Journal of Political
Science 50, 998–1012.
* Hanmer, M.J., 2007. An alternative approach to estimating who is most
likely to respond to changes in registration laws. Political Behaviour
29, 1–30.
* Heath, A., 2000. Where traditional labour voters disillusioned with new
labour? Abstention at the 1997 general election. Journal of Elections,
Public Opinion and Parties 10, 32–46.
* Heath, O., 2004. Modelling the components of political action in Britain:
the impact of civic skills, social capital and political support. Journal
of Elections, Public Opinion and Parties 14, 75–93.
* Heath, O., 2007. Explaining turnout decline in Britain, 1964–2005:
Party identification and the political context. Political Behaviour
29, 493–516.
* Highton, B., 2000. Residential mobility, community mobility, and electoral
participation. Political Behaviour 22, 109–120.
* Hillygus, D.S., 2005a. Campaign effects and the dynamics of turnout
intention in election 2000. The Journal of Politics 67, 50–68.
* Holbrook, A.L., Krosnick, J.A., Visser, P.S., Gardner, W.L., Cacioppo, J.T.,
2001. Attitudes toward presidential candidates and political parties:
Initial optimism, inertial first impressions, and a focus on flaws.
American Journal of Political Science 45, 930–950.
* Huber, G.A., Arceneaux, K., 2007. Identifying the persuasive effects of
presidential advertising. American Journal of Political Science 51,
957–977.
* Huddy, L., Khatib, N., 2007. American patriotism, national identity, and
political involvement. American Journal of Political Science 51, 63–77.
* Jackson, R.A., 2003. Differential influences on Latino electoral participation.
Political Behaviour 25, 339–366.
* Jang, S.J., 2009. Get out on behalf of your group: electoral participation of
Latinos and Asian Americans. Political Behaviour 31, 511–535.
* Johnston, R., Matthews, J.S., Bittner, A., 2007. Turnout and the party
system in Canada, 1988–2004. Electoral Studies 26, 735–745.
* Johnston, R.J., Pattie, C.J., Dorling, D.F.L., MacAllister, I., Tunstall, H.,
Rossiter, D.J., 2001. Housing tenure, local context, scale and voting in
England and Wales, 1997. Electoral Studies 20, 195–216.
* Karp, J.A., Banducci, S.A., 2008. Political efficacy and participation in
twenty-seven democracies: how electoral systems shape political
behaviour. British Journal of Political Science 38, 311–334.
* Killian, M., Schoen, R., Dusso, A., 2008. Keeping up with the joneses: the
interplay of personal and collective evaluations in voter turnout.
Political Behaviour 30, 323–340.
* Lyons, W., Alexander, R., 2000. A tale of two electorates: generational
replacement and the decline of voting in presidential elections. The
Journal of Politics 62, 1014–1034.
* Malhotra, N., Krosnick, J.A., 2007. The effect of survey mode and sampling
on inferences about political attitudes and behavior: comparing
the 2000 and 2004 ANES to internet surveys with nonprobability
samples. Political Analysis 15, 286–323.
* Mc Kenzie, B.D., 2008. Reconsidering the effects of bonding social capital:
a closer look at black civil society institutions in America. Political
Behaviour 30, 25–45.
* McDonald, M.P., 2008. Portable voter registration. Political Behaviour 30,
491–501.
* Mondak, J.J., Halperin, K.D., 2008. A framework for the study of personality
and political behaviour. British Journal of Political Science 38, 335–362.
* Mughan, A., Lacy, D., 2002. Economic performance, job insecurity and
electoral choice. British Journal of Political Science 32, 513–533.
* Mutz, D.C., 2002. The consequences of cross-cutting networks for political
participation. American Journal of Political Science 46, 838–855.
* Pattie, C.J., Johnston, R.J., 2009. Conversation, disagreement and political
participation. Political Behaviour 31, 261–285.
* Pierce, R., 2003. Modelling electoral second choices: thwarted voters in
the United States, France, and Russia. Electoral Studies 22, 265–285.
* Pierce, R., 2004. Thwarted voters at the 2000 US presidential election.
Electoral Studies 23, 601–621.
* Sanders, D., Clarke, H.D., Stewart, M.C., Whiteley, P., 2007. Does mode
matter for modeling political choice? Evidence from the 2005 British
election study. Political Analysis 15, 257–285.
* Schildkraut, D.J., 2005. The rise and fall of political engagement among
Latinos: the role of identity and perceptions of discrimination. Political
Behaviour 27, 285–312.
* Shaw, D., de la Garza, R.O., Lee, J., 2000. Examining Latino turnout in
1996: a three-state validated survey. American Journal of Political
Science 44, 338–346.
* Sides, J., Karch, A., 2008. Messages that mobilize? Issue publics and the
content of campaign advertising. The Journal of Politics 70, 466–476.
* Sondheimer, R.M., Green, D.P., 2010. Using experiments to estimate the
effects of education on voter turnout. American Journal of Political
Science 54, 174–189.
* Staton, J.K., Jackson, R.A., Canache, D., 2007. Dual nationality among
Latinos: what are the implications for political connectedness? The
Journal of Politics 69, 470–482.
* Stein, R.M., Vonnahme, G., 2008. Engaging the unengaged voter: vote
centers and voter turnout. The Journal of Politics 70, 487–497.
* Stevens, D., Sullivan, J., Allen, B., Alger, D., 2008. What’s good for the
goose is bad for the gander: negative political advertising, partisanship,
and turnout. The Journal of Politics 70, 527–541.
* Tenn, S., 2005. An alternative measure of relative education to explain
voter turnout. The Journal of Politics 67, 271–282.
* Timpone, R.J., 2002. Estimating aggregate policy reform effects: new
baselines for registration, participation and representation. Political
Analysis 10, 154–177.
* Vowles, J., 2010. Electoral system change, generations, competitiveness
and turnout in New Zealand, 1963–2005. British Journal of Political
Science 40, 875–895.
* Wass, H., 2007. The effects of age, generation and period on turnout in
Finland 1975–2003. Electoral Studies 26, 648–659.
* Wielhouwer, P.W., 2000. Releasing the fetters: parties and the mobilization
of the African-American electorate. The Journal of Politics 62,
206–222.
* Yoo, S., 2010. Two types of neutrality: ambivalence versus indifference
and political participation. The Journal of Politics 72, 163–177.
K. Smets, C. van Ham / Electoral Studies 32 (2013) 344–359 359