<<

. 14
( 35 .)



>>

I do not report the results from the cross-sectional in¬‚ation model here due to the
small number of observations in that speci¬cation. However, the results from that
model are nearly identical to the pooled in¬‚ation model (#4).
95% Confidence Interval




76
Marginal Effect of Proximate Presidential Marginal Effect of Proximate Presidential
(a) (b)
Elections on Inflation as ENPC Changes (XS) Elections on ENPNAT as ENPC Changes (XS)
10
.2

0 5

“.2
0
“.4
“5




Presidential Elections
“.6




Presidential Elections




Marginal Effect of Proximate
Marginal Effect of Proximate
“10
“.8
0 1 2 3 4 5 0 1 2 3 4 5
Effective Number of Presidential Candidates
Effective Number of Presidential Candidates

Marginal Effect of Proximate Presidential Marginal Effect of Proximate Presidential
(c) (d)
Elections on Inflation as ENPC Changes (Pooled) Elections on ENPNAT as ENPC Changes (Pooled)
10
.5

5


0 0


“5




Presidential Elections
Presidential Elections




Marginal Effect of Proximate
“.5
Marginal Effect of Proximate



“10
0 2 4 6 0 2 4 6
Effective Number of Presidential Candidates Effective Number of Presidential Candidates


¬gure 3.3. Marginal Effect of Proximate Presidential Elections as the Effective Number of Presidential Candidates Changes
table 3.7. Reelection Restrictions and the Effective Number of Presidential Candidates

Cross-Sectional Analyses “ Dependent Variable Pooled Analyses “ Dependent Variable

Explanatory 1 (ENPnat) 2 (ENPres) 3 (ENPres) 5 (ENPnat) 6 (ENPres) 7 (ENPres)
Variables (Golder-leg) (Golder-pres) (Golder-pres) (Golder-leg) (Golder-pres) (Golder-pres)
4 (in¬‚ation)
0.87** 0.10*** 1.29***
ENPres
(0.30) (0.02) (0.20)
log_govage À0.03**
(0.01)
0.01
log_districts
(0.01)
0.23 1.69***
log_avemag
(1.06) (0.24)
ef*log_avemag À0.37 À0.83*
(1.52) (0.46)
0.27 0.47 0.89* 0.15* 1.67** 0.12 0.98
ef
(2.83) (0.36) (0.46) (0.07) (0.72) (0.47) (0.63)
0.43 0.42* 0.55** 0.53*
no_reelection
(0.26) (0.22) (0.27) (0.27)
0.11 0.53 0.53*
runoff À0.42
(0.39) (0.37) (0.55) (0.28)
0.84 2.03*
ef*runoff
(0.89) (1.16)
0.04 0.11
no_reelection*
runoff (0.47) (0.45)
0.57 2.04*** 1.84*** 2.28*** 1.88***
Constant À0.12 À1.32*
(1.89) (0.31) (0.32) (0.09) (0.71) (0.30) (0.39)
R-squared .30 .32 .21 .77 .51 .24 .19
13 31 31 38 104 104 104




77
Observations
* p < .1, ** p < .05, *** p < .01; standard errors in parentheses
Building Party Systems in Developing Democracies
78

number of presidential candidates, especially in the pooled analysis.
But how do we account for the nature of the presidential electoral
system? Do we observe more presidential candidates in the face
of reelection restrictions when the presidential electoral system is
permissive, and fewer when it is restrictive? To test whether this is the
case, I interact no_reelection with runoff and display the results of these
analyses in columns 3 and 7. (Recall that runoff is coded 1 if majority
runoff, a permissive electoral rule, is used to elect the president, 0
otherwise.) No_reelection has a positive and signi¬cant effect on the
number of presidential candidates, even when the electoral rule is
restrictive (runoff equals 0). By contrast, Figure 3.4 suggests that under
majority runoff, the marginal effect of no-reelection is no longer
distinguishable from zero. It appears then, that reelection restrictions
only have a discernable in¬‚ationary effect on the number of parties
where the electoral system is restrictive. Where a permissive electoral
system (i.e., majority runoff) already allows for a large number of
candidates the effect of reelection restrictions is super¬‚uous.
In summary, the data support the hypotheses laid out in this
section. Proximate presidential elections are associated with better
aggregation (and fewer parties), but this effect is conditional on the
number of presidential candidates. In addition, when presidential and
legislative elections are concurrent, the effective number of presi-
dential candidates is positively related to both party system in¬‚ation
and the effective number of electoral parties, as hypothesized. There is
also some support for the hypothesized link between bans on presi-
dential reelection and the effective number of candidates, even when
controlling for the effect of the presidential electoral system and
ethnic heterogeneity. Finally, the marginal effect of the reelection
does appear to be conditional on the permissiveness of the electoral
system.


3.6 social heterogeneity and aggregation
Even though social heterogeneity is not the focus of this study, it is
worth taking a moment to think more carefully about the role it plays in
shaping the incentives and capability of candidates to coordinate across
districts. Throughout the preceding analyses, I controlled for the social
heterogeneity (operationalized as ethnic fractionalization [ef]) but now
Testing the Theory 79

95% Confidence Interval
Marginal Effect of No Relection on
(a)
ENPC as the Electoral System Changes (XS)
1.5
Marginal Effect of No Relection




1



.5



0



“.5
0 .5 1
Electoral System (Runoff or no?)

Marginal Effect of No Relection on
(b)
ENPC as the Electoral System Changes (Pool)
1.5
Marginal Effect of No Relection




1




.5




0


0 .5 1
Electoral System (Runoff or no?)

¬gure 3.4. Marginal Effect of No Reelection as the Electoral System
Changes
Building Party Systems in Developing Democracies
80

investigate how social heterogeneity might interact with the size of the
aggregation payoff to affect aggregation. In Chapter 2, I argued that by
itself social heterogeneity is neither necessary nor suf¬cient to produce
poor aggregation. Instead, the effect of heterogeneity on aggregation is
conditional on the size of the aggregation payoff. A large payoff may
mitigate the negative effect of social heterogeneity on aggregation, and
likewise, a small payoff may undermine aggregation incentives, even in
the face of social homogeneity. Aggregation should be poorest when
there is both a high degree of social heterogeneity and a small aggre-
gation payoff. This suggests the following hypothesis.

Social Heterogeneity*Payoff Hypothesis: The marginal effect of
social heterogeneity on in¬‚ation is conditional on the size of the
aggregation payoff. Social heterogeneity increases in¬‚ation only
where the aggregation payoff is not large.

I test this hypothesis with both cross-sectional and pooled analyses
once again using OLS with robust standard errors clustered by country
for the pooled analyses. I continue to operationalize social heteroge-
neity as ethnic fractionalization (ef) and use the decentralization index
as a proxy for the size of the aggregation payoff. Table 3.8 displays
the results. For comparative purposes, I™ve reproduced the additive
models found in Table 3.4 as columns 1 and 3 in this table. Entered
separately, both decentralization and ef have signi¬cant positive effects
on in¬‚ation. When interacted, however, we see that the effects of
ethnic fractionalization are conditional on the degree of decentraliza-
tion. When power is centralized (decentralized ¼ 0), the coef¬cient on
ethnic fractionalization, while still positive, is no longer signi¬cant.
This suggests that a high degree of centralization can still induce
aggregation, even in the face of social heterogeneity. Alternatively, the
positive and signi¬cant coef¬cient on decentralization implies that even
when a polity is ethnically homogenous (ef ¼ 0) a small aggregation
payoff can still be suf¬cient to undermine aggregation. The marginal
effect of each variable on in¬‚ation increases in the other (as the coef-
¬cient on the interaction variable signi¬es). The marginal effects graphs
in Figure 3.5 show that, as hypothesized, the marginal effect of ethnic
fractionalization is signi¬cant only for values of decentralization
greater than 0.
table 3.8. Social Heterogeneity and Aggregation (Dependent Variable: In¬‚ation)

Cross-Sectional Analyses Pooled Analyses

Explanatory Variable 1 2 3 4
0.06** 0.06* 0.07*** 0.06*
decentralization

<<

. 14
( 35 .)



>>