<<

. 19
( 35 .)



>>

candidate wins the most votes and so wins the seat. Now add two seats
to that district and give the voters in that district two additional votes. If
voters truly cast their votes according to party label, then they will cast
their additional votes for the two additional NAP candidates, and the
NAP will win all three seats. If on the other hand party labels are not the
primary cue for voters when casting their votes, then they may well cast
their additional votes for candidates from other parties. If this is the
case then the M þ 1 rule should apply, with larger seat districts having
more parties.
How important were party labels in Thailand? As the preceding section
demonstrated, labels were generally weak “ party label meant little to
either candidates or voters. This suggests that, given Thailand™s pre-1997
block vote electoral system, we should expect the average effective num-
ber of parties in each district (ENPavg) to be near M þ 1 for all districts.
However, if it is the case that party labels are in fact meaningful to voters,
then the number of parties should be near two in all districts, regardless of
district magnitude. These expectations are summarized in Figure 4.2.

The Southern Exception
Before presenting the data, one additional observation is useful. Recall
that in addition to polling data, voters might rely on party label or can-
didate/party electoral history as signals of which parties and candidates
are viable in a given district. In most Thai districts, there is no party with
a signi¬cant electoral history, nor do party labels carry with them a
habitual allegiance on the part of voters or candidates. The exception to
this general statement can be found in Thailand™s Southern region. The
South has long been the stronghold of the Democrat Party “ Thailand™s
oldest party. Of all the parties, the Democrats have traditionally had the
strongest party label.34 This party label, combined with the history of
Democrat strength in the South, means that voters, parties, and candi-
dates should have an easier time identifying viable candidates in Southern

34
This is relative to other Thai party labels. In fact, the Democrat Party label is not as
strong as one might expect given the party™s long history due in part to numerous
intra-party factional con¬‚icts throughout much of the party™s history (Somporn 1976;
Surin 1992).
Building Party Systems in Developing Democracies
106


M=1
ENPavg near 2

M=2
ENPavg near 3

M=3
ENPavg near 4


If party labels are valuable, ENPavg should be
near 2 across all district magnitudes.


¬gure 4.2. Expectations for ENPavg


districts compared to other regions of the country. As a result, one would
expect the effective number of parties in the South to consistently be
among the lowest in Thailand. One would also expect there to be very
little difference between one-, two-, and three-seat constituencies in terms
of the effective number of parties “ all should be near two, with one
caveat. During the mid- to late 1980s, the Democrat Party was riddled
with factional problems, culminating in a split in the party and the for-
mation of a breakaway party in 1988, the Prachachon Party. Thus one
would expect ENPavg in the South to be near or less than two and lower
than ENPavg in other regions of the country with the exception of the
1986 and especially the 1988 elections.


4.3.2 Testing the Expectations

In order to test these expectations, I collected district-level electoral data
for ¬ve general elections: 1986, 1988, September 1992, 1995, and 1996.
Elections prior to 1986 were excluded from the dataset because of some
differences in the electoral laws prior to 1986.35 The March 1992
elections were also excluded due to incomplete district level electoral data.
For each district I calculated the effective number of parties in that district.
In multiseat districts, vote totals for candidates of the same party were

35
One of the biggest differences was that prior to the 1986 election candidates were not
required to belong to a political party.
Thailand: Aggregation, Nationalization, Number of Parties 107

table 4.7. Regression Results: Effect of District Magnitude on the
Effective Number of Parties at the District Level

Dependent Variable: Effective Number of
Parties at the District Level (ENPavg) 1986“1996 1986“2005
District Magnitude (M) .26** .50***
(.03) (.03)
Constant 2.56 1.91
(.21) (.06)
R-squared .02 .15
Number of Observations 725 1525
**Signi¬cant at the .001 level; ***Signi¬cant at the .000 level; Standard errors in
parentheses



added together and used to calculate a party vote share.36 The data from
the ¬ve elections were then combined in a single dataset. Standard OLS
regressions were run to determine whether there was a relationship
between district magnitude and the effective number of parties. The
statistical analysis reveals a signi¬cant positive relationship between
district magnitude and the effective number of parties at the district level
(Table 4.7).
As expected, the ENPavg varies by district magnitude. Higher district
magnitudes are associated with more parties. This is additional evidence
that party label is not the primary guide for Thai voters when they cast
their votes. (The results are even stronger when one includes constituency
results from the 2001 and 2005 elections, which used only single-seat
districts.) Figure 4.3 graphs the effective number of parties averaged across
one-, two-, and three-seat districts over the period. Again, the difference
between one-, two-, and three-seat districts is evident. The effective
number of parties is also near where we would expect given the M þ 1 rule.
In three-seat districts ENP is 4 or less in every election. In two- and single-
seat districts, ENP is slightly higher than the M þ 1 limit in the ¬rst couple
of elections, and then falls to within the M þ 1 range in later elections.37

36
In two cases the election results failed to list a vote total for a candidate. In both cases
I calculated an estimated vote total by splitting the difference in the vote totals for the
next highest and next lowest candidates. In both 1986 and September 1992 four
districts had to be thrown out because of missing data.
37
Note that it is harder to draw inferences about single-seat districts due to the
relatively small number of single-seat districts in each election.
Building Party Systems in Developing Democracies
108

4

3.5
Number of Parties
Average Effective




3

2.5

2
M=3
M=2
1.5
M=1
1
1986 1988 1992b 1995 1996
Election Year
¬gure 4.3. Average Effective Number of Parties by District Magnitude


Note in Figure 4.3 that ENPavg falls from 1988 onward for all three
district magnitudes. What™s behind this steady fall in the number of
parties? One might argue that the decline in the number of parties over
time is evidence that party labels have become more important signals to
voters and candidates as parties have built up an electoral history at the
district level. As a result, it has become easier over time for voters, can-
didates, and parties to distinguish frontrunners from the also-rans. Indeed,
this may be the case in certain constituencies, especially in the South as is
discussed in more detail later. However, the fortunes of most political
parties still vary greatly from election to election. This suggests that the
decline in the number of parties over the period is a result of other factors
quite apart from an increase in the value of party labels. For example, the
electoral history of certain candidates, incumbents, or factions may be
built up over time even if they switch parties from election to election
(Napisa 2005). Thus, although a given candidate™s party label may
communicate little information about the candidate™s chances, the can-
didate™s (or his/her faction™s) previous showing in the district might be all
the information voters, parties, and other actors need to coordinate as the
M þ 1 rule predicts. Another possible explanation for the decline in
ENPavg is new electoral system shock. When a new electoral system is
adopted, it may take candidates and voters time to understand the
incentives of the new system and adjust their behavior accordingly. Thus
one would expect ENPavg to decline over time.38

38
Although the block vote electoral system had been used in Thai elections before
1979, 1979 marks the beginning of regular elections. Also, 1986 marked the ¬rst
Thailand: Aggregation, Nationalization, Number of Parties 109

9
8
Effective Number


7
of Parties


6
5
4
3
2
1
0
1986 1988 1992b 1995 1996
Election Year
Bangkok Northeast South
Central North
¬gure 4.4. Average Effective Number of Parties per District by Region


To summarize, the evidence is fairly consistent with the M þ 1 rule.
ENPavg varies by district magnitude and is at or near the M þ 1 level.
In other words, the intra-district dynamics seem to be working as
expected.

Regional Variation
Figure 4.4 displays the average effective number of parties by region.
As expected, in the last three elections Southern districts had, on average,
a much lower effective number of parties than districts elsewhere in
Thailand “ 2.4 over the period compared with 3.3 for the rest of the
country. For most elections the Southern ENPavg re¬‚ected the strength of
the Democrat label in the region “ around two or less. However, one can
also see the factional con¬‚ict in the Democrat party re¬‚ected in the 1986
and 1988 elections. While the Southern ENPavg for 1986 is still one of the
lowest in the country, it is still much higher than it is throughout the 1990s.
In 1988, the year in which Democrats and former Democrats ran against
each other in several constituencies, the effective number of parties in the
Democrat™s Southern stronghold is the second highest in the country.
Excluding the 1988 election yields an ENPavg of 2.1 for the South.
In summary, the large number of parties in Thailand at the national
level does not appear to be a function of the electoral system or

election under the new constitution where candidates were required to be members
of political parties.
Building Party Systems in Developing Democracies
110

district-level coordination failures. The average effective number of
parties at the constituency level is quite small (3.2 on average),
much smaller than the effective number of parties nationally. How, then,
do we account for the large number of parties at the national level?



4.4 the national party system: cross-district
coordination (aggregation)
I argued in previous chapters that Duverger™s law and Cox™s M þ 1 rule
operate at the district level. The result is numerous district-level party
systems each with its own effective number of parties. How do the local
party systems map onto the national party system? How well do parties
coordinate or aggregate across districts? Recall that if the same parties are
the frontrunners in all districts nationwide, then the effective number of
parties nationally should be equivalent to the average number of parties in
each district. The difference, then, between the effective number of parties
nationally and the average effective number of parties at the district level is
a measure of the extent of aggregation between the local and national
party systems (D ¼ ENPnat À ENPavg). The larger the difference is, the
poorer the aggregation will be (Cox 1999; Chhibber and Kollman 1998,
2004). ENPavg and ENPnat for Thai elections are displayed in Figure 4.5.
Note that even though ENPavg averages 3.2, the effective number
of parties nationally (ENPnat) averages 7.2. We can convert this difference
into Cox™s in¬‚ation measure using the formula I ¼ 100*[(ENPnat À
ENPavg)/ ENPnat].39 As discussed in Chapter 1, the in¬‚ation score com-
municates the percentage of the effective number of parties nationally that
is due to poor aggregation “ higher numbers re¬‚ect worse aggregation.
Figure 4.6 displays the in¬‚ation scores for each of Thailand™s elections.
As Figures 4.5 and 4.6 indicate, aggregation was poor between
the local and national party systems during the pre-1997 period. The
average effective number of parties at the district level ranges from 2.4 to
3.7, but the effective number of parties nationally goes from 4.6 to 9.8
(Figure 4.5). In other words, between 48 and 62% of the national party

<<

. 19
( 35 .)



>>