Implementation Intentions, Information, and Voter Turnout:
An Experimental Study
Cameron D. Anderson
Western University
Peter J. Loewen
University of Toronto
R. Michael McGregor
Ryerson University
Are citizens more likely to vote when they are asked to make plans about how they will cast their ballots? Such
planning—typically described as “implementation intentions”—has been shown to increase many types of
desirable behaviors, including exercising and healthy eating, receiving vaccinations, physical rehabilitation,
and recycling. Important earlier work in political science suggests voter turnout can also be inﬂuenced by
implementation intention interventions, whereby electors are prompted to “make a plan” to vote (Nickerson &
Rogers, 2010), though this ﬁnding has gone largely unreplicated. At the same time, elections management
bodies (EMBs) in many contexts regularly conduct informational campaigns in the period leading up to
elections, though little is known about the effects of such efforts upon turnout. Using data from an online
experiment conducted at the time of the 2015 Canadian Federal Election, we demonstrate that implementation
intention interventions can improve voter turnout but that this effect is conditional upon electors being exposed
to informational materials about how to vote in the election. When survey respondents were provided with
information on voting requirements and methods, and then prompted with questions forcing them to
contemplate the act of casting their ballots, we observe a sizable increase in turnout rates.
KEY WORDS: voter turnout, implementation intentions, voter information, election management bodies, online
experiment
Voter participation is widely regarded as central to electoral democracy (Barber, 2003; Cavanagh,
1981; Pateman, 1970; Salisbury, 1975). Accordingly, a substantial amount of research has explored
why people vote (or not) (e.g., Blais, 2001; Green & Gerber, 2008; Wolﬁnger & Rosenstone, 1980;
Verba, Schlozman, Brady, & Brady, 1995). The most widely supported conclusions are that turnout
decisions are inﬂuenced by sociodemographic characteristics (including income, education, age,
gender, race/ethnicity), attitudes and beliefs (partisanship, political interest, efﬁcacy, duty), and mobilization
efforts by parties and interest groups. In relatively recent developments in this literature,
experimental designs have been utilized to explore the effects of arguments and mobilization techniques
(e.g., Brown, Perrella, & Kay, 2010; Davenport et al., 2010; Gerber, Green, & Larimer, 2008).
1
0162-895X VC 2017 International Society of Political Psychology
Published by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, 9600 Garsington Road, Oxford, OX4 2DQ,
and PO Box 378 Carlton South, 3053 Victoria, Australia
Political Psychology, Vol. xx, No. xx, 2017
doi: 10.1111/pops.12467
bs_bs_banner
Within this emerging segment of the literature on turnout, work on the effects of asking voters to
make a plan to vote—known as implementation intention interventions—represents a recent and still
largely unreplicated contribution.
The focus of this study is to experimentally assess the effects on turnout of implementation intention
prompts, messages deployed by an Election Management Body (EMB), and the joint effect of
these two factors. Is it the case that voter mobilization campaigns are more effective when accompanied
by implementation intention messages? Conversely, are implementation intention messages
more potent when accompanied by informational messages of the type contained in educational materials
created and disseminated by EMBs?
To answer these questions, we use data from a two-wave online survey of Canadian electors, conducted
in the fall of 2015. The ﬁrst wave was conducted shortly before the 2015 Canadian Federal
Election and the second immediately afterwards. Speciﬁcally, we consider the effectiveness of two
experimental treatments upon turnout intention (measured before the election) and reported turnout
(measured after). The ﬁrst treatment involves exposure to Elections Canada’s mobilization materials
about “being ready” to vote and how to vote—messages which are clearly linked to the concepts of
implementation of intentions and planning. The second treatment is based upon two forms of implementation
intention questions, the ﬁrst of which is based upon Nickerson and Rogers’ (2010) study
and the second related to the information materials produced by an EMB. The combination of these
treatments allows us to consider if and how implementation intention interventions interact with exposure
to ofﬁcial information from an EMB to affect voter turnout, enabling us to add to the nascent literature
on implementation intentions in the ﬁeld of political science.
Literature Review
Often termed the intention-behavior gap, studies in social psychology have long shown a disjuncture
between intended desired outcomes and actual goal attainment (e.g., Gollwitzer & Oettingen,
2013; Gollwitzer & Sheeran, 2006). One means of reducing this gap is through the use of implementation
intentions. Implementation intentions have been described as “if-then” plans, where actors are
asked to spell out in advance how they will undertake some action (Gollwitzer & Oettingen, 2013).
Articulating such a “plan” is thought to create cognitive links in memory between anticipated future
actions and the behaviors required to achieve these actions, making it more likely that the behavior
will take place (Rogers, Milkman, John, & Norton, 2015). The literature on the phenomenon revolves
heavily around the idea of “goal attainment,” and researchers have used the concept to increase the
likelihood of a number of socially desirable behaviors. On a variety of fronts, including exercising
and healthy eating (Achtziger, Gollwitzer, & Sheeran, 2008; Rothman et al., 2015), scholastic performance
(Bayer & Gollwitzer, 2007), being vaccinated (Leventhal, Singer, & Jones, 1965), physical
rehabilitation (Kersten, McPherson, Kayes, Theadom, & McCambridge, 2015), and recycling
(Holland, Aarts, & Langendam, 2006), formulating a plan and thinking through the actions required
to achieve some future goal have been shown to increase the likelihood of “success.”
In order to reduce the intention-behavior gap, the literature differentiates between two types of
intentions: goal intentions and implementation intentions (e.g., Gollwitzer, 1996, 1999). Goal intentions
specify a particular behavior or outcome deemed desirable. By contrast, implementation intentions
specify when, where, and how a goal intention will be realized. They have also been shown to
facilitate the attainment of speciﬁed goals (Bayer & Gollwitzer, 2007). Implementation intention plans
provide a “structure in which people (1) identify key opportunities or, obstacles to, taking action, (2)
specify a way to respond to each opportunity and obstacle, and then (3) formalize a link between the
opportunity or obstacle and the response: ‘If (opportunity/obstacle) arises, then I will (respond in this
way)!’” (Rothman et al., 2015).
2 Anderson et al.
In the ﬁeld of political science, implementation intentions have been applied (though to a limited
degree) to the study of voter turnout (Loewen, Rubenson, & Martin, n.d; Nickerson & Rogers, 2010).
In this context, the act of voting is viewed as the desirable “goal” to be attained. This attainment can
be hampered, however, by a variety of factors, including illness, poor weather, a lack of time, or simply
a lack information on where or when to vote (Pammett & LeDuc, 2003). Might it be the case that
“making a voting plan” increases the likelihood that electors will overcome these issues and turnout
on Election Day?
Nickerson and Rogers (2010) provide the strongest evidence of the effect of implementation
intentions on voter turnout. In their study, conducted during the 2008 Democratic Party primary election
in Pennsylvania, 287,000 people were randomly assigned to several experimental groups. The
ﬁrst treatment group received phone calls reminding them of the upcoming election, encouraging
them to vote, and asking if they intended to turn out. Electors in the second, or “plan-making” group,
were provided with the same information and asked the same questions, but they were also presented
with three additional questions about when they would vote, how they would get to their polling place,
and where they would be coming from before voting. As compared to the control group, the ﬁrst treatment
group saw a bump of 2.0 percentage points in turnout, while those in the “plan-making” group
were 4.1 points more likely to vote than the control. Thus, asking respondents about their plan for voting
was twice as effective as the standard get-out-the-vote prompts.1
The broader implementation intentions literature considers how effects may be moderated
(Sheeran, 2002). In the extant psychological literature, Sheeran identiﬁes four types of factors (behavior
types, intention type, properties of intentions, and personality and cognitive variables) that appear
to inﬂuence the degree of consistency between the presence of intentions and behavior. In the context
of political science and turnout, little is known about if, and how, the effect of implementation intention
interventions upon turnout may be moderated. Nickerson and Rogers (2010) ﬁnd that the size of
the household has such an effect, but other than that, other potential moderating factors have yet to be
identiﬁed.
Accordingly, the present work considers the potentially moderating role of information campaigns
conducted by election management bodies (or EMB), aimed at boosting voter turnout. In
Canada, Elections Canada regularly conducts information campaigns aimed at mobilizing turnout in
Federal elections (the comparable provincial organizations do the same thing for Provincial elec-
tions).2
To date, however, there is little work that assesses the effectiveness of these information campaigns,
and what work that has been done suggests mixed or at least modest effects. In their study of
Elections Canada’s 2006 and 2008 voter mobilization campaigns, McGregor and Anderson (2014)
ﬁnd that mobilization messages are only effective at increasing turnout among particular segments of
the population with traditionally low turnout rates (Aboriginals, ethnic minorities, and disabled voters).
For most electors, however, the information campaigns had no discernible impact upon turnout.
Theory
We think that the question of the effect of implementation intention interventions on turnout
deserves scrutiny because of the exceptional nature of participation in intermittent elections. Consider
two other activities which have been shown to increase with implementation intentions: healthy eating,
exercising, and recycling. These activities share two traits which are not shared by voting. First,
they are frequent. Individuals make choices about food every day. Recycling is likewise an action
1
These estimates are derived from Nickerson and Rogers (2010) as well as later accounts of their work (see Green &
Gerber, 2008).
2
We note here the uniformity of the administration of federal elections across Canada. Unlike the highly localized and
varied administration of national elections in the United States, federal elections in Canada are administered by a single
party which engages in uniform messaging and polling practices across the country.
Implementation Intentions, Information, and Voter Turnout 3
which includes everyday components (i.e., disposing of items in a recycling receptacle rather than a
garbage receptacle) and a typically weekly or biweekly component (taking out the receptacle for
pickup). Second, individuals usually know the minimally required information to undertake these
actions, for example, where a grocery store is located, the nutritional value of fresh fruit and vegetables
over processed foods, or what types of goods might be recyclable and where they can be
disposed.
Voting shares neither of these qualities. First, it is not a frequent action. The average Canadian
elector will have the chance to participate in a federal, provincial, or municipal election perhaps once
a year. Second, because elections are administered by different electoral management bodies at each
level of government, how and where an individual votes will differ between elections. As such, voting
does not resemble the types of familiar and simple acts that have been previously shown to be affected
by implementation intention interventions. While turnout is an infrequent activity, prior work does
show that such interventions can affect success in the completion of even one-time activities
(Dholakia & Bagozzi, 2003).3
As we discuss in greater detail below, however, our theoretical intuition
is that the effectiveness of implementation intention interventions for such rare events will be most
effective—and perhaps only effective—among voters who also possess the basic information required
to actually make a plan about where, when, and how to cast a ballot.
The current study thus examines three related research questions. First, does exposure to an
EMB’s informational materials increase turnout? Second, does exposure to implementation intention
interventions increase turnout? Finally, does exposure to Elections Canada’s informational materials
and implementation intention questions act jointly to increase turnout?
We have expectations regarding each of these three questions. First, there is some very modest
evidence that Elections Canada’s advertisements inﬂuence turnout (McGregor & Anderson [2014]
have shown that the ads only affected speciﬁc segments of the population).4
At the provincial level,
we know of only one EMB that has attempted to measure the impact of their advertising campaigns
upon turnout. A survey conducted following the 2012 provincial election in Quebec revealed that
34% of voters polled credited a media campaign run by the Directeur General des Elections du
Quebec for their decision to cast a ballot.5,6
Though such a simple approach to measuring the relationship
between attitudes towards advertising and reported turnout has obvious limitations (including the
possibility that two variables might be spuriously related to one another, or that responses may be
inﬂuenced by normative judgments on the part of respondents regarding the proper role of government
in encouraging turnout), we nevertheless begin with the following expectation:
H1: Exposure to informational materials from EMBs will be positively associated with
turnout.
Next, based upon Nickerson and Rogers’ (2010) study, we do expect to see a positive effect upon
turnout for the ﬁrst group of implementation intention questions. Prior work shows that actively generating
implementation intentions for even one-time behaviors is effective at fostering the cognitive
3
The context referenced in this work is short-fuse behaviors—actions that must be enacted within a limited window
for success. An example from this work includes picking up course reading materials at a particular time and place
(Dholakia & Bagozzi, 2003).
4
The aforementioned study also considers the effects of a different set of advertisements. Prior to the 2015 campaign,
Elections Canada was tasked with actively encouraging voter turnout. Legislation passed shortly before the election,
however, changed the mandate of the organization, which is permitted now only to run informational campaigns but
is to stop short of speciﬁcally urging electors to vote.
5
This is according to Denis Dion of le Directeur General des Elections du Quebec, as quoted in La Presse (February
11, 2014).
6
To our knowledge, there has yet to be any research into the mechanism(s) through which informational materials provided
by EMBs affect turnout. This would seem a fruitful avenue for future research.
4 Anderson et al.
links between an anticipated situation and behavioral intent (Dholakia & Bagozzi, 2003). So we posit
that exposure to implementation intentions prompts will establish effective cognitive links for turnout
decisions. We have a weaker assumption of success for the second, untested, group of questions
(related to Elections Canada information materials). As for the question of which set of messages
should have the greatest effect upon turnout, we have no expectations. Regardless, this leads us to
expect the following:
H2: Exposure to implementation intentions questions will be positively associated with
turnout.
Finally, we expect to see an interactive effect between exposure to informational materials and
implementation intention interventions. In particular, we suggest that implementation intention
prompts will be more effective for turnout among those respondents who are also exposed to Elections
Canada’s informational ﬂyers. We suggest that exposure to information materials has two effects
upon the efﬁcacy of implementation intention interventions. First, the provision of information regarding
the materials needed to vote (identiﬁcation requirements, dates for advance voting, etc.) lessens
the costs required to collect information to form a turnout plan. Second, respondents prompted to
devise a turnout plan will be more certain of how to carry out their plan based on the information provided
in the ﬂyer prompts. Both of these factors may serve to make respondents more conﬁdent about
their ability to formulate a high-quality plan and to make it immediately. This should make them
more likely to actually turn out. We might also state that while turning out to vote is a comparatively
simple task and relatively little is asked of the voter in terms of the act, the decision criteria that goes
into the content of the choice once at the ballot box is far from a simple task. For this reason, the combined
effects of implementation intention interventions and information from ﬂyers contribute to lessening
the overall cognitive demands of navigating a complex environment within which a vote is cast.
Those individuals who have a relatively high level of information of the behavior we are trying to
encourage should be inﬂuenced more by such interventions than will those with less information.
We note as well that previous work on implementation intentions suggests that information on
how to successfully carry out the action is in fact required to form an effective plan. As Rogers et al.
(2015) note, the ﬁrst notable implementation intention experiment (Leventhal et al., 1965) provided
subjects with the information required to form an effective plan to receive an inoculation. Closer to
our study, Rogers and Nickerson’s (2010) control condition, to which treatments were added, appears
to contain at least some of the information required to know when to vote. More generally, Rogers
et al. (2015) note that effective plan making is more likely when individuals have information on the
“when, where, and how” of the action they wish to undertake. This leads us to our ﬁnal expectation:
H3: The effect of either implementation intention messages will be increased by the
provision of information about how to vote.
Data and Methodology
Our study relies on a two-wave online panel survey7
in which we embed two different sets of
experimental treatments. Survey respondents, of whom there were 1,454, were interviewed in the
days leading up to the October 19th Canadian Federal Election, and 67% of respondents also
7
The survey was coded through the Qualtrics interface. The sample was purchased from and the data were collected
on our behalf by Survey Sampling International.
Implementation Intentions, Information, and Voter Turnout 5
completed the postelection wave of the survey, administered shortly after Election Day.8
Our design
was executed in the context of a real election campaign using materials provided by the national
EMB itself. Accordingly, our experiment has arguably high construct and external validity.
The case of Canada and the mobilization campaign by Elections Canada in 2015 are particularly
well suited to the study of implementation intention interventions. The thrust of the information campaign
was “being ready” to vote; its goal was to equip voters with the necessary information to be
able to cast a ballot. Such a goal was particularly germane in the 2015 electoral cycle because of
changes introduced by the Fair Elections Act. Passed in 2014 by the governing Conservative Party,
the Fair Elections Act introduced a number of changes to the regulatory environment of federal elections
in Canada. Of particular note for individual voters, the Act demanded that voters prove their
identity by requiring government-issued photo identiﬁcation conﬁrming their address and ending the
use of “vouching” for someone’s identity or simply using a voter information card.9
These changes
were controversial, well-publicized, and may well have contributed to a general sense of confusion
over what were approved sources and means of proving one’s identity at the polling station. For these
reasons, Elections Canada’s mobilization and information efforts as well as implementation intentions
to vote are particularly worthy of study.
Respondents were randomly assigned on the basis of two manipulations. The ﬁrst was exposure
to two informational fact sheets or ﬂyers created by Elections Canada. Half of respondents were
assigned to see the two ﬂyers, while the other half were kept in control and thus saw nothing. The second
manipulation, independent of the ﬁrst, was to one of three implementation intention conditions:
implementation intention message 1, implementation message 2, or control. We thus have a 2 3 3
design with roughly evenly sized groups. A multinomial logistic regression of treatment condition on
sociodemographics—age, gender, education, income, and visible minority status—suggests balance
between the groups.10
We are thus conﬁdent that any observed differences in the turnout patterns of
these groups are a result of our experimental treatments.
In our ﬁrst treatment condition, respondents were shown informational materials from Elections
Canada, the arms-length government agency responsible for the administration of federal elections in
this country. Respondents were exposed to two posters, in the language of their choice (French or
English): The ﬁrst encourages respondents to “get ready to vote” (preparez-vous a voter) and provides
information on how to register, reminds voters to watch for their voter information card in the mail
and to ensure that they bring the correct ID to the polling booth. The second ﬂyer informs electors
that there are four potential ways to vote (on Election Day, in advance polls, by mail, or at an Election
Canada ofﬁce outside of advance polling days). These materials were publicly available on Elections
Canada’s website. To maximize message reception and acceptance, participants were asked to summarize
the central point of the posters in a single statement.11
Our second manipulation is meant to evaluate the effect of implementation intention questions
upon turnout. The ﬁrst treatment group was presented with the same series of questions asked by
Nickerson and Rogers (2010): “If you decide to vote in the upcoming federal election: Around what
time do you expect you will head to the polls on Election Day? Where do you expect you will be
coming from when you head to the polls? What do you think you will be doing before you head out
to the polls?” The second treatment group was presented with three questions, this time related
8
The ﬁrst wave was administered October 9–17, and the second from October 21 to October 28.
9
The practice of vouching allowed for a registered and conﬁrmed voter to “vouch” for the identity of another voter
who did not possess the correct pieces of identiﬁcation to conﬁrm their identity and residence.
10
None of the variables included in the model reach statistical signiﬁcance in any of the group-wise comparisons. The
chi-square value of the multinomial model’s likelihood ratio is 20.00, p 5 0.75, and R-squared value is a 0.0043. Full
model results found in Appendix B.
11
More precisely, respondents were prompted with the following (and an open-ended text box): “During each federal
election, Elections Canada runs informational campaigns. In the box below, please type what you believe to be the
main message of the ad below.”
6 Anderson et al.
speciﬁcally to Elections Canada’s voter-information campaign. “If you decide to vote in the
upcoming federal election: What piece(s) of ID are you going to take with you to the polls? Will
you vote on Election Day, at advance polls, by mail, or at an Elections Canada ofﬁce? Will you
read your voter information card prior to heading to the polls?” These implementation intention
scripts are intended to prompt individuals into thinking about how they will overcome common
obstacles to voting (such as time taken to vote, identiﬁcation requirements, and information related
to where and how to vote).
The variation between these two treatments is intended to determine whether contemplating
some of the factors covered by Elections Canada’s informational campaign has a different effect than
the standard implementation intention questions. All questions were open ended; for our purposes, it
is not the answers that were important, simply the fact that respondents answered at all. We consider
below whether the treatments had effects independently or in interaction with one another.
Our sample thus consists of six groups, with variation on exposure to the Elections Canada ﬂyers
and implementation intention questions. Table 1 shows a 2 3 3 matrix representing the experimental
groups (entries report the sample sizes of each group).12
We measure the effect of these two types of treatments upon two separate, though related,
outcome variables: turnout intention and reported turnout.13
Turnout intention was measured in the
preelection wave of the survey, immediately after exposure (or not) to treatments. Reported turnout
is measured in the second wave of the survey, conducted after Election Day. Considering
both outcome variables allows us to determine if treatment effects are immediate or delayed in
nature.
As a ﬁnal note, we recognize that voters tend to be overrepresented in election studies of this
nature, and our dataset is no exception (the actual turnout rate in the 2015 Canadian election was
68.3%, as compared to 88.9% in our postelection sample). We suspect that this overrepresentation
may be due to several factors common to all studies like ours, including sampling bias (voters are
more likely to participate in a survey of this nature), a Hawthorne effect (whereby participation in a
political study increases the likelihood of voting), or misreported voting (nonvoters succumbing to
social desirability pressures and stating that they voted). We see no reason why these factors would
inﬂuence the effects of our treatments upon reported turnout. Sampling bias should be irrelevant in
this regard, as all respondents agreed to participate in the study long in advance of receiving treatments,
while we expect that Hawthorne effects would apply equally to all survey participants of a
political survey of this nature. Finally, social desirability pressures should not differ between groups,
as the treatments are all value-neutral with respect to the act of voting. As such, we proposed that any
observed differences between our experimental groups to be the result of treatments, rather than methodological
artefacts.
Table 1. Experimental Setup
Implementation Intentions
Nickerson and Rogers (2010) Questions Elections Canada Questions Control
Exposure to ﬂyers Yes 261 234 232
No 230 254 243
12
Sample sizes are from a preelection questionnaire (fewer respondents completed the postelection questionnaire). Postelection
sample sizes can be found in Table 4.
13
Both turnout variables are based upon the following question: “In each election, we ﬁnd that a lot of people are not
able to vote because they were not registered, they were sick or they did not have the time. On Election Day, are you
planning to vote (turnout intention)/did you vote (reported turnout)?”
Implementation Intentions, Information, and Voter Turnout 7
Results
To begin, we examine the independent effects of exposure to the Elections Canada ﬂyers and
implementation intention questions upon intended and reported turnout, testing Hypothesis 1. To that
end, Table 2 shows the rates of turnout on the basis of exposure to the Elections Canada ﬂyers and
reveals the difference in intended and reported turnout rates between the two experimental groups (we
estimate the statistical signiﬁcance of these differences using chi-square tests).14
Put succinctly, Table 2 reveals that exposure to the two pieces of Elections Canada literature
employed here, on their own, had no effect upon turnout in the 2015 Canadian federal election. There
is no difference between the treatment and control groups with respect to either turnout intention or
reported turnout, suggesting that the ﬂyers alone had neither short- nor long-term effects.15
Our data
thus provide no support for Hypothesis 1. This is not to rule out the possibility that Elections Canada’s
larger information campaign (which included a variety of advertisements through various media) had
no impact upon turnout. However, we ﬁnd no evidence that the two speciﬁc pieces of literature to
which survey respondents were exposed had an independent effect.
To evaluate Hypothesis 2, we determine the effect of our implementation intention treatments
independent of the effect of the Elections Canada ﬂyers. Table 3 shows the results of this analysis,
again reporting values for the turnout variables for each experimental group, as well as the differences
between groups. The ﬁrst implementation intention treatment group was exposed to the same questions
used by Nickerson and Rogers (2010), while the second was asked the aforementioned questions
speciﬁcally relevant to Elections Canada’s informational materials.
Table 2. Turnout (Intention and Reported) by Exposure to Elections Canada Flyers
Flyers Control Difference
Turnout intention 78.95 (N 5 727) 79.09 (N 5 727) 20.14
Reported turnout 88.91 (N 5 514) 88.96 (N 5 489) 0.05
* p < 0.10, ** p < 0.05.
Table 3. Turnout (intention and reported) by Exposure to Implementation Intention Questions
Group 1 Nickerson and
Rogers (2010) Questions
Group 2 Elections
Canada Questions
Group 3
Control
Difference
(1–2)
Difference
(1–3)
Difference
(2–3)
Turnout intention 81.26 (N 5 491) 79.92 (N 5 488) 75.79 (N 5 475) 1.34 5.35** 4.13*
Reported turnout 89.88 (N 5 326) 91.18 (N 5 340) 85.76 (N 5 337) 21.30 4.12* 5.42**
* p < 0.10, ** p < 0.05.
14
As we have directional expectations for our effects, all analysis is based upon one-tailed tests of signiﬁcance.
15
Readers will notice that reported turnout rates are higher than turnout intention. We expect that this is due to survey
attrition, with voters more likely than abstainers to complete the second (postelection) wave of the survey. We do not
expect this difference to bias our analysis of treatment effects. Readers may also notice that, despite our use of random
selection, the size of the “reported turnout” groups in Table 1 varies, with the control group having roughly 5%
fewer respondents than the treatment group. Given our assumption that voters are more likely to complete the postelection
survey, such a difference might cause one to doubt the null results for the reported turnout results in Table 1;
in essence, if nonvoters in the control group were particularly likely to drop out, this might conceivably lead us to
overestimate turnout among the control group. For two reasons, we are conﬁdent that, even if such an effect does
exist, it is not enough to account for the observed null results. First, there is no statistically signiﬁcant difference in
the rate of completion of the postelection survey and exposure to the Elections Canada ﬂyers. Thus the difference in
sample size is statistically insigniﬁcant. Second, we estimated the rate of turnout for the control group in this category
if it had the same sample size as the treatment, assuming that the additional respondents had voted at the national
average rate of 68.3%. The estimated turnout rate decreases slightly to 88.0%, a value which remains statistically
indistinguishable from the turnout rate for the treatment group.
8 Anderson et al.
In contrast to the null results in Table 2, Table 3 reveals statistically signiﬁcant effects for both
treatment groups and for both outcome variables. As compared to the control group, both sets of
implementation intention questions were associated with increases in turnout intention and reported
turnout. The size of the effects were the same for both treatments (just over 4 percentage points), with
no statistically signiﬁcant differences observed on either turnout measure, and are strikingly similar to
the 4.1-point effect observed by Nickerson and Rogers (2010). Of note, we ﬁnd that both types of
messages affect both intended and reported turnout. Such a ﬁnding suggests that implementation
intention questions had an immediate effect and that this effect lasted until Election Day. The data are
thus congruent with Hypothesis 2.
So what then about the interaction of the ﬂyers and implementation intention questions? Table 2
suggests that, on their own, Elections Canada’s informational ﬂyers had no effect. Might it be the case
that they do if combined with implementation intention questions? To consider this, we present Table
4, which shows rates of turnout (intention and reported) for each combination of the experimental
treatments.
Simply put, the results in Table 4 suggest that the implementation intention treatments only had
an effect upon those participants exposed to the Elections Canada ﬂyers, providing support for
Hypothesis 3. Among those participants who saw the informational materials, the effect of both implementation
intention treatments was over 7 percentage points, on both turnout intention and reported
turnout. In the absence of exposure to the ﬂyers; however, the implementation treatment had no effect
on either indicator of turnout. This ﬁnding suggests that the effect of implementation intentions is conditional
upon having sufﬁcient information to plan effectively for the act in question.16
We wish to note one ﬁnal consideration about the longevity of the observed effects. Our study
employs two cross-sectional dependent variables, but it is conceivable that treatments may lead to
behavioral changes between survey waves. One possibility is that exposure to our treatments might
Table 4. Turnout by Exposure to Flyers and Implementation Intention Questions
Group 1 Nickerson and
Rogers (2010)
Questions
Group 2
Elections Canada
Questions
Group 3
Control
Difference
(1–2)
Difference
(1–3)
Difference
(2–3)
Turnout
intention
Flyers 80.84 (N 5 261) 82.05 (N 5 234) 73.71 (N 5 232) 21.21 7.13* 8.34**
No ﬂyers 81.74 (N 5 230) 77.95 (N 5 254) 77.78 (N 5 243) 3.79 3.96 0.17
Difference 20.9 4.1 24.07
Reported
turnout
Flyers 90.91 (N 5 176) 92.22 (N 5 167) 83.63 (N 5 171) 21.31 7.28** 8.59**
No ﬂyers 88.67 (N 5 150) 90.17 (N 5 173) 87.95 (N 5 166) 21.5 0.72 2.22
Difference 2.24 2.05 24.32
* p < 0.10, ** p < 0.05.
16
We recognize the possibility that some survey respondents who viewed the Elections Canada materials may have
anticipated that our study was aimed at identifying the relationship between exposure to the materials and turnout.
Such individuals may theoretically be more likely to state that they plan to vote or have voted, which would provide
an alternative explanation for our ﬁndings. We are unable to disentangle such effects from those theorized in our article.
For two reasons, however, we are skeptical that such an effect is responsible for the observed results. First, even
if such an effect existed, we would expect it to be strongest in the ﬁrst wave of the survey (when respondents were
exposed to the ads). Given that the second survey was conducted several days later, we should anticipate such an
effect to weaken, and possibly disappear, by the time the second wave of the questionnaire is administered (particularly
since both waves consisted of dozens of additional questions not considered in the present study, so respondents
may not easily be able to determine the purpose of this particular survey experiment). If we were to observe an effect
with turnout intention, but not reported turnout, there would therefore be some cause for concern. As the results stand,
however, this is not the case—the same relationship was observed for both intended and reported turnout. Second,
there is already a well-known social-desirability bias to report that one plans to vote or has voted. Survey respondents
who hoped to “impress” the researchers would already be expected to respond to this pressure, regardless of exposure
to the educational materials.
Implementation Intentions, Information, and Voter Turnout 9
increase the likelihood of responding immediately to the turnout intention questions, but this effect
may weaken, or “wear off,” by Election Day. The results above seem to suggest that our effects are
long lasting, but this type of “conversion” is worth examining in greater detail. We conducted a brief
analysis to test for such effects, replicating the results from Tables 2 and 3 above, limiting the sample
only on those respondents who stated that they intended to vote in the preelection wave. Among this
subset of survey participants, neither of our treatments was associated with a difference in turnout, as
reported in the postelection survey. In other words, the effects of the treatments did not appear to
“wear off” by Election Day. The effects of our experimental treatment were thus relatively long-
lasting.
We ran a similar longitudinal analysis focusing only upon those respondents who stated in the
preelection survey that they did not intend to vote. This time, the purpose was to determine if exposure
to treatments had a “delayed effect,” perhaps affecting reported turnout more strongly than turnout
intentions. On this matter our data are suggestive, though sample-size limitations make us hesitant
to draw authoritative conclusions. Our dataset includes only 190 respondents who stated in the preelection
survey that they did not intend to vote and for whom we also have postelection turnout
responses. Among these respondents, we ﬁnd that exposure to the Elections Canada materials is not
associated with postelection reported turnout (thus indicating no delayed effect). However, there is
limited evidence that the implementation intention questions (pooling these respondents) are associated
with increased levels of turnout (p 5 0.094). Such a ﬁnding points to the importance of further
research into the temporal dimensions of implementation intentions and political behavior. Put another
way, more research is required to determine if the effects of such treatment may take some time to
“sink in.”
Conclusions
Drawing upon insights from the psychological literature on the concept of implementation intentions,
our experimental study adds to the growing scholarship on the effects of government-run public-education
programs upon voter turnout. More speciﬁcally, we have discovered that exposure to the
Elections Canada informational materials, on their own, have no discernible effect upon turnout.
However, when also asked to engage with questions about the process of voting, individuals who saw
the ﬂyers did report higher rates of turnout. We also ﬁnd that the two types of implementation intention
questions had the same effect upon turnout, despite the fact that one group of questions was
designed speciﬁcally to complement the Elections Canada ﬂyers. Such a result may suggest that the
very act of consciously reﬂecting on the voting process affects turnout rates, regardless of which
aspects of the process are contemplated. The effects of these treatments were both immediate and
long lasting.
Why is it that our experimental treatments only have effects when both are present? Why might
we only observe effects upon turnout among those individuals who have seen the Elections Canada
informational materials and who are asked questions about their intentions to vote? One possible
answer to this question lies in the complexity of the action serving as our outcome variable. At ﬁrst
glance, the act of voting may seem relatively simple; you go to the polling booth, show identiﬁcation,
and then cast a ballot. As noted above, however, changes were made to the registration process and
identiﬁcation requirements in the months leading up to the 2015 federal election. These changes
received a great deal of attention in the media. Indeed, the informational materials developed by Elections
Canada, and presented to survey respondents here, were meant to address some of these
changes.
Accordingly, our implementation intention treatments might only have an effect when the action
being considered is fully understood by respondents. Those participants exposed to the information
treatment will have greater information of the voting process than those who are not. When the former
10 Anderson et al.
group is asked to contemplate the process of voting through the implementation intention questions,
their deliberations will be made on the basis of a relatively high level of information about the action.
This difference may indeed lead to deliberations having a greater impact upon the likelihood of undertaking
the act of voting. In contrast, if little is known about the act under consideration, deliberations
will be made on the basis of less information and thus may be less effective. Future research is
required to test our explanation in more detail and to identify the exact mechanisms through which
implementation intention treatments affect behaviors in high- and low-information settings. Our
results nevertheless point strongly to one factor that might inﬂuence the effectiveness of implementation
intention considerations: information about the act under consideration.
Regardless of the explanation for our ﬁndings, however, the results of our study have implications
for actors tasked with conducting informational campaigns in the period leading up to elections (particularly
if the goal of such campaigns is to encourage voter turnout). Our experiment, conducted in
the context of a real election and using actual materials from a government agency, suggests that the
effectiveness of such informational campaigns hinges upon the extent to which electors actively contemplate
the act of voting and do so with accurate information. One question for those interested in
increasing turnout therefore becomes how to encourage such contemplation. Regardless of the answer
to this question, however, our data suggest that educating the public about the requirements of voting
is a necessary condition for such contemplation to have an effect. In a context such as the 2015
Canadian federal election, where electoral rules had recently been changed, providing electors with
information on the voting process appears to have been an important, though insufﬁcient, condition
for boosting voter turnout.
We wish to conclude with a discussion of the generalizability of our ﬁndings. We have argued
above that our use of information materials from an actual EMB, and data collected during the course
of a real national election, endows our ﬁndings with high external validity. Still, we do not assume
that our results will hold in all contexts. We do, however, expect that they would hold in elections
where eligibility or identiﬁcation or registration rules have recently changed or are unclear. Such elections
are far from rare, with many American states, for example, debating the merits of changing identiﬁcation
requirements in recent years. In the face of such uncertainty about voting rules, we expect
exposure to information and implementation intention interventions to increase voter turnout. The
applicability of our ﬁndings to other settings should be tested with replication studies.
ACKNOWLEDGMENTS
This project was supported by funding from the Bishop’s University Senate Research
Committee and the Samuel Clark Research Fund in Social Science, Western University. Correspondence
concerning this article should be addressed to Cameron D. Anderson, Department of Political
Science, Social Science Centre, Western University, London, Ontario, Canada, N6A 5C2. E-mail:
cander54@uwo.ca
REFERENCES
Achtziger, A., Gollwitzer, P. M., & Sheeran, P. (2008). Implementation intentions and shielding goal striving from
unwanted thoughts and feelings. Personality and Social Psychology Bulletin, 34, 381–393.
Barber, B. R. (2003). Strong democracy: Participatory politics for a new age. Berkeley, CA: University of California
Press.
Bayer, C., & Gollwitzer, P. M. (2007). Boosting scholastic test scores by willpower: The role of implementation
intentions. Self and Identity, 6, 1–19.
Blais, A. (2000). To vote or not to vote? Pittsburgh, PA: University of Pittsburgh Press.
Implementation Intentions, Information, and Voter Turnout 11
Brown, S. D., Perrella, A. M., & Kay, B. J. (2010). Revisiting local campaign effects: An experiment involving literature
mail drops in the 2007 Ontario election. Canadian Journal of Political Science, 43, 49–67.
Cavanagh, T. E. (1981). Changes in American voter turnout, 1964–1976. Political Science Quarterly, 96, 53–65.
Davenport, T. C., Gerber, A. S., Green, D. P., Larimer, C. W., Mann, C. B., & Panagopoulos, C. (2010). The enduring
effects of social pressure: Tracking campaign experiments over a series of elections. Political Behavior, 32,
423–430.
Dholakia, U. M., & Bagozzi, R. P. (2003). As time goes by: How goal and implementation intentions inﬂuence enactment
of short-fuse behaviors. Journal of Applied Social Psychology, 33, 889–922.
Elections Canada. (2015). Election Canada advertisements ﬂyer 1 and ﬂyer 2. Ottawa: Elections Canada.
Gerber, A. S., Green, D. P., & Larimer, C. W. (2008). Social pressure and voter turnout: Evidence from a large-scale
ﬁeld experiment. American Political Science Review, 102, 33–48.
Gollwitzer, P. M. (1996). The volitional beneﬁts of planning. In P. M. Gollwitzer & J. Bargh (Eds.), The psychology of
action: Linking cognition and motivation to behavior (pp. 287–312). New York, NY: Guilford.
Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54,
493–503.
Gollwitzer, P. M., & Oettingen, G. (2013). Implementation intentions. In M. Gellman and J. R. Turner (Eds.),
Encyclopedia of behavioral medicine (pp. 1043–1048). New York, NY: Springer-Verlag.
Gollwitzer, P. M., & Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis of effects
and processes. Advances of Experimental Social Psychology, 38, 69–119.
Green, D. P., & Gerber, A. S. (2008). Get out the vote: How to increase voter turnout. Washington, DC: Brookings
Institution Press.
Holland, R. W., Aarts, H., & Langendam, D. (2006). Breaking and creating habits on the working ﬂoor: A ﬁeld
experiment on the power of implementation intentions. Journal of Experimental Social Psychology, 42, 776–783.
Kersten, P., McPherson, K. M., Kayes, N. M., Theadom, A., & McCambridge, A. (2015). Bridging the goal intention–
action gap in rehabilitation: A study of if-then implementation intentions in neurorehabilitation. Disability and
Rehabilitation, 37, 1073–1081.
Leventhal, H., Singer, R., & Jones, S. (1965). Effects of fear and speciﬁcity of recommendation upon attitudes and
behavior. Journal of Personality and Social Psychology, 2, 20–29.
Loewen, P., Rubenson, D., & Martin, A. (n.d.). Implementation intention messages do not outperform motivational
messages: Evidence from a massive ﬁeld experiment. Working Paper, University of Toronto.
McGregor, M., & Anderson, C. 2014. The effectiveness of elections Canada’s campaign period advertising. Canadian
Journal of Political Science, 47, 813–826.
Nickerson, D. W., & Rogers, T. (2010). Do you have a voting plan? Implementation intentions, voter turnout, and
organic plan making. Psychological Science, 21, 194–199.
Pammett, J., & LeDuc, L. (2003). Explaining the turnout decline in Canadian federal elections: A new survey of
non-voters. Ottawa: Elections Canada.
Pateman, C. (1970). Participation and democratic theory. Cambridge, United Kingdom: Cambridge University Press.
Rogers, T., Milkman, K. L., John, L. K., & Norton, M. I. (2015). Beyond good intentions: Prompting people to make
plans improves follow-through on important tasks. Behavioral Science & Policy, 1, 33–44.
Rothman, A. J., Gollwitzer, P. M., Grant, A. M., Neal, D. T., Sheeran, P., & Wood, W. (2015). Hale and hearty policies
how psychological science can create and maintain healthy habits. Perspectives on Psychological Science, 10,
701–705.
Salisbury, R. H. (1975). Research on political participation. American Journal of Political Science, 19, 323–341.
Sheeran, P. (2002). Intention—behavior relations: A conceptual and empirical review. European Review of Social
Psychology, 12, 1–36.
Wolﬁnger, R. E., & Rosenstone, S. J. (1980). Who votes? New Haven, CT: Yale University Press.
Verba, S., Schlozman, K. L., Brady, H. E., & Brady, H. E. (1995). Voice and equality: Civic voluntarism in American
politics. Cambridge, MA: Harvard University Press.
12 Anderson et al.
Appendix A
Elections Canada Advertisements
Flyer #1
Implementation Intentions, Information, and Voter Turnout 13
Flyer #2
14 Anderson et al.
Appendix B
Experimental Group Balance Check: Multinomial Logistic Regression
Table B1 shows the results of a multinomial logistic regression model to conﬁrm balance
between experimental groups. Each series of entries in the table represents a comparison between two
groups (for a total of 15 group-wise comparisons). Rows of groups represent the baseline category, to
which the other group (represented by the columns) is compared.
Legend
Group 1 No ﬂyers, no message
Group 2 No ﬂyers, Nickerson & Rogers messages
Group 3 No ﬂyers, Elections Canada messages
Group 4 Flyers, no message
Group 5 Flyers, Nickerson & Rogers messages
Group 6 Flyers, Elections Canada messages
Model statistics
Likelihood ratio Chi2
5 20.00
Likelihood ratio p-value 5 0.75
Pseudo R2
5 0.0043
N 5 1,306
Table B1. Comparison of Experimental Groups on the Basis of Sociodemographic Characteristics
Group 2 Group 3 Group 4 Group 5 Group 6
Base Group 1 Age (years) 20.01 (0.01) 0.00 (0.01) 20.00 (0.01) 0.00 (0.01) 0.00 (0.01)
Gender (female) 20.10 (0.20) 0.16 (0.19) 0.10 (0.19) 0.13 (0.19) 20.04 (0.20)
Education (University educated) 0.05 (0.20) 20.22 (0.19) 0.16 (0.20) 0.10 (0.19) 20.17 (0.20)
Income (above median) 0.18 (0.34) 20.16 (0.36) 0.06 (0.34) 0.40 (0.32) 0.21 (0.35)
Visible minority 0.29 (0.20) 0.01 (0.20) 0.03 (0.20) 0.04 (0.20) 0.14 (0.20)
Constant 0.11 (0.38) 20.02 (0.37) 0.03 (0.37) 20.23 (0.37) 20.21 (0.38)
Group 2 Age 0.01 (0.01) 0.00 (0.01) 0.01 (0.01) 0.01 (0.01)
Gender 0.26 (0.20) 0.20 (0.20) 0.23 (0.19) 0.05 (0.20)
Education 20.28 (0.20) 0.11 (0.20) 0.04 (0.20) 20.22 (0.20)
Income 20.35 (0.35) 20.12 (0.33) 0.22 (0.31) 0.04 (0.34)
Visible minority 20.28 (0.20) 20.26 (0.21) 20.25 (0.20) 20.16 (0.32)
Constant 20.13 (0.37) 20.08 (0.38) 20.34 (0.37) 20.32 (0.38)
Group 3 Age 20.01 (0.01) 0.00 (0.01) 0.00 (0.01)
Gender 20.06 (0.19) 20.03 (0.19) 20.21 (0.20)
Education 0.38 (0.20) 0.32 (0.19) 0.06 (0.20)
Income 0.23 (0.36) 0.57 (0.34) 0.38 (0.36)
Visible minority 0.02 (0.20) 0.03 (0.20) 0.13 (0.20)
Constant 0.05 (0.37) 20.21 (0.37) 20.19 (0.38)
Group 4 Age 0.01 (0.01) 0.01 (0.01)
Gender 0.03 (0.20) 20.15 (0.20)
Education 20.06 (0.20) 20.33 (0.20)
Income 0.34 (0.32) 0.15 (0.34)
Visible minority 0.01 (0.20) 0.11 (0.21)
Constant 20.26 (0.37) 20.24 (0.38)
Group 5 Age 0.00 (0.01)
Gender 20.18 (0.19)
Education 20.26 (0.20)
Income 20.19 (0.32)
Visible minority 0.10 (0.20)
Constant 0.02 (0.38)
Note. Entries report coefﬁcients and standard errors (in parentheses). * p < 0.05.
Implementation Intentions, Information, and Voter Turnout 15