Swiss popular votes: Does turnout increase when the outcome

Swiss popular votes: Does turnout increase when the outcome is expected to be tight?
Paper prepared for the 4th International Conference on Democracy as Idea and Practice,
University of Oslo, January 10 – 11, 2013
Sveinung Arnesen
UNI Rokkan Centre for Social Research
Nygårdsgaten 5
5015 Bergen
Tel.: +47 5558 3885
Email: [email protected]
Oliver Strijbis
Institut für Politikwissenschaft
Universität Hamburg
Allendeplatz 1
20146 Hamburg
Tel.: + 49 0(40) 428 38 32 33
Email: [email protected]
Abstract
This article analyzes the effect of expected closeness of the outcome and turnout for 18
national and cantonal Swiss referenda and initiatives. The study makes three main
contributions. First, it is the first study to operationalize the expected closeness with data from
prediction markets. Second, it adds to the small literature on the effect of expected closeness
on turnout in direct democratic decisions. Third, it is the first one to investigate the
interrelationships between expected closeness and turnout at different levels within a country.
Our results suggest that the closeness of the closest national ballot has a considerable effect on
turnout for all proposals of the same date for popular votes at the national and the subnational
level.
1
Introduction
Direct democratic decision making is on the rise around the world (Altman 2010; LeDuc
2003). According to many, this is a positive trend since direct democracy has positive effects
on policy representation, the economy, and probably even on life satisfaction.1 One of the few
major normative problems with direct democracy, however, is that it is often associated with
low turnout (e.g. Freitag and Stadelmann-Steffen 2010). Hence, the question on what impacts
on turnout in direct democratic ballots is of formidable importance. The aim of this article is
to test whether expected closeness of the outcome does also trigger turnout for direct
democratic ballots.
From the individual voter’s perspective, the benefit of voting increases when the election is
expected to be tight, since the voter’s probability of being pivotal then rises. Also, in the
context of a close election, the competing political actors will spend more resources on
convincing the population. They thereby reduce the voters’ costs of acquiring the knowledge
they need to make an informed decision in the voting booth. Hence, turnout should be higher
when the race is expected to be tight than when the outcome is expected to be clear because
both the benefits for the voters are increased and the costs are reduced. In the context of
elections the empirical support for this argument is solid; according to an extensive literature
review by Blais (2006) it is one of the most consistent findings in cross-national electoral
research that tight elections have a positive effect on turnout numbers (also Geys 2006).
So far little research on direct democracy has been devoted to study whether the expected
closeness of ballots in popular referenda and initiatives does increase turnout. Furthermore,
the few studies that have tried to answer this question have come to competing conclusions.
An important reason for the lack of cumulative findings lies in the difficulty to measure the
expected closeness of the outcomes. The expected closeness of the outcomes of direct
democratic decisions has so far been measured with data on the actual outcome while research
on election turnout has started to make use of polls. In this article we go beyond both types of
studies by making use of data from prediction markets that we organized in relation to these
referenda.
1
For reviews see (Budge 2006; Lupia and Matsusaka 2004; Maduz 2010; Papadopoulos
2001).
2
Prediction market data are the most valid data to use as measure for expected closeness. In
contrast to polls and actual outcomes, prediction markets directly measure the anticipation
that a vote will be close or not. Polls only measure the current public opinion at some given
time point ahead of the referendum, and are per se not making any future predictions. This
point is especially important in the context of direct democratic choices, since opinions can
change relatively late in direct democratic campaigns. Also, using data from the actual
outcome is of little use since the result is not known to the voters at the time they need to
make a decision on voting or staying at home.
A second weakness of current research on turnout in referenda and popular initiatives is that it
neglects the multi-level nature of many ballots. In particular in the case of Switzerland ballots
at the cantonal and national level take place simultaneously and might impact on each other.
Hence, this study seeks to investigate the mutual mobilization effects of ballots at the national
and subnational level. This article seeks to fill this gap by analyzing 18 national and cantonal
Swiss referenda and initiatives.
The article has three main contributions. First, we add to the small literature on the effect of
expected closeness on turnout in direct democratic decisions. Second, we are the first ones' to
operationalize the expected closeness with data from prediction markets. Third, we are the
first ones to investigate the interrelationships between expected closeness at different levels
within a country. Our results suggest that the closeness of the closest national ballot has a
considerable effect on turnout for all proposals of the same date be it at the national or at the
subnational level.
The article is structured as follows. In the first section we summarize the current literature on
the expected closeness of election outcomes on turnout. In the second section we focus more
closely on the literature on turnout in direct democratic ballots and derive our hypotheses. We
then turn to empirical part of the paper by presenting our data. In the fourth section we discuss
our results. In the final section we summarize our results and point to future avenues of
research.
Expected closeness and turnout – theory and empirical findings
3
Why people use their right to vote is a central and persistent question within political science.
In the vast literature on electoral turnout, there are many empirically and theoretically
founded determinants of turnout. One of them, which resonates well both with theoretical
arguments and the empirical findings, is that competitive elections increase turnout. Indeed,
there are two theoretical arguments that favor the positive relationship between expected
closeness and turnout. The decision hypothesis states that from the individual voter’s
perspective, the benefit of voting increases when the election is expected to be tight, since the
voter’s probability of being pivotal then rises (Downs 1957; Ledyard 1984; Palfrey and
Rosenthal 1983). The mobilization hypothesis argues that, in the context of a close election,
the competing political actors will spend more resources on convincing the population (Key
1949). They thereby reduce the voters’ costs of acquiring the knowledge they need to make an
informed decision in the voting booth. Hence, turnout should be higher when the race is
expected to be tight than when the outcome is expected to be clear because both the benefits
for the voters are increased and the costs are reduced.2 The empirical support for this
argument is solid; indeed it is one of the most consistent findings in cross-national electoral
research that tight elections have a positive effect on turnout numbers {Blais, 2006 #419119}.
Rational choice models view the individual voter as a utility maximizing individual that
calculates the costs and the benefits of voting, before making the decision to vote or not.
When the benefits outweigh the costs, the rational voter will go to the polling station. In
practice, the direct costs seem to always outweigh the benefits of voting. The probability that
one single vote will be pivotal is microscopic (Funk 2010; Matsusaka and Palda 1993), which
leads to the question why anyone would want to vote at all. We know that they do, hence,
viewing the voters’ decision as a product of a cost/benefit analysis is not satisfactory to
explain the voters’ electoral behavior.
One way to save the rational choice model is to stretch the meaning of costs and benefits by
pointing to psychological benefits of the procedure of voting. Voting, then, is understood as
an expressive act (Riker and Ordeshook 1968). The act of voting, then, implies benefits - for
instance through the satisfaction of a duty to vote - which are weighted against its costs. By
introduction of normative motivations such an explanation derives considerably from classical
rational choice approaches and comes close to social psychological and cultural explanations.
2
It is not our intention to determine which of the theoretical arguments is the correct one.
4
And it shares the same problem with these theories which is the question why turnout varies
so dramatically between elections within countries and across time when voting is associated
with psychological benefits or social norms that can hardly be assumed to vary in the short
run (Blais 2006, 113; Kirchgässner and Schulz 2005, 6).
Due to these weaknesses of the rational choice models scholars have come up with alternative
explanation for a correlation between the closeness of the outcome and turnout. Most
importantly they have argued that in the context of a narrow election campaign intensity will
be higher, which might lead to higher mobilization of the voters (Key 1949). In particular in
the context of first-past-the-post elections the relationship between expected closeness and
campaign intensity is well established. In the context of direct democratic decisions it
remains, however, unclear whether the expected closeness of an outcome has a relevant
impact on campaign intensity.
In particular in the context of popular initiatives the goal is often not to achieve a relative
majority for the proposal. Instead, the goal of popular initiatives is often in setting certain
topics on the political agenda. The goals of the initiative takers are sometimes even achieved
before their proposal is voted on. This is the case when the opponents on the initiative change
legislation in the direction of the initiative in order to reduce the probability of a success of
the latter (Kriesi 2005; Papadopoulos 2001). This can explain why in Switzerland the
organization of initiatives is a popular tool although only very few of them are approved by
the voters.3 Hence, in the context of votings it remains unclear whether the "mobilization
hypothesis" holds, i.e. whether the expected closeness of an outcome has a relevant effect on
turnout.
Besides theoretical considerations the study on the relationship between the closeness of
outcomes and turnout has been confronted with empirical issues. One point of discussion is
that while the empirical relationship between the narrowness of the outcome and turnout is
well established, the magnitude of the impact is often shown to be small. According to Blais
(2006) a reason might be that the variable is not adequately measured. The standard indicator
is the vote gap between the first and the second party. A conceptual issue here is whether the
absolute or the relative difference in the votes should be taken. According to theory it should
3
The rate of aproval is less than 10% at the national level.
5
be the absolute difference since this is the value that determines the individual's vote impact
or the efficiency of mobilization efforts by the elites.4
A more fundamental problem of indicators calculated from the closeness of the outcome is
that they estimate the perceived narrowness ex-post (for an early review see Matsusaka and
Palda 1993; also Ashworth, Geys, and Heyndels 2006; Kirchgässner and Schimmelpfennig
1992; Endersby, Galatas, and Rackaway 2002). The problem is that the voters might not
correctly perceive the narrowness of an election. For instance, under the impression of media
coverage voters might overestimate the narrowness of the vote. As a response to this problem
several attempts to arrive at better indicators for the voters perception of the narrowness of the
election have been considered.
Most importantly, research on election turnout has turned to polls (Kunce 2001; Shachar and
Nalebuff 1999). Although the use of polls in order to operationalize the expected outcome is a
major advancement it is not a perfect solutions. This is because polls only measure the current
public opinion at some given time point ahead of the referendum, and are per se not making
any future predictions. This point is especially important in the context of direct democratic
choices, since opinions can change relatively late in direct democratic campaigns.5 An
alternative operationalization of the expected outcome is to take the first round result of an
election in two round ballots. Fauvelle-Aymar and François (2006) for French and Simonovits
(2012) for Hungarian two-round elections show that the closeness of the first round has a
positive effect on turnout in the second round. Obviously, this result is restricted to the
context of particular electoral law and is not applicable to direct democratic decision making.
Expected closeness and turnout in popular initiatives and referenda
4
From the mobilization perspective this is in some instances less clear since it depends on the
assumption whether campaigns can be entirely canalized to specific electoral districts or
whether campaigning efforts diffuse across constituencies (for example through national
media coverage).
5
Polls that are made one month before voting day are generally not considered predictions
since the share of undecided voters is often huge.
6
The finding that the expected closeness of the outcome triggers turnout is well-established in
electoral research. This is the reason why current research is less interested in the overall
effect of expected closeness than on the mechanisms with which closeness triggers turnout.
The situation is very different for research on turnout in referenda and popular initiatives. In
this context it remains undecided whether the expected closeness of the outcome has a
relevant impact on turnout or not.
In one of the first and still the largest study of the effect of expected closeness on turnout
Matsusaka (1993) investigates this relationship for 885 Californian ballot propositions from
1912 through 1990. He fails to find a systematic link although he suggests that ballot
propositions should be a most likely situation for such a relationship to be established.
Two more recent studies systematically investigate the relationship between the expected
closeness of the outcome of votings and its turnout. Kirchgässner und Schulz (2005)
separately test the effect of the closeness of an election outcome and the campaign
expenditure on the turnout in Swiss national ballots. They use data on 142 Swiss referenda
and initiatives from 1981 to 1999. The empirical results suggest that the closeness of the
outcome has some impact on the (financial) campaign intensity, which itself strongly triggers
turnout. The closeness of the outcome, however, has no direct effect on turnout. This suggests
that it is the elites' mobilization efforts and not the individuals' cost-benefit calculations that
impact on the propensity to vote (Kirchgässner and Schulz 2005, 32-33).
The centrality of the campaign intensity to the turnout is in line with recent research that
emphasizes the relevance of direct democratic campaigns for the activation of the voters
(Bernhard 2012; Kriesi 2012; Nai 2013).6 A question here is whether it is the closeness of the
most contested outcome of the same voting day that matters or whether it is the average
closeness of the propositions that is relevant. This is hard to judge from the results of the
analysis of Kirchgässner and Schulz. It also remains unclear from the perspective of the
mobilization effort. According to the "motor proposal hypothesis" what matters is if at voting
day there is one proposal that mobilizes the voters with diffusion effects for all other ballots
of the same day (Joye and Papadopoulos 1994). However, one might alternatively argue with
the "voter segment hypothesis" that several campaigns mobilize different segments of the
electorate and trigger high turnout only in combination.
6
And less so for their conversion.
7
The result from Kirchgässner and Schulz contrast with the analysis of Søberg and Tangerås
(2007), which analyze voter turnout using data from 309 local Norwegian language
referendums carried out during 1965–2005. They find that turnout is positively correlated
with the closeness of the outcome. Since campaign intensity in these local referenda can be
assumed to be very modest their result seems to corroborate a direct link between the
closeness and turnout.7 This is plausible insofar as in local referenda each voter's probability
to cast the decisive ballot is considerably higher than in national referenda with large numbers
of voters. Also Aguiar-Conraria and
agalh es (2010) find in the context of referenda on the
accession to European treaties a positive effect of closeness of the outcome and a negative
effect of the size of the electorate on turnout. However, since they do not control for
campaign intensity it is hard to judge whether the closeness of the outcome had a direct effect
on turnout or whether it was moderated by the elites' campaign efforts such as suggested by
Kirchgässner and Schulz.
It is premature to draw conclusions from these few studies of the relationship between the
expected closeness of the outcome and turnout. First, all studies on expected closeness and
turnout in direct democratic ballots use the actual outcome as a proxy for the predicted
outcome. As argued above this operationalization might be heavily misleading and so doing
invalidate research on the topic.
Second, it remains unclear whether the results from Kirchgässner and Schulz are compatible
with the ones from Søberg and Tangerås. The reason is that they investigate the relationship
between closeness of the ballot and turnout at two very different levels. While Kirchgässner
and Schulz investigate voting turnout at the national level, Søberg and Tangerås investigate it
at the local level. The fact that the absolute narrowness of the race varies systematically
between the two studies—at the national level absolute closeness is systematically larger than
at the local level—makes the results only partly comparable. In fact it is possible that the
expected closeness of the result matters such as predicted by the decision hypothesis and the
mobilization hypothesis in both cases. In the case of national ballots the mobilization
hypothesis explains why a relevant share of the voters goes to the polls while the decision
hypothesis explains why the campaigns are not sufficient to mobilize the entire electorate. At
7
However, they also find a negative correlation of turnout with electoral size, which clearly
speaks against the rational choice model. This is so since in the case of small electorates the
probability that a single citizens vote is determinant for the outcome is high.
8
the level of the local ballots the decision hypothesis can explain the relevant share of voters
that turn out while the low levels of mobilization explain why turnout at the local level is not
higher than at the national level.8
Since this interpretation is based on two cases only—national referenda in Switzerland and
local referenda in Norway—this is speculation. Hence, what is needed is to simultaneously
investigate the relationship between the expected closeness of the outcome and turnout at the
national and the local level. Only with such a research design it is possible to contrast
situations in which the expected outcome is clear but campaign intensity high (national
uncontested ballots) and situations in which the outcome is expected to be close while
campaign intensity is low (local contested ballots).
The comparison of turnout at the national and local level also points to an additional research
problem that has been neglected entirely: the mutual relationships between ballots of the same
day at different levels. The question here is whether turnout at the national level also triggers
turnout at the subnational level such as would be expected from the motor proposal
hypothesis, or whether ballots at the subnational level can also increase turnout for national
ballots, such as we might expect from the voter segments hypothesis. This is also an open
question because we cannot extrapolate from research on turnout in subnational ("second
order") elections (Reif and Schmitt 1980). This is because with elections there is no
segmentation of the electorate according to issues such as in direct democratic decision
making.
To summarize, both the decision and the mobilization hypothesis predict that the closer the
expected outcome of a ballot the higher turnout. This is our first hypothesis:
H1: The closer the expected outcome the higher turnout
If we assume that both the decision and the mobilization hypothesis are true it does not matter
on the level the ballot takes place. However, if we think that only the decision hypothesis is
true, we expect the closeness only to have a positive impact on turnout in the case of small
electorates:
8
In the Norweigan local referenda turnout was around 60% while the turnout in Swiss
national referenda is typically around 40%.
9
H2: The expected closeness of the outcome has a stronger effect in subnational ballots than in
national ones
From the perspective of the mobilization hypothesis, however, the closeness of the outcome
should only trigger turnout in contexts where campaigns are generally intense. Hence,
according to this hypothesis we expect:
H3: The expected closeness of the outcome has a stronger effect in national ballots than
subnational ones
Additionally we might distinguish between the "motor voting" and the "voter segment"
hypothesis. While the former states that the turnout among all ballots at the same date might
be determined by the expected closeness of one proposition, according to the "voter segment"
hypothesis we can expect the number of ballots and their diversity to positively impact on
turnout. If we apply the above to the relationship between the closeness of the outcome and
turnout we can hypothesize according to the motor proposal hypothesis:
H4: The more close the closest expected outcome the higher turnout of all proposals
According to the voter segment hypothesis it is not one single proposal that matter, but the
average. Hence, according to this hypothesis we can state:
H5: The more close the average expected outcome the higher turnout of all proposals
Data and method
In order to make turnout rates comparable across levels our dependent variable consists of the
turnout for both national and cantonal ballots among the electorate of the Canton of Zurich.
As Kirchgässner and Schulz (2005) propose turnout is measured as the number of Yes and No
votes divided by the people entitled to vote. Such an operationalization can better capture
turnout than the official numbers which include blank votes since it is less of an effort to vote
blank than not to vote for a specific proposal if one votes for at least one alternative proposal.9
9
This is because all proposals at one level (national, cantonal, local) are tied together on one
piece of paper that must be divided into parts if one decides not to vote on some proposals but
wants to cast the vote on others.
10
The aggregate turnout figures have been retrieved from the Statistical Office from the Canton
of Zurich.
The absolute expected closeness has been operationalized as the predicted outcome from the
prediction market (in percent) times the electorate. The prediction market data have been
gathered from our own prediction market software.10 Prediction markets as defined by Berg,
Nelson, and Rietz (2003) are markets run with the purpose of using the information content in
market values to make predictions about specific future events. They are also known as
information or decision markets. The intention of these markets is to show the traders’ best
collective guess about future events such as election results, stock prices, and movie revenues.
In these markets, values of traded contracts depend directly on future outcomes and, hence,
prices give information about what traders believe will be the most probable outcomes (Berg
et al. 2003:79). Today there are a number of prediction markets, but perhaps the best known
prediction market is the Iowa Electronic Markets (IEM), which was first established at Iowa
University for experimental purposes in connection with the 1988 US presidential election.
Since then, it is generally established that prediction markets have predicted election
outcomes more accurately than comparable polls, both in the US and elsewhere (Berg et al.
2008a; Berg et al. 2008b; Arnesen 2011b). 11
Prediction market data have previously been used to study partisan impacts on the economy
(Snowberg et al. 2007, 2006), to study the effect of campaign events on election outcomes
(Arnesen 2011a), and has been suggested used as information for political decision makers on
their likely consequences of their decisions (Berg and Rietz 2003; Wolfers and Zitzewitz
2009). These kinds of data have however never been used to study the effect of expected
closeness, even though they have a natural place in that literature. As such, this is an
innovation of this paper.
The political market that was used for the Swiss referenda are electronic real-time exchanges
where traders buy and sell futures contracts. The trading procedure is conducted using an
automated market maker which always ensures liquidity in the market (see Hanson (2007) for
10
Practically, establishing and running prediction markets in Switzerland has been organized
by Principe Consulting, in which both the authors of this paper are involved.
11
Note that Erikson and Wlezien have challenged the prediction markets’ superiority over polls. See (Erikson
and Wlezien 2008) and (Erikson and Wlezien 2012)
11
a detailed description of how the market maker works). Payoffs are based on the ultimate vote
outcomes. Because real money is used, traders are subject to the monetary risks and returns
that result from their trading behaviour. The data utilized for the analysis are from vote-share
markets. The predictions from vote-share markets may be read as a direct prediction of the
share of yes-votes the referendum in question is estimated to receive.
If structured correctly, the prices should reflect the expected future outcome, given the
information available at the time:
Though simple in concept, such markets act as complex, dynamic, interactive systems
that incorporate information in new ways. Through the action of traders, prediction
markets aggregate information from individuals, incorporate polls and other sources of
information and weight all of this information through the price formation process.
Berg, Nelson, and Rietz (2003:3)
In all, there are prediction data for 18 referenda and initiatives that have been conducted in
2012. 12 of them are at the national level, and six of them at the cantonal level (Zürich). The
relatively low number of observations does not let us reach strong conclusions. However,
given the strong theoretical and empirical foundations on which the assumption is based, we
find this study relevant as a first step in establishing whether or not competitive campaigns
also matter in a direct democratic context. All the data are available in the online appendix. 12
12
Online appendix url:
http://folk.uib.no/st03889/4th%20international%20conference%20on%20Democracy/
12
Analysis
The first column in table 1 below lists the names of the various initiatives/referenda. They
have for simplicity been shortened, for their full description we refer to the online appendix.
The next column displays the expected share of yes-votes the day before the voting has taken
place. As we can see, they range from a low of 39.1% to a high of 78.9%. Many of the votes
are anticipated to be quite close. If we use +/-five percentage points’ deviation from a 50/50
vote as the cutting point to separate close referenda from the others, then 11 out of the 18 are
expected to be close. Five of the tight outcomes are expected at the cantonal level, and six of
them are at the national level.
The turnout has a lower variation range, from 32.0% to 42.8%. Strikingly, all referenda held
on the same date tend to have similar levels of turnout. The turnout in most referenda are
within one percentage points of each other, and the highest deviation between two referenda
on the same day only marginally exceeds two percentage points (Gegenvorschlag Grundstufe
at 32.0% and Transparente Mieten at 34.1%).
13
Table 1: Overview of the referenda and their characteristics.
Name
Predicted yes-
Turnout
Date
vote, percent
Zurich, percent
Governmental
Type of
level
vote
Zweitwohnungen
46.7
41.7 March 2012
national
initiative
Bauspar
48.5
40.9 March 2012
national
initiative
Ferien
39.1
42.3 March 2012
national
initiative
Geldspiele
68.4
40.3 March 2012
national
referendum
Buchpreis
45.6
40.4 March 2012
national
referendum
Staatsverträge
43.6
39.3 June 2012
national
initiative
Krankenversicherung
43.9
39.2 June 2012
national
referendum
Bausparen
46.2
39.2 June 2012
national
initiative
Kunde ist König
47.8
39.8 June 2012
cantonal
initiative
Schulwahl
43.0
39.3 June 2012
cantonal
initiative
Kulturland
54.1
38.0 June 2012
cantonal
initiative
Jugendmusik
76.9
41.8 Sep 2012
national
referendum
Sicheres Wohnen im
49.3
Sep 2012
national
initiative
Sep 2012
national
initiative
42.1
Alter
Schutz vor
46.0
42.8
Passivrauchen
Tierseuchengesetz
78.9
32.5 Nov 2012
national
referendum
Einführung Grundstufe
45.0
32.3 Nov 2012
cantonal
referendum
Gegenvorschlag
54.3
Nov 2012
cantonal
counter
32.0
proposal
Grundstufe
Transparente Mieten
46.8
34.1 Nov 2012
cantonal
initiative
Source: Own data, Statistisches Amt Kanton Zürich (2013)
14
Figure 1 shows that there is very little difference in turnout when the expected close votes –
that is outcomes within the 45/55 percentage range – are separated from those who are not
expected to be close. The correlation coefficient between turnout and expected closeness also
reveals a weak relationship between the two variables with Pearson’s R standing at -0.12. 13 It
does have correct negative sign which indicates that the lower the distance to a tied vote, the
higher the turnout. The bivariate regression output however shows that the relationship is
insignificant (t-value of closeness variable is -0.47).
Figure 1: Differences in mean turnout in Swiss referenda, between those votes that are
expected to be close and those who are not. Turnout figures are in percent and from the
Zürich canton.
Turning to national vs. cantonal referenda, figure 2 shows that national referenda are
considered more important and consequentially has a higher turnout than cantonal referenda.
It shows that the citizens in Zürich turn out to a greater degree when the question is at the
national level than when the subject is voted on at a cantonal level. In our sample, the
cantonal referenda have a turnout of almost five percentage points lower than the national
referenda.
13
For the correlation analysis, the closeness variable is measured as absolute distance from a tied vote at 50%.
15
Figure 2: Differences in mean turnout in Swiss referenda, depending on which governmental
level the vote is held. Turnout figures are in percent and from the Zurich canton.
Having found that the general participation level is different between cantonal and national
levels, do the voters also react differently to competitive referenda on the two levels? We do
find in our data set that the only voting day that saw a substantial decrease in turnout concurs
with the only day that there was no competitive referendum at the national level (see table 1).
In November 2012, there were three contested cantonal referenda, and one uncontested
national referendum. As displayed in figure 3, the turnout for all the referenda dropped from
around 40 percent to the low 30’s. Could the turnout for all the referenda in November have
decreased due to the same reason, namely that there were no competitive national referenda
on that date? If so, this finding lends support to the motor voting hypothesis in the sense that
the absence of an important referendum, turnout figures will drop. And to be viewed as
important, a referendum must satisfy two conditions: It must be national, and it must be
competitive.
16
Figure 3: Referenda turnout on 2012 voting dates. Turnout numbers are from the Zürich
canton.
50%
45%
42.8
42.3
41.8
41.7
40.9
40%
40.3
42.1
40.4
39.3
39.8
39.2
39.2
39.3
38.0
35%
34.1
32.5
32.3
32.0
30%
25%
M
ar
M
ar
M
ar
M
ar
M
ar
Ju
n
Ju
n
Ju
n
Ju
n
Ju
n
Ju
n
se
p
se
p
se
p
no
v
no
v
no
v
no
v
20%
Figure 4 shows the difference in turnout between those referenda that are held on a day when
one or more of the national ones are expected to be close, and those that are held when there
is no national referendum or it is expected to have a clear outcome.
17
Figure 4: Differences in mean turnout in Swiss referenda, depending on whether there is at
least one close national vote or not. Turnout figures are in percent and from the Zürich
canton.
One should be wary of making too drastic inferences from a data set of aggregated data with
few observations. The results may be coincidental, and there might be unidentified factors that
explain the outcome. For now, suffice it to say that the bivariate relationship is quite strong
given the data we have at hand. The average turnout figures rise about seven and a half
percentage points when a national vote is presumed to be close on the voting day. With more
future research, it will be possible to establish whether or not this is the uncovering of the
impact from a national referendum motor.
Summary
This paper has sought to contribute to the low number of studies that have analyzed the
impact of expected contested referenda on turnout figures. While this relationship has been
found to be solid with regards to elections, only a handful of studies have been undertaken for
referenda and initiatives. Using prediction market data, this paper has introduced what we
argue is a more valid measure of expected closeness. The analysis of 18 referenda from 2012
indicates that motor voting might be a considerable driver of turnout in the context of Swiss
18
direct democracy. The motor referenda are those that are at the national level and are expected
to have a close outcome.
19
References
Aguiar-Conraria, Luís, and Pedro C. agalh es. 2010. “Referendum design, quorum rules
and turnout.” Public Choice 144(1-2): 63–81.
Altman, David. 2010. Direct Democracy Worldwide. Cambridge University Press.
Arnesen, Sveinung. 2011a. "How prediction markets help us understand events' impact on the
vote in US presidential elections." Journal of Prediction Markets 5 (3):42-63.
———. 2011b. "Informasjon, motivasjon, prediksjon: Eit forsøk med prediksjonsmarknad før
stortingsvalet 2009." Tidsskrift for samfunnsforskning (1):19.
Ashworth, John, Benny Geys, and Bruno Heyndels. 2006. “Everyone likes a winner: An
empirical test of the effect of electoral closeness on turnout in a context of expressive
voting.” Public Choice 128(3-4): 383–405.
Berg, Joyce E., and Thomas A. Rietz. 2003. "Prediction Markets as Decision Support
Systems." Information Systems Frontiers 5 (1):79-93.
Berg, Joyce, Robert Forsythe, Forrest Nelson, and Thomas Rietz. 2008a. "Results from a
Dozen Years of Election Futures Markets Research." In Handbook of Experimental
Economic Results. Amsterdam: Elsevier.
Berg, Joyce, Forrest Nelson, and Thomas Rietz. 2003. Accuracy and Forecast Standard Error
of Prediction Markets. Working Draft, Henry B. Tippie College of Business and
Administration, University of Iowa, Iowa City.
———. 2008b. "Prediction Market Accuracy in the Long Run." International Journal of
Forecasting 24:285-300.
Bernhard, Laurent. 2012. Campaign Strategy in Direct Democracy. Palgrave Macmillan.
Blais, André. 2006. “What Affects Voter Turnout?” Annual Review of Political Science 9(1):
111–125.
Budge, Ian. 2006. “Direct and Representative Democracy: Are They Necessarily Opposed?”
Representation 42(1): 1–12.
Downs, Anthony. 1957. An economic theory of democracy. Harper.
Endersby, James W., Steven E. Galatas, and Chapman B. Rackaway. 2002. “Closeness
Counts in Canada: Voter Participation in the 1993 and 1997 Federal Elections.”
Journal of Politics 64(2): 610–631.
Erikson, Robert S., and Christopher Wlezien. 2008. "Are Political Markets Really Superior to
Polls as Election Predictors?" Public Opinion Quarterly 72:1-24.
———. 2012. "Markets vs. polls as election predictors: An historical assessment." Electoral
Studies 31 (3):532-9.
20
Fauvelle-Aymar, Christine, and Abel François. 2006. “The impact of closeness on turnout: An
empirical relation based on a study of a two-round ballot.” Public Choice 127(3-4):
461–483.
Freitag, Markus, and Isabelle Stadelmann-Steffen. 2010. “Stumbling block or stepping stone?
The influence of direct democracy on individual participation in parliamentary
elections.” Electoral Studies 29(3): 472–483.
Funk, Patricia. 2010. “Social Incentives and Voter Turnout: Evidence from the Swiss ail
Ballot System.” Journal of the European Economic Association 8(5): 1077–1103.
Geys, Benny. 2006. “Explaining voter turnout: A review of aggregate-level research.”
Electoral Studies 25(4): 637–663.
Hanson, Robin. 2007. "Logarithmic market scoring rules for modular combinatorial
information aggregation." Journal of Prediction Markets 1 (1):1-15.
Joye, D., and Y. Papadopoulos. 1994. “Votations moteurs: les logiques du vote blanc et de la
participation.” Papadopoulos, Yannis (Hrsg.): 245–276.
Key, Valdimer Orlando. 1949. Southern politics in State and Nation. A. A. Knopf.
Kirchgässner, Gebhard, and Jörg Schimmelpfennig. 1992. “Closeness counts if it matters for
electoral victory: Some empirical results for the United Kingdom and the Federal
Republic of Germany.” Public Choice 73(3): 283–299.
Kirchgässner, Gebhard, and Tobias Schulz. 2005. “Was treibt die Stimmbürger an die Urne?
Eine empirische Untersuchung der Abstimmungsbeteiligung in der Schweiz, 19811999.” Swiss Political Science Review 11(1): 1–56.
Kriesi, Hanspeter. 2005. Direct Democratic Choice: the Swiss Experience. Lexington Books.
———. 2012. Political Communication in Direct Democratic Campaigns: Enlightening Or
Manipulating? Palgrave Macmillan.
Kunce,
itch. 2001. “Pre-Election Polling and the Rational Voter: Evidence from State Panel
Data (1986–1998).” Public Choice 107(1-2): 21–34.
LeDuc, Lawrence. 2003. The Politics of Direct Democracy: Referendums in Global
Perspective. University of Toronto Press.
Ledyard, John O. 1984. “The pure theory of large two-candidate elections.” Public Choice
44(1): 7–41.
Lupia, Arthur, and John G. atsusaka. 2004. “DIRECT DE OCRACY: New Approaches to
Old Questions.” Annual Review of Political Science 7(1): 463–482.
aduz, Linda. 2010. “Direct Democracy.” Living Reviews in Democracy 2(0).
http://democracy.livingreviews.org/index.php/lrd/article/view/lrd-2010-1 (Accessed
January 7, 2013).
21
atsusaka, John G. 1993. “Election closeness and voter turnout: Evidence from California
ballot propositions.” Public Choice 76(4): 313–334.
Matsusaka, John G., and Filip Palda. 1993. “The Downsian voter meets the ecological
fallacy.” Public Choice 77(4): 855–878.
Nai, Alessandro. 2013. “What really matters is which camp goes dirty: Differential effects of
negative campaigning on turnout during Swiss federal ballots.” European Journal of
Political Research 52(1): 44–70.
Palfrey, Thomas R., and Howard Rosenthal. 1983. “A strategic calculus of voting.” Public
Choice 41(1): 7–53.
Papadopoulos, Y. 2001. “How Does Direct Democracy atter? The Impact of Referendum
Votes on Politics and Policy‐ aking.” West European Politics 24(2): 35–58.
Reif, Karlheinz, and Hermann Schmitt. 1980. “Nine Second-Order National Elections – a
Conceptual Framework for the Analysis of European Election Results.” European
Journal of Political Research 8(1): 3–44.
Riker, William H., and Peter C. Ordeshook. 1968. “A Theory of the Calculus of Voting.”
American Political Science Review 62(01): 25–42.
Shachar, Ron, and Barry Nalebuff. 1999. “Follow the Leader: Theory and Evidence on
Political Participation.” The American Economic Review 89(3): 525–547.
Simonovits, Gábor. 2012. “Competition and turnout revisited: The importance of measuring
expected closeness accurately.” Electoral Studies 31(2): 364–371.
Snowberg, Erik, Justin Wolfers, and Eric Zitzewitz. 2006. "Party Influence in Congress and
the Economy." National Bureau of Economic Research Working Paper Series No.
12751.
———. 2007. "Partisan Impacts on the Economy: Evidence From Prediction Markets and
Close Elections*." Quarterly Journal of Economics 122 (2):807-29.
Søberg, orten, and Thomas P. Tangerås. 2007. “Voter turnout in small referendums.”
Electoral Studies 26(2): 445–459.
Statistisches Amt Kanton Zürich. 2013. “Wahlen & Abstimmungen.”
http://www.statistik.zh.ch/internet/justiz_inneres/statistik/de/wahlen_abstimmungen.ht
ml (Accessed January 7, 2013).
Wolfers, Justin, and Eric Zitzewitz. 2009. "Using Markets to Inform Policy: The Case of the
Iraq War." Economica 76 (302):225-50.
22
23