. 29
( 35 .)


in¬‚ation score was 9.8 “ in other words less than 10% of the size of the
national party system was due to aggregation failures. This stands in stark
contrast to the post-Marcos in¬‚ation score of 32 (Table 6.4). Aggregation
has clearly declined in the recent democratic period, and it is this failure
to coordinate across districts that is primarily responsible for the larger
effective number of parties nationally.22

Kasuya ¬nds that post-Marcos aggregation failures have been most pronounced
between regions (Kasuya 2001b). Before 1972 the Nationalista and Liberal Parties
were consistently able to garner support across all regions of the country. After 1986,
a more regionalized party system emerged, with parties and candidates unable to
coordinate across regions. She argues that this re¬‚ects the failure of post-Marcos
presidential candidates to win cross-regional support. I explain why this failure has
occurred later in this chapter.
Building Party Systems in Developing Democracies

In the next two sections, I consider why aggregation has deteriorated
since 1986. I ¬rst review existing explanations for the expansion of the
national party system. I then argue that changes to the rules regarding
presidential reelection are responsible for the undermining of aggrega-
tion incentives and the consequent in¬‚ation of the party system.

6.3.2 Existing Explanations

A variety of explanations have been offered in an attempt to explain the
in¬‚ation of the national party system since 1986. These include a change in
the structure of local politics in the Philippines, the decreased importance
of the board of elections, and the advent of synchronized local and national
elections.23 A common explanation for both the stability of the two-party
system pre-1972 and the rise of multipartism after Marcos is the structure
of local politics in the Philippines. Before martial law, bifactionalism at the
local level was the norm (Land 1965, 1971; Wolters 1984; Kimura 1997).
As far back as the Spanish colonial period the political elite in each
area tended to divide itself into two major factions (Hollnsteiner 1963).
This local bifactionalism continued as the norm after independence and
prevented the emergence of viable third parties. Each local faction would
align itself with one of the two major parties, leaving third parties with no
organizational base to rely on at the local level (Land 1971, 103“4).
Eventually, though, bifactionalism began to break down in the
Philippines. This occurred ¬rst in urban areas where multifactionalism
had begun to displace bifactionalism by the 1960s (Laquian 1966; Nowak
and Snyder 1974; Kimura 1997). Bifactionalism continued to deteriorate
throughout the Marcos years so that by the time democracy returned
multifactionalism was the norm in many localities (Kimura 1992, 49).
Even though the shift from local bifactionalism to multifactionalism is
an interesting phenomenon, it cannot fully account for the growth of the
national party system. First, by itself bifactionalism locally does not
necessarily predict bipartism nationally. Indeed, it is quite possible to
imagine that local factions from different localities might back different
sets of parties. In other words, there is no reason to assume, a priori, that

The write-in ballot and the plurality rule for presidential elections have also come
under ¬re for causing multipartism (Velasco 1999, 173“4). However, since these two
features of the electoral system were the same both before and after Marcos, they
cannot explain the growth in the number of parties.
Philippines: Term Limits, Aggregation Incentives, No. of Parties 167

local factions will back the same two parties in every locality. A theory of
aggregation, such as the one presented in this book, is needed to explain
why actors might have the incentive to coordinate across different local-
ities to produce a national two-party system. Second, the shift toward
multifactionalism locally has not been associated with a large increase in
the effective number of parties at the district level (ENPavg). If the factional
structure of local politics were really driving the growth in the number of
parties, one would expect a large increase in ENPavg, and one would also
expect the growth of ENPavg to be primarily responsible for the increase in
the national effective number of parties (ENPnat). However, as already
discussed, this is not the case. ENPavg grew only slightly from 2 to 2.3 and
is responsible for only a small portion of the growth of ENPnat. In short,
despite the rise of local multifactionalism, intra-district coordination
continues to be quite good (as expected given the SMDP electoral system).
It is aggregation across districts that has broken down.
A second existing explanation for the advent of multipartism since
1986 is the decreased importance of party representation on precinct-level
election-monitoring bodies (the Board of Elections Inspectors and Board
of Election Canvassers) (Carlos 1997a; Velasco 1999). These bodies
consist of four members including one representative from the govern-
ment party and one from an opposition party (Carlos 1998a; Omnibus
Election Code). Before 1972, these party representatives played an active
and important role in protecting the votes for their party and monitoring
the fairness of the electoral process (Velasco 1999, 173). Candidates were
reluctant to join third parties because they would lose the right to party
representation on the inspection board (Carlos 1997a, 18). However, the
advent of independent electoral watchdog organizations since 1986 has
reduced the advantage of having party representation on the of¬cial Board
of Election Inspectors and Board of Election Canvassers. Thus third
parties are no longer deterred from entering (Carlos 1997a, 18). The
problem with this argument is the same as the problem with the bifac-
tionalism argument. If the declining importance of party representation
on precinct boards is really driving party system growth, then we should
see evidence of that growth occurring at the precinct and district levels.
But, once again, the source of more national parties post-1986 is primarily
poorer aggregation, not more parties winning votes at the local level.
A third explanation for the end of bipartism in the Philippines is the
adoption of synchronized local and national elections beginning in 1992.
Building Party Systems in Developing Democracies

Before martial law, local and national elections were held in different
years. The result was that during national elections local politicos would
support national candidates, and national politicians would return the
favor when local elections came around. Synchronized elections
disturbed this exchange relationship. In the words of one scholar:

[P]arties are adversely affected, as simultaneous elections weaken party links
between national and local candidates. Before 1972, when local and national
elections took place at different times, local leaders could devote their full
energy to supporting candidates for national of¬ce and vice versa: this they
cannot do any longer, as they have to ¬ght their own electoral battles during
the same period. (Velasco 1999, 173)

It is probably the case that the move to synchronized elections did
indeed make electoral coordination between the national and local levels
more complicated. However, there is good reason to suspect that this is
not the complete answer. The ability and willingness of local candidates to
coordinate with national candidates in synchronized elections is itself a
function of aggregation incentives. The stronger the aggregation incen-
tives, the more likely candidates are to do what is necessary to overcome
the added challenges associated with synchronized elections. To the
extent that synchronized elections increase the size of the payoff to
coordination, we might actually expect better aggregation during such
elections. In fact, there were three instances before 1972 where national
and local elections were held at the same time, and in no case did this lead
to a proliferation of parties at the national level. In 1947, 1951, and 1971,
elections for the Senate were run jointly with local elections. Even though
there are important differences between House and Senate elections, it is
instructive that the effective number of parties in the Senate was actually
lower when elections were synchronized “ the average effective number of
Senate parties for the 1946“71 period was 2.2 versus 2.0 in 1947, 1.99 in
1951, and 1.99 in 1971.

6.3.3 Presidential Term Limits and Aggregation Incentives

How and why, then, did aggregation incentives change post-Marcos?
When democratic government made its return to the Philippines,
the rules and institutions in place before martial law were largely
re-adopted. Within the national government the distribution of power
Philippines: Term Limits, Aggregation Incentives, No. of Parties 169

remained relatively concentrated in the hands of the president.24 The
national government also retained its dominant position over subna-
tional units (though subnational governments did receive new guar-
antees of budgetary support). In short, the payoff to being the largest
party was fairly constant across the two democratic periods.
Nevertheless, even though there was a high degree of institutional
continuity before and after martial law, the 1987 Constitution did
introduce one important change “ a ban on reelection for the president.
Before 1972, Philippine presidents were limited to two terms. In the wake
of the Marcos dictatorship, the constitution drafters opted to limit pres-
idents to a single term. If the theory outlined in Chapter 2 is correct, the
introduction of a reelection ban should boost the effective number of
presidential candidates in post-Marcos elections and thereby undermine
aggregation incentives and the reductive effect of concurrent elections.
Prior to 1972, incumbent Filipino presidents regularly marshaled
the resources and in¬‚uence of the presidency to back their reelection
bids. All but Marcos were unsuccessful in their bid for a second term;
nevertheless, the costs associated with challenging a sitting president
weeded out all but the most serious of challengers and enabled voters to
easily distinguish the frontrunners from the also-rans. This changed
with the introduction of the reelection ban. The ban lowered the bar-
riers to entry for presidential contenders and undermined the incentives
for sitting presidents to invest in party building. The result has been a
large increase in the number of viable presidential candidates.25 This is
clear from a comparison of the effective number of presidential can-
didates. During the 26 years before martial law, the average effective
number of presidential candidates (ENPres) was 2.2. By contrast, the
effective number of presidential candidates in the 1992 and 1998
presidential elections was 5.8 and 4.3, respectively (see the ENPres
column in Table 6.5). This is consistent with the theory and with the
large-N results in Chapter 3. Where there is no incumbent, the effective
number of presidential candidates is larger.

In a response to the excesses of the Marcos era, a few of the president™s powers were
curtailed, including the ability to declare a state of emergency and the ability to
transfer “saved funds” between governmental departments.
Choi (2001) also draws the connection between term limits and an increase in the
effective number of presidential candidates post-Marcos but does not discuss the
implications for the legislative party system.
Building Party Systems in Developing Democracies

The 2004 presidential election affords a unique opportunity to test the
relationship between reelection bans and the number of viable presi-
dential candidates. In 2001, President Joseph Estrada was forced from
of¬ce less than half-way through his term in the wake of a corruption
scandal. The vice-president, Gloria Macapagal Arroyo, took Estrada™s
place as president for the remainder of his term. Since she was not elected
to the of¬ce, President Arroyo was eligible to run for her own term as
president in the 2004 election. After initially promising she would not
run, Arroyo eventually decided to enter the race. In effect, then, the 2004
election was a race with an incumbent president as a candidate, akin to
the norm before martial law. If the hypothesis about the effect of
incumbency is correct, the presence of an incumbent in the race (albeit a
very weak and vulnerable one) should raise the barriers to entry for
prospective candidates and make the job of distinguishing between the
frontrunners easier, thereby reducing the effective number of candidates.
Indeed, this was the case. The effective number of presidential candidates
fell to 3.2 in the 2004 election, down from an average of 5.1 in the
previous two presidential contests (see the ENPres column in Table 6.5).
To determine whether the increase in ENPres post-Marcos has the
hypothesized effect on aggregation, we can compare the total effective
number of electoral parties (ENPnat) pre“ and post“martial law. As is
clear from Table 6.5, the rise the effective number of presidential can-
didates since 1986 has corresponded with a rise in ENPnat, as hypoth-
esized. Pre“martial law the effective number of electoral parties is 2.4, as
opposed to 4.0 after martial law.26 Comparing the three post-Marcos
presidential elections provides additional evidence of a relationship
between the number of viable presidential candidates and aggregation
incentives. Compare ENPres for the 1992 and 1998 elections. Note that
although ENPres for 1998 is quite high (4.3), it is still lower than ENPres
for 1992 (5.8). If the theory is correct, fewer viable candidates should
translate into better aggregation in 1998. Likewise, in the 2004 election,
the effective number of presidential candidates was lower than in either
1992 or 1998 (due to the presence of an incumbent in the race). If the
theory is correct, fewer viable candidates should translate into better
aggregation in 2004 compared to the previous two elections.

Unfortunately, party vote share data are as yet unavailable for 2004 so it is not
possible to calculate ENPnat.
Philippines: Term Limits, Aggregation Incentives, No. of Parties 171

Ideally I would measure the extent of aggregation in post-Marcos
elections as I have throughout this book “ by using party vote share data to
calculate the in¬‚ation score (I). However, for the 2004 election party vote
share, data at both the district level and national level are not yet available,
making it impossible to calculate ENPavg, ENPnat, and the in¬‚ation score.
So in addition to the in¬‚ation score for 1992 and 1998, I also use the
effective number of legislative parties (ENPseat) as an imperfect proxy for
the level of aggregation. ENPseat is calculated in the same way as ENPnat
but substitutes party seat shares in place of vote shares. Table 6.5 places
the ENPseat and ENPnat side by side. We can see that even though ENPseat
is always smaller than ENPnat (as expected), the two almost always move
in tandem.27 Thus, if the theory is correct, we would expect a decrease in
the number of presidential candidates (ENPres) to produce improved
aggregation and a decline in ENPnat and ENPseat. Table 6.5 demonstrates
that the number of electoral and legislative parties does indeed decline as
hypothesized “ fewer presidential candidates in 1998 and 2004 are
associated with fewer parties, whether measured by seat or vote shares.
Figure 6.1 incorporates the post-Marcos data in Table 6.5 with
additional information on the effective number of parties at the district
level (ENPavg) and the in¬‚ation score (I) for 1992 and 1998 (the only years
for which the requisite vote share data are available). As expected, the
effective number of presidential candidates tracks very closely to the
number of parties at the national level. Figure 6.1 demonstrates that we
cannot ascribe solely changes in the effective number of parties locally
(ENPavg). Looking just at the 1992 and 1998 election for which district-
level data are available, we can see that even though ENPavg does drop
between the elections, this decline in the average size of the local party
systems is not enough to account for the change in ENPnat. ENPavg falls by
less than 19%, while the national party system contracts by 38%. Better
aggregation accounts for the majority of the decline in the number of
parties in 1998. This is re¬‚ected in a decline in the in¬‚ation score from 45
in 1992 to 28 in 1998.
The upward pressure on the number of viable presidential candidates
since the introduction of the reelection ban has also served to undermine
the concurrency effect, as hypothesized. In fact, since 1987 the

The sole exception is between 1957 and 1961 where ENPnat decreases slightly and
ENPseat increases slightly.


. 29
( 35 .)