Voter turnout is the percentage of eligible voters who cast a
ballot in an election. After increasing for many decades, there has been a
trend of decreasing voter turnout in most established
democracies since the 1980s. In general, low turnout is attributed
to disillusionment, indifference, or a sense of futility (the perception that
one’s vote won’t make any difference). Low turnout is usually considered to be
undesirable since it can lead to unequal representation among various parts of
the population. Consequently, there have been numerous studies which aim was to
identify the determinant factors of turnout, in order to encourage
participation in the political process. Despite significant study into the matter,
academics are divided on the causes and consequences for the changes in voting
turnout. Its cause has been attributed to a wide range of economic, demographic,
cultural, technological and institutional factors. Among these factors, of
relevant importance is the impact of the electoral system on voter turnout. The present work wishes to contribute to the
literature that tries to explain voter turnout (Downs 1957, Riker and Ordeshook
1968), in particular to the literature that studies the link between the
electoral system and voter turnout.
Blais and Carty
(1990) estimate the impact of the proportional representation on voter turnout
of western democracies and identify higher turnout rates in PR.
Sanz (2015) studies
the effect of electoral system on voter turnout in Spain where municipalities follow different electoral
systems depending on their population size. He compares turnout under closed
list proportional representation and under an open list, plurality-at-large
system where voters can vote for individual candidates from the same or
different party lists. His results suggest that open list systems, which
introduce competition both across and within parties, are conducive to more voter
turnout. On the same line, with the present work the aim is to assess the
impact of electoral system on voter turnout in Italy. I will exploit the unique
institutional framework of local elections where municipalities follow
different electoral systems depending on their population size, namely the
single ballot plurality system and the dual ballot majority system, as required
by a national law. More precisely, the electoral systems in question allow
citizens to express their vote according different rules, depending on the
number of inhabitants in each municipality.
Several empirical studies have attempted to explain
the consequence of having more freedom on the choice of a candidate on voter
turnout. Among them, there are contrasting
views in the literature about how the use of Open List vs. Closed List systems
should affect voter turnout. On the one hand, it has been argued that OL
systems should increase turnout. Mattila (2003) claims that since voters can
choose the candidate they wish to vote for, they are likely to feel more
satisfied with the act of voting. Similarly, Karvonen (2004) shows that voting
for individual candidates makes the election more personal and concrete and
that both elements should provide an incentive for active electoral
In contrast, Robbins (2010) hypothesizes that
turnout should be higher in CL systems. His argument is that in OL systems
parties may not exert the same level of resources to lobby support or mobilize
voters as in CL systems. Mattila’s (2003) empirical results, with data from
elections to the European Parliament, show that a variable indicating a CL
system is not significant. In their empirical analyses, Blais and Aarts (2006),
Dos Santos (2007) and Karvonen (2004) also conclude that there is not
satisfactory evidence to support the hypothesis of a positive correlation
between preferential voting and electoral participation.
Finally, the empirical analysis of Farrell and
McAllister (2006) find that preferential voting systems, where voters are given
more freedom in completing the ballot paper, lead to higher satisfaction with
Overall, the empirical evidence is non-conclusive:
the effect of the ballot structure on turnout is still an open question.
Hence, the present work aims to contribute to the
already existing collection of works, hoping to put more light on the
consequences that the freedom of choice of voters has on turnout. Moreover, to
my knowledge there are no other studies which use a quasi-experimental design
to estimate how the two electoral systems in question affect voters’ casting
choice and consequently voter turnout of Italian municipalities.
Electoral system in Italy
administrative divisions in Italy are: Regions, Provinces and Municipalities.
Municipalities are autonomous and independent administrative divisions. They
are included in the Province. However, they can entertain direct relations with
the regions and the State due their autonomy. As of February 2017, there were
7,981 municipalities in Italy. The political organs of a municipality include municipal
council (a legislative body) and the municipal committee (an executive body),
both organs assist the mayor. Elections are held every five years.
implementation of the law n.81 of 1993, the mayor is directly elected by
citizens which happens simultaneously with the election of the municipal
council. Moreover, such law requires that municipalities are subject to two
different electoral rules, according to their population size.
with a population of less than 15000 residents apply the single ballot plurality
system. Residents in this category
vote for both the municipal council members and the mayor. Voting for a mayoral
candidate means voting also for the list supporting the candidate. It will be
elected mayor the candidate that gets the plurality of votes. The election
takes place in a single round, except for the eventuality in which the two most
voted candidates have obtained the same number of votes. In this case, the
second ballot will proceed and if a further parity were to be registered, it
will be elected the eldest one among them. The list of the elected mayor will
get a two-thirds majority of the seats in the municipal council, whereas
one-third of the seats will be proportionally distributed among the defeated
other hand, municipalities with a population of more than 15000 residents apply
the dual ballot majority election rule. If no candidate has the
majority of votes (i.e. more than half) in the first round, then only the two
candidates with the most votes proceed to a second round (called “closed
ballot”), which will be held two weeks later. In the second ballot, because
there are only two candidates, one candidate will attain with certainty a
majority. In this category, each candidate is supported by
one or more lists. Besides that, each
voter is entirely free to change the candidate he votes for, even if his
preferred candidate has not yet been eliminated but he has simply changed his
mind (called “panachange vote” or “disjoint vote”). The “panachange
vote” allows citizens of municipalities with more than 15000 residents to
vote, in the first round, in three possible ways:
can vote for a list. By voting a list they also vote the candidate associated
to that list.
can vote for a list and they can specify two candidates of that list, provided
they are of opposite sex.
can vote a candidate. By voting this way, their vote is valid only for the
candidate and not for the list or lists supporting the candidate.
second round, voters can vote only between the two most voted mayoral
candidates and cannot vote for any list. The two candidates can choose to be
supported by a different list than the one they were initially supported by.
The winner is the candidate that gets the majority of votes. Once the winners
are announced, seats will be assigned to each list or group of lists with the
proportional d’Hondt method.
municipal council is elected in a single round with a complex system that we
can define as mixed in some specific cases, while it can be proportional in
others. In the latter case it is called a “mixed system”, since it can be
partly proportional and partly majoritarian. More precisely, we can have the following
is elected in the first round (by majority), but the list or lists supporting
the mayor do not get 50% of votes. In this case seats are assigned in a
is elected in the first ballot by majority, but the list or lists supporting
the mayor gets 40% of votes and no other list gets more than 50% of votes. In
this case the list supporting the mayor gets 60% of the seats in the council
and the remaining seats are assigned proportionally.
is elected in the second ballot (by plurality), but the list supporting the
mayor already got in the first ballot 40% of the votes and no other list got
more 50% of votes. Also in this case, the list supporting the mayor gets 60% of
seats and the remaining seats are assigned proportionally. Lists that have not
achieved at least 3% of votes are not assigned any seats.
Overall, this is an “anomalous” electoral
system, both with respect to the Italian institutional history (as for the
first time it is also directly voted for an executive body) and for what
concerns the broad categories of the electoral systems themselves. The anomaly
it consists in the fact that, with a single contextual and connected vote, both
the mayor and the council (i.e. the executive and legislative bodies) are
An important thing to keep in mind, is that the preference
vote was maintained in both municipality groups. Moreover, in both cases,
expressing a unique preference allows the voter to change the order of the
candidates in the list and to establish their ranking. The main difference
between larger and smaller municipalities, is therefore, the possibility of
expressing a disjoint vote in the
Data and Methodology
In order to estimate the causal effect of the two
electoral rules on voter turnout, I will collect a rich data set from the
official interior ministry site, with results from all elections since the
implementation of the national rule (data available from 1993 to 2017). The
data set contains detailed information on candidates and voters in each
election year, both at the aggregate level and at the municipality-individual
level (number of electors and of voters, number of invalid votes, sex of voters
and candidates, etc).
Moreover, I will use publicly available data from
the National Statistics Institute (ISTAT) to investigate the validity if the
identification strategy. ISTAT attributes to each existing municipality a
six-digit code. The first three digits of the code identify the province to
which the municipality belongs, while the subsequent three identify the
municipality within the province. Following each administrative change, the
list of Italian municipalities and their codes is updated. It also takes care
of an archive containing the list of the municipalities in force, the list of
deleted municipalities, the territorial and administrative changes of the
municipalities since 1991 and a table of codes and names of geographical areas,
regions and provinces.
To assess the impact of the electoral system on
voter turnout I will exploit the regression discontinuity design.
The control group will be represented by the
municipalities with a population less than 15000, while the treatment group
will be represented by the municipalities with a population higher than 15000. I will consider the following variables: the
percent of voter turnout, the percent of invalid votes and the share of votes for the leading parties in
Italy. The main outcome of interest will be Turnout.
Moreover, I will add to the basic equation
municipality and year fixed effects.
A suitable model could be the following:
Therefore, the identification strategy entails that,
conditional on municipality and
year fixed effects, there are no
unobservable factors that may simultaneously affect voter turnout and whether a
municipality’s population is just above or just below the cut-off. Hence, we
can let the value of the unobserved variables be different in the treatment and
in the control group, as long as they remain fixed over time and at individual
level. Error terms will be clustered at municipality level.
Thanks to its peculiarities, the Italian
institutional context of local elections enables to trustfully identify the
differences in outcomes between treatment and control group, as the causal
effect of the electoral system. Meaning, the number of municipalities and the
years of all elections since the adoption of the national law are very high (data
about elections available from 1993 to 2017). Accordingly, the number of
observations will also be high which will lead to increased precision
in estimates. Moreover, all the
municipalities under any of the electoral rules follow the same electoral
system and no other rule changes at the cut-off besides that. Thus, the
variable of interest will identify the average treatment effect of the
electoral system on voter turnout for municipalities close to the threshold.
Additionally, I will restrict my analysis to only
those municipalities where elections did not have a second ballot, for several
First, as a consequence of the different nature of
the two electoral systems, the control group and the treatment group will have
different probabilities to proceed to the second turn of elections. Therefore,
the estimates will be more precise if we only compare municipalities similar with
respect to the prior probability of having a second turn of elections. But,
knowing this, candidates would behave differently if they were in a
municipality with risk of run-off, regardless of the electorate’s inclination. For
example, smaller party lists may decide to present themselves with their own
candidate and measure their consent in the first round, significantly fragmenting
the electoral offer. Eventually, in a second moment, they could form a
coalition with one of the two competitors for the run-off, in order to also
enjoy the eventual majority prize.
On the other hand, voters themselves
can set up strategic behaviors, casting their vote only for the candidate-mayor
with higher possibility of victory and not for the list supporting that
candidate (disjoint vote).
In both cases, estimates would be biased.
Therefore, by considering exclusively municipalities with no second ballot
elections, we reduce such biases.
design can be invalid if individuals can precisely manipulate the assignment
variable. A way to
test the imprecision of control over the assignment variable is to examine its density.
If the density of the assignment variable for each individual is continuous, then
the marginal density of the assignment variable over the population should be
continuous as well. A jump in the density at the threshold is probably indicative
of some degree of sorting around the threshold, which invalidates the
suitability of the RD design. We could also examine if the treatment and the
control group are similar conditional on observable variables, most of which
should be the same. Finally, we could implement a falsification test. One would
expect there to be continuity in predetermined variables at the treatment
cut-off. Since these variables were determined before the treatment decision,
treatment status should have no effect on them. If a discontinuity in
predetermined variables is present at the treatment cut-off or in other points,
then the RD design is put into question.
Another problem might be that of endogeneity.
However, as already said, the electoral system a municipality must follow is
determined by a national law as a function of the its population size, which
reduces considerably the problem of endogeneity.
As we know, another key issue when using RD design
is the choice of the bandwidth, since there is a trade-off between precision
and bias, with both increasing as the bandwidth increases. In order to choose
the most appropriate bandwidth, we can provide results for different portions
of it and see how sensitive the results are to these different adoptions.
The biggest limitation of this approach is that
only municipalities with single-ballot elections will be considered in the
analysis. This implies a restriction in the number of observations, but more
importantly, it could mean that we are only considering those municipalities
where voters are less indifferent to certain choices than to others.
This together with the possibility of strategic
behavior of candidates and voters, as already discussed, would bias our