You are on page 1of 20

Anti-Government Protests in Democracies: A Test of Institutional Explanations

Author(s): Yen-Pin Su
Source: Comparative Politics, Vol. 47, No. 2 (January 2015), pp. 149-167
Published by: Comparative Politics, Ph.D. Programs in Political Science, City University of
New York
Stable URL:
Accessed: 16-01-2018 10:20 UTC

Linked references are available on JSTOR for this article:
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

Comparative Politics, Ph.D. Programs in Political Science, City University of New

York is collaborating with JSTOR to digitize, preserve and extend access to Comparative

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Anti-Government Protests in Democracies

A Test of Institutional Explanations

Yen-Pin Su

Social protests have played an important role in the political development of many
countries. For instance, since the 1990s, popular protests in Latin America have
played a crucial role in removing elected presidents from office before their ends of
terms.1 Additionally, the economic crises in Europe since 2008 have mobilized large-
scale protests in Greece and Spain, which have greatly influenced the stability of their
respective governments.2
Previous research has shown that a higher level of democratization reduces the
mobilization costs for civil society actors and thus is more conducive to protests.3
However, other studies argue that democratization might lead to a decline of social
protest activities because a more democratic political system provides more institu-
tionalized channels for political participation and grievance resolution.4 These two
arguments about the effects of democratization on social protests pose a puzzle:
why do some democratic countries experience more protests than the others?
Empirical studies have provided explanations about the national economy, showing
that large-scale collective protests were often driven by poor national economic perfor-
mance.5 Moreover, a higher level of economic development might provide more mobili-
zation resources for social actors to transform their grievances into protest action.6 While
these studies underline the importance of macro-level factors in explaining protests,
they fail to uncover how institutional factors link democratic regimes with protests.
This study empirically tests two institutional explanations against existing expla-
nations of protests. The first explanation deals with certain political institutions as
structural determinants that shape protest activities, while the second explanation con-
siders opposition parties as agents of protest mobilization. Drawing on data of anti-
government protest events and legislative elections in 107 democratic countries from
1990 to 2004, the empirical analyses show that the second explanation holds more
weight. The results show that the mobilization capacity of opposition parties matters
for understanding the cross-national variation in protests; specifically, a larger opposition

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Comparative Politics January 2015

camp fosters more anti-government protests only if this opposition camp is more united.
Moreover, I find that the mobilization capacity of opposition parties matters for anti-
government protests in developing countries but not for those in developed countries.

Institutional Theories of Social Protests

Based on rational-choice institutionalising the political science literature suggests tw

different institutional approaches for studying political behavior.8 The first approach
conceives of institutions as rules and procedures that shape political actors' strategi
calculus,9 while the second approach conceives of certain kinds of institutions as ration
agents with strategic goals.10 Both institutional approaches have been present in th
literature on social protests. For example, studies that use the framework of politic
opportunity structure have considered certain institutional environments for explainin
protests.11 In addition, studies that adopt the insights of resource mobilization theor
have shown that parties are important institutional actors that can strategically coordinat
with social movements to mobilize protests for pushing certain political agendas.12 Belo
I will discuss both institutional approaches and generate testable hypotheses.

Institutions as Structural Factors that Shape Protests Given that parties perfor
essential functions for democratic representation,13 scholars have argued that countr
with better partisan representation tend to experience fewer anti-government protests
Quality of partisan representation is, in part, determined by the level of party syste
institutionalization. In general, a highly institutionalized party system indicates th
parties within the system have stable organizations and that different political prefe
ences are stably structured in the inter-party competition over time.15 An institutio
alized party system can "help groups express their interests... channel politi
demands and can dampen political conflicts."16 In contrast, a poorly institutionalize
party system implies that parties are not well connected to society and that govern
ments are less responsive and accountable. As a result, a poorly institutionalized par
system "creates a political vacuum, producing a more conducive environment f
greater levels of mobilization."17
Using party system fragmentation as an indicator for the level of party system
institutionalization, Arce finds that a highly fragmented party system leads to mo
protest activities.18 However, Arce and Rice find that a fragmented party system mig
help reduce the incidence of protests because such a party system suggests greater
electoral competition, which may encourage the governing party to make policies t
prevent protests.19 In short, these studies indicate that the relationship between par
system fragmentation and protests might not be clear.
In addition, some studies have found that a country experiences fewer protests
it has institutions that can better ensure political representation for diverse social groups
such as the parliamentary system or the proportional representation system.20 The ma
theoretical reasoning is that such political institutions provide more incentives for variou

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Yen-Pin Su

social groups to form new political parties to access seats in the legislature and pro-
vide more opportunities for these groups to hold posts in the government.21 As a
result, the institutions that are better able to encompass more political minorities
should discourage certain political elites from mobilizing extra-institutional protest
activities to pursue political goals.
Another important institutional factor that can help explain incidences of protest
is the presence or absence of influential political elite allies in the government.22 Jung
shows that social protests in Western European countries tend to occur less often when
leftist governments are in power.23 Some studies provide evidence that social move-
ment organizations tend to lead to an increased number of protest activities when
the legislature has more members that share a similar ideological position with the
movements.24 However, other studies show that with more elite allies in the legisla-
ture, the incentives for protest might be reduced because there is a higher probability
that the movement's demands can be addressed through institutionalized means.25
Last, the timing of electoral cycles, such as the general election year and the
"honeymoon" year (the first year of the new government), might be an important
structural factor that shapes protest activities. Bruhn argues that more protests tend
to occur during the general election year because protest groups expect that their
voices are more likely to be heard during the electoral campaign.26 However, Meyer
and Minkoff find that protest incidence is neither higher nor lower during the election
year.27 Finally, while it is possible that movements might reduce protest activities in
the "honeymoon year" because they may be willing to give the new government more
time to respond to their demands, Bruhn finds that movements actually tend to
mobilize more anti-local government protests in the honeymoon year.28

Institutional Actors as Mobilizing Agents for Protests Political parties are impor-
tant institutional actors that can influence protests. To better understand such impacts,
it is necessary to consider the division of governing party and opposition party as a
salient political cleavage. In general, a government tends to maintain political stability,
and thus the party or parties that control the government would try to demobilize
activities that could threaten political stability. Opposition parties might also pursue
political stability, but in many countries these parties often use both institutional
and non-institutional means to influence the policy-making process.29
As Maguire argues, strong interaction between parties and movements tends to be
prevalent "when parties are in opposition and are building social coalitions for elec-
toral purposes."30 This implies that there exist mutual benefits for both social move-
ments and opposition parties.31 On the one hand, opposition parties might view certain
movements as their constituency;32 on the other hand, coordinating with opposition
parties might enable movements to obtain financial support, broaden their mobiliza-
tion networks, and enhance their media visibility.33
Given that opposition parties and movements mutually benefit by coordinating
protests to challenge the government, under what conditions are large-scale coordinated
protest campaigns more likely to occur? This study argues that the mobilization capacity

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Comparative Politics January 2015

of opposition parties is a key factor. Specifically, my core theoretical intuition suggests

that when opposition parties are strong and united, they are more able to mobilize
greater constituencies for large-scale collective protest actions. I measure opposition
mobilization capacity along two dimensions: the first is the size of opposition, captur-
ing the amount of resources that opposition parties are potentially able to mobilize for
protests; the second is the unity of opposition, capturing the degree to which opposi-
tion parties are able to coordinate as a collective actor.
Almeida demonstrates that the frequency of protests increases with higher popular
support for opposition parties.34 However, Boulding shows that a larger opposition
party might discourage protests because "participation will be channeled into electoral
routes if those routes are promising."35 As LeBas argues, a larger opposition camp may
either indicate that the camp is effective in mobilizing voters or simply suggest high
levels of discontent against the incumbent party and have little to do with the strength
of opposition parties.36 Thus, another important dimension of opposition mobilization
capacity - the opposition unity - must be taken into account.37
My theory suggests that the frequency of protests is a multiplicative function of
opposition size and opposition unity. In general, the number of protests is likely to
increase when the opposition camp is large, which means that there are more potential
resources that opposition parties can mobilize. However, a large opposition camp does
not guarantee the mobilization of large-scale protests if the camp is highly fragmented.
Protests may still occur when the opposition camp is fragmented, but a high level of
fragmentation makes it more difficult for the various groups to coordinate for mobi-
lizing larger collective protests. This coordination problem also makes people skeptical
about the possibility of protest success and thus may discourage them from joining the
protests. In contrast, a large and united opposition camp indicates that the camp has
both sufficient resources and the capacity to coordinate large protests. In this case,
the expected probability of a successful mobilization will be higher, and large-scale
protests are therefore more likely.
Take the relations between opposition parties and social movements in Peru,
for example.
After the fall of Fujimori, Peruvian opposition parties together captured 74 percent
of the votes in the 2001 legislative election and 79 percent in the 2006 legislative elec-
tion. Aggregated, the data suggest that the opposition camp in Peru had a considerably
strong presence, and it may be expected that massive protest mobilization would
occur more often. However, while regional- level protests did occur occasionally
(e.g., Arequipazo in 2002), large-scale protests did not.
As some scholars have argued, in Peru, parties' lack of strong organizational
roots in society38 and the high level of party system fragmentation39 are important
factors that impede political parties' capacity for collective action in this country. The
absence of strong parties also implies that social movements lack influential allies
for coordinating large-scale protest mobilization. As Arce describes,40 many Peruvian
protests took place spontaneously without a coordinating body, and most of them
were geographically segmented in peripheral regions.

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Yen- Pin Su

Overall, the case of Peru in the post-Fujimori period implies that a large opposi-
tion camp does not necessarily guarantee a higher mobilization capacity if it lacks
unity. In order to better explain why some countries experience more anti-government
protests than others, it is necessary to take into account the interaction effects of
opposition size and opposition unity. In short, the second institutional explanation
suggests a testable hypothesis: a large opposition camp will encourage more anti-
government protests only if the camp is more united.

Opposition Mobilization and Protests in Different Contexts So far, the theory of

opposition mobilization capacity argues that the level of protests depends on the level
of opposition parties' capacity for mobilization. However, the effect of opposition
mobilization capacity on protests might depend on different socio-economic contexts.
Here I argue that the level of modernization is an important factor that influences
the role of opposition parties in mobilizing anti-government protests.
Studies have shown that there is a distinct trend for certain countries becoming
"movement societies,"41 in which protest activities are expanding42 and becoming a
conventional form of political participation along with lobbying and voting.43 More
importantly, several studies have shown that movement societies are common in post-
industrial societies.44 Why do countries with a higher level of modernization experi-
ence more protests? Inglehart proposes a cultural explanation, arguing that a deeper
modernization process will result in a cultural change from materialism to post-
materialism.45 The core of postmaterialism is the emphasis on self-expression values,
such as social tolerance, which "involves an inherently antidiscriminatory orientation"
and "provides people with a strong motivation to engage in social movements."46
Inglehart and Welzel find that a country with stronger self-expression values tends
to experience more protests,47 and one consequence is that "the bureaucratic organiza-
tions that once controlled the masses, such as political machines, labor unions, and
churches, are losing their grip."48
By implication, it is expected that in a highly modernized postindustrial society,
citizens will tend to actively engage in protest activities to affect policy-making pro-
cess regardless of the strength of mobilizing agents such as political parties. Mobili-
zation capacity of the opposition parties might still matter, as in 1999 when the
Christian Democratic Union of Germany and the Christian Social Union of Bavaria
(CDU/CSU ) successfully launched an influential protest campaign in Germany.49
However, opposition parties might not matter in other cases. A recent example is
the 2011 anti-state-government protests in Wisconsin in the United States. While
many protesters were Democrats, the Democratic Party did not actively mobilize
the movement, which was described as "a mix of spontaneity and organizing ... led
not by a single organization or coalition but rather by a variety of different groups,
organizations, and coalitions, not all of which were in contact with one another."50
In contrast, I argue that opposition mobilization capacity plays a very important
role in collective protests in developing countries. First, in many developing countries
that are also new democracies, protests are less institutionalized as a conventional form

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Comparative Politics January 2015

of political participation because of the legacies of the former authoritarian regime51

and the lack of democratic learning experiences.52 This implies that the role of mobi-
lizing agents, such as unions and parties, is much more crucial for protest mobilization
in developing countries. In other words, without being mobilized by these agents,
citizens in developing countries might not participate in protests as actively as those
citizens in developed countries.
Moreover, because individual political parties and social movements in develop-
ing countries are generally weaker and less institutionalized than those in developed
countries, a strong and united opposition camp is crucial for large-scale collective pro-
tests.53 Without being mobilized by political parties, citizens in developing countries
might still go to the streets to protest, but it is expected that the scale and political
impact of the protests will be limited. In short, I hypothesize that a higher level of
mobilization capacity for opposition parties encourages anti-government protests in
developing countries but not in developed countries.

Data, Research Design, and Operationalization

To test the two institutional explanations, I conduct statistical analyses of anti-go vemment
protests, legislative elections, and opposition parties in 107 democratic countries from
1990 to 2004 (the information about these cases is available in the appendix54). The unit
of analysis is country-year. To ensure the electoral results are relatively reliable for
measuring the level of opposition mobilization capacity, a country-year observation is
included in the analysis when its Polity IV55 score is greater than or equal to five during
the period under study.56 The national lower house elections in countries with bicameral
legislatures and the elections of the national assembly/congress in countries with uni-
cameral legislatures are considered. The country sample is selected to include a diversity
of democracies in terms of levels of development, size, and geographical locations.

Dependent Variable The dependent variable in this study is the frequency of

"Anti-Government Protests," operationalized as the annual number of protest events
that were initiated by domestic actors against their domestic government in a country.57
To code this variable, I use King and Lowe's 10 Million International Dyadic Events
database,58 which relies on Reuters Global News Service from 1990 to 2004. 59 An
important feature of King and Lowe's dataset is that it codes an event in a "who did what
to whom, when, and where" manner. For the purpose of this study, an anti-government
protest event is retrieved from this database if domestic government-related agents are the
"whom"60 and protest activities are the "what."61 I do not specify the particular domestic
actors who conducted the protest event in order to encompass as many observations
as possible for the empirical analyses and to test the generalizability of my theory.
King and Lowe's data must be used with due caution. For instance, one prob-
lem with using newspapers as a source of data on contentious events is bias toward
disproportionate reporting of larger events.62 It is clear that King and Lowe's data do

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Yen-Pin Su

not contain information on all protest incidences that have taken place in a country, but
only, or largely, the protest events that are important enough to capture international
media attention.63 Moreover, the information generated by the automated procedures
might be limited and thus further undermine the quality of the data.64 Although King
and Lowe's database may underestimate the number of small-scale protest events,
it is sensitive to the anti-government protests that are likely to have greater political
impacts. Thus, while the database does not allow me to explore how opposition parties
affect the frequency of all protests, it facilitates the analyses about the frequency of
larger and more important protests.

Explanatory Variables To test the first institutional explanation, several institu-

tional variables are considered. First, I include "Party System Fragmentation^" in
the model, measured as the index of the effective number of parties (ENP) based
on the vote data of national legislative elections.65 In addition, I include "Parliamentary
System," "PR System," and "Leftist Government."66 To take into account the possible
effects of temporal dimensions on protests, I include two dummy variables: "General
Election Year" and "Honeymoon Year" (the year after the general election year).
To capture the effects of elite presence in the government, I include "Percentage of
Opposition Party Members in the Legislature," which is constructed by using various
electoral data sources (see below).
The major explanatory variables of the second institutional explanation pertain to
the mobilization capacity of opposition parties. For the purpose of the analyses, an
opposition party is defined as a party that does not control or share the executive
power.67 The mobilization capacity of opposition parties are measured using three vari-
ables. The first explanatory variable is "Opposition Size," measured as the vote share
of opposition parties in an election.68 The second explanatory variable is "Opposition
Unity," measured as the inverse value of the effective number of opposition parties
in an election,69 which is calculated as:

U = E (Ovj/ZOv)2
where U is the degree of opposition unity, Ov¡ is the vote share of the ith opposition
party, and XOv is the sum of all opposition parties' vote shares in the election. The
upper bound of U is 1 , meaning that the opposition camp is unified as a single oppo-
sition party. The closer that U approaches 1, the more united the opposition is. The
third explanatory variable is "Opposition Size*Opposition Unity." Based on my dis-
cussion, this variable should have a positive and significant effect, indicating that the
size of the opposition camp will have a greater influence on increasing the number of
anti-government protests as this camp is more united.
As I have discussed in the theoretical section, however, I contend that the role
of opposition parties in mobilizing protests should be more important in devel-
oping countries than in developed democracies. Categorizing the entire sample into
two subsamples - developed countries and developing countries - I expect that the
multiplicative term "Opposition Size*Opposition Unity" should have no significant

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Comparative Politics January 2015

effect on anti-government protests in developed countries, while the effects of

"Opposition Size*Opposition Unity" on anti-government protests should be positive
and strong in developing countries.70
These three explanatory variables are constructed relying on various sources for
the electoral data.71 Ideally, I should consider all opposition parties appearing on the
ballot; however, most available election results cluster parties that received very few
votes under a residual "other parties" category. For this reason, I restrict the measure
of opposition mobilization capacity considering only the parties that receive at least
0.5 percent of the vote.72
In general, the values of the three opposition mobilization capacity variables for a
country vary by election cycles. For two kinds of cases, the values of these variables
may change within an electoral cycle: 1) when a presidential democracy has a non-
concurrent schedule for legislative and presidential elections; and 2) when a parliamen-
tary democracy experienced a change in the coalition government within an electoral
cycle. Last, because opposition parties and protests may affect one another in
the election year, I use lagged electoral data to construct the independent variables
to account for possible endogeneity with the dependent variable.73

Control Variables To capture the short-term economic impacts on protests, I control for
"GDP Growtht_!"and "Inflation^." "Inflation^" is operationalized as the logged value of
the inflation rate for the year prior to the election year.74 To control for the possibility
that a higher level of economic development helps provide more resources for protest
mobilization, I include "GDP Per Capita^. "75 "Freedom House Index^" is included to
control for the possibility that a higher level of democratization would encourage more
protests.76 Last, I include the logged value of a country's "Population"77 to control for the
possibility that a larger country may experience more protests (the summary statistics of
the variables used for the empirical analyses are available in an online appendix).

Estimation Technique

This study utilizes zero-inflated negative binomial regression (ZINB) to test the impact
of opposition mobilization capacity on the frequency of anti-government protest events.
Given the discrete and nonnegative properties of the dependent variable, using OLS
regression is inappropriate. The summary statistics show that the variance of my
dependent variable (56.2) is much larger than its mean (3.8), which suggests that
the data is over-dispersed and the use of negative binomial regression is justified as
a more appropriate estimation technique than Poisson regression.78 However, because
the alpha parameters are statistically significant in all models and the Vuong statistic79
is significant,80 I use zero-inflated negative binomial regressions instead of a standard
negative binomial regression model.
Several additional methodological issues must be addressed. First, because my
data include multiple observations from the same country over time, observations

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Yen-Pin Su

within countries may not be truly independent. Therefore, I employ HuberAVhite/

sandwich robust variance estimators to obtain robust standard errors. Second, to
account for period-specific impacts (e.g., global economic crisis), I include a series
of year dummies in the full model.81 Third, although the tests for autocorrelation do
not exhibit a clear pattern of temporal dependence in my data,82 I include a lagged-
dependent variable in the full model as a robustness check.

Empirical Results

In this study, the ZINB model estimates the effects of the explanatory variables
on 1) the absence of the anti-government protests and 2) on the count of the anti-
government protests in a single model by combining a logit distribution (inflation
stage) with a negative binomial distribution (count stage). While the results in the
inflation stage may show interesting information, it is only the negative binomial
regression that provides a test of my hypotheses. It is because the explanations to
be tested in this study mainly concern the frequency rather than the absence (or pres-
ence) of protests in a given country. Therefore, my analytical attention will con-
sequently focus on the results of the count stage of the ZINB model (i.e., negative
binomial regression).
Table 1 reports the results for four separate zero-inflated negative binomial
models. Model 1 includes only the explanatory variables of the first institutional
explanation, which considers institutions as structural determinants. Model 2 includes
only the explanatory variables of the second institutional explanation, which con-
siders opposition parties as institutional actors that mobilize anti-government protests.
Model 3 jointly tests the above two institutional explanations. Model 4 specifies a
full model by including all explanatory variables, control variables, and year dummies.
The results in Model 1 and Model 3 show that a country that adopts the PR system
for electing the majority of the legislative seats tends to experience fewer anti-
government protests. However, in Model 4 where all control variables are included,
the effect of PR system becomes statistically insignificant. In contrast, the results of
Model 2, Model 3, and Model 4 show that the second institutional explanation works
better. The coefficient of the multiplicative term "Opposition Size*Opposition Unity"
is positive and its effect is statistically significant across different model specifications.
Figure 1 presents how the relationship between anti-government protests and
opposition size varies across different levels of opposition unity.83 When the opposi-
tion camp is fragmented, the marginal effect of "Opposition Size" on anti-government
protests does not reach statistical significance. However, when the "Opposition Unity"
is larger than 0.8, namely, the effective number of opposition parties is lower than
1.25, the marginal effect of "Opposition Size" on anti-government protests becomes
positive and statistically significant. Overall, this result supports the hypothesis of
opposition mobilization capacity, suggesting that a large opposition camp encourages
more anti-government protests only if this camp is more united.

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Comparative Politics January 2015

Table 1 Zero-Inflated Negative Binomial Models for Anti-Government Protests in

Democratic Countries (1990-2004)

Variables Model 1 Model 2 Model 3 Model 4

7 ; 0.031 0.068 0.002
Party system frag
„ 0.286 0.128 0.061
„ Parliamentary system (0 2gl) (0 256) (() ]()4)
-0.846*** -0.706** -0.045
PR system (0.314) (0.286) (0.123)
. 0.202 0.177 -0.072
Leftist . government (0 J69) (0 ]46) (0 0g7)
Percentage of Opposition Party Members 0.004 -0.003 -0.004
in the Legislature (0.007) (0.011) (0.007)
, , . -0.069 -0.080 -0.111
General , election , . year {Q m) (0 107) (0 J0())
-0.080 -0.109 -0.071
Honeymoon year (0 og3) (0.090) (0.102)
-0.030*** -0.045*** -0.010
Opposition size (0.011) (0.015) (0.009)
-4.815*** -6.396*** -2.204**
Opposition unity (ļ 399) (ļ 6ļļ) (0 g99)
0.122*** 0.154*** 0.044**
Opp. size*Opp. Unity (0 032) (Q ^ m)
. J -0.082**
Freedom House index,./ . J ^

GDP growth,./ (0 013)


Inflation,., (In) (0.047)

GDP per capita,./ (In) (0.067)
Population (In) (0.059)

Anti-government protest,./ ^
Year Dummies No No No Yes
1.358*** 2.568*** 3.364*** -4.717***
Constant (0.489) (0.566) (0.818) (1.135)
N 1,335 1,335 1,335 1,274
Wald Chi-square 15.96 15.97 37.40 631.63
Prob > Chi-square 0.026 0.001 0.000 0.000
Alpha 2.587*** 2.470*** 2.310*** 0.523***
Vuong Test

Robust standard error in parenthesis: *

Note: Reported estimates are coefficients
predicts the probability that a country ne
same set of independent variables that app
available in an online appendix.

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Yen-Pin Su

Figure 1 Marginal Effects of Opposition Size at Different Levels of Opposition Unity

Ö ~


Q> S "
s §:
» § i- o » o < »

!i: ° il" ° " "

1 gl 9 I
LU 9

I 8 I
Jļ 9 -

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Opposition Unity

Note: The dots are point estimates and t

The coefficient of "Opposition S

Model 3, while the coefficient of "O
Models 2, 3, and 4. The interpretati
of the inclusion of the interaction te
change of "Opposition Size" has a neg
equals zero, and a positive change
protests when "Opposition Size" e
the results suggests that a country
to experience fewer anti-governme
united. Moreover, a country with a
experience fewer anti-government p
Many variables based on the first i
nificance.First, the levels of party sy
tests in my sample. Second, whether
significantly affect the frequency o
that the frequency of anti-governm
there are elite allies in the governm
and "Honeymoon Year" do not matte
In Model 4, some control variables
protests while others do not. For in

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Comparative Politics January 2015

of economic development tends to experience more protests. In addition, a country with a

higher level of democratization is likely to experience fewer anti-government protests,
which implies that a more democratized country provides more institutionalized chan-
nels for articulating popular demands and thus discourages anti-government protests.
Surprisingly, none of the indicators regarding national economic performance
attain statistical significance, suggesting that the frequency of anti-government protests
is not greatly influenced by short-term economic conditions in my sample. In Model 4,
both "Population" and "Anti-Government Protest^" are positive and statistically sig-
nificant. While the former indicates that larger countries tend to experience more
anti-government protest mobilization, the latter suggests that a country would expe-
rience more anti-government protests if there were more anti-government protests in
the previous year.
The coefficients of a negative binomial regression are difficult to interpret because
of the log-linear nature; therefore, I use the exponential calculation to discuss the
substantive effects of opposition mobilization capacity on protests. Table 2 presents
the predicted number of anti-government protests derived by exponentiating the sum
of the product of each independent variable's estimated coefficient and a chosen
value based on the count stage of Model 4. With other continuous variables held
at their means and dichotomous variables held at their modes,84 the predicted counts
of "Anti-Government Protests" are computed for different levels of "Opposition
Size" and different levels of "Opposition Unity." In order to understand whether a
higher level of unity helps a larger opposition camp to mobilize more protests, I set
the chosen values for opposition size as the entire opposition camp obtains 50 per-
cent, 55 percent, and 60 percent popular support, respectively. The chosen values for
opposition unity are set 0.125, 0.5, and 1. Put differently, these three different levels
of opposition unity indicate that the opposition camp is composed of eight, two, and
one effective opposition parties, respectively.

Table 2 Predicted Number of Anti-Government Protests for Different Levels of

Opposition Mobilization Capacity

Levels of Opposition Unity 50% 55% 60%

0.125 (8 opposition parties) 1.07 1.05 1.03
0.5 (2 opposition parties) 1.08 1.14 1.21
1 (1 opposition party)

Note: The predict

product of each
of Model 4 in Table 1 .

Table 2 demonstrates that whether an opposition camp that received sizable popular
support is able to mobilize more anti-government protests depends on how united
the opposition camp is. For a large opposition camp (capturing 60 percent of popular
support) of eight opposition parties, the annual predicted number of anti-government

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Yen-Pin Su

protests is 1.03, while the predicted number of protests increases to 1.21 when there are
only two parties in the opposition camp. When there is only one opposition party that
captures 60 percent of popular support, the predicted number of protests increases to
1.52. Put it differently, the results indicate that , holding all other variables constant,
the predicted number of anti-government protest events increases by 48 percent
as the opposition camp moves from a low level of unity to a high level of unity. In
addition, the results show that for an opposition camp that captures 55 percent of
popular support, the predicted number of anti-government protest events increases
22 percent more as the opposition camp moves from a low level of unity to a high level
of unity, holding all other variables constant.

Robustness Checks Table 1 has provided strong evidence to support the second
institutional explanation: "Opposition Size*Opposition Unity" has a positive and sig-
nificant effect on anti-government protests. To check the robustness of this finding,
I conduct several re-estimations based on the specification of Model 4 in Table 1 (results
available on request). First, to test whether my findings are driven by the cutoff
point choice that excludes from the analyses the observations that are not democratic
countries, I use different cutoff points (Polity IV score of 4 and Polity IV score of 6)
to re-select observations. The results in either sample remain substantively unchanged.
In addition, recall that the opposition mobilization capacity variables included in
the empirical tests are constructed based on electoral results of parties that received
more than 0.5 percent of votes. To test whether the statistically significant result
of "Opposition Size*Opposition Unity" is driven by this specific operationalization,
I re-construct "Opposition Size" and "Opposition Unity" by considering all of the vote
share that is not captured by the governing party or parties. Given that most electoral
data lump tiny parties and independents into an "Other" category, I consider this Other
category as another opposition party. The re-estimated results using the re-constructed
explanatory variables remain virtually unchanged.

Opposition and Protests: Developed Countries vs. Developing Countries So far

the statistical results in Table 1 have shown that opposition mobilization capacity matters
for democratic countries in general. However, as I have discussed before, the patterns
of opposition mobilization and protests might be different for developed countries and
developing countries. As Table 3 demonstrates,85 "Opposition Size*Opposition Unity"
has no significant effect on anti-government protests in the sample of developed countries
(Model 5). This finding indicates that in developed countries, anti-government protests
can be higher or lower regardless of the level of opposition mobilization capacity. In
contrast, "Opposition Size*Opposition Unity" has a positive and statistically significant
effect on anti-government protests in developing countries (Model 6). This finding sug-
gests that in developing countries, whether a large opposition camp is able to mobilize
more anti-government protests depends on the level of opposition unity. Overall, the
mobilization capacity of opposition parties matters in explaining the frequency of anti-
government protests in developing countries but not in developed countries.

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Comparative Politics January 2015

Table 3 Zero-Inflated Negative Binomial Models for Anti-Government Protests in

Developed Countries and Developing Countries (1990-2004)

Model 5 Model 6
Variables (Developed (Developing
Countries) Countries)
7 ; ÕÕ43 0.020
Party system fragmentation,., (0 055) (0 Q5())
„ -1.396** 0.273
„ Parliamentary system (0 62?) (0 320)
-0.307 -0.583*
PR system (0.435) (0.340)
T 0.217 -0.156
T Leftist government (0 ]65) (0 2]6)
Percentage of Opposition Party Members -0.040** 0.023*
in the Legislature (0.016) (0.012)
, , . -0.097 -0.040
General , election , . year (0 17?) (0.l24)
-0.201 -0.040
Honeymoon year (0 ]43) (0 J25)
0.033 -0.052***
Opposition size {Q m) (0.0,6)
_ . -1.202 -5.472***
Opposition _ unity . (3.939) (1.885)
. ^ _ _ 0.048 0.088**
Opp. size*Opp. . ^ _ Unity _ (0 0?6) (0 Q42)
2.846 3.356***
Constant (1.766) (0.976)
N 344 991
Wald Chi-square - 24.67
Prob > Chi-square - 0.006
Alpha 0.973*** 2.439***
Vuong Test

Robust standard
Note: Reported est
predicts the proba
same set of indepe
available in an on

There are also

in Table 3. Fir
protests in dev
in developed c
reduce anti-go
for protests in
is the effect o
tion parties in

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Yen-Pin Su

developed countries. However, in developing countries the frequency of anti-government

protests increases with the number of legislators of opposition parties.


While previous literature has examined how the presence of protest is influenced
various factors, such as democratization and national economic performance, this st
argued that political institutions and opposition parties are crucial factors that sh
protests in democracies.86 Using a global sample of democratic countries, this stud
empirically tested different institutional explanations on anti-government protes
I found strong empirical support for a theoretical innovation: a large opposition cam
is able to encourage more anti-government protests only if it has a higher level o
unity. The results additionally indicate that opposition mobilization capacity matt
for protests in developing countries but not in developed countries.
This study not only facilitates a better understanding of the dynamics of soci
protest and party politics, but also offers insights for social movement activ
to craft effective strategies for interacting with political parties. The results sug
that merely having a large opposition camp is insufficient for mobilizing large-sc
anti-government protests. A unified opposition with strong support is a key to pr
ducing a larger number of protests. Moreover, this research has important implicat
for democracy and representation. For example, given that groups with effective p
tisan representation tend to work through the legislative process, but groups exclud
from the legislative process tend to turn to street protests, it is important that po
makers design institutions that better connect parties and movements.
Last, but not least, this study points to some directions for future research. F
instance, it will be interesting to test whether the institutional explanations of prot
can travel to different levels of analysis (e.g., subnational level). It is also importa
to analyze how opposition parties with different political ideologies affect differe
types of protest actions.87 Moreover, while this study focuses on the interaction eff
of opposition size and unity on the frequency of anti-government protests, futur
studies should also explore how such effects influence other features of protest acti
(e.g., size and duration) or certain policy outcomes.


I thank Steven Finkel, John Markoff, Scott Morgenstern, Aníbal Pérez-Liñán, Ya-Wen
Zigerell, and the journal's anonymous referees for their helpful comments. All remaining errors

1. Aníbal Pérez-Liñán, Presidential Impeachment and the New Political Instability in La

(Cambridge: Cambridge University Press, 2007).
2. Anna Bosco and Susannah 'ferney, "Electoral Epidemic: The Political Cost of Econo
Southern Europe, 2010-11," South European Society and Politics , 17 (June 2012), 129-54.

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Comparative Politics January 2015

3. Jack A. Goldstone, "More Social Movements or Fewer? Beyond Political Opportunity Structures to
Relational Fields," Theory and Society, 33 (June- August 2004), 333-65.
4. Patricia L. Hipsher, "Democratization and the Decline of Urban Social Movements in Chile and
Spain," Comparative Politics , 28 (April 1996), 273-97.
5. John Walton and Charles Ragin, Global and National Sources of Political Protest: Third World
Responses to the Debt Crisis," American Sociological Review ; 55 (December 1990), 876-90.
6. Moisés Arce, "Parties and Social Protest in Latin America's Neoliberal Era," Party Politics ,
16 (April 2010), 676.
7. Kenneth A. Shepsle, "Studying Institutions: Some Lessons from the Rational Choice Approach,"
Journal of Theoretical Politics , 1 (April 1989), 131-47.
8. Aníbal Pérez-Liñán, "Neoinstitutional Accounts of 'foter Turnout: Moving beyond Industrial
Democracies," Electoral Studies , 20 (June 2001), 282.
9. Peter A. Hall and Rosemary C. R. Taylor, "Political Science and the Three New Institutionalisms,"
Political Studies, 44 (December 1996), 936-57.
10. B. Guy Peters, Institutional Theory in Political Science : The New Institutionalism , 3rd ed. (New York:
Continuum, 2012), 155; see also Kaare Strom, "A Behavioral Theory of Competitive Political Parties,"
American Journal of Political Science, 34 (May 1990), 565-98.
11. David S. Meyer, "Protest and Political Opportunities," Annual Review of Sociology, 30 (2004), 125-45.
Nella 'àn Dyke, "Protest Cycles and Party Politics: The Effects of Elite Allies and Antagonists on Student
Protest in the United States, 1930-1990," in Jack A. Goldstone, ed., States, Parties and Social Movements
(Cambridge: Cambridge University Press, 2003), 226-45.
12. John D. McCarthy and Mayer N. Zald, "Resource Mobilization and Social Movements: A Partial
Theory," American Journal of Sociology, 82 (May 1977), 1212-41. Paul Almeida, "Social Movement Partyism:
Collective Action and Oppositional Political Parties," in Nella 'àn Dyke and Holly J. McCammon, eds., Stra-
tegic Alliances: Coalition Building and Social Movements (Minneapolis: University of Minnesota Press, 2010),
13. Elmer E. Schattschneider, Party Government (New York: Holt, Rinehart and Winston, 1942);
Giovanni Sartori, Parties and Party Systems: A Framework for Analysis (Cambridge: Cambridge University
Press, 1976).
14. Arce, 2010.
15. Scott Mainwaring and Timothy R. Scully, "Introduction: Party Systems in Latin America," Building
Democratic Institutions: Party Systems in Latin America (Stanford: Stanford University Press, 1995), 5.
16. Ibid., 23.
17. Arce, 2010, 671-72.
18. Ibid.
19. Moisés Arce and Roberta Rice, "Social Protest in Post-Stabilization Bolivia," Latin American
Research Review, 44 (2009), 88-101.
20. $. Ilgii Ozler, "Political Institutions and Protest: A Comparative Analysis,' Representation , 49 (2013),
135-54; Tavishi Bhasin, Democracy and Dissent: Explaining Protest and State Response, Ph.D. Dissertation
(Emory University, 2008).
21. Bhasin, 58.
22. Sidney Tarro w, Power in Movement: Social Movements and Contentious Politics, 3rd ed. (Cambridge:
Cambridge University Press, 2011); '&n Dyke.
23. Jai Kwan Jung, "Disentangling Protest Cycles: An Event-History Analysis of New Social Movements
in Western Europe," Mobilization, 15 (February 2010), 25-44.
24. J. Craig Jenkins, David Jacobs, and Jon Agnone, Political Opportunities and Atrican-Amencan
Protest, 1948-1997," Mobilization, 109 (September 2003), 277-303; '&n Dyke, 240-41.
25. Hanspeter Kriesi, "The Political Opportunity Structure ot New Social Movements: Its Impact on
Their Mobilization," in J. Craig Jenkins and Bert Klandermans, eds., The Politics of Social Protest:
Comparative Perspectives on States and Social Movements (Minneapolis: University of Minnesota Press,
1995), 167-98.
26. Kathleen Bruhn, Urban protest in Mexico and Brazil (Cambridge: Cambridge University Press, 2UUo).
27. David Meyer and Debra Minkoff, "Conceptualizing Political Opportunity, Social Forces, 82
(June 2004), 1472.
28. Bruhn, 6-7.
29. Adrienne LeBas, From Protests to Parties : Party-Building and Democratization in Africa (Oxford:
Oxford University Press, 2011); Diarmuid Maguire, "Opposition Movements and Opposition Parties: Equal


This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Yen-Pin Su

Partner or Dependent Relations in the Struggle for Power and Reform?" in J. Craig Jenkins and Bert
Klandermans, eds., 199-297.
30. Maguire, 199.
3 1 . Jack A. Goldstone, "Introduction: Bridging Institutionalized and Noninstitutionalized Politics, in Jack
A. Goldstone, ed., States, Parties, and Social Movements (Cambridge: Cambridge University Press, 2003),
32. Goldstone, "Introduction," in Jack A. Goldstone, ed.; see also Paul Almeida, "Opportunity Organi-
zations and Threat-Induced Contention: Protest Waves in Authoritarian Settings," American Journal of
Sociology ; 109 (September 2003), 350.
33. Thomas R. Rochon, Mobilizing for Peace: the Antinuclear Movements in Wèstern Europe (Princeton:
Princeton University Press, 1988).
34. Paul Almeida, "Subnational Opposition to Globalization," Social Forces , 90 (June 2012), 1051-72.
35. Carew E. Boulding, "NGOs and Political Participation in Weak Democracies: Subnational Evidence
on Protest and '6ter Turnout from Bolivia," Journal of Politics, 72 (April 2010), 464.
36. Lebas, 26.
37. Studies have demonstrated the importance ot opposition unity to mobilize protests, ror example,
Florian Bieber, "The Serbian Opposition and Civil Society: Roots of the Delayed Transition in Serbia,"
International Journal of Politics, Culture , and Society ; 17 (Fall 2003), 73-90.
38. Martin Tanaka, Democracia sin partidos, Perú, 2000-2005 (Lima: Instituto de Estudios Peruanos, 2005).
39. Steven Levitsky and Maxwell A. Cameron, "Democracy without Parties? Political Parties and Regime
Change in Fujimori's Peru," Latin American Politics and Society 45 (Autumn 2003), 15-16.
40. Moises Arce, The Repoliticization of Collective Action After Neoliberalism in Peru," Latin
American Politics and Society, 50 (Fall 2008), 40-41.
41. David S. Meyer and Sidney Tarrow, "A Movement Society: Contentious Politics for the New
Century," in David S. Meyer and Sidney Tarrow, eds., The Social Movement Society: Contentious
Politics for the New Century (Lanham, MD: Rowman and Littlefield, 1998), 1-28.
42. Pippa Norris, Democratic Phoenix: Reinventing Political Activism (Cambridge: Cambridge University
Press, 2002), 2 1 1 ; Ronald Inglehart and Christian Welzel, Modernization, Cultural Change, and Democracy:
The Human Development Sequence (Cambridge: Cambridge University Press, 2005), 122-23; Russell J.
Dalton, Citizen Politics: Public Opinion and Political Parties in Advanced Industrial Democracies
(Washington, D.C.: CQ Press, 2008), 52.
43. Dieter Rucht and Friedhelm Neidhardt, "Towards a 'Movement Society'? On the Possibilities of
Institutionalizing Social Movements," Social Movement Studies, 1 (2002), 7-30; Tarrow, 111-17.
44. Dalton, 48-52.
45. Ronald Inglehart, Modernization and Postmodernization: Cultural, Economic, and Political Change
in 43 Societies (Princeton: Princeton University Press, 1997).
46. Inglehart and Welzel, 293.
47. Ibid., 123-25.
48. Ibid., 294; see also Norris, 23.
49. Alice Holmes Cooper, "Party-Sponsored Protest and the Movement Society: The CDU/CSU
Mobilises against Citizenship Law Reform," German Politics, 11 (August 2002), 88-104.
50. Patrick Barrett, "An Interview with Ben Manski of Wisconsin Wave," in Mari Jo Buhle and Paul
Buhle, eds., It Started in Wisconsin: Dispatches from the Front Lines of the New Labor Protest (London:
'ferso, 2011), 67.
51. J. Craig Jenkins, Michael Wallace, and Andrew S. Fullerton, "A Social Movement Society? A Cross-
National Analysis of Protest Potential," International Journal of Sociology, 38 (Fall 2008), 15.
52. Dalton, 50.
53. Almeida, 2010.
54. Due to space constraints, the Appendix is not in the print version of this article. It can be viewed in
the online version, at
55. Monty G. Marshall, Keith Jaggers, and Ted R. Gurr, Polity IV Project: Political Regime Charac-
teristics and Transitions, 1800-2010 (College Park: Center for International Development and Conflict
Management, University of Maryland, 20 1 1 ).
56. This cutoff point follows Stephen L. Quackenbush and Michael Rudy, "Evaluating the Monadic
Democratic Peace," Conflict Management and Peace Science, 26 (July 2009), 268-85. The Polity IV
database lacks entries for the Bahamas, Barbados, Belize, Cape 'èrde, Grenada, Iceland, Luxembourg, Malta,
Saint Lucia, St. Vincent and the Grenadines, Samoa, Sao Tome and Principe, Suriname (2000-2004), and


This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Comparative Politics January 2015

'&nuatu in my sample. However, I include these cases in the analyses because they are categorized as
"free" countries by Freedom House (
57. I acknowledge that protest event also varies in many other aspects, such as size and duration. These
properties of protests may be quantified for particular research goals. Due to my research design and
theoretical concerns, nevertheless, the statistical tests will focus on the frequency of anti-government protests.
58. Gary King and Will Lowe, "10 Million International Dyadic Events," available at http://hdl.handle.
net/1902. 1/FYXLAWZRIA UNF:3:dSE0bsQK2o6xXlxeaDEhcg== IQSS Dataverse Network [Distributor] V3
['èrsion]; See also Gary King and Will Lowe, "An Automated Information Extraction Tool For International
Conflict Data with Performance as Good as Human Coders: A Rare Events Evaluation Design," International
Organization , 57 (Summer 2003), 617-42.
59. Another possible data source is Arthur S. Banks and Kenneth A. Wilson, Cross-National Time-Series
Data Archive (Jerusalem, Israel: Databanks International, 2012), which relies on the New York Times. How-
ever, I do not use this database because it has been criticized for its lack of information completeness and
arbitrary coding criteria for protest events. See Taehyun Nam, "What You Use Matters: Coding Protest
Data," PS: Political Science and Politics , 39 (April 2006), 281-87.
60. I consider nine types of domestic governmental taigets defined in King and Lowe s codebook: the
national executive, the judiciary, legislators, legislations, government agents, national government officials,
subnational government officials, the military, and the police.
61. I consider seven types of events as protest activities denned in King and Lowe s codebook: sit-ins,
hunger strikes, protest demonstrations, picketing, defacing properties for protest, labor strikes, and riots.
62. Roberto Franzosi, "The Press as a Source of Socio-historical Data: Issues in the Methodology of Data
Collection from Newspapers," Historical Methods , 20 (Winter 1987), 5-16.
63. Another potential problem of relying on the information from Reuters is that international news media may
generally pay more attention to report news events that occurred in geopolitically more important countries.
64. I thank an anonymous reviewer for this point.
65. Party system fragmentation is calculated as the inverse of the sum of squares of each party s vote
share in an election, following Markku Laakso and Rein Taagepera, "Effective Number of Parties: a Mea-
sure with Application to West Europe," Comparative Political Studies, 12 (April 1979), 3-27. The data for
this variable are mainly from Michael Gallagher and Paul Mitchell, eds., The Politics of Electoral Systems
(Oxford: Oxford University Press, 2008), coupled with my own calculation for the observations missing
from Gallapher and Mitchell's dataset.
66. The data of these variables are from Thorsten Beck, George Clarke, Alberto Gron, Philip Keeter, and
Patrick Wàlsh, "New Tools in Comparative Political Economy: The Database of Political Institutions," Wbrld
Bank Economic Review, 15 (2001), 165-76. "Parliamentary System" and "Leftist Government" are dummy var-
iables. "PR system" is coded 1 if the majority of legislative seats is elected using a PR system, and 0 otherwise.
67. In parliamentary systems, an opposition party is a party that was not included in the cabinet. In
presidential systems, an opposition party is a non-presidential party or a party that was excluded from
the coalition government formed by the president. In semi-presidential systems, an opposition party is a
party that is not the presidential party or a party in the cabinet.
68. This variable and the variable of Percentage of Opposition Party Members in the Legislature capture
different theoretical concepts, and they are not perfectly collinear. The correlation coefficient of these two
variables is 0.73 in my dataset.
69. Ko Maeda, "Divided We Fall: Opposition Fragmentation and the Electoral rortunes ot Governing
Parties," British Journal of Political Science , 40 (April 2010), 419-34.
70. In this study, I consider developed countries to be the high-income OLCD countries, denned by the
World Bank (http://data.worldbank.Org/about/country-classifications/country-and-lending-groups#OECD_
members). I exclude the Czech Republic, Estonia, Hungary, Israel, South Korea, Poland, Slovakia, and
Slovenia from the group of developed countries because these eight countries joined the OECD more
recently, and most of them (except Israel) are new democracies. The developed countries in my dataset
include: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland,
Italy, Japan, Luxembourg, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the
United Kingdom, and the United States of America.
71. The sources include: Adam Carr, Adam Carrb Election Archive , available at http://psephos.adam-carr.
net/; Albert C. Nunley, African Elections Database , available at; Daniel
Bochsler, Territory and Electoral Rules in Post-Communist Democracies (Houndmills: Palgrave, 2010);
Dieter Nohlen, ed., Elections in the Americas: A Data Handbook, vols. 1-2 (New York: Oxford University
Press, 2005); Dieter Nohlen, Bernard Krennerich, and Michael Thibaut, Elections in Africa: A Data

This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to
Yen-Pin Su

Handbook (Oxford: Oxford University Press, 1999); and Dieter Nohlen, Florian Grotz, and Christof
Hartmann, Elections in Asia and the Pacific: A Data Handbook (Oxford: Oxford University Press,
2001); Electoral Institute for the Sustainability of Democracy in Africa, Election Archive, available at; Holger Döring, "The Collective Action of Data Collection: A Data
Infrastructure on Parties, Elections and Cabinets," European Union Politics , 14 (March 2013), 161-78,
available at; Inter-Parliamentary Union, PARLINE Database on National Parlia-
ments , available at; and Norwegian Social Science Data Ser-
vices (NSD), European Election Database , available at
(Note: NSD are not responsible for the analyses/interpretation of the data presented here). I use data on
legislative seats for countries where the vote data are unavailable; these countries include Benin ( 1 999 elec-
tion), Central African Republic, Guinea-Bissau (1999 election), Haiti, Madagascar, Niger (1995, 1996, and
1999 elections), and the Solomon Islands.
72. I do not consider independent candidates because this study focuses on examining how opposition
parties affect anti-government protests.
73. For instance, the values of the opposition mobilization capacity variables calculated using the
2000 Mexican legislative elections are assigned for the observations of Mexico-2001, Mexico-2002,
and Mexico-2003.
74. The logged inflation rate is used for preventing hyperinflation cases from skewing the resu
Following Marcus J. Kurtz and Sarah M. Brooks, "Embedding Neoliberal Reform in Latin Amer
Wbrld Politics , 60 (January 2008), 231-80, I assign 1 to the observations with an inflation rate low
than 1 so that the logged inflation value for these observations is 0.
75. The data for these economic indicators are from International Monetary Fund, Wbrld Econom
Outlook Database , available at
76. I use the Freedom House Index data from Steven E. Finkel, Aníbal Pérez-Liñán, and Mitchell A
Seligson, "The Effects of US Foreign Assistance on Democracy Building, 1990-2003," Wbrld Polit
59 (April 2007), 404-39. The value of the Index ranges from 1 to 13, with a higher value indica
that a country has a higher level of democratization.
77. The population data of all countries except Taiwan are from World Banks, Wbrld Developm
Indicators, available at Taiw
population data are from IMF's World Economic Outlook Database.
78. William H. Greene, Econometric Analysis (New York: Prentice Hall, 2012), 805-07; J. Scott Long, Reg
sion Models for Categorical and Limited Dependent Variables (Thousand Oaks, CA: Sage, 1997), 236-3
79. Quang H. Vuong, "Likelihood Ratio Tests for Model Selection and Non-Nested Hypothese
Econometrica, 57 (March 1989), 307-33.
80. In general, zeros in my protest event count dataset might be either a country that did not experie
anti-government protests in this given sample, or a country that never experiences any anti-governm
protests at all by nature because of country characteristics such as a small population or particular cultu
factors. If both groups are present in a sample, estimating a Poisson or Negative Binomial model w
overestimate the theoretical probability of zero (Konstantinos Drakos and Andreas Gofas, "In Search of t
Average Transnational Terrorist Attack 'fcnue," Defense and Peace Economics, 17 (2006), 81).
8 1 . Ideally, I should control for the country-specific unobserved heterogeneity. However, the model
not converge with the inclusion of country dummies.
82. I have conducted an Arellano-Bond test of autocorrelation on all the models using Stata's
module after estimating regression models for each of the four models in Table 1. All the tests for
(1) do not reach statistical significance.
83. The conditional coefficients and their associated confidence intervals are calculated using the lin
command in Stata.
84. The mode for general election year, honeymoon year, leftist government, and year dummies is 0,
while the mode for parliamentary system and PR system is 1.
85. I conduct the analyses using Model 3 instead of Model 4 for two reasons. First, substantively, it
makes sense to focus on testing the two proposed institutional approaches for these subsamples for the
purpose of this paper. Second, the model does not converge in Stata when Model 4 is estimated on the
subsample of developing countries and the subsample of developed countries.
86. Doug McAdam and Sidney Tarrow, "Ballots and Barricades: On the Reciprocal Relationship between
Elections and Social Movements," Perspectives on Politics, 8 (June 2010), 529-42.
87. I thank an anonymous reviewer for this point.


This content downloaded from on Tue, 16 Jan 2018 10:20:12 UTC
All use subject to