Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
Share It or Lose It? Authoritarian Power-Sharing
and Regime Stability in Multi-Ethnic States.
Yannick I. Pengl1 and Ilyas Saliba2,3
1International Conflict Research, ETH Zurich2Berlin Graduate School of Social Sciences, Humboldt University
3Department Democracy and Democratization, Berlin Social ScienceCenter (WZB)
June 17, 2015
Abstract
In this paper, we examine the impact of ethnic cooptation on authoritar-ian stability. We argue theoretically and demonstrate empirically that theinclusion of elites from politically relevant ethnic groups significantly reducesthe risk of authoritarian breakdown. We also showhow this e↵ect varies acrossdi↵erent forms of regime change and type. While a prominent explanatoryfactor in conflict research, inter-ethnic competition for state power has sofar been understudied within the literature on authoritarian regime survival.Although accounts of cooptation strategies have recently gained prominence,most remain structuralist and focus on formal institutions such as authori-tarian parties, parliaments and elections that facilitate cooperation between adictator and coopted elites. We go beyond institutionalist analyses and shiftthe attention to the identity of those who need to be coopted, the degree towhich cooptation is actually achieved and the contexts in which cooptationworks best. We present three main findings: 1.) Ethnic cooptation stabilizesauthoritarian regimes. 2.) It reduces the risk of irregular and autocratic butnot of regular and democratizing forms of regime change. 3.) It is most e↵ec-tive where formal institutions are weak, namely in personalist dictatorships.
1
Introduction
In this paper, we examine how ethnic cooptation strategies a↵ect authoritarian
stability. While cooptation in general has figured quite prominently in the recent
comparative authoritarianism literature, strategies that address specifically ethnic
challenges to regime survival remain understudied. Most recent work on the role
of cooptation in authoritarian regimes focus on how formal, often semi-democratic
institutions such as regime parties, authoritarian legislatures, and more or less com-
petitive elections tie in potential challengers and facilitate monitoring and commit-
ment among the dictator and his ruling elite (see e.g. Svolik, 2012; Gandhi and
Przeworski, 2006).1
While providing novel insights in the sources of authoritarian stability, analyses of
formal cooptation institutions often remain quite detached from the actors involved.
As Pepinsky (2014) has recently argued “the institutionalist turn in comparative
authoritarianism” may all to liberally gloss over how institutions are shaped by the
preferences of and power balance among the most important political agents. In
this vein, formal institutions undoubtedly provide opportunities for the cooptation
and inclusion of potential regime challengers. However, it remains opaque who
these challengers are and whether this cooptation potential is realized. Only by
identifying the most relevant social actors and gauging the extent to which they
are coopted one can judge whether formal institutions actually perform the task
they are assumed to perform in the authoritarian institutions literature. Shifting
the attention to relevant social groups and their representation in political regimes
thus seems promising to improve our theoretical understanding and empirical tests
of cooptation mechanisms in authoritarian regimes.
In this context, we regard politicized ethnicity as particularly relevant to the
question of authoritarian stability. Where political competition is structured along
ethnic lines, included and excluded groups struggle for material resources and po-
litical power, quite frequently by violent means (Olzak, 1983; Horowitz, 1985; Ce-
derman, Gleditsch and Buhaug, 2013). If ethno-political competition is linked to
political instability in the form of violent conflict, it may also matter in determining
the stability and instability of authoritarian regimes. Focusing on ethno-political
power struggles between included and excluded groups thus allows us to move one
step closer to the actual locus of agency than has been the case in previous analyses
of formal institutional contexts that may or may not serve cooptation purposes. In
short, we see ethnic elites excluded from the regime as especially important potential
challengers. Thus, ethnically targeted cooptation strategies may be needed to foster
1Other relevant studies highlighting the role of formal cooptation institutiuons include Levitskyand Way (2002), Levitsky and Way (2010), Lust-Okar (2004), Brownlee (2007), and Wright andEscriba-Folch (2012).
2
authoritarian stability.
In further developing and testing this argument, we borrow concepts and opera-
tionalizations from seminal work on ethnic conflict (Cederman, Wimmer and Min,
2010) as well as from the literature on neopatrimonialism (Bratton and Van de Walle,
1994) and apply them to the study of autocratic resilience. Our approach provides
a novel and we believe more suitable and actor-centered way of modeling the role
of cooptation in authoritarian regimes. In what follows, be briefly summarize the
theoretical building blocks that we borrow from the previous literature before we
turn to the main theoretical part of the paper.
Theoretical Building Blocks
Let us briefly discuss the most important concepts and insights from the three
literatures upon which we base our theoretical argument.
Comparative Authoritarianism: First and as already discussed above, com-
parative authoritarianism research provides the most natural point of departure for
our study. In line with recent work from this strand of literature, we regard coop-
tation as an essential pillar of authoritarian stability (Gerschewski, 2013; Wintrobe,
1998). The other most important pillars are repression and legitimacy (Gerschewski
et al., 2013). Cooptation is often defined as the strategic targeting of resources, gov-
ernment posts and policy concessions to potential regime challengers (Gandhi and
Przeworski, 2007). The aim is to buy their loyalty and provide them with a vested
stake in the continued survival of the regime (Frantz and Kendall-Taylor, 2014).
Authoritarian institutions may facilitate this in two main ways. First, they provide
posts and forums, in which coopted elites can be placed (Malesky and Schuler, 2010;
Schedler, 2009; Gandhi and Przeworski, 2006). Second, they reduce uncertainty and
enhance the credibility of authoritarian powersharing arrangements. Formalized
procedures and repeated interaction within legislative bodies and regime parties are
expected to alleviate ruling coalition members’ fears of a dictator defecting from
a powersharing deal (Geddes, 1999; Brownlee, 2007; Boix and Svolik, 2013; Maga-
loni, 2008). Recent quantitative analyses indicate that formal institutions such as
authoritarian legislatures and regime parties e↵ectively reduce the risk of at least
some forms of regime failure (Wright and Escriba-Folch, 2012; Boix and Svolik,
2013).
We see two main reasons why getting a more fine-grained understanding of au-
thoritarian cooptation strategies requires moving beyond formal institutions. First,
the mere absence or presence of specific institutions does not tell much about the
degree to which cooptation is actually achieved. Even in the absence of authori-
3
tarian legislatures and parties, potential challengers may be coopted through more
informal channels, as e.g. the literature on neopatrimonialism suggests. Second,
most institutionalist analyses do not identify the most relevant social actors that
need to be coopted. In many authoritarian regimes, ethnicity is one societal fault
line along which political competition takes place and challenges to the incumbent
regime are organized.
Neopatrimonialism: Referring back to Max Weber’s work, the concept of pat-
rimonialism re-entered mainstream political science literature in the 1970s and has
mainly been applied to the Global South in general and Sub-Saharan African polit-
ical systems in particular (see e.g. Lemarchand, 1972). In his seminal contribution
Eisenstadt (1973, p. 14) concludes that “patterns of political organization” that
are partially yet insu�ciently penetrated by institutionalized formal relationships
among political elites result in authoritarianism and clientelism.
In similar vein, Clapham (1996, p. 48) defines neopatrimonialism is “a form of
organisation in which relationships of a broadly patrimonial type pervade a political
and administrative system which is formally constructed on rational-legal lines.”
Authority is based on exchanging material benefits for loyalty through largely in-
formal clientelist networks that range from the very top to the bottom of political
hierarchy (Jackson and Rosberg, 1984; Bratton and Van de Walle, 1994). Patron-
age, clientelism, nepotism, and corruption are thus seen as characteristic features
of neopatrimonial regimes (Erdmann and Engel, 2007). In this sense, neopatrimo-
nial patronage politics represent informal forms of cooptation that may be just as
important as cooptation via formal, quasi-democratic institutions such as parties,
parliaments and elections. Despite their informality, neopatrimonial patron-client
exchanges “take place within the framework of, and with the claim to, legal-rational
bureaucracy or ‘modern’ stateness” (Erdmann and Engel, 2007, p. 105). Thus,
access to the state apparatus is key to secure a share of patronage rents.
Ethnic Politics: While the clientelist networks just described do not have to fol-
low ethnic lines, in practice they frequently do (Clapham, 1996). Wherever neopat-
rimonial or other forms of political relations are ethnically structured, intense com-
petition for controlling the state ensues (Bates, 1974; Olzak, 1983). Especially in
weakly institutionalized multi-ethnic developing countries, the state tends to be the
main source of economic rents and opportunities (Wimmer, 1997).Those in power
may channel public spending to their ethnic constituency and only grant selective
access to positions in government, the bureacracy or the armed forces.
Recent empirical work shows that ethnic favoritism is a common theme, espe-
cially so in Sub-Saharan Africa (Franck and Rainer, 2012; Kramon and Posner,
4
2012; Burgess et al., forthcoming). This may motivate challenges to the regime in
the name of groups that are excluded from state power. Cederman, Gleditsch and
Buhaug (2013) show how a lack of ethnic representation leads to a higher risk of eth-
nic civil war. In what follows, we outline how the cooptation of ethnic elites at the
executive level may a↵ect the risk of other, more common threats to authoritarian
regime stability.
Our Argument & Hypotheses
The brief review of the literature suggests how concepts and mechanisms from studies
on neopatrimonialism and ethnic politics may further improve our understanding of
cooptation in authoritarian regimes. Moving beyond the focus on formal institutions
seems key to more fully grasp the extent and e↵ectiveness of autocratic cooptation
strategies. The more informal character of neopatrimonial patronage networks as
well as the frequently observed structuring of political competition along ethnic lines
motivate our emphasis on ethnic elites and their inclusion at the executive level. In
short, we argue that ethnically inclusive cabinets indicate a greater extent of ethnic
cooptation and thus contribute to a reduced risk of regime failure.
Before we outline our reasoning behind this claim in greater detail, it needs to be
clarified what we refer to as autocratic regime breakdown and/or stability. Along the
lines of Geddes (2003, p. 44) Geddes, we conceive of a political regime as a “set of
formal and informal rules” determining who will rule, how the rulers will be chosen
and how the basic distributive decisions will be made. It is autocratic, if it comes into
existence in a non-democratic way, if it suspends democratic rules once achieving
power, or if it emerges from nominally competitive elections in which one or more
parties with realistic chances of success are banned (Geddes, Wright and Frantz,
2014, p. 317). Autocratic regimes survive as long as the given set of key formal and
informal rules remains in place. They are replaced by regular or irregular means.
Regular regime change occurs in two main ways. First, an incumbent dictator or the
ruling party may be defeated at the ballot box, comply with the electoral verdict, and
allow the election winners to assume o�ce. Second, the incumbent ruling coalition
may reform a regime to such an extent “that the identity of the group from which
leaders can be chosen or the group that can choose major policies changes” (Geddes,
Wright and Frantz, 2014, p. 318). Irregular regime change comes in the form of
coups, successful rebellions, popular uprisings, foreign invasions, or similar events.
The common characteristic of these forms of regime failure is the coerced and extra-
institutional nature of political change (ibid.). A further distinction can be made
according to whether authoritarian breakdown results in a democratic or autocratic
successor regime.
5
What role does ethnic cooptation or the lack thereof play in preventing autocratic
regimes from collapsing in general or in one of the specific ways just outlined? In
contexts where ethnic cleavages are present and salient, organized challenges to the
incumbent regime frequently emanate from specific ethnic groups and/or the elites
that (claim to) represent them. Ethnicity provides both opportunities and motiva-
tion to challenge the current rulers. As far as opportunities are concerned, ethnic
identities may serve as a powerful tool to mobilize support and organize collective
action. First, co-ethnics often live in geographically concentrated settlement areas
within a country and share a common language, religion or culture. This should
make it easier to organize politically in order to influence where public goods spend-
ing is targeted, what language policies are chosen and which religious or cultural
practices and symbols are allowed or even actively promoted (Hardin, 1997; Bates,
1974). Second, a shared language and common understandings and norms facilitate
communication and in-group cooperation (Habyarimana et al., 2007). Third, denser
social networks within than across ethnic groups facilitate the identification and pun-
ishment of potential defectors (Fearon and Laitin, 1996). Finally, in-group norms
of cooperation and reciprocity also contribute to a higher organizational capacity of
ethnically homogeneous regime challengers (Habyarimana et al., 2007).
Turning to the motivational side, a lack of access to the regime provides ethnic
elites and their followers with material and non-material incentives to change their
lot. In many developing-world autocracies, the state apparatus is the key source of
economic rents and opportunities (Olzak, 1983; Arriola, 2009). As clientelist net-
works in neopatrimonial systems are often structured along ethnic lines, securing
access to the pool of resources that is the executive apparatus is highly important
for actual and prospective ethnic patrons and their clients (Clapham, 1996). Less
tangible, yet arguably equally important are ethnonationalist ideologies making the
political domination by ethnic others appear particularly intolerable (Cederman,
Gleditsch and Buhaug, 2013). Thus, both hard material interest and ethnonation-
alist grievances motivate ethnic elites outside of the regime to challenge the existing
order. (Cederman, Wimmer and Min, 2010) have prominently shown that exclusion
from state power is associated with a higher risk of ethnic civil war. In principle,
however, both the opportunity and the motivation mechanisms may facilitate other
forms of regime challenges as well, regardless of whether they take the form of elec-
toral competition by an opposition party that appeals to specific groups, a popular
uprising, or a coup plot in the name of a particular group.
Against this backdrop, coopting ethnic elites into the ruling coalition turns
regime outsiders into insiders and makes them invested in the maintenance of the
current order (Geddes, 1999; Gandhi and Przeworski, 2006, 2007). Both the in-
dividual rents and opportunities that ethnic elites secure from holding o�ce and
6
their ability to distribute patronage toward and politically represent their ethnic
constituency depend on the further existence of the regime (Arriola, 2009; Magaloni
and Kricheli, 2010; Brancati, 2014). As Arriola (2009) has pointed out, granting
access to executive power, or more specifically the cabinet is arguably more im-
portant to credibly promise the continued flow of patronage than inviting outsiders
to join authoritarian parties or legislatures. Even where the access to patronage is
quite limited or policies are still far from what coopted elites and their constituencies
want, these elites tend to prefer trying to change the system from within over openly
challenging it in a risky gamble such as a rebellion or a coup attempt (Gandhi and
Przeworski, 2006). Thus, we expect ethnic cooptation at the executive level to be
an e↵ective strategy to increase the stability of authoritarian regimes. In contrast to
much of the institutionalist literature on authoritarian cooptation that highlights the
role of parties and parliaments in managing the power-sharing arrangement between
a dictator and his ruling coalition (see e.g. Svolik, 2012), our argument relates to
the previous step. Instead of fostering facilitating the monitoring and enforcement
of an already given power-sharing deal, ethnic cooptation determines the identity of
those included in power-sharing in the first place.
The need to coopt potential ethnic challengers depends on the political salience of
ethnicity and the number of politically relevant ethnic cleavages in a given society.
In line with Posner (2004, p. 852), we argue that what matters here are “the
groups that are actually doing the competition over policy, not the ones that an
ethnographer happens to identify as representing distinct cultural units.” Taken
together, this line of reasoning leads to our first testable hypothesis:
Hypothesis 1 The more politically relevant ethnic groups are represented at the
executive level of an authoritarian regime, the lower the likeli-
hood of regime failure.
Quite obviously, this general hypothesis might gloss over important di↵erences
between di↵erent types of regime failure as well as di↵erent authoritarian regimes.
In fact, some accounts of authoritarian survival strategies have questioned the un-
ambiguously stabilizing e↵ects of cooptation (see e.g. Brancati, 2014). Including
potential ethnic challengers may backfire if it empowers them politically and equips
them with resources and opportunities that may make them more likely to challenge
the seating incumbent at some point in the future (Wright and Escriba-Folch, 2012).
The former Zimbabwean Prime Minister Morgan Tsvangirai is but one example of
a powerful rival and opposition leader who has acquired political capital, resources
and some of his allies during his time in the ruling regime party (Brancati, 2014).
In addition, coopted elites may become more powerful over time, forging alliances
within the ruling coalition and pressing the dictator to implement institutional re-
forms credibly ensuring his continued adherence to ethnic power-sharing (Acemoglu
7
and Robinson, 2005). If such reforms are su�ciently far-ranging, they might amount
or at least contribute to a process of elite-led and negotiated regime change. In both
cases, we would expect coopted elites to first work through existing political channels
and aim at a regular, i.e. an electoral or elite-led process of regime change.
When it comes to irregular forms of regime change, Roessler (2011) has promi-
nently argued that ethnic inclusion reduces the risk of ethnic rebellion but, on the
other hand, heightens the threat of coups. While inter-ethnic coups staged by insid-
ers of ethnic power-sharing arrangements are an empirically relevant phenomenon,
as Roessler (2011) convincingly shows, the question of how general of a pattern
they are remains unanswered. Horowitz (1985) regards situations where some eth-
nically defined faction of the military is dissatisfied with a lack of representation in
the civilian regime as particularly dangerous. This suggests that carefully targeted
ethnic cooptation might actually reduce the risk of inter-ethnic coups. In addi-
tion, Roessler’s (2011) analysis also shows that the risk of coups decreases with the
number of included groups. In more inclusive ethnic power-sharing regimes, the co-
ordination of coup attempts arguably requires cooperation across group boundaries.
Moreover, the dictator’s long-term commitment to ethnic power-sharing seems more
credible the more groups are included. This should reduce the incentive of ethnic
elites to plot a coup in order to pre-empt being excluded by a ruler defecting from
ethnic power-sharing (Roessler, 2011).
On these grounds, we expect ethnic cooptation to be a particularly e↵ective strat-
egy against irregular forms of autocratic regime failure. The political empowerment
e↵ect of ethnic cooptation may weaken or even counteract the regime-stabilizing
e↵ect when it comes to regular instances of regime change. Stated as hypothesis:
Hypothesis 2 Ethnic cooptation more e↵ectively reduces the risk of irregular
than regular forms of autocratic regime failure
A similar point can be made with respect to the distinction between democratic
transitions and autocratic regime changes. Most irregular regime breakdowns (75%)
result in subsequent autocracy or state failure whereas most regular regime changes
result in a democratic successor regime (84%) (Geddes, Wright and Frantz, 2014).
Acemoglu and Robinson (2005) have famously argued that democratic concessions
may be the best or even only institutional reform enabling the dictator to credibly
commit not to defect from power-sharing or promised policy reforms. Thus potential
ethnic elites may use their new influence to push for democratic or at least semi-
democratic reforms. On the other hand, Acemoglu and Robinson (2005) claim that
ruling elites are more likely to acquiesce to democratization when it preserves some
of their power and privileges. Forging an inclusive ethnic coalition that stands a
realistic chance of winning even truly democratic elections may thus make democ-
8
ratization more palatable to the incumbent rulers (Bormann, 2014). This leads to
our third hypothesis:
Hypothesis 3 Ethnic cooptation more e↵ectively reduces the risk of autocratic
than of democratizing forms of regime failure.
Finally, the e↵ect of ethnic cooptation may depend on the specific type of author-
itarian regime. Geddes (1999) has prominently distinguished between personalist,
party-based, and military dictatorships, depending on who controls access to politi-
cal power, key policy decisions, and the executive apparatus including the country’s
security forces. In her account of regime stability and breakdown, Geddes (1999)
argues that authoritarian power-sharing is more challenging in military than in per-
sonalist or party-based regimes. For members of a personalist dictator’s clique or
a regime party, personal career prospects, political power and access to resources
depend to a larger extent on the continuation of the regime than for a military junta.
Quite frequently, o�cers remain in high-ranking military positions even after
episodes of regime change away from military dictatorship. As a result, breakdowns
of military regimes often originate in internal splits whereas in the case of person-
alist and party-based regimes, disintegration is initiated by external events such as
economic crises, rebellions, popular protest, and the like (Geddes, 1999). This sug-
gests that military regimes face a lower need for cooptation in the first place. In
addition, the inherent fragility of power-sharing under military rule may counteract
any stabilizing e↵ects of cooptation in such regimes. It is arguably in this context
that Roessler’s 2011 “internal security dillemas” and the associated risk of coups
are most pronounced. The fact that military dictatorships are the most repressive
authoritarian regime type (Escriba-Folch, 2013) may further reduce the likelihood
and e↵ectiveness of cooptation strategies. Frantz and Kendall-Taylor (2014) argue
that repression and cooptation are substitutes.
We have argued above that ethnic cooptation strategies are particularly relevant
in neopatrimonial contexts where informal patronage networks are both important
and follow ethnic lines. The degree to which the informal trading of material benefits
for political loyalty pervades the system most likely depends on the institutional-
ization of a particular authoritarian regime. In line with Geddes (1999), we thus
expect neopatrimonial clientelism to be a characteristic feature of personalist dicta-
torships. It is less important in party-based regimes and in military dictatorships
that broadly conform to what O’Donnell (1973) has famously dubbed “bureaucratic
authoritarian” systems. Mostly found in the Latin American context of the 1970s,
bureaucratic military regimes practice a more professionalized, rule-based and ra-
tional type of rule than the typical neopatrimonial personalist regime.
In addition, formal cooptation institutions such as parties and legislatures may
fail to perform their intended functions in personalist regimes (Wright, 2008; Wilson
9
and Wright, forthcoming). Authoritarian power-sharing between the dictator and
his ruling elite needs to be made credible in order to remain intact (Svolik, 2009,
2012). In this sense, cooptation institutions may serve as a credible commitment
on behalf of the dictator to maintain power-sharing and to refrain from arbitrarily
reducing the size of the ruling coalition or changing key policies (Gehlbach and
Keefer, 2012; Wright, 2008). Moreover, parties and legislatures reduce information
asymmetries allowing the dictator and the ruling elites to monitor each other. This
makes threats to detect and punish defections from the agreed-upon power-sharing
arrangement much more credible (Boix and Svolik, 2013; Svolik, 2012). In order to
perform these functions, authoritarian institutions need to e↵ectively constrain the
dictator. This is unlikely to be the case in personalist regimes, where power over
key policies and appointments is concentrated in the hands of the leader.
The large power di↵erential between the dictator and the elites represented in
the legislature or the regime party prevents the latter from credibly threaten the
dictator with punishment, e.g. in the form of a coup (Svolik, 2009). Recent em-
pirical work on growth, investment and expropriation risk in authoritarian regimes
shows that authoritarian legislatures indeed fail to constrain dictators in personal-
ist regimes (Wilson and Wright, forthcoming; Gehlbach and Keefer, 2012; Wright,
2008). On these grounds, we expect other forms of cooptation to be more important.
Where parties and legislatures are not enough, access to more high-level positions
in the executive apparatus ensuring direct control over at least some budgets and
appointments may be needed to successfully coopt potential external challengers.
Taken together, these mechanisms motivate our final hypothesis:
Hypothesis 4 Ethnic cooptation more e↵ectively reduces the risk of regime fail-
ure in personalist than in party-based autocracies. The e↵ect is
even weaker in military regimes.
In short, we expect ethnically inclusive authoritarian regimes to be more stable
than their more exclusive counterparts (H1) and predict this e↵ect to di↵er across
regime and failure types (H2–H4).2 Testing these hypotheses requires us to identify
the most relevant data sources and employ adequate empirical strategies. In what
follows, we briefly outline our methodological approach.
Data & Methods
In order to test our hypotheses about general regime stability (H1 and H4), we
run simple time-series cross-section logistic regression models including controls for
temporal dependence as recommended by Carter and Signorino (2010). While more
2We have no clear theoretical expectation of how ethnic cooptation plays out in monarchies.
10
advanced survival analysis methods could be performed, the simple method chosen
here should yield equivalent results in the absence of strong assumptions about
the time dependencies of the baseline hazard of regime breakdown as well as the
temporal e↵ects of the various predictors (Beck, Katz and Tucker, 1998). To test H3
and H4, we use multinomial logit models with duration controls to assess the e↵ect
of our predictors on the competing risks of regular vs. irregular and democratic
vs. autocratic regime change respectively. Our sample consists of all authoritarian
regimes in multiethnic countries and covers the period 1946–2010. The unit of
analysis is the country-year. In identifying autocracies, we rely on the data set
provided by Geddes, Wright and Frantz (2014). This data set codes all regimes
conforming to the definition of autocracy cited above, records the date and the
type of regime failure and classifies regimes as personalist, party-based, military, or
monarchical. To identify multiethnic autocracies, we use the Ethnic Power Relations
(EPR) data set (Cederman, Wimmer and Min, 2010). The EPR data identifies all
politically relevant ethnic groups in independent states with more than 500’000
inhabitants for the period 1946–2013 and codes whether these groups have access
to executive state power or not. A group is deemed politically relevant if it is either
represented by at least one political organization active at the national level or
“subjected to state led discrimination” (Cederman, Wimmer and Min, 2010).
Our dependent variable in the logit models is an authoritarian regime breakdown
dummy that is coded one for country-years in which an authoritarian regime col-
lapses and zero otherwise. Country-years in which more than one regime collapsed
are also coded as one. This means that we lose some information, but if anything,
ignoring very short-lived regimes should produce somewhat more conservative es-
timates of the e↵ects of our main independent variables. In the multinomial logit
models testing H2, we use categorical dependent variables coded 1 in the case of
regular regime failure, 2 in the case of irregular regime failure, and 0 otherwise. The
dependent variable distinguishing between democratic and autocratic regime failure
(H3) is constructed equivalently.
The focus on authoritarian regimes instead of indiviudal leaders or authoritarian
spells is the theoretically most appropriate one and well in line with recent work
by Svolik (2012) and Geddes, Wright and Frantz (2014). Cooptation institutions
and strategies should have an e↵ect on the longevity of a particular regime, not
necessarily on the time in o�ce of one particular dictator. Quite obviously, ethnic
inclusion does not a↵ect the odds of a dictator dying in a given year. Neither is it
likely to prevent a certain rotation scheme, e.g. in a military governing council, from
operating. A focus on authoritarian spells would be equally misleading, since they
end, by definition, in a democratic transition or the collapse of statehood. These
are, of course, only some and not even the most frequent forms of authoritarian
11
breakdown Svolik (2012). Thus, we look at authoritarian regimes, which are defined
as “a set of formal and/or informal rules for choosing leaders and policies” (Geddes,
Wright and Frantz, 2014, p. 315). As soon as these rules are replaced by another
set of rules, an authoritarian regime breakdown is coded.
Our main independent variable simply counts the number of included groups and
is taken from the EPR data set (Cederman, Wimmer and Min, 2010). Since this
variable codes group-level access to the highest echelons of the executive apparatus,
it seems a useful proxy for the cooptation of ethnic elites. We expect ethnically
motivated challenges to the regime to become less likely the more groups are coopted.
Quite obviously, including three groups in a country populated by four politically
relevant ethnic groups implies a higher degree of inclusiveness than including the
same number of groups in a country, which is home to, say, nine ethnic groups. To
control for this di↵erence, we include the total number of politically relevant ethnic
groups in all models that use the number of included groups measure as a proxy for
the degree of inclusion. To test whether the e↵ect of ethnic inclusion a↵ects di↵erent
authoritarian regime types in di↵erent ways (H4), we interact our ethnic cooptation
variable with regime-type dummies indicating whether we observe a party-based
regime, a military dictatorship or a personalist regime Geddes, Wright and Frantz
(2014).
To avoid biased estimates, we need to accommodate for alternative cooptation
explanations as well as other factors a↵ecting regime stability that might correlate
with ethnic power-sharing. Our choice of control variables is losely based on Ger-
schewski’s (2013) three-pillar model that regards the broad factors of cooptation,
repression and legitimiation as main sources of authoritarian stability. We control
for cooptation institutions highlighted in the previous literature by including dum-
mies for the presence of a legislature and/or a single regime party. The data comes
from Cheibub, Gandhi and Vreeland (2010). As mentioned before, the presence of
a legislative body or political parties is expected to serve as a source of institution-
alized cooptation of relevant elites. Concerning repression, we include an indicator
measuring the extent to which physical integrity rights are violated. We employ the
latent physical integrity rights score developed by Fariss (2014). Fariss’ (2014) ap-
proach uses a whole range of pre-existing human rights data sets and combines them
into a score that o↵ers greater temporal coverage (1946-2010) and is better suited
to take changes in the standard of accountability into account. Successful coop-
tation and repression heavily depend on the availability of resources to any given
authoritarian regime (see e.g. Beck, 2009). Especially windfall revenues from natural
resources are often thought to stabilize authoritarian regimes (Wright, Frantz and
Geddes, 2015). We thus control for the logged value of annual petroleum rents per
capita. The data comes from Ross (2012) and covers revenues from oil and natural
12
gas production.
Lastly, we have to take the legitimacy of authoritarian regimes into account.
Conceptually, political legitimacy consists of two main dimensions: output and input
legitimacy, generating specific and di↵use regime support on behalf of the population
(Easton, 1967; Gerschewski, 2013). Specific support can be expected where a regime
satisfies the material needs of the country’s population. Economic well-being as
well as current economic performance are particularly relevant here. To control for
output legitimacy, we thus include a logged variable of the country’s per capita
GDP and the annual rate of economic growth. The input dimension of legitimacy
or di↵used support are defined in less materialist terms and include factors such
as political ideologies, nationalism, or charismatic leadership (Gerschewski, 2013).
These concepts are notoriously di�cult to pin down in quantitative analyses and
tend to di↵er starkly across countries. In light of this operationalizational ambiguity,
we are currently not able to explicitly control for the di↵used support dimension of
political legitimacy. If ethnic inclusion is chosen primarily by authoritarian regimes
already enjoying high levels of di↵use support, any statistical association between
inclusion and regime stability may be spurious, since both factors are explained
by legitimacy. If, on the other hand, cooptation via ethnic inclusion is chosen to
compensate for low levels of di↵use regime support, our inability to control for this
dimension of legitimacy will lead us to underestimate the ‘true’ e↵ect of inclusion
on regime stability.
In order to reduce the impact of this and other unobserved variables a↵ect-
ing both the likelihood of regime failure and ethnic power-sharing, we also esti-
mate models including country-specific fixed or random e↵ects. Since fixed-e↵ects
models restrict the analysis to within-country variation in both the dependent and
independent variables, they imply dropping all countries that have never seen an
instance of regime failure. Moreover, the e↵ects of time-invariant or only weakly
time-varying covariates cannot be adequately estimated. Random-e↵ects models do
not have these problems yet make the assumption that the country-specific random
error term is uncorrelated with the other right-hand side variables of the model.
This assumption is most likely violated in our case. To balance the advantages
and disadvantages of these approaches to modeling unobserved heterogeneity across
countries, we estimate pooled, random-e↵ects, and fixed-e↵ects models and compare
their results.3 In all models, time-varying predictors are lagged by one year in order
to limit potential simultaneity bias. Table A1 in the Appendix provides summary
statistics of all the variables just described.
3See Clark and Linzer (2015) and Bell and Jones (2015) for discussions of fixed and random-e↵ects models.
13
Results & Interpretation
Before we turn to the results of our multivariate regression analysis, we take a
brief descriptive look at the data. Table 1 cross-tabulates our dependent variable
regime breakdown against an ethnic cooptation dummy that is coded one for all
country-years in which an authoritarian regime is includes more than one politically
relevant ethnic group at the executive level. In our sample, we observe 2244 country
years without and 1614 country years with such powersharing arrangements. The
failure rate in powersharing regime years is 4.29% and thus somewhat lower than in
ethnic monopoly regime years. Based on this simple cross-tabulation, the cooptation
of elites from one or more politically relevant ethnic groups is associated with a
reduction of the failure rate by roughly 17%. This is broadly in line with our
theoretical expectations reflected in H1.
Table 1: Ethnic Powersharing & Regime Failure, 1946–2010
Ethnic Powersharing
No Yes Total
Regime BreakdownNo 2123 (94.61%) 1541 (95.48%) 3664 (94.97%)Yes 121 (5.39%) 73 (4.52%) 194 (5.03%)
Total 2244 1614 3858
Percentages in parantheses sum to 100 across columns
Table 2 illustrates the distribution of ethnic powersharing across di↵erent types
of authoritarian regimes as well as the failure rates of power-sharing and non-
powersharing monarchies, party-based, personalist, and military regimes. Quite
strikingly, military regimes appear as the odd man out. While we observe ethnic
powersharing in more than half of all other regime years, military dictatorships em-
ploy ethnic inclusion in less than 22% of the observed country years. In addition,
the failure rate of powersharing military regimes is almost twice as high as the one
of their less inclusive counterparts. In personalist and party-based regimes, power-
sharing is associated with lower rates of regime failure and the decrease seams most
pronounced in the case of personalist autocracies. This is in line with our theo-
retical reasoning underlying H4, in which we highlight the potentially destabilizing
e↵ects of powersharing in military regimes as well as a particular relevance of ethnic
cooptation in personalist autocracies.
The mere bivariate relationship between ethnic cooptation and regime stability
might be spurious if other factors at least partially determine both ethnic powershar-
ing and the risk of authoritarian breakdown. In our multivariate regression analyses,
we control for a host of such potential omitted variables and address the issue of un-
observed heterogeneity by including country-specific random or fixed e↵ects. Table
3 reports the results of our random e↵ects models of authoritarian stability. Re-
14
Table 2: Regime Type, Ethnic Powersharing & Failure Rates
No Powersharing PowersharingRegime Type N Share Fail Rate Share Fail Rate
Personalist 952 57.25% 7.34% 42.75% 5.65%Party-based 1961 54.67% 2.99% 45.33% 2.47%Military 496 78.43% 11.05% 21.57% 20.56%Monarchy 449 53.01% 2.52% 47.99% 2.84%All Autocracies 3858 58.16% 5.39% 41.84% 4.52%
sults from population-averaged and fixed-e↵ects models can be found in Table A2
and Table A3 in the Appendix. If anything, the fixed-e↵ects models show stronger
e↵ects of our cooptation variable than the results reported here. As our dependent
variable is regime breakdown, positive coe�cients indicate destabilizing e↵ects while
negative coe�cients imply a stabilizing impact on authoritarian regimes. All four
models include controls for repression, output legitimacy,natural resource rents, and
politically relevant ethnic diversity. Model 2-4 introduce dummies for authoritarian
legislatures and regime parties.
The coe�cient on our ethnic cooptation proxy (No. Included Groups) is negative
and significant at the 10% confidence level in Model 1 and remains so in Model 4
that also includes the institutional cooptation variables. These results lend support
to our first hypothesis postulating that the cooptation of ethnic elites is associated
with higher regime stability, ceteris paribus. Models 2 and 3 test the e↵ect of
cooptation institutions. While the coe�cient on the single-party dummy is negative
and significant, the legislature coe�cient has the expected sign yet fails to reach
conventional significance levels. Higher levels of repression seem to correlate with a
greater risk of regime failure, although only Model 2 produces a significant estimate.
While this seems somewhat counterintuitive, it may be due to a selection e↵ect
in the sense that that repression is more intensely used by inherently unstable,
e.g. military, regimes (Escriba-Folch, 2013). Economic growth is associated with
a significantly lower probability of regime breakdown, while the level of per capita
income does not reach significance. Resource rents are associated with lower odds of
regime breakdown, although the coe�cient on our petroleum rents variable misses
significance in Model 3.
Figure 1 plots the average predicted probabilities of regime failure across the
observed range of our cooptation variable. It is based on Model 1 in Table 3 assuming
that the random e↵ect is zero. Excluding one group from a three-group ethnic
ruling coalition as it happened in the final years of the long-lasting Ivorian post-
independence regime increases the risk of regime breakdown by 17% (from 3.6% to
4.2%). This di↵erence is statistically significant at the 95% confidence interval.4 Its
4calculation based on Stata’s lincom command, not reported here.
15
Table 3: Random E↵ects Logit Models of Regime Failure
Model 1 Model 2 Model 3 Model 4
No. Included Groups �0.164+ �0.150+
(0.085) (0.085)
Legislature �0.328 �0.289(0.238) (0.238)
Single Party �0.575⇤ �0.544⇤
(0.242) (0.240)
Repression 0.183 0.198+ 0.192 0.135(0.118) (0.115) (0.119) (0.119)
GDP p.c. (log) 0.033 0.043 �0.002 �0.003(0.138) (0.137) (0.142) (0.138)
GDP Growth �4.951⇤⇤⇤ �5.095⇤⇤⇤ �5.086⇤⇤⇤ �5.131⇤⇤⇤
(1.058) (1.068) (1.072) (1.065)
Petroleum Rents p.c. (log) �0.097+ �0.091+ �0.086 �0.092+
(0.055) (0.053) (0.056) (0.054)
No. of Groups �0.025 �0.032 �0.030 �0.020(0.024) (0.024) (0.026) (0.024)
Constant �2.885⇤⇤ �3.134⇤⇤ �2.926⇤⇤ �2.489⇤
(1.053) (1.037) (1.087) (1.061)
Observations 3248 3199 3199 3199Duration Polynomials Yes Yes Yes YesClustered SE (Country) No No No NoRandom E↵ects (Country) Yes Yes Yes YesCountry-Fixed E↵ects No No No No
Standard Errors in Parentheses
+p < 0.10,⇤ p < 0.05,⇤⇤ p < 0.01,⇤⇤⇤ p < 0.001
16
Figure 1: Failure Risk across Number of Coopted Groups
substantive size is comparable to the e↵ect of increasing a country’s economic growth
rate from 0% to 3% and amounts to almost a third of the e↵ect of introducing a
single regime party.5
Distinguishing Types of Regime Change: Turning to the distinction between
di↵erent types of regime failure, Table 4 presents the results of our competing risks
analysis of regular vs. irregular forms of authoritarian breakdowns. H2 above pre-
dicts ethnic cooptation to be e↵ective when it comes to preventing coerced and
extraconstitutional forms of regime changes such as coups, rebellions, and popular
uprisings. It may be a less potent tool when it comes more regular, i.e. electoral and
elite-led instances of regime change. Model 5 is a simple multinomial logit model,
whereas Model 6 includes country-random e↵ects to account for unobserved hetero-
geneity. The results provide clear support for our hypothesis that ethnic cooptation
primarily safeguards against irregular challenges initiated by excluded ethnic elites
and their followers. In both models, the coe�cient on our main explanatory variable
is negative and significant at the 95% confidence interval.
Figure 2 depicts the predicted probabilities of regular and irregular regime change
based on the random-e↵ects specification (Model 6). While Increasing the number
of politically relevant groups included in the ruling coalition has no discernable
e↵ect on the risk of regular regime change, it markedly reduces the probability of
irregular regime change.Substantively, moving from a three-group to a two-group
ruling coalition increases the risk of irregular regime failure by 27% (from 1.9% to
5calculations based on Model 1 and Model 4 in Table 3, respectively
17
Figure 2: Risk of Regular vs. Irregular Regime Failure
Table 4: Competing Risks (Regular vs. Irregular Change)
Model 5 Model 6Type of Regime Change Regular Irregular Regular Irregular
No. Included Groups 0.034 �0.202⇤ 0.021 �0.254⇤
(0.108) (0.092) (0.124) (0.112)
Legislature �1.022⇤⇤⇤ 0.085 �0.944⇤⇤ 0.213(0.308) (0.307) (0.351) (0.309)
Single Party �2.352⇤⇤ 0.045 �2.456⇤⇤⇤ �0.021(0.746) (0.222) (0.739) (0.270)
Repression �0.047 0.199 0.001 0.277+
(0.165) (0.151) (0.172) (0.148)
GDP p.c. (log) 0.198 �0.151 0.240 �0.218(0.201) (0.154) (0.196) (0.178)
GDP Growth �1.645 �6.107⇤⇤⇤ �1.817 �6.667⇤⇤⇤
(1.804) (1.130) (1.882) (1.260)
Petroleum Rents p.c. (log) �0.070 �0.097+ �0.086 �0.111(0.073) (0.058) (0.074) (0.073)
No. of Groups �0.094 �0.002 �0.093 �0.006(0.059) (0.018) (0.070) (0.026)
Constant �4.068⇤⇤ �1.853 �5.068⇤⇤ �1.853(1.484) (1.250) (1.587) (1.333)
Observations 3199 3199Duration Polynomials Yes YesClustered SE (Country) Yes NoRandom E↵ects (Country) No YesCountry-Fixed E↵ects No No
Standard Errors in Parentheses
+p < 0.10,⇤ p < 0.05,⇤⇤ p < 0.01,⇤⇤⇤ p < 0.001
18
Table 5: Comepting Risks (Democratization vs. Autocratic Change)
Model 7 Model 8Type of Regime Change Democratic Autocratic Democratic Autocratic
No. Included Groups 0.002 �0.288⇤⇤ �0.024 �0.341⇤
(0.094) (0.091) (0.108) (0.134)
Legislature �0.692⇤ �0.103 �0.592+ 0.020(0.293) (0.295) (0.323) (0.324)
Single Party �1.647⇤⇤⇤ 0.072 �1.733⇤⇤⇤ 0.009(0.481) (0.232) (0.487) (0.295)
Repression 0.004 0.112 0.051 0.184(0.141) (0.163) (0.156) (0.159)
GDP p.c. (log) 0.122 �0.110 0.131 �0.170(0.163) (0.148) (0.176) (0.192)
GDP Growth �3.778⇤ �5.300⇤⇤⇤ �4.197⇤⇤ �5.784⇤⇤⇤
(1.477) (1.208) (1.533) (1.364)
Petroleum Rents p.c. (log) �0.082 �0.103+ �0.095 �0.111(0.063) (0.056) (0.068) (0.078)
No. of Groups �0.050 �0.045+ �0.058 �0.056(0.035) (0.025) (0.049) (0.048)
Constant �3.627⇤⇤ �1.715 �4.166⇤⇤ �1.667(1.227) (1.164) (1.400) (1.457)
Observations 3199 3199Duration Polynomials Yes YesClustered SE (Country) Yes NoRandom E↵ects (Country) No YesCountry-Fixed E↵ects No No
Standard Errors in Parentheses
+p < 0.10,⇤ p < 0.05,⇤⇤ p < 0.01,⇤⇤⇤ p < 0.001
2.5%).
Our third hypothesis predicts a similar pattern when it comes to distinguishing
between democratic transitions and regime failures that result in subsequent autoc-
racies. We expect ethnic cooptation to predominantly a↵ect the likelihood of the
latter form of authoritarian breakdown. Table 5 presents the results of the models
testing H3. Again, the first two columns report the estimates from a population-
averaged multinomial logit model, whereas columns three and four show results from
a random-e↵ects model. The coe�cients on our main explanatory variable suggest
that higher levels of ethnopolitical inclusiveness are associated with significantly
lower odds of autocratic regime failure. The e↵ect of ethnic cooptation on the likeli-
hood of democratic transitions, on the other hand, is statistcially indistinguishable
from zero. Figure 3 facilitates the substantive interpretation of these results. The
predicted probability curve plotted in the right-hand panel now looks even steeper
than it was the case in Figure 2. According to the estimates derived from Model
19
Figure 3: Risk of Democratization vs. Autocratic Regime Failure
8, moving from three to two coopted ethnic groups on average raises the risk of
autocratic regime failure by 39% (from 1.3% to 1.9%).
Distinguishing between Regime Types: To test whether the e↵ect of ethnic
cooptation found in the first few models di↵ers across regime types, we run two
models – one pooled, the other with random e↵ects – in which we interact the
count of politically included ethnic groups from the EPR data set with dummies
for personalist, party-based, and military regimes as coded by Geddes, Wright and
Frantz (2014). Monarchies constitute the baseline category. Table 6 reports the
coe�cients, standard errors, and significance levels from these models: In both
specifications, the ethnic cooptation proxy, all regime type dummies and interaction
terms are statistically significant. To facilitate the interpretation of the interactions
between cooptation and regime type, we plot the predicted probabilities of regime
failure across the range of ethnic inclusion for all four regime types. All plots are
based on the random-e↵ects specification (Model 10).
In line with our theoretical expectations, in the top-left panel, we see that only
in personalist autocracies, regime stability can be markedly enhanced by making
the ruling coalition ethnically more inclusive. the other panels suggest a moderately
stabilizing yet statistically insignificant e↵ect of ethnic cooptation in party-based
regimes and destabilizing, yet imprecisely estimated e↵ects in military dictatorships
and monarchies. Taken together, our findings suggest that in ethnically diverse
countries, ethnically inclusive authoritarian regimes are less prone to breakdown
than their more exclusive counterparts. However, this e↵ect is not homogeneous
across di↵erent forms of regime change. As predicted by our theory, cooptation
of ethnic elites seems e↵ective when it comes to preventing irregular challenges
emanating from underrepresented ethnic groups that would, if successful result in
subsequent autocracy instead of democratic transition. Moreover, ethnic cooptation
works best in personalist dictatorships in which patronage is more important and
institutional forms of cooptation are less e↵ective than in other regime types.
20
Table 6: Authoritarian Breakdown by Regime Type
Model 9 Model 10� s.e. � s.e.
No. Included Groups 0.797⇤⇤⇤ (0.206) 0.801⇤ (0.339)
Personalist Regime 2.713⇤⇤⇤ (0.816) 2.750⇤⇤ (0.861)
Personalist x Incl. Groups �1.024⇤⇤⇤ (0.246) �1.042⇤⇤ (0.381)
Party-Based Regime 1.763⇤ (0.694) 1.779⇤ (0.801)
Party-Based x Incl. Groups �0.877⇤⇤⇤ (0.214) �0.884⇤ (0.354)
Military Regime 3.037⇤⇤⇤ (0.789) 3.071⇤⇤⇤ (0.864)
Military x Incl. Groups �0.644⇤⇤ (0.220) �0.653+ (0.370)
Legislature �0.008 (0.251) 0.000 (0.239)
Single Party �0.372+ (0.212) �0.373 (0.232)
Repression 0.018 (0.104) 0.022 (0.105)
GDP p.c. (log) �0.039 (0.111) �0.048 (0.125)
GDP Growth �5.009⇤⇤⇤ (1.049) �5.056⇤⇤⇤ (1.051)
Petroleum Rents p.c. (log) �0.079+ (0.043) �0.079+ (0.046)
No. of Groups �0.024 (0.023) �0.024 (0.023)
Constant �4.802⇤⇤⇤ (1.203) �4.808⇤⇤⇤ (1.161)
Observations 3199 3199Duration Polynomials Yes YesClustered SE (Country) Yes NoRandom E↵ects (Country) No YesCountry-Fixed E↵ects No No
Standard Errors in Parentheses
+p < 0.10,⇤ p < 0.05,⇤⇤ p < 0.01,⇤⇤⇤ p < 0.001
21
Conclusion
Our analysis highlights the importance of ethnic cooptation strategies in multi-
ethnic autocracies. Including elites from politically relevant ethnic groups at the
executive level may indeed influence authoritarian regime survival. The empirical
results confirm a strong e↵ect of ethnic cooptation across a number of models and
specifications. We have also demonstrated how this e↵ect varies across di↵erent
regime types and how it only reduces the risk of irregular and autocratic forms
of regime change. Our results are robust to including controls for other typically
mentioned explanations of regime stability. In sum, our analysis suggests that ethnic
inclusion at the executive level not only reduces the risk of ethnic rebellion, as e.g.
Cederman, Wimmer and Min (2010) have shown, but also prevents other threats to
regime stability.
These findings complement previous cooptation research, which predominantly
highlights the role of formal institutions instead of focusing on the most relevant
actors that need to be coopted. In addition, moving beyond seemingly democratic
institutions such as legislatures and political parties enables the study of less formal,
more clientelistic or neopatrimonial cooptation strategies. We regard both of these
points as promising avenues for future research. Methodologically, our large-n re-
search design allows us to show that ethnic inclusion matters yet remains too crude
to pin down the precise mechanisms of elite cooptation in multi-ethnic autocracies.
Moreover, it seems extremely challenging to rule out all potential sources of omitted
variable bias and reverse causality in quantitative models of authoritarian stability
(Pepinsky, 2014). In order to better understand whether our hypotheses and the
underlying causal mechanisms hold true, process-tracing case studies may fruitfully
complement the large-n analysis presented in this paper.
22
References
Acemoglu, Daron and James A. Robinson. 2005. Economic Origins of Dictatorshipand Democracy. Berkeley, CA: Cambridge University Press.
Arriola, Leonardo R. 2009. “Patronage and Political Stability in Africa.” Compar-ative Political Studies 42(10):1339–1362.
Bates, Robert H. 1974. “Ethnic Competition and Modernization in ContemporaryAfrica.” Comparative Political Studies 6(4):457–484.
Beck, Martin. 2009. Rente Und Rentierstaat Im Nahen Osten. In Der Nahe Ostenim Umbruch: Zwischen Transformation und Autoritarismus, ed. Martin Beck,Cilja Harders, Annette Juenemann and Stephan Stetter. Berlin: Springer.
Beck, Nathaniel, Jonathan N. Katz and Richard Tucker. 1998. “Taking Time Se-riously: Time-Series-Cross-Section Analysis with a Binary Dependent Variable.”American Journal of Political Science 42(4):1260–1288.
Bell, Andrew and Kelvyn Jones. 2015. “Explaining fixed e↵ects: Random e↵ectsmodeling of time-series cross-sectional and panel data.” Political Science Researchand Methods 3(01):133–153.
Boix, Carles and Milan W Svolik. 2013. “The Foundations of Limited AuthoritarianGovernment: Institutions, Commitment, and Power-Sharing in Dictatorships.”Journal of Politics 75(2).
Bormann, Nils-Christian. 2014. The Causes and Consequences of Ethnic Power-Sharing. Doctoral Dissertation: ETH Zurich.
Brancati, Dawn. 2014. “Democratic Authoritarianism: Origins and E↵ects.” AnnualReview of Political Science 17:313–326.
Bratton, Michael and Nicolas Van de Walle. 1994. “Neopatrimonial Regimes andPolitical Transitions in Africa.” World Politics pp. 453–489.
Brownlee, Jason. 2007. Authoritarianism in an Age of Democratization. CambridgeUniversity Press.
Burgess, Robin, Remi Jedwab, Edward Miguel, Ameet Morjaria et al. forthcoming.“The Value of Democracy: Evidence from Road Building in Kenya.” AmericanEconomic Review (forthcoming) .
Carter, David B. and Curtis S. Signorino. 2010. “Back to the Future: ModelingTime Dependence in Binary Data.” Political Analysis 18:271–292.
Cederman, Lars-Erik, Andreas Wimmer and Brian Min. 2010. “Why Do EthnicGroups Rebel?: New Data and Analysis.” World Politics 62(1):87–119.
Cederman, Lars-Erik, Kristian S. Gleditsch and Halvard Buhaug. 2013. Inequality,Grievances, and Civil War. New York, NY: Cambridge University Press.
Cheibub, Jose Antonio, Jennifer Gandhi and James Raymond Vreeland. 2010.“Democracy and dictatorship revisited.” Public Choice 143(1-2):67–101.
23
Clapham, Christopher S. 1996. Third World Politics: An Introduction. 2. ed.Madison, WI: Univrsity of Wisconsin Press.
Clark, Tom S and Drew A Linzer. 2015. “Should I use fixed or random e↵ects?”Political Science Research and Methods 3(02):399–408.
Easton, David. 1967. A Systems Analysis of Political Life. New York: Wiley &Sons.
Eisenstadt, Shmuel Noah. 1973. Traditional patrimonialism and modern neopatri-monialism. Vol. 1 Sage Publications (CA).
Erdmann, Gero and Ulf Engel. 2007. “Neopatrimonialism reconsidered: criticalreview and elaboration of an elusive concept.” Commonwealth & ComparativePolitics 45(1):95–119.
Escriba-Folch, Abel. 2013. “Repression, political threats, and survival under autoc-racy.” International Political Science Review 34(5):343–60.
Fariss, Christopher J. 2014. “Respect for human rights has improved over time:Modeling the changing standard of accountability.” American Political ScienceReview 108(02):297–318.
Fearon, James D and David D Laitin. 1996. “Explaining interethnic cooperation.”American political science review 90(04):715–735.
Franck, Raphael and Ilia Rainer. 2012. “Does the Leader’s Ethnicity Matter? EthnicFavoritism, Education, and Health in Sub-Saharan Africa.” American PoliticalScience Review 106(02):294–325.
Frantz, Erica and Andrea Kendall-Taylor. 2014. “A dictator’s toolkit.” Journal ofPeace Research 51(3):332–346.
Gandhi, Jennifer and Adam Przeworski. 2006. “Cooperation, cooptation, and re-bellion under dictatorships.” Economics & Politics 18(1):1–26.
Gandhi, Jennifer and Adam Przeworski. 2007. “Authoritarian Institutions and theSurvival of Autocrats.” Comparative Political Studies 40(11):1279–1301.
Geddes, Barbara. 1999. “What do we know about Democratization after TwentyYears?” Annual Review of Political Science 2(1):115–144.
Geddes, Barbara. 2003. Paradigms and Sand Castles: Theory Building and ResearchDesign in Comparative Politics. Ann Arbor: University of Michigan Press.
Geddes, Barbara, Joseph Wright and Erica Frantz. 2014. “Autocratic breakdownand regime transitions: A new data set.” Perspectives on Politics 12(02):313–331.
Gehlbach, Scott and Philip Keefer. 2012. “Private Investment and the Institution-alization of Collective Action in Autocracies: Ruling Parties and Legislatures.”The Journal of Politics 74(02):621–635.
Gerschewski, Johannes. 2013. “The three pillars of stability: legitimation, repres-sion, and co-optation in autocratic regimes.” Democratization 20(1):13–38.
24
Gerschewski, Johannes, Wolfgang Merkel, Alexander Schmotz, Christoph H Stefesand Dag Tanneberg. 2013. “Warum uberleben Diktaturen?” Politische Viertel-jahresschrift 53.
Habyarimana, James, Macartan Humphreys, Daniel N. Posner and Jeremy M. We-instein. 2007. “Why Does Ethnic Diversity Undermine Public Goods Provision?”American Political Science Review 101(4).
Hardin, Russell. 1997. One For All: The Logic of Group Conflict. Princeton, NJ:Princeton University Press.
Horowitz, Donald L. 1985. Ethnic Groups in Conflict. Berkeley, CA: University ofCalifornia Press.
Jackson, Robert H. and Carl G. Rosberg. 1984. “Personal rule: theory and practicein Africa.” Comparative Politics 16(4):421–442.
Kramon, Eric and Daniel N Posner. 2012. “Ethnic favoritism in primary educationin Kenya.” Unpublished manuscript .
Lemarchand, Rene. 1972. “Political Clientelism and Ethnicity in Tropical Africa:Competing Solidarities in Nation-Building.” American Political Science Review66(01):68–90.
Levitsky, Steven and Lucan A Way. 2010. Competitive authoritarianism: Hybridregimes after the cold war. Cambridge University Press.
Levitsky, Steven and Lucan Way. 2002. “The rise of competitive authoritarianism.”Journal of democracy 13(2):51–65.
Lust-Okar, Ellen. 2004. “The Management and Manipulation of Political Opposi-tion.” Comparative Politics 36(2):159–179.
Magaloni, Beatriz. 2008. “Credible Power-Sharing and the Longevity of Authoritar-ian Rule.” Comparative Political Studies 41(4-5):715–741.
Magaloni, Beatriz and Ruth Kricheli. 2010. “Political Order and One-Party Rule.”Annual Review of Political Science 13:123–43.
Malesky, Edmund and Paul Schuler. 2010. “Nodding or Needling: Analyzing Dele-gate Responsiveness in an Authoritarian Parliament.” American Political ScienceReview 104(03):482–502.
O’Donnell, Guillermo A. 1973. Modernization and bureaucratic-authoritarianism:Studies in South American politics. Berkeley: University of California Press.
Olzak, Susan. 1983. “Contemporary ethnic mobilization.” Annual Review of Soci-ology 9(1):355–374.
Pepinsky, Thomas. 2014. “The institutional turn in comparative authoritarianism.”British Journal of Political Science 44(03):631–653.
Posner, Daniel N. 2004. “Measuring Ethnic Fractionalization in Africa.” AmericanJournal of Political Science 48(4):849–863.
25
Roessler, Philip. 2011. “The Enemy Within: Personal Rule, Coups and Civil Warin Africa.” World Politics 63(2):300–346.
Ross, Michael L. 2012. The Oil Curse: How Petroleum Wealth Shapes the Develop-ment of Nations. Princeton, NJ: Princeton University Press.
Schedler, Andreas. 2009. “Electoral Authoritarianism.” The SAGE Handbook ofComparative Politics p. 381.
Svolik, Milan W. 2009. “Power Sharing and Leadership Dynamics in AuthoritarianRegimes.” American Journal of Political Science 53(2):477–494.
Svolik, Milan W. 2012. The Politics of Authoritarian Rule. New York: CambridgeUniversity Press.
Wilson, Matthew Charles and Joseph Wright. forthcoming. “Autocratic Legislaturesand Expropriation Risk.” British Journal of Political Science .
Wimmer, Andreas. 1997. “Who Owns the State? Understanding Ethnic Conflict inDost-Colonial Societies.” Nations and Nationalism 3(4):631–666.
Wintrobe, Ronald. 1998. The Political Economy of Dictatorship. Vol. 6 New York:Cambridge Univ Press.
Wright, Joseph. 2008. “Do Authoritarian Institutions Constrain? How LegislaturesA↵ect Economic Growth and Investment.” American Journal of Political Sciencepp. 322–343.
Wright, Joseph and Abel Escriba-Folch. 2012. “Authoritarian Institutions andRegime Survival: Transitions to Democracy and Subsequent Autocracy.” BritishJournal of Political Science 42(02):283–309.
Wright, Joseph, Erica Frantz and Barbara Geddes. 2015. “Oil and autocratic regimesurvival.” British Journal of Political Science 45(02):287–306.
26
Appendix
Table A1: Summary Statistics
count mean sd min max
Regime Failure 3889 0.050 0.219 0 1Irregular Failure 3889 0.032 0.176 0 1Regular Failure 3889 0.018 0.134 0 1Democratic Transition 3889 0.023 0.149 0 1Autocratic Change 3889 0.026 0.158 0 1No. Included Groups 3858 1.964 1.457 1 13Legislature 3764 0.806 0.396 0 1Single Party 3764 0.264 0.441 0 1Repression 3780 0.492 0.947 -2.091 3.134GDP p.c. (log) 3807 7.625 1.115 4.703 11.626GDP Growth 3781 0.021 0.078 -0.646 0.661Petroleum Rents p.c. (log) 3334 2.357 2.859 0 10.903No. of Groups 3858 5.754 7.402 2 57
Table A2: Pooled Logit Models of Regime Failure
(1) (2) (3) (4)
No. Included Groups �0.135⇤ �0.120+
(0.067) (0.065)
Legislature �0.433+ �0.390+
(0.234) (0.230)
Single Party �0.519⇤ �0.490⇤
(0.202) (0.203)
Repression 0.121 0.153 0.139 0.083(0.098) (0.101) (0.100) (0.105)
GDP p.c. (log) 0.052 0.061 0.004 0.016(0.115) (0.121) (0.123) (0.120)
GDP Growth �4.519⇤⇤⇤ �4.680⇤⇤⇤ �4.580⇤⇤⇤ �4.735⇤⇤⇤
(0.941) (0.938) (0.949) (0.955)
Petroleum Rents p.c. (log) �0.084+ �0.083+ �0.069 �0.081+
(0.044) (0.045) (0.045) (0.044)
No. of Groups �0.021 �0.029 �0.030 �0.019(0.016) (0.023) (0.026) (0.020)
Constant �2.615⇤⇤ �2.802⇤⇤ �2.467⇤ �2.197⇤
(0.878) (0.909) (0.968) (0.934)
Observations 3248 3199 3199 3199Duration Polynomials Yes Yes Yes YesClustered SE (Country) Yes Yes Yes YesRandom E↵ects (Country) No No No NoCountry-Fixed E↵ects No No No No
Standard Errors in Parentheses
+p < 0.10,⇤ p < 0.05,⇤⇤ p < 0.01,⇤⇤⇤ p < 0.001
27
Table A3: Fixed-E↵ects Logit Models of Regime Failure
(1) (2) (3) (4)
No. Included Groups �0.428⇤ �0.464⇤
(0.185) (0.189)
Legislature 0.123 0.171(0.274) (0.275)
Single Party �0.568+ �0.625⇤
(0.311) (0.311)
Repression 0.502⇤⇤ 0.483⇤⇤ 0.474⇤⇤ 0.507⇤⇤
(0.173) (0.176) (0.175) (0.179)
GDP p.c. (log) �0.435 �0.461 �0.451 �0.370(0.480) (0.486) (0.481) (0.495)
GDP Growth �4.564⇤⇤⇤ �4.782⇤⇤⇤ �4.834⇤⇤⇤ �4.693⇤⇤⇤
(1.213) (1.229) (1.239) (1.233)
Petroleum Rents p.c. (log) �0.106 �0.113 �0.125 �0.120(0.112) (0.112) (0.113) (0.114)
No. of Groups 0.287⇤ 0.242+ 0.224+ 0.300⇤
(0.142) (0.135) (0.134) (0.144)
Observations 2349 2325 2325 2325Duration Polynomials Yes Yes Yes YesClustered SE (Country) No No No NoRandom E↵ects (Country) No No No NoCountry-Fixed E↵ects Yes Yes Yes Yes
Standard Errors in Parentheses
+p < 0.10,⇤ p < 0.05,⇤⇤ p < 0.01,⇤⇤⇤ p < 0.001
28