Professional Documents
Culture Documents
Golosov - 2010 - Party Politics - The Effective Number of Parties A New Approach
Golosov - 2010 - Party Politics - The Effective Number of Parties A New Approach
Golosov - 2010 - Party Politics - The Effective Number of Parties A New Approach
http://ppq.sagepub.com/
Published by:
http://www.sagepublications.com
On behalf of:
Political Organizations and Parties Section of the American Political Science Association
Additional services and information for Party Politics can be found at:
Subscriptions: http://ppq.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://ppq.sagepub.com/content/16/2/171.refs.html
What is This?
Grigorii V. Golosov
ABSTRACT
Introduction
The ‘effective’ number of parties is a very simple general concept that entails
a complex methodological problem. Conceptually, the effective number of
parties is simply the number of ‘viable’ or ‘important’ or (to put it most
radically) ‘real’ political parties in a party system that includes parties of
unequal size. In comparative political party research, we need to distinguish
between a four-party constellation with party vote-shares (0.52, 0.45, 0.02,
0.01), which is ‘effectively’ a two-party system, and a four-party constella-
tion with party vote-shares (0.40, 0.25, 0.20, 0.15), which is ‘effectively’ a
case of multipartism. There should be a way to discount very small parties.
The methodological problem is that there are none readily available. Of
course, it is possible to set an arbitrary threshold of exclusion, discounting
1354-0688[DOI: 10.1177/1354068809339538]
all parties that fail to reach 1 or 3 or 5 percent of the vote.1 Yet it is clear
that a 5 percent party might be unimportant if other parties enjoy massive
support, and it might be quite important if other parties are comparably
weak; for instance, if the largest party’s vote-share is 15 percent. Then we
need to quantify the idea of the ‘effective’ number of parties in a systematic
way that allows for taking into account the relative sizes of parties, which
is impossible without using a measure expressed as a mathematical formula.
One such measure, the Laakso–Taagepera effective number of parties
(Laakso and Taagepera, 1979), is most widely used in comparative research
on political parties and party systems (Amorim Neto and Cox, 1997; Cox,
1997; Lijphart, 1984, 1994; Ordeshook and Shvetsova, 1994). In this
analysis, I attempt to demonstrate that while the Laakso–Taagepera index
(NLT) is indeed a very good measure, a better measure still is possible. The
purpose of this article is to propose such a measure. In the first section of
the article, I briefly overview the relationship between NLT and the index
that serves as its computational core, the Herfindahl–Hirschman index of
industrial concentration, thus placing NLT within a wider context of aggre-
gate quantities used in social science research. In the second section, I
analyse the properties of NLT. Then I discuss the critiques of NLT and the
proposed alternative or supplementary measures, which will help identify
major problems that have to be solved by an alternative index. In the fourth
section, I propose a new measure and discuss its properties in relation to
those of NLT. In the final section, I illustrate these properties on a limited
set of real-life party constellations.
∑ (s − (1/ n))
i
2
(2).
V= σ = 2 1
n
172
Here, si is the size of the i-th component expressed as its decimal (absolute)
share of the total, and σ stands for the standard deviation. The notation
above and below the summation sign demonstrates that it involves all
components from the largest (S1) to the smallest (Sx).
Thus formula (1) can be rewritten as follows:
x
1
HH = –– +
n Σ1 (si – (1/n))2 (3).
x x
Since, under the given conditions, Σ1 si = 1 and Σ1 1 = n, it can be further
mathematically reduced:
x x
⎛ 2si 1 ⎞ 1 x 2 2 n x
HH = –1
n– + Σ (si – (1/n))2 = –1
– + Σ ⎜si2 – –––
n n + ––2⎟ = –n– + Σ si – –n– + ––2 = Σ si2.2
1 1 ⎝ n⎠ 1 n 1
In the end, we get the formula that is much simpler than formula (1) and
that is standard for HH:
x
HH = Σ1 si
2 (4).
the index (in other words, its theoretical maximums and minimums) at any
given size of the largest component, s1. The upper limit values of the index
are achieved when, in addition to s1 of a specified size, the given party
constellation includes an infinite number of infinitely small components. For
such constellations:
1
NLT = ––2 (6).
s1
The lower limit values of the index are achieved for party constellations
in which all components are equal to s1, while the smallest component is
the remainder. If s1 > 0.50, these are necessarily two-component constella-
tions where the second component is the remainder. Equal-component
constellations involve zero-size remainders. The general formula for the
lower limit values of NLT is:
1
NLT = ––––––––––––––––– (7),
Zs1(s1 – 2 + Zs1) + 1
where Z is the integer quotient of 1 and s1. It can easily be established that
if 1/s1 = Z, formula (7) is equivalent simply to 1/s1. This is not only the NLT
formula for all equal-component constellations, but also the only conceiv-
able formula satisfying the first basic requirement for such constellations.
A useful property of NLT is that it can be calculated not only on the basis
of the decimal shares of components, as in all formulae above, but also on
the basis of the raw numbers of votes or seats received by each of the parties.
In order to achieve that, we have to use a different formula:
2
⎡ x ⎤
⎢ ∑ si ⎥ (8),
NLT = ⎣ 1x ⎦
∑ si2 1
where si is the number of votes or the number of seats received by the i-th
party. This formula yields values equivalent to those yielded by formula (5),
yet it avoids minor deviations stemming from rounding. Of course, for a
user of contemporary software who can afford not to round the absolute
shares of seats or votes, no discrepancies between the values yielded by the
two formulae occur.
It is also noticeable that when introducing NLT, Laakso and Taagepera
(1979) described it as part of a wider family of indices whose generic
formula is:
1/(1 – a)
⎡x ⎤ (9).
Na = ⎢∑ sia⎥
⎣1 ⎦
175
1
NLT = ––––––––––– (10).
2(s12 – s1) + 1
178
179
x (si /2)
N= ∑ (16).
1 (si /2) + log(s1 + 1) – log(si + 1)
180
d = s12 – si2.
Then, with b = 1, we can write the following formula for the new index:
x
si
Np = ∑ (17).
1 si + s12 − si2
where Z is the integer quotient of 1 and s1. Yet it is clear that if Z = 1 the
formula reduces simply to 1 + (1 – s1)/s1 = 1/s1, which is what I sought to
receive.
The formula for the upper limit values of NP is:
s12 + 1 – s1
Np = –––––––– (19).
s12
These values are therefore always by (1 – s1)/s1 smaller than those of NLT.
The lower limit values of the new index, however, are not necessarily smaller
or equivalent. They exceed the values of NLT at 0.303 < s1 < 0.33, then at
0.224 < s1 < 0.25, then at 0.179 < s1 < 0.2, and so on, the pattern reiterating
itself at every equal-component constellation. Starting with s1 < 1/18, the
lower limit values of NP are always equal to those of NLT or greater than
them, which, of course, does not matter much for any practical research
purposes. However, it is practically useful to note that at very high levels of
fragmentation the distance between NP and NLT is not as well articulated
as between NB and NLT. For instance, the values of NLT, NB and NP for the
extremely fragmented constellation with s1 = 0.2 and 80 components at 0.01
each are 20.83, 12.91 and 17.03, respectively. In general, unlike with NB,
the mathematical form of the new index does not preclude it from assuming
values closely approximating those of the Laakso–Taagepera index or even
exceeding them, unless at the upper limit. But this happens only if s1 is
relatively small. For illustrative purposes, Table 1 reports the values of the
three indices most extensively discussed above, NLT, NB and NP, for eight
181
even though it might take an additional effort to identify the largest compo-
nent as such.
Like NLT, the new index can be calculated not only on the basis of the
decimal shares of components, as in all formulae above, but also on the
basis of the raw numbers of votes or seats received by each of the parties.
For this, we have to use the formula that is derivative from formula (20):
x
x ∑s i
Np = ∑ x
1
(21).
1
∑ si + (s12/si ) − si
1
x
This might seem excessively complex, but note that ∑s1
i is simply the
Indices remain abstract until they are tested on real-life data. To cite a recent
example, Dunleavy and Boucek (2003) test their index on a set of 102 post-
war elections in seven countries, including Canada, Japan and five countries
of West Europe. Laakso and Taagepera (1979) tested their index exclusively
on West European cases. Thus, with a few exceptions,9 the testing grounds
for measures of party system fragmentation were provided by the world of
well-established democracies. This is problematic. For a test of an effective
number of parties measure to be adequate, the set of real-life cases has to
include at least some cases indicating how the measure under examination
performs at the extremes, that is, in those situations when party systems are
either extremely fragmented or extremely concentrated. Few such situations
are available in the world of well-established democracies. For instance, of
the 102 elections tested by Dunleavy and Boucek (2003), in only 12 was the
vote-share of the largest party smaller than 30 percent, including nine in
183
France, two in Italy and one in Japan. At the same time, there is no single
case with s1 > 0.6. In my view, this concentration on well-established democ-
racies should be rejected for two reasons. The methodological reason has
already been stated above: we need to know how indices perform at the
extremes, not in situations that are bound to generate moderate scores. In
such situations even indices with significant deficiencies, such as the Molinar
index, assume quite intuitively appealing values. However, they are not very
different from what other indices indicate.
The second reason is more practical. In 1979, when Laakso and Taagepera
proposed their effective number of parties, the world of well-established
democracies was almost synonymous with the world of democracy in general.
Since then, however, we have witnessed a massive ‘third wave’ of democ-
ratization (Mainwaring, 1999), and, as a setback, a wide spread of what is
sometimes called ‘electoral authoritarianism’, when quasi-democratic insti-
tutions are used to legitimize, for internal use, and conceal, for external use,
the authoritarian practices of government (Schedler, 2006). Hence the emer-
gence of two distinct types of party systems that are often characterized by
extremely high and extremely low levels of fragmentation, respectively: the
‘inchoate’ party systems of those new democracies that struggle for demo-
cratic consolidation (Mainwaring and Scully, 1995: 22–8), and those party
systems in which, while some opposition exists, the dominance of the largest
party is exercised at scales unthinkable even in the ‘uncommon’ democracies
of the 20th century (Pempel, 1990; Rimanelli, 1999). In order to describe
such systems adequately and to incorporate them within the mainstream of
comparative research on political parties, we need an index that works well
at the extremes.
These are the reasons why I test the new index on a dataset from two
regions that are generally recognized as representative of the two types
identified above. The party systems of East Central Europe and the former
Soviet Union (ECE/FSU) are used as an example of ‘inchoate’ party systems,
primarily because one of their most salient characteristics is high fragmenta-
tion (Lewis, 2000), while the emerging party systems of sub-Saharan Africa
(SSA) are often characterized by the existence of a ‘predominant single
party’ (Manning, 2005: 716).10 In order to make the test even more pointed
to the extremes, I use different bases for computing effective numbers of
parties, votes for ECE/FSU, which allows for registering higher levels of
fragmentation, and seats for SSA, which allows for registering lower levels
of fragmentation. Thus constructed, the set of cases includes 42 national
vote distributions from nine countries of ECE/FSU (1990–2006) and 38
seat distributions in lower or single chambers of national legislatures in the
same number of countries of SSA (2000–7), which makes 80 observations
overall.11
Table 2 gives the average effective numbers of parties, calculated using
three different measures (NLT, NB and the new index, NP) for groups of cases
defined by geography and the sizes of the largest components. The first
184
Sub-Saharan s1>0.7 12 1.56 1.60 1.17–1.95 1.42 1.46 1.13–1.68 1.31 1.33 1.09–1.54
Africa 0.5<s1<0.7 17 2.17 2.11 1.80–3.09 1.89 1.88 1.64–2.52 1.70 1.72 1.49–2.35
185
s1<0.5 9 4.25 3.89 2.74–8.22 3.54 3.13 2.42–6.25 3.46 3.07 2.28–7.55
All SSA cases 38 2.47 2.09 1.17–8.22 2.13 1.87 1.13–6.25 2.12 1.70 1.09–7.55
East Central s1>0.35 13 3.73 3.83 2.80–5.36 3.02 3.05 2.46–4.02 2.99 2.88 2.32–4.14
Europe / 0.3<s1<0.35 12 5.43 5.54 4.18–7.16 4.27 4.30 3.55–5.18 4.65 4.55 4.14–5.67
Former s1<0.3 17 7.16 6.94 5.33–10.06 5.65 5.47 4.35–8.09 6.46 6.40 4.97–10.61
Soviet Union All ECE / 42 5.60 5.54 2.80–10.06 4.44 4.38 2.46–8.09 5.00 4.91 2.32–10.61
FSU cases
inference that can be made from the data is that NP does indeed work better
on highly concentrated party systems. Most people would say that if the
largest party takes more than 70 percent of seats in the assembly, as happens
in as many as 12 countries of sub-Saharan Africa, then we observe a single-
party dominant system. Yet the Laakso–Taagepera index typically registers
one-and-a-half party format, and it does not preclude such systems from
being described as cases of bipartism. The Dunleavy–Boucek index offers
a partial improvement. The new index, however, says what is intuitively
apparent: it counts the largest party as 1 and the rest as about one-third.
For systems with 0.5 < s1 < 0.7, the values of NLT are greater than 2.0 in
the majority of SSA cases, and in one case a three-party system is registered.
The values of NB are more plausible in that they are smaller, yet NP offers
a better solution again: in fact, it attributes the value greater than 2.0 to
only one party constellation within this category, with s1 = 0.51. Turning to
the cases of ECE/FSU, we can observe how the three indices describe more
fragmented constellations. In so far as s1 remains relatively large, the pattern
observed above remains in place, even though the distance between the
values of NB and NP becomes reduced to a minimum due to the presence
of constellations with s1 < 0.51. With s1 within the (0.3–0.35) range, the
rank-ordering of index values changes: now, while the values of the Laakso–
Taagepera index remain greater than those of other indices, the values of
the new index take second place, and the values of NB third place. If s1 <
0.3, the values of NP are not only greater than those of NB; in four cases
they also exceed the values of NLT, even though the mean and median values
of the Laakso–Taagepera index remain greater.
Figure 1 scatterplots the values of two indices, NLT and NP, against the
largest party’s vote-share. NB is omitted because, while making the figure
less readable, the presence of its values would not add much information:
they are always smaller than those of the Laakso–Taagepera index, the
distance slowly increasing as s1 goes down. With the new index, the picture
is more complex. In so far as the size of the largest component exceeds 50
percent, the values of NLT and NP remain spatially separated, with the group
of NP values being located well below the group of NLT values. As the largest
parties’ shares decrease, the two groups mix up, so that the levels of frag-
mentation registered by the two indices become similar. One might be
tempted to suggest that since the values of NP are sometimes even greater
than those of the Laakso–Taagepera index, the new index is excessively
sensitive to the presence of very small parties, which would make a good
argument for using NB instead. To test the validity of this argument, it is
useful to have a closer look at the party constellation that generates the
greatest value of NP, 10.60, which exceeds not only NB (8.09) but also NLT
(9.61). This is the constellation of the 1995 national legislative elections in
Latvia. In these elections, the largest party received 15.2 percent of the
vote, three smaller parties, more than 10 percent of the vote each, and six
still smaller parties, more than 3 percent of the vote each, the remaining 7.4
186
Conclusions
While stating that there was no perfect measure of the effective number of
parties became commonplace in scholarly papers dealing with the matter,
and while the validity of such claims cannot be denied on philosophical
grounds, a possible extension of this philosophy is that there is always room
for improvement. If the effective number of parties, as defined by Laakso and
Taagepera, tends to produce unrealistically high scores for very concentrated
187
party systems, thus failing on intuitive content, why not develop a measure
that is devoid of such a shortcoming? And, even though a partial solution to
this problem is already available, in the form of the Dunleavy–Boucek index,
why not eliminate the problem completely? This is what I have attempted
x
1
to do in this study. The new index, defined as Np = ∑ , solves
1 1 + ( s12/si ) − si
several problems faced by those who need to count the effective number of
parties. First, it satisfies all basic requirements of indices of this kind. Second,
it produces reasonably small scores for party constellations that appear to
have few important parties. Third, it registers many parties in those constel-
lations where there seems to be many parties. Fourth, it minimizes unde-
sirable side effects such as the ‘kink’ effect. This combination of properties
makes the proposed index superior to the earlier ways of counting the effec-
tive number of parties.
For East Central Europe/the former Soviet Union, a large general popu-
lation of country cases was not available. This made me use several election
results from each of the countries. For a dataset based on vote rather than
seat distributions, it was also essential to provide for greater data homogen-
eity, which was achieved by excluding elections held by non-proportional
rules, the non-proportional tiers of mixed electoral systems and whole coun-
tries that held their national legislative elections exclusively or primarily by
non-proportional rules. In those cases when the electoral system, while
being generally proportional, made provisions for independent candidacies,
each of such candidates was counted as an individual party. The 42 cases
included are Bulgaria (1991, 1994, 1997, 2001 and 2005), the Czech
Republic (the Czech National Council in 1990 and 1992, the national legis-
lature in 1996, 1998, 2002 and 2006), Hungary (1990, 1994, 1998, 2002
and 2006), Latvia (1993, 1995, 1998, 2002 and 2006), Lithuania (1992,
1996, 2000 and 2004), Moldova (1994, 1998, 2001 and 2005), Russia
(1993, 1995, 1999 and 2003), Slovakia (the Slovak National Council in
1990 and 1992, the national legislature in 1994, 1998, 2002 and 2006) and
Ukraine (1998, 2002 and 2006). For the majority of elections, the data are
from an Internet source: http://www2.essex.ac.uk/elect/database/.
For the most recent elections, the data are from the Internet sites of the
national electoral commissions:
http://www.2005izbori.org/results/index.html (Bulgaria),
http://www.volby.cz/pls/ps2006/ps2?xjazyk=EN (the Czech Republic),
http://www.valasztas.hu/parval2006/outroot/vdin1/en/l22.htm (Hungary),
http://www.cvk.lv/cgi-bin/wdbcgiw/base/saeima9.GalRezS9.vis (Latvia),
http://www3.lrs.lt/rinkimai/2004/seimas/rezultatai/rez_l_20.htm (Lithuania),
http://www.elections2005.md/results/total/ (Moldova),
http://gd2003.cikrf.ru/ (Russia),
http://www.statistics.sk/nrsr_2006/angl/obvod/results/tab3.jsp (Slovakia),
http://www.cvk.gov.ua/vnd2006/w6p001e.html (Ukraine).
All Internet sites were accessed in April 2007.
Notes
I am grateful to the anonymous referees of the journal for criticism and comment.
All errors of fact and interpretation are entirely my own.
in political science, and a crucial advantage of NLT is that it has already entered
a number of them (Taagepera, 2007). Yet it is the intuitive content of the index
that makes it useful for model-building, not vice versa. For instance, an index
might fit perfectly into a model that predicts bipartism at certain constellations
of societal and institutional influences, but if this index describes situations in
which the largest party takes more than 60 percent of the vote as cases of
bipartism, then there might be something wrong with the model.
5 Most authors who have tried to deal with the effective number of parties axio-
matically have normally pointed only to the first of these properties as the funda-
mental one (Lijphart, 1994: 68). The second requirement seems self-evident.
However, we will see that one index discussed in the literature actually failed to
fulfil this requirement. As for indices violating the third requirement for constella-
tions with more than two parties, many of them are theoretically available, and one
(Molinar, 1991) was proposed as a major alternative to NLT. The credit for making
this requirement axiomatic should be given to Dunleavy and Boucek (2003).
6 It is interesting to note, as an example of the intuitive appeal of such an approach,
that Barnea and Rahat (2007: 384) introduce the expression s2/s1 as an ad hoc
competitiveness index.
7 The idiosyncratic dynamics of NLT between constellations D and E is also notice-
able.
8 It is also worth mentioning that unlike NLT, NP cannot be meaningfully estab-
lished for constellations involving zero-size components. Such constellations,
however, are of mere theoretical interest in political science research.
9 See, for instance, Coppedge (1997) and Taagepera and Shugart (1989: 81–3).
10 Of course, it would be too far-fetched to claim that any very concentrated party
system is necessarily associated with ‘electoral authoritarianism’. For different
views on African party systems, see Bogaards (2004) and Mozaffar et al. (2003).
11 See the Appendix for more details of case selection and for the description of
data sources.
References
Amorim Neto, Octavio and Gary W. Cox (1997) ‘Electoral Institutions, Cleavage
Structures, and the Number of Parties’, American Journal of Political Science 41:
149–74.
Barnea, Shlomit and Gideon Rahat (2007) ‘Reforming Candidate Selection Methods:
A Three-level Approach’, Party Politics 13: 375–94.
Bogaards, Matthijs (2004) ‘Counting Parties and Identifying Dominant Party Systems
in Africa’, European Journal of Political Research 43: 173–97.
Coppedge, Michael (1997) ‘District Magnitude, Economic Performance, and Party-
System Fragmentation in Five Latin American Countries’, Comparative Political
Studies 30: 156–85.
Cox, Gary W. (1997) Making Votes Count: Strategic Coordination in the World’s
Electoral Systems. Cambridge: Cambridge University Press.
Dumont, Patrick and Jean-François Caulier (2003) ‘The “Effective Number of Rele-
vant Parties”: How Voting Power Improves Laakso–Taagepera’s Index’, Working
Paper 2003/7. Brussels: Center for Research in Economics (CEREC), Facultés Uni-
versitaires Saint-Louis.
190
Dunleavy, Patrick and Françoise Boucek (2003) ‘Constructing the Number of Parties’,
Party Politics 9: 291–315.
Feld, Scott L. and Bernard Grofman (2007) ‘The Laakso–Taagepera Index in a Mean
and Variance Framework’, Journal of Theoretical Politics 19: 101–6.
Laakso, Markku and Rein Taagepera (1979) ‘“Effective” Number of Parties: A
Measure with Application to West Europe’, Comparative Political Studies 12: 3–27.
Lewis, Paul G. (2000) Political Parties in Post-Communist Eastern Europe. London:
Routledge.
Lijphart, Arend (1984) Democracies: Patterns of Majoritarian and Consensus
Government in Twenty-One Countries. New Haven, CT: Yale University Press.
Lijphart, Arend (1994) Electoral Systems and Party Systems: A Study of Twenty-
Seven Democracies. Oxford: Oxford University Press.
Mainwaring, Scott (1999) Rethinking Party Systems in the Third Wave of Democ-
ratization: The Case of Brazil. Stanford, CA: Stanford University Press.
Mainwaring, Scott and Timothy R. Scully (1995) ‘Introduction: Party Systems in
Latin America’, in Scott Mainwaring and Timothy R. Scully (eds) Building Demo-
cratic Institutions: Party Systems in Latin America. Stanford, CA: Stanford Uni-
versity Press.
Manning, Carrie (2005) ‘Assessing African Party Systems after the Third Wave’,
Party Politics 11: 707–27.
Molinar, Juan (1991) ‘Counting the Number of Parties: An Alternative Index’,
American Political Science Review 85: 1383–91.
Mozaffar, Shaheen, James R. Scaritt and Glen Galaich (2003) ‘Electoral Institutions,
Ethno-political Cleavages, and Party Systems in Africa’s Emerging Democracies’,
American Political Science Review 97: 379–90.
Niemi, Richard G. and John Fuh-sheng Hsieh (2002) ‘Counting Candidates: An
Alternative to the Effective N (with an Application to the M + 1 Rule in Japan)’,
Party Politics 8: 75–99.
Ordeshook, Peter C. and Olga V. Shvetsova (1994) ‘Ethnic Heterogeneity, District
Magnitude, and the Number of Parties’, American Journal of Political Science 38:
100–23.
Pempel, T. J. (ed.) (1990) Uncommon Democracies: The One-Party Dominant
Regimes. Ithaca, NY: Cornell University Press.
Rae, Douglas W. (1967) The Political Consequences of Electoral Laws. New Haven,
CT: Yale University Press.
Rimanelli, Marco (ed.) (1999) Comparative Democratization and Peaceful Change
in Single-Party-Dominant Countries. New York: St. Martin’s Press.
Schedler, Andreas (ed.) (2006) Electoral Authoritarianism: The Dynamics of Unfree
Competition. Boulder, CO and London: Lynne Rienner.
Taagepera, Rein (1999) ‘Supplementing the Effective Number of Parties’, Electoral
Studies 18: 497–504.
Taagepera, Rein (2002) ‘Limiting Frames of Political Games: Logical Quantitative
Models of Size, Growth and Distribution’, paper 02’04. University of California,
Irvine: Center for the Study of Democracy.
Taagepera, Rein (2005) ‘Conservation of Balance in the Size of Parties’, Party
Politics 11: 283–98.
Taagepera, Rein (2007) Predicting Party Sizes: The Logic of Simple Electoral Systems.
Oxford: Oxford University Press.
191
Taagepera, Rein and Bernard Grofman (1981) ‘Effective Size and Number of Com-
ponents’, Sociological Methods and Research 10: 63–81.
Taagepera, Rein and James L. Ray (1977) ‘A Generalized Index of Concentration’,
Sociological Methods and Research 5: 367–84.
Taagepera, Rein and Matthew S. Shugart (1989) Seats and Votes: The Effects and
Determinants of Electoral Systems. New Haven, CT: Yale University Press.
Ware, Alan (1995) Political Parties and Party Systems. Oxford: Oxford University
Press.
Warren-Boulton, Frederick R. (1990) ‘Implications of US Experience with Horizon-
tal Mergers and Takeovers for Canadian Competition Policy’, in Frank Mathewson,
Michael Trebilcock and Michael Walker (eds) The Law and Economics of Compe-
tition Policy. Vancouver, BC: The Fraser Institute.
192