. 7
( 35 .)


These are the United States, India, Canada, and the United Kingdom.
A Theory of Aggregation Incentives 35

level, these actors have an incentive to coordinate across districts
nationally in a bid to capture that prize. By contrast, where subnational
governments control substantial resources, candidates and party leaders
may focus their attention on winning these prizes and eschew cross-
district coordination on a national scale.12 Ultimately, however, vertical
centralization is not suf¬cient to produce strong aggregation incentives.
The other component of the centralization equation is also important.

Horizontal Centralization
Aggregation incentives are greatest where political power is concen-
trated not only vertically but horizontally (within the national govern-
ment) as well. To assess the degree to which power is horizontally
centralized, we want to consider the extent to which the largest legisla-
tive party controls all the reins of government. Where power and control
of resources are concentrated horizontally, a party that successfully
coordinates can expect to face relatively few checks on its exercise of
national power. However, where power is dispersed at the national level,
the largest legislative party may still ¬nd it is unable to fully control or
direct the resources of government. The greater is the degree of hori-
zontal centralization, the stronger are the incentives for actors to coor-
dinate across districts to try to win control of government.
A variety of factors affect the degree to which power is horizontally
centralized. Here I focus on three: (a) the separation of power “ specif-
ically the presence of a second legislative chamber; (b) the degree of party
cohesion; and (c) the existence of reserve domains. Note that the number
of parties or partisan veto players is missing from this list. At ¬rst blush
this may seem an oversight “ the addition of another partisan veto player
other than the largest party would certainly represent a potential check
on that party™s authority. However, the distinction between partisan veto
players and veto gates is crucial in this context. Veto gates are the points
in the policy process where approval must be granted in order to change
the status quo (cf. Tsebelis™s institutional veto points “ Tsebelis 1995).
Veto players are those actors, single or collective, who sit at each veto
gate. I assume that, for most elections, candidates, party leaders, and
voters treat the number of veto gates at the national level as exogenous.

With reference to the model, a high degree of vertical decentralization raises the
expected utility of remaining a small, subnational party, ceteris paribus.
Building Party Systems in Developing Democracies

Ultimately we know that institutions are at least partially endogenous to
the con¬gurations of power at their inception but most elections do not
occur during founding institutional moments and so it seems reasonable
to assume that actors generally take the formal, institutional distribution
of veto authority as given. The same assumption cannot be made about
the number of partisan veto players. The number of partisan veto players
is partially endogenous to aggregation. Control of government by a
single elected veto player (e.g., a majority party) is itself likely evidence of
successful cross-district coordination. Likewise, multiple partisan veto
players (as in coalition governments) may re¬‚ect poor aggregation across
districts.13 It is inaccurate, therefore, to treat the number of veto players
as an exogenous explanatory variable.

Separation of Power “ Bicameralism
Where there is a separation of power and authority at the national level
(e.g., where power is spread among multiple veto gates), there is always a
risk that the largest legislative party will have to share control of gov-
ernment with other actors. Given this risk, aggregation incentives should
be negatively correlated with the number of veto gates within the
national government, ceteris paribus. The diffusion of power can arise
from a variety of institutional sources, including the division of the leg-
islature into two chambers. Where there is an upper chamber with veto
authority, the party with a lower chamber majority may still not control
all the points of power. The upper chamber is most likely to represent a
check on the largest party in the lower chamber where district bound-
aries for the lower and upper chamber are not the same, where upper and
lower chamber elections are not concurrent, and where upper chamber
support is required to pass new legislation.14
Diffusion of power may also result from the separation of executive
and legislative power as occurs in presidential systems. However, the
effects of presidentialism are conditional on the proximity of presidential
and legislative elections, which is discussed later in more detail.15

It may also re¬‚ect a permissive electoral system.
For more on bicameralism and its effects, see Diermeier and Myerson (1999); Heller
(2001); Druckman and Thies (2002); Rogers (2003); Tsebelis and Money (1997); and
an excellent review by Cutrone and McCarty (n.d).
The rules that govern interactions between different veto gates and veto players (e.g.,
agenda-setting powers, reversion points, veto powers) can also shape the degree of
A Theory of Aggregation Incentives 37

Party Cohesion
Party cohesion can also affect the degree of power concentration within
the national government. Where parties are not cohesive (e.g., when
parties are factionalized), the payoff to being the largest party in gov-
ernment is reduced. When the party is cohesive, it is, in effect, a hier-
archical and unitary actor. Party leaders can be reasonably con¬dent of
the support and cooperation from the party™s other members. However,
when the party is more like a coalition of factions (Laver and Scho¬eld
1990) than a cohesive political hierarchy, it ceases to be a unitary actor,
and each party faction becomes a potential veto player. Thus even if the
party controls all the reins of government, the power (and payoff)
associated with control of government will have to be shared between
various subparty factions. The head of a majority party might still ¬nd
his or her power checked by rival internal party factions. All else equal
then, party factionalism should discourage attempts at aggregation by
would-be party leaders/chief executives.
I am assuming here that party cohesion is exogenous to aggregation.
For example, party factionalism has been linked to the nature of the
electoral system, speci¬cally the presence of intra-party competition
(Katz 1986; Shugart and Carey 1992; Lijphart 1994; Reed 1994; Hicken
2002). Rules that govern party leaders™ ability to control backbenchers
(and are exogenous to the parties themselves) such as rules about party
switching or campaign ¬nance regulations can also affect party cohesion.
In short, to the extent that such systemic factors exist, it seems reasonable
to expect that actors will treat party factionalism as exogenous “ at least
in the short term. However, I acknowledge that factionalism may be at
least partially endogenous to aggregation. Stronger incentives to coor-
dinate across districts may induce smaller groups/parties to ally under the

horizontal centralization and hence aggregation incentives, particularly the power of
the chief executive relative to other actors at the national level. I do not explore the
role presidential power plays in shaping aggregation incentives in this book for both
theoretical and practical reasons. First, I expect that the effect of presidential powers
is secondary to the in¬‚uence presidentialism exerts through the timing of elections
and the number of presidential candidates. Second, reliable comparable data on
presidential powers are dif¬cult to come by, though efforts are underway to try to
collect such data by a number of scholars (e.g., Tsebelis and Aleman 2005; Tsebelis
and Rizova 2007). As data become available, it is worth looking more closely the role
presidential power plays in shaping aggregation incentives. See Hicken and Stoll
(2006) for some initial work along these lines.
Building Party Systems in Developing Democracies

banner of a larger party. The net effect of this may be an increase in intra-
party factionalism. I attempt to control for this potential endogeneity in
the large-N tests in Chapter 3 by using the features of the electoral system
as my proxy for party cohesion rather than a more direct measure.

Reserve Domains
Finally, in transitioning democracies, it is useful to keep in mind the
possible effects of reserve domains “ institutional or policy domains
controlled by actors who are not directly accountable to elected of¬cials
(Valenzuela 1992). These might include appointed legislative seats (or
chambers)16 as well as cabinet positions or policy areas that are widely
considered off limits to elected of¬cials (e.g., control over military
appointments and budgets). For example, in Thailand an appointed
Senate was traditionally stacked with representatives of Thailand™s
military and bureaucracy “ conservative forces often at odds with
Thailand™s elected representatives. The appointed Senate was not
replaced with an elected body until 1997. In newly democratic Indonesia,
large portions of the military™s operations and budget are still considered,
by custom and necessity, off limits to the country™s elected leaders.
Pinochet™s 1980 Chilean constitution famously provided for nine non-
elected senators (out of a total of 26) including eventually Senator-for-life
Pinochet. It also denied the president the power to remove the com-
mander in chief of the armed forces and chief of police. Though Pinochet
lost a plebiscite and stepped down as president in 1988, these provisions
remained in place until 2005. In short, by guaranteeing that even
majority parties will face checks on their authority, the existence of
reserve domains effectively reduces the payoff to being the largest party
in government, and hence undermines aggregation incentives.
Reserve domains operate in concert with party cohesion and
bicameralism to shape the size of the aggregation payoff. This payoff
should be highest where party cohesion is high, reserve domains are
absent, and the legislature is unicameral. The aggregation payoff
should be smallest where reserve domains exist along side a bicameral
legislature populated by factionalized political parties.
It is possible to imagine a counterargument to the theory as I have
described it thus far “ one that turns my hypothesis about the link

Conditional, of course, on the speci¬c appointment and removal procedures.
A Theory of Aggregation Incentives 39

between horizontal centralization and aggregation on its head. Rather
than horizontal decentralization serving as a deterrent to coordination
attempts, one might argue that horizontal decentralization should spur
greater aggregation as parties endeavor to capture all of the relevant
veto gates and obtain enough power to eliminate any lingering reserve
domains. Even though this alternative hypothesis still does not address
the issue of party cohesion, it would serve as falsifying evidence for this
portion of the theory should I ¬nd that horizontal decentralization is
associated with better rather than worse aggregation.
To summarize, I argue a chain of variables together determine the
payoff to being the largest party in the legislature “ the larger the payoff,
the stronger the aggregation incentives. This payoff is greatest where the
national government dominates subnational units (vertical centraliza-
tion) and where power is concentrated within the national government
(horizontal centralization). If there is a high degree of decentralization on
either dimension, then the payoff to being the largest party in government
is sharply reduced. For example, even if the largest national party could
rule unchecked at the national level, national of¬ce would not be worth
organizing for if the power and resources were really controlled by sub-
national units. Likewise, a national government with great power means
little if other actors regularly block the largest party from exercising that

2.2.2 Probability of Capturing the Payoff

Even though a suf¬ciently large payoff is necessary to induce aggrega-
tion, it is not suf¬cient. There must also be a reasonable probability that
the largest party will be able to capture that prize. Of chief concern is
whether or not the largest legislative party is also able to win control of
the executive of¬ce (Cox 1997). If the largest legislative party has no
chance of capturing the reins of government, then even a potentially large
payoff will not be enough to induce national coordination. In contrast,
where the largest legislative party also captures executive power with a
high degree of probability, the expected utility of coordinating to form a
large party will be greater.
The relationship between legislative and executive power varies
greatly from country to country, but the most fundamental distinction
is between presidential and parliamentary regimes. In this section, I
Building Party Systems in Developing Democracies

discuss the way in which the probability of capturing executive of¬ce
operates in each.

Presidential Systems
In presidential systems, the probability that the largest legislative party
will also win control of the chief executive of¬ce is a function of three
variables: (a) whether or not presidential and legislative elections are
held concurrently; (b) whether or not voters cast a single, fused vote for
the executive and legislature;17 and (c) the number of viable presidential
candidates.18 Because fused votes are relatively rare, I will focus on the
remaining two variables.19
Where presidential and legislative elections are concurrent, the elec-
toral stakes are magni¬ed, and it is likely that the same parties will be the
frontrunner in both presidential and legislative contests (Shugart 1995).
In effect, when elections are concurrent, the presidential contest casts a
shadow over legislative contests in the eyes of both voters and candidates
(Shugart 1995; Cox 1997). The issues and parties that are in contention
in the nationwide presidential race tend to migrate down the ballot and
in¬‚uence voter choices for legislative elections. In effect, voters use the
presidential campaign as an information shortcut to help guide their
choice of legislative candidates (Samuels 2003; Golder 2006). Candi-
dates, for their part, recognize that voters will rely on the presidential
contest as a cue in concurrent elections, and so face strong incentives to
try to coordinate their campaigns with one of the front-running presi-
dential candidates. There are also economies of scale to be had from such
coordination (Samuels 2002; Golder 2006). Presidential campaigns
typically involve strong, national campaign organizations and command
the bulk of media and donor attention. By coordinating their campaign

Under a fused vote, voters cast a single vote in the legislative contest, which is also
counted as a vote for that party™s presidential candidate. When fused votes occur, the
probability that the largest party will capture both the legislature and the executive is
virtually 100 percent. See Jones (2000) for more information on the fused vote.
See Cox (1997, 190). My list differs from Cox™s in one respect. He uses the “strength
of the presidential election procedure” as his third variable; I instead use the number
of viable presidential candidates.
For work on the link between presidential elections and fewer legislative parties, see
Shugart and Carey (1992), Jones (1994, 1999), Shugart (1995), Cox (1997), Mozzaffar,
Scarritt, and Galaich (2003), Hicken (2005), and Golder (2006). For an opposing
view see Filippov, Ordeshook, and Shvetsova (1999) and Coppedge (2002).
A Theory of Aggregation Incentives 41

with the campaign of a leading presidential candidate (usually from the
same party), legislative candidates expect to enhance their own electoral
This rallying around the leading presidential contender occurs
both within districts, leading to a smaller effective number of candi-
dates/parties in each districts, and across districts, leading to better
aggregation and fewer national political parties. In short, the more


. 7
( 35 .)