Turnout and Closeness: Evidence from 60 Years of Bavarian Mayoral Elections

• DOI: http://doi.org/10.1111/sjoe.12241
• Date: 01 April 2018
• Author: Felix Arnold
• Published date: 01 April 2018

©The editors of The Scandinavian Journal of Economics 2017.

Scand. J. of Economics 120(2), 624–653, 2018

DOI: 10.1111/sjoe.12241

Turnout and Closeness: Evidence from 60

Years of Bavarian Mayoral Elections*

Felix Arnold

DIW Berlin and FU Berlin, DE-10117 Berlin, Germany

farnold@diw.de

Abstract

One prediction of the calculus of voting is that electoral closeness positively affects

turnout via a higher probability of one vote being decisive. I test this theory with

data on all mayoral elections in the German state of Bavaria between 1946 and 2009.

Importantly, I use constitutionally prescribed two-round elections to measure electoral

closeness and thereby improve on existing work that mostly uses ex post measures that

are prone to endogeneity. The results suggest that electoral closeness matters: a one

standard deviation increase in closeness increases turnout by 1.27 percentage points,

which corresponds to 1/7 of a standard deviation in this variable. I also evaluate how

other factors such as electorate size or rain on election day affect turnout differentially

depending on the closeness of the race. While rain decreases turnout on average, this

effect is mitigated in close elections, as indicated by a positive interaction effect of

the two variables.

Keywords: Germany; mayoral elections; two-round ballot

JEL classification:D72; H70

I. Introduction

The well-known calculus of voting (Downs, 1957; Riker and Ordeshook,

1968) postulates that the probability of influencing the electoral outcome is

an important determinant in the turnout decision. It predicts that electoral

participation should increase with the closeness of the race. Put differently,

when “more is at stake”, people have an extra incentive to go to the polls.

However, it is not straightforward to measure electoral closeness in

empirical applications. How do people know whether the election is

competitive or not? Several approaches are used in the literature. Most

existing studies rely on ex post realized closeness and assume rational

expectations of the voters. Apart from making a strong assumption,

additional endogeneity concerns arise. If vote shares appear on the right-

*I would like to thank Ronny Freier, Viktor Steiner, Peter Haan, participants at the Spring

Meeting of Young Economists in Ghent and the ifo Political Economy Workshop in Dresden,

and the two anonymous referees for their many helpful comments. All remaining errors are

my own.

F. Arnold 625

hand side and turnout on the left-hand side of the empirical model,

the total number of votes cast appears on both sides of the equation.

Turnout and ex post closeness may, therefore, be spuriously correlated

(Cox, 1988). As a consequence, other researchers employ ex ante historic

closeness measures from previous elections to avoid these issues. However,

elections are not necessarily comparable across time, especially if the set

of candidates changes. The longer the electoral period, the less useful the

information from a previous election. A third approach therefore relies

on pre-election opinion polls to gauge voters’ perceptions of electoral

closeness. However, these data are typically only available for large-scale

federal elections. It is unclear whether electoral closeness exists in this

context at all, because even a razor-thin victory in a large electorate might

translate into vote margins of tens of thousands of votes between winner

and runner up.

In this paper, a different approach is used. I employ two-round elections

to identify closeness effects. Mayoral elections in the German state of

Bavaria provide an ideal institutional setting: according to the local

constitution in Bavaria, if no mayoral candidate is able to get an absolute

majority of the vote in the first round, a second (runoff) election is held

two weeks later with the two leading candidates from the first round.

Now, closeness of the first round can be taken as a proxy for closeness

of the second round. This measure, which I call “revealed closeness”,

avoids the drawbacks of other commonly used measures in the literature.

Revealed closeness is both predetermined and recent at the same time,

whereas other ex ante (ex post) measures only fulfill the first (second)

criterion. Another advantage of studying Bavarian mayoral elections lies

in the small size of the electorate. Municipalities in Bavaria can have as

few as 17 eligible voters. The largest city in the data still has fewer than

40,000 eligible voters. Thus, I can evaluate closeness effects in a setting

where the probability of being the decisive voter is non-negligible, which

is rarely the case in large-scale federal elections.

Using data on all mayoral elections in 2,031 Bavarian municipalities

between 1946 and 2009, I find that closeness does indeed matter for

electoral participation: a one standard deviation increase in closeness

increases turnout by 1.27 percentage points, which corresponds to 1/7

of a standard deviation in this variable. Importantly, I show that revealed

closeness, which makes use of two-round elections, yields approximately

the same results as the common ex post measures employed in the

literature so far. Due to the panel structure of the data, I can also evaluate

a historical ex ante measure of closeness. I contribute to the literature

by offering a first comparison of three differently calculated indicators

(ex ante,ex post, and two-round) for electoral closeness. Importantly, this

comparison is conducted within one institutional framework.

©The editors of The Scandinavian Journal of Economics 2017.