. 30
( 35 .)


Building Party Systems in Developing Democracies

Effective Number of Parties/Candidates


4 Inflation

Inflation Score
2 ENPseat


1992 1998 2004
¬gure 6.1. Post-Marcos Concurrent Elections Compared

Philippines has experienced a counter-concurrency effect “ midterm
elections for the House and Senate produce fewer national parties than
do concurrent elections (see Table 6.6). This is understandable in the
Philippine context. The payoff to belonging to the president™s party is
large in both concurrent and mid-term elections, due to the power of the
president and the resources she controls. However, the level of uncer-
tainty varies greatly between concurrent and mid-term elections. In
concurrent elections, the large number of viable presidential candidates
since Marcos reduces the probability that the largest legislative party will
also be the party of the president, undermining aggregation. However,
the rules allow politicians to switch parties at virtually any time without
penalty. As I showed previously, large numbers of politicians at all levels
of government take advantage of these rules to switch to the president™s
party after elections. In other words, uncertainty regarding the presi-
dential frontrunners means that aggregation is both a pre- and post-
electoral phenomenon.
The situation is very different in mid-term elections. There is no
uncertainty about who the president will be in mid-term elections, thus
increasing the probability that the president™s party will also be the
largest legislative party. As a result, aggregation improves in mid-term
elections with a corresponding decrease in the number of national
Philippines: Term Limits, Aggregation Incentives, No. of Parties 173

table 6.6. Counter-Concurrency in Post-Marcos Philippine

Presidential Elections Mid-term Elections
(1992, 1998, 2004a) (1995, 2001a)
ENPnat 2.6
In¬‚ation 25
ENPseat 2.6 2.3
ENP 3.0 2.0
In 2001 and 2004, formal national electoral alliances (the PPC in 2001 and K-4 in
2004) are counted as a single party.
2004 election excluded due to lack of data.
Sources: Author™s calculations from COMELEC (various years); Hartmann et al.
(2001); Tehankee (2002).

parties. Table 6.6 compares aggregation and the number of parties in
the three post-Marcos concurrent elections (1992, 1998, 2004) and two
midterm elections (1995, 2001).28 (Once again, the lack of party vote
share data requires me to drop the 2001 and 2004 elections from the
calculations of the in¬‚ation score and ENPnat.) Looking ¬rst at aggre-
gation, one can see that in¬‚ation is higher during presidential elections
years (I ¼ 36) than in mid-term elections (I ¼ 25). The effective number
of electoral parties correspondingly rises in presidential election years
(ENPnat ¼ 4.0) and falls during mid-term elections (ENPnat ¼ 2.6). Of
course it is dif¬cult to draw strong inferences from only the 1998 mid-
term election, as one is forced to do if we rely solely on the in¬‚ation
score and ENPnat. However, an analysis of the effective number of
legislative parties (ENPseat) in all elections reveals a similar pattern “
more parties in presidential elections than in mid-term elections.
Likewise, Senate elections since 1986 display a similar counter-
concurrency pattern.29 During the presidential election years, the
average effective number of parties in the Senate is 3 compared to 2 in
mid-term elections. Table 6.6 summarizes the counter-concurrency
pattern for both the House and Senate.

The ¬rst post-Marcos legislative election in 1987 is excluded due to the lack of
reliable information about the distribution of vote and seat shares.
For the Senate, the challenge is not inter-district coordination (aggregation) since all
Senators are elected from a single nationwide district.
Building Party Systems in Developing Democracies

The counter-concurrency effect described previously is also useful for
addressing an as yet undiscussed alternative explanation for the post-
Marcos party system “ namely, that the in¬‚ation of the party system
post-Marcos simply re¬‚ects the uncertainty among voters, candidates,
and parties that is often present in early elections in new democracies.
While this argument cannot be dismissed out of hand, there are reasons to
think that the shock of a new system is not a suf¬cient explanation. First,
the Philippines had a long history of democratic elections prior to the
imposition of martial law “ 26 years as an independent state and nearly
40 years as a colony of the United States compared to 14 years of martial
law. Thus democratic elections were not novel. Much of the electorate
and most of the political elite had participated in the democratic elections
prior to martial law. Second, the electoral system the country adopted
after the fall of Marcos was familiar. It was the same basic system that
had been used throughout the Philippines™ electoral history (single-
member district plurality for the House and plurality rule for the presi-
dency). Finally, the counter-concurrency effect suggests that even if
actors are learning (leading presumably to fewer parties over time), they
are also responding to aggregation incentives. When those incentives are
the strongest (i.e., in midterm elections), actors coordinate on fewer
parties than when aggregation incentives are relatively weaker (i.e., in
presidential election years).
To summarize, the switch to a single term limit in 1987 affected
aggregation in a manner consistent with the theoretical expectations “
namely, the effective number of presidential candidates increased
dramatically from what had been the norm prior to martial law. More
viable presidential candidates made it dif¬cult for legislative candidates
and voters to clearly distinguish the frontrunners from the also-rans
and increased the probability that the largest legislative party would
not control the executive branch. This undermined the incentive to
aggregate across districts, leading to a larger number of more localized/
regionalized parties. Finally, the large number of viable presidential
candidates offset the effect of concurrent elections on the number of
parties. A partial exception to this pattern is the 2004 election where
the presence of the an incumbent in the race worked to lower the
effective number of candidates, increase aggregation incentives, and
lower the number of parties to a level closer to the norm in the pre“
martial law period.
Philippines: Term Limits, Aggregation Incentives, No. of Parties 175

6.4 comparing aggregation the number
of parties in the philippines and thailand
Returning to the comparison with Thailand that opened this chapter, we
see that in many respects the Thai and Philippine party systems are very
similar. In both countries, parties tend to be temporary electoral alliances
of locally focused politicians. These parties elicit low levels of discipline
and cohesion from their members and little loyalty from voters. The two
party systems do differ in one important respect, however. Until recently
Thailand has been home to many more political parties than the
Philippines. What explains this divergence when other aspects of the two-
party systems look so alike? One place to begin searching for an expla-
nation is the electoral system. To the extent their electoral systems differ,
we would expect the size of their local/district party systems to diverge.
And, as local party systems are the building blocks of the national party
system, more parties locally would mean more parties nationally, ceteris
paribus. So, how do the Thai and Filipino electoral systems differ? Recall
that Thailand, prior to the 1997 reforms, used the block vote with district
magnitudes of 2 or 3 to elect the House of Representatives. The Philippines,
by contrast, has primarily relied on a single member district plurality
(SMDP) system throughout its democratic history. Given the difference in
district magnitude one would expect the average effective number of
parties (ENPavg) to be slightly lower in the Philippines. In fact, this is the
case. ENPavg for all postindependence House elections in the Philippines is
2.1 compared to 3.2 in pre-1997 Thailand (see Table 6.7). This difference
holds even if we separate pre“ and post“martial law elections in the
Philippines “ the average number of district parties post-Marcos is 2.3
versus 2.0 pre“martial law. However, once Thailand switches to a system
where 80% of House seats are ¬lled using SMDP, Thailand™s ENPavg falls
to the level we observe in the Philippines. In short, some of the differences in
the Thai and Philippine party systems can indeed be traced to the different
district electoral systems in each country prior to 1997.
Note, however, that the difference between the number of parties in
each district is quite modest “ on average there are only about 1.2 more
parties in each district in pre-reform Thailand compared to the average in
all Philippine elections.30 This is in sharp contrast to the national level

If we focus on just post-Marcos elections, the difference is an even smaller 0.9.
Building Party Systems in Developing Democracies

table 6.7. Aggregation in the Philippines and Thailand

Average Average
Country (election year) ENPavg ENPnat Average I
Philippines (1946“69) 2.0 2.3 9.8
Thailand (1986“1996) 3.2 7.2 54
Philippines (1992“98) 2.3 3.6 32
Thailand (2001“2005) 2.4 3.1 23
Philippines (1946“69; 1992“98) 2.1 2.6 17
Thailand (1986“2005) 3.0 6.0 45
Sources: Author™s calculations from Ministry of Interior, election reports (1986, 1988, 1992a,
1992b, 1995, 1997); COMELEC (various years); Hartmann et al. (2001); Kasuya 2001b.

where there are on average 4.6 more parties in pre-reform Thailand than
in the Philippines (7.2 versus 2.6).31 As these numbers indicate, prior to
the recent reforms Thai voters, parties, and candidates did a much poorer
job coordinating across districts than did their Filipino counterparts. The
poorer aggregation in Thailand is re¬‚ected in the in¬‚ation scores “ the
average in¬‚ation score for the Philippines is 17 compared to pre-1997
Thailand™s 54 (Table 6.7).32

6.4.1 Accounting for Aggregation Differences

What explains why aggregation has traditionally been so much better in
the Philippines than in Thailand? To begin with, differences in social
heterogeneity cannot account for the difference in cross-district coordi-
nation. Most measures of social heterogeneity score the Philippines as
more diverse than Thailand. For example, the Philippines™ score on the
ethno-linguistic fractionalization index is 0.84 compared to 0.57 for
Thailand (Krain 1997). Similarly, if we look at religious fractionalization
the Philippines is also more diverse with a score of 0.29 versus 0.15 for
Thailand (Annett 2000).33 The Philippines™ greater social diversity is also

Using only post-Marcos elections, the difference is 3.6 parties.
For post-Marcos elections, the in¬‚ation score is 32, still much below Thailand™s 54.
An exception is Fearon™s ethnic fractionalization score based on linguistic diversity,
which reports Thailand as more diverse than the Philippines (Fearon 2003).
Philippines: Term Limits, Aggregation Incentives, No. of Parties 177

reinforced by geography with the country™s various groups spread out
among the country™s 7,107 islands. In short, if we look solely at social
heterogeneity, we would expect aggregation to be more dif¬cult in the
Philippines rather than less.
In Chapter 2, I argued that aggregation was primarily a function of
(a) the payoff to being the largest party at the national level and (b) the
probability that the largest party will be able to capture that payoff.
Two factors shape the largest party payoff: the distribution of power
between the national and subnational level (vertical centralization) and
the distribution of power within the national government (horizontal
centralization). Thailand and the Philippines look very similar in terms
of vertical centralization. In 1992 (the last year for which reliable data
are available for the Philippines), the subnational government™s share
of expenditures and revenues was 8.7 and 1.9, respectively, in the
Philippines versus 8.4 and 1.4 in Thailand (World Bank Group n.d.). In
both countries subnational governments are highly dependent on the
central government™s largesse.34
Within the national government, however, power has been relatively
more concentrated in the Philippines than in Thailand (again, until the
recent Thai reforms). Both countries have bicameral legislatures, but
unlike the Thai Senate, which was appointed and somewhat outside of
the control of the prime minister, the Philippine Senate is elected and is
typically controlled by the president™s party. Between 1946 and 2004,
the president™s party failed to capture at least 50% of the Senate seats
only twice. It is important to note that just because the president™s party
nominally controls the Senate it does not mean that the Senate is not an
important veto gate. Filipino parties are notoriously short on discipline


. 30
( 35 .)