. 3
( 35 .)


Vanuatu: “We have to stop the disorganization caused by too many political
parties.” (Saribo 2003)
The degree of nationalization is also related to the kinds of electoral strategies can-
didates and parties employ as well as the types of appeals to which voters respond
(Schattschneider 1960; Hicken 2002; Jones and Mainwaring 2003).
Introduction 9

consolidation in divided societies is more likely where parties compete
for nationwide votes as opposed to votes from a narrow group or region
(Horowitz 1985, 1991; Diamond 1988; Reynolds 1999; Reilly 2001).
On the other hand, if the priority is a party system that preserves and
protects the preferences of small, subnational constituencies (e.g.,
regions or geographically concentrated ethnic/religious groups), then
less nationalization is better. From this perspective, nationalized party
systems are more likely to under-represent the interests of potentially
powerful subnational groups leading to diminished democratic
responsiveness (Lijphart 1977; Powell 2000).

1.4 arguments of the book: a question
of coordination
What explains the type of party systems that emerge in democracies? To
answer this question, it is useful to think of the party system as the
outcome of various coordination opportunities. Voters and candidates
may successfully coordinate on a small number of parties, or such
coordination may fail, leading to a proliferation of parties. Candidates
may choose to coordinate across districts to form large national parties,
or they may eschew such cross-district coordination.
In the chapters that follow, I explore the coordination successes
and failures at the heart of democratic party systems. Speci¬cally, I
examine factors that encourage and discourage greater coordination
between voters, candidates, and parties. I argue that coordination
incentives are often not conducive to a party system with a modest
number of national parties “ especially in developing democracies. In the
case study analyses, I discuss how various historical and societal factors
helped shape the development of the party system. However, the focus of
the argument is on the role of institutional factors. Rules and institutions
such as the electoral system, the manner of selecting the chief executive,
and the distribution of power between different branches of government
have profound and predictable impacts on the development of parties
and party systems.
The features of the party system of interest in this study “ the number
of parties and nationalization “ are the product of coordination (or
coordination failures) among voters, candidates, and party leaders
within electoral districts (intra-district coordination) and across districts
Building Party Systems in Developing Democracies

(inter-district coordination or aggregation). I argue that coordination
failures at the district level are not the primary explanation for why party
systems in many democracies often diverge from expectations. Though
there are certainly exceptions (see Backer and Kollman 2003; Chhibber
and Kollman 2004), voters, candidates, and parties can and do coordi-
nate locally in response to electoral incentives, even in new democracies
(Clark and Golder 2006 and Chapters 4 and 6 in this study). Cross-
district coordination, however, is another matter. Cross-district coordi-
nation requires that political elites from many localities cooperate by, for
example, forging an electoral alliance (a party) and compromising on a
set of policy goals and priorities. It is these cross-district attempts at
coordination or aggregation that often fail, particularly in developing
democracies, with two consequences. First, where aggregation failures
regularly occur, national parties do not develop. Second, poor aggrega-
tion is associated with the in¬‚ation of the number of political parties.
By way of a brief illustration, consider the relationship between
coordination (within and across districts) and the size of the party sys-
tem. (I will discuss this in more detail later in Chapter 2). The bulk of the
existing work on the determinants of the number of parties focuses on the
electoral system “ speci¬cally the electoral formula (e.g., plurality versus
proportional representation) and the number of seats open for compe-
tition (district magnitude) (Duverger 1954; Taagepera and Shugart
1989; Cox 1997). Cox™s M þ 1 rule is useful as a generalized statement of
the relationship between the electoral system and the number of political
parties. M equals the number of seats in a district, and the M þ 1 rule
predicts that no more than M þ 1 candidates or parties are viable in any
single seat districts (M ¼ 1), and that no more than M þ 1 parties are
viable in multiseat districts (M > 1). In other words, as the number of
seats in a district increases, we expect more parties, ceteris paribus.11
Often lost in discussions of electoral systems is the fact that although
these institutions allow predictions about the number of parties in each
individual electoral district, they do not enable one to anticipate the
number of parties that will arise nationally. Electoral rules directly
affect the nature of coordination within electoral districts. Why

The M þ 1 rule is an upper limit on the number of political parties. Whether a party

system is at or below the M þ 1 threshold is a function of the degree of social het-
erogeneity (Amorim-Neto and Cox 1997; Clark and Golder 2006).
Introduction 11

candidates might (fail to) coordinate across districts is a separate, but
equally important question. Despite observations by numerous scho-
lars that the size of the national party system can diverge quite sharply
from what we observe in individual districts (see, for example, Riker
1982; Sartori 1986; Kim and Ohn 1992), the issue of cross-district
coordination or aggregation has received relatively little attention.12
Yet, understanding aggregation is crucial if we are to explain why party
systems look as they do.
Imagine two countries, A and B in Figure 1.1. The result of within-
district coordination in each country is numerous district-level party
systems each with its own effective number of parties. In this case, the
electoral system induces coordination on two parties within each elec-
toral district. Thus the average effective number of parties locally
(ENPavg) is 2 in both country A and B.13 How, though, do these
numerous local party systems map onto the national party system? Does
a unique set of parties run in each district or are the same few parties the
frontrunners in most districts? Countries A and B differ in the degree of
aggregation “ that is, the extent to which candidates coordinate across
districts under a common party label. In country A, each electoral district
contains a different set of parties “ Yellow and Green in district 1, Purple
and Blue in district 2, and Red and Orange in district 3. At the other
extreme, the same two parties “ Yellow and Green “ are the frontrunners
in all the districts in country B. The difference in the level of aggregation
between the two countries has profound implications for the national
party system. In country A, the effective number of parties nationally
(ENPnat) is 6 “ much larger than the average effective number of parties at
the district level (ENPavg). By contrast, the national effective number of
parties in country B equals 2 “ re¬‚ecting exactly what we see in each of
the districts.
Now consider a third country, country C. Here a permissive electoral
system allows for a large number of parties in each district (ENPavg ¼ 6).
However, coordination across districts is extensive such that when we

Riker (1982) notes that single-member district plurality systems may not generate
two national parties when third parties nationally are continually one of two parties
locally. Sartori (1986) and Kim and Ohn (1992) also argue that single-member dis-
trict plurality systems will not lead to a two-party system if the electorate is comprised
of geographically concentrated minorities.
Assuming the two parties split the vote equally.
Country A Country B Country C


1 2 3 1 2 3 1 2 3
Party PB YG YG
¬gure 1.1. Comparing Aggregation
Introduction 13

aggregate to the national level, the national party system mirrors the
local party systems in both size and composition (ENPnat ¼ 6).
As these examples make clear, a country™s party system may be large
or in¬‚ated due to (1) a large number of parties winning seats in each
district (country C), (2) poor aggregation across districts (country A), or
(3) a combination of 1 and 2. Looking at the national effective number
of parties in isolation tells us nothing about whether party system
in¬‚ation is due to coordination failures within districts, across districts,
or both. Instead, it is necessary to compare the size of the national party
system to the local party system in order to separate out aggregation
from district level processes.
Table 1.1 displays the information on the size of the local party systems
(ENPavg) and national party systems (ENPnat) in 16 countries from around
the world.14 The difference between the effective number of parties
nationally (ENPnat) and the average effective number of parties in the
district (ENPavg) is a measure of the extent of aggregation across districts
(D ¼ ENPnat À ENPavg). Higher differences re¬‚ect worse aggregation.
One useful way to capture the extent of aggregation is to convert this
difference into a percentage measure of how much larger the national
party system is than the average district-level party system. This party
system in¬‚ation measure (I) is computed by dividing the difference
between ENPnat and ENPavg (D) by ENPnat and then multiplying by 100
(Cox 1999:17). I ¼ 100(ENPnat À ENPavg)/ENPnat. The resulting in¬‚ation
score tells us what portion of the size of the national party system is due to
poor aggregation, and what percentage re¬‚ects the extent of coordination
within districts. Based on this calculation, if I is10, then 10% of the size of
the national party system can be attributed to different parties garnering
votes in different parts of the country (poor aggregation), with the other
90% ascribable to the average number of parties at the district level. The
larger the in¬‚ation score, the poorer the aggregation. Note that in Table
1.1, for the countries with a large number of parties, poor aggregation is
responsible for one-third or more of party system in¬‚ation.15 In short, in

These 16 countries were chosen from a sample of 46 countries used in Chapter 3. The
list of all 46 countries and their in¬‚ation scores is available in Chapter 3.
In addition, electoral rules and institutions appear to be having the expected effect on
the coordination within districts. In almost all of the countries listed in Table 1.1,
ENPavg is within the range we would expect given the average number of seats in each
district. The only exceptions are some of the single“member-district cases (e.g., India,
Building Party Systems in Developing Democracies

table 1.1. The Size of the Local and National Party System Compared

Country ENPavg ENPnat In¬‚ation
Ecuador (1979“88) 3.5 8.3 58
Thailand (1986“2001) 3.1 6.6 53
Belgium (1971“99) 4.1 8.0 49
India (1971“99) 2.4 4.7 49
Switzerland (1971“95) 4.0 6.4 38
Philippines (1992“98) 2.3 3.6 36
Brazil (1986“2002) 5.4 7.8 31
South Korea (1988“2000) 2.8 3.9 28
Germany (1972“98) 2.5 3.2 22
Canada (1972“2000) 2.6 3.3 21
Argentina (1983“99) 2.7 3.2 16
United Kingdom (1970“97) 2.5 3.0 17
Botswana (1994“99) 2.1 2.4 13
United States (1972“2000) 1.8 2.0 10
Venezuela (1973“83) 2.9 3.1 6
Denmark (1971“98) 5.1 5.2 2
Notes: ENPavg is the average effective number of parties in each district. ENPnat is the
effective number of parties nationally (as measured by a party™s vote share). Both terms
are averaged across all elections within the speci¬ed time period.
Source: Author™s calculation.

many countries, the lack of coordination across districts accounts for a
substantial share of the party system™s size, in some cases the lion™s share.
Understanding the role aggregation plays in shaping a country™s party
system is crucial not just because of its effect on the number of parties. The
degree of aggregation also communicates important information about
the nature of parties in a given party system. Where aggregation is good,
parties will tend to have larger, broader, more national constituencies
than where aggregation is poor. The fact that aggregation is extremely
poor in a place like Thailand, for example, supplies important clues about
the interests and orientation of Thai parties (and their members).
Given the role aggregation plays in shaping the party system,
ignoring aggregation carries considerable theoretical and practical
risks. The failure to take aggregation into account might lead one to
misinterpret the results of hypothesis tests (for example, concluding

the Philippines, and Canada) where there is slightly more than the expected two
parties. For more on violations of Duverger™s law and the M þ 1 rule in single-seat
districts, see Backer and Kollman (2003).
Introduction 15

erroneously that electoral rules failed to produce the predicted
number of parties). The neglect of aggregation might also lead to a
misdiagnosis of the causes of an in¬‚ated party system and the subse-
quent prescription of an ineffective or inappropriate remedy.

Explaining Aggregation

Thanks to the rich literature in comparative electoral studies, we know a
good deal about what shapes intra-district coordination (Duverger 1954;
Taagepera and Shugart 1989; Lijphart 1994; Cox 1997, 1999; Clark and
Golder 2006). We know much less about the factors that in¬‚uence cross-
district coordination or aggregation. A key contribution of this book is
the development and testing of a theory of aggregation incentives.
Even though there is a lack of theorizing about aggregation, some
students of comparative elections have acknowledged its role in shap-
ing the national party system. Maurice Duverger, for example, con-
sidered the question of how local two-party systems become a national
two party system, stating:

[T]he increased centralization of organization within the parties and the
consequent tendency to see political problems from the wider, national
standpoint tend of themselves to project on to the entire country the localized
two-party system brought about by the ballot procedure. (1954, 228)


. 3
( 35 .)