A Priori versus Empirical Voting Power in the EU Council of Ministers
| DOI | 10.1177/1465116504040446 |
| Date | 01 March 2004 |
| Author | Antti Pajala,Mika Widgrèn |
| Published date | 01 March 2004 |
| Subject Matter | Journal Article |
04 040446 (to/d) 9/1/04 11:20 am Page 73
European Union Politics
A Priori versus Empirical
DOI: 10.1177/1465116504040446
Volume 5 (1): 73–97
Voting Power in the EU
Copyright© 2004
SAGE Publications
Council of Ministers
London, Thousand Oaks CA,
New Delhi
Antti Pajala
University of Turku, Finland
Mika Widgrén
Turku School of Economics, The Research Institute of the Finnish
Economy, Centre for Economic Policy Research and CESifo
A B S T R A C T
Numerous scholarly articles have investigated member
states’ a priori voting power in past, present and possible
future Councils of the European Union (EU). This article
introduces three empirically oriented variants of the stan-
dardized Banzhaf index and compares them with the stan-
dardized Banzhaf index itself. The comparison suggests that
the averages of the empirically oriented indices converge
towards the Banzhaf index, and hence that there is no domi-
nant ordering of the member states or the issues and, conse-
quently, no stable minimal winning coalitions over time.
However, statistical analysis reveals that the country-by-
country average difference from the Banzhaf index,
considering all issues, remains significant in all three vari-
ants of empirical power.
K E Y W O R D S
Banzhaf index
EU Council
power index
voting
voting power
7 3
04 040446 (to/d) 9/1/04 11:20 am Page 74
7 4
European Union Politics 5(1)
Introduction
The first studies on voting power in the EU Council of Ministers appeared
some years after the first enlargement of the Union in 1973 (e.g. Johnston,
1977). Today, after four enlargements and with the fifth just taking place, the
Council has probably become the most popular target for voting power
research (see Widgrén, 1994, 1995; Pajala, 2002: 16–17). The motivation behind
the studies has been the relationship between member states’ varying
numbers of votes and voting power. It is well known and easy to show that
this relationship is not straightforward (Felsenthal and Machover, 1998: 4–5).
In the past, almost no information about the Council’s internal negotia-
tions was publicly available. In recent years, however, the situation has
changed somewhat. The first scholarly works known to us applying such
information were the edited volume by Bueno de Mesquita and Stokman
(1994) and the roll-call analysis by Mattila and Lane (2001). There are also
some attempts in the literature to explain decision-making outcomes more
indirectly using power indices. Studies of the relationship between budget
transfers from the EU to member states are a good example (see Baldwin et
al., 2000, 2001). In these studies, power indices indeed have explanatory
power. The Decision-making in the European Union (DEU) data set, in turn,
covers controversial Commission proposals between 1995 and 2000, includ-
ing detailed information about negotiations within the Council. The data set
opens up new avenues for voting power analysis in the Council. The DEU
data set is applied here (and is explained in more detail in the introduction
to this special issue by Thomson and Stokman).
Power indices, such as the (standardized) Banzhaf (1965) index or the
Shapley–Shubik (1954) index, measure individual voters’ capabilities to
control the outcome of a vote. The indices are by nature a priori and take into
account only constitutional aspects, voting weights and the vote threshold of
a voting body. Proponents of the voting power approach do not usually add
any information to their evaluation and, what is more striking, power
measures are very rarely tested against empirical data (see, however, Baldwin
et al., 2001). Therefore, our intention is to examine one of the classical a priori
power measures – the normalized Banzhaf index – against empirically
oriented variants of the index.
An increasing number of studies over recent years have introduced
empirically oriented indices. The aim is to obtain a more realistic picture of
a voting body by including additional information in the index calculation.
This information comprises data about voter preferences, which are shown
as negotiating positions. However, other similar ideas have been presented
as well (Pajala, 2002: 74–5). The works of Kirman and Widgrén (1995),
04 040446 (to/d) 9/1/04 11:20 am Page 75
Pajala and Widgrén
A Priori versus Empirical Voting Power
7 5
Widgrén (1995), Bilal and Hosli (2000), Colomer and Hosli (1997), Edelman
(1997), Hosli (1999, 2002) and Perlinger (2001) are methodologically close to
this article since they define very similar swing sets to the ones outlined
below. The sets are variations of the Shapley–Shubik index and the Banzhaf
index (Shapley and Shubik, 1954; Banzhaf, 1965).1
The main research problem in this study is whether the positioning of
member states on controversial issues leads to systematically or significantly
different power distributions than the Banzhaf index, where the impact of
policy positions is disregarded. In our study we shall consider only decision-
making in the Council, and the research question is relevant only for issues
that are decided upon using qualified majority voting (QMV) – 62 out of 87
votes. Moreover, we assume decision-making in connected coalitions in a
unidimensional policy space throughout the study. We shall define three new
power indices, which are based on the Banzhaf index. They measure funda-
mentally different swing possibilities of member states in unidimensional
connected coalitions. A comparison of the swing possibilities with the ones
in the Banzhaf index gives us information on power gain or loss on the basis
of the positioning of the member states. For the empirical investigation of the
research question we shall use the DEU data set.
Assessing a small fraction of the DEU data set, Pajala (2002) speculates
that the different values of the Banzhaf index and the variations might
converge in the long run. In the absence of systematic differences in actors’
policy positions, the values should converge according to the law of large
numbers. A further research question is thus: do the differing values between
the Banzhaf index and the variations converge in the long run? And if they
do, how and why?
Lastly, the results reveal possible patterns in member states’ negotiating
behaviour. Did a member state have a tendency to be an extreme or a centrist
actor in Council negotiations? Which member state had the most strategically
advantageous negotiating position?
A discussion on the philosophical foundations of voting power and
power indices can be found in recent forum sections of this journal (Albert,
2003, and this issue; Felsenthal et al., 2003). As regards the illustrative expla-
nations, or ‘stories’, of the power indices, some researchers tend to prefer the
Shapley–Shubik, others prefer the Banzhaf, and some researchers like neither
of them. The criticism seems to be always the same: the power indices assume
too little and too much at the same time: ‘too little’ means that, in order to
analyse a voting body, only the constitutional aspects are taken into account
(voters’ votes and the threshold); ‘too much’ is assumed if we consider coin
toss voting (the probability of a voter voting yes/no is 50%) and especially
when we expect each voter combination or permutation to be equally likely.
04 040446 (to/d) 9/1/04 11:20 am Page 76
7 6
European Union Politics 5(1)
In this study we try to overcome some of these problems. To start with, we
have preference data for the voters, which we can add to the constitutional
information. This, in turn, enables us to restrict the set of allowable coalitions.
In our model, however, yes/no voting is unchanged. In sum, this study is an
attempt to make the standardized Banzhaf index more realistic when suitable
data are available, but, as is shown below, all this comes with a cost.
The article is organized as follows. We first present the basic concepts of
voting power together with some new additional concepts that are required
for the measurement of empirically oriented voting power. We also introduce
the Banzhaf index, together with the three new empirically oriented varia-
tions of the index and a brief discussion on the behaviour of the variations.
The subsequent section starts with a short description of the applied data and
then reports and discusses the results of the empirical investigation of the
research problems. Finally, we present some concluding remarks on the basis
of the results.
Voting power, the Banzhaf index and the variations
Concepts of voting power
Decision-making situations in which individuals or groups of individuals
must make collective decisions can be found everywhere. In our case, in the
EU Council the member states must collectively decide how to deal with a
Commission proposal. In formal decision-making bodies such as the Council,
there usually exist specific rules for decision-making; i.e. the rules must deter-
mine what kind of coalition can ensure acceptance. The decision rule deter-
mines how many votes there must be in favour of a proposal to guarantee its
acceptance. More than one decision rule exists in the Council, depending on
the type and contents of a proposal.
Formally, a voting body can be represented as a set N of actors. The set
W includes all winning subsets, i.e. subsets that can ensure acceptance of a
randomly chosen proposal. The notion of a simple game was introduced by
von Neumann and Morgenstern (1944). In simple games (using zero–one
normalization), a coalition S has the value 0 or 1. If S wins it has the value of
1, otherwise 0. The characteristic function v indicates the...
To continue reading
Request your trial