REVISITING THE DEMOCRATIC CIVIL PEACE HYPOTHESIS

REVISITING THE DEMOCRATIC CIVIL PEACE HYPOTHESIS:
ELECTORAL DEMOCRACY AND CIVIL CONFLICT*
Henrikas Bartusevičius
Department of Government
University of Essex
[email protected]
Svend-Erik Skaaning
Department of Political Science
Aarhus University
[email protected]
Abstract
The democratic civil peace hypothesis has generated considerable debate in conflict research. This
debate has centered on three general claims: democracies have lower risk of civil conflict; autocracies
have as low a risk of civil conflict as democracies; and “anocracies” (or hybrid regimes) have the
highest risk of civil conflict. This study presents an argument that more democracy—understood in
minimalist, electoral sense—always leads to more civil peace. Whether legislative or executive,
elections constitute a major institutional venue for potential contenders to channel their political or
socio-economic grievances. The more this venue is restricted, the more the contenders will consider
alternative options, including political violence. We test this claim in a global statistical analysis that
spans 1816–2008 and employs the newly constructed Lexical Index of Electoral Democracy (LIED).
The index is based on binary coding of institutional features of political regimes that are aggregated
into a cumulative scale with seven levels, where each level identifies a unique and theoretically
meaningful regime type. In line with our argument, we show that hybrid regimes do not have the
highest risk of civil conflict. Instead, we demonstrate that electoral regimes outperform non-electoral
regimes when it comes to civil peace, and that the effect of electoral democracy is strongest at the stage
of minimally competitive elections.
*
Paper prepared for the 46th Annual Meeting of the Danish Political Science Association, Vejle, Denmark, 23-24 October
2014. First draft, please do not cite.
1
Introduction
The idea that democracy leads to civil peace has generated considerable debate in conflict research
(Hegre 2014). This debate has centered on three general claims. First, democracies have lower risk of
civil conflict, as democratic regimes generally create fewer political grievances and have institutional
means to accommodate potential contenders in non-violent ways. Second, autocracies can be as
peaceful—or even more peaceful—than democracies, as autocratic regimes have means to repress
potential contenders in violent ways. Finally, “anocracies” (or hybrid regimes) are the most conflictprone, as they lack both established institutions to accommodate contenders and sufficient means to
repress them (see Auvinen 1997; Fearon and Laitin 2003; Gleditsch, Hegre, and Strand 2009; Gurr
2000; Hegre et al. 2001; Hegre 2014; Muller and Weede 1990; Reynal-Querol 2002).
These claims have received considerable support in empirical research. A number of studies
have found that the relationship between political regime and the outbreak of civil conflict follows a
pattern of an inverted-U, suggesting that the risk of civil conflict is indeed lowest at both ends of the
political regime spectrum (i.e., in full democracies and full autocracies) and highest in the middle (see,
e.g., Fearon and Laitin 2003; Hegre et al. 2001; Muller and Weede 1990).
However, the inverted-U hypothesis has been challenged in a few studies (Collier and Hoeffler
2004; Elbadawi and Sambanis 2002; Vreeland 2008). Most forcefully, Vreeland (2008) has shown that
the findings of the previous studies have been affected by endogeneity problems, since the middle
values of the Polity index (Marshall, Gurr, and Jaggers 2013), which has been used in most previous
studies to identify anocracies, by construction includes regimes already plagued by civil conflicts (see
also Strand 2007).
We posit that more democracy–understood in minimalist, electoral sense–alwayys leads to more
civil peace. In line with Gleditsch and Ruggeri (2010) and Goldstone et al. (2010), we argue that
2
previous studies on regime type and civil conflict have failed to appropriately account for the “middle
ground” of the political regime spectrum (see also Fjelde 2010). Most of previous research has relied
on measures representing the overall level of democracy such as the Polity index (Marshall et al. 2013)
or the Political Rights and Civil Liberties indices (Freedom House 2014) that do not link particular
regime characteristics to the different levels of the measures (Skaaning, Gerring, and Bartusevičius
2014; Ravallion 2013). These measures fail to account for the characteristics of regimes theorized to
influence civil conflict. As a consequence, previous research has largely failed to provide clear
conclusions on the relationship between regime type and civil conflict.
Joining few recent studies, that have shifted focus to particular qualities of regimes (Goldstone
et al 2010; Fjelde 2010), we center on the relationship between electoral characteristics and onset of
civil conflict. In our attempt to shed new light on this relationship, we employ the new Lexical Index of
Electoral Democracy (LIED). The LIED classifies regimes into seven ordinal categories based on their
electoral characteristics, where all levels signify a particular combination of regime features, with “0”
representing no elections and “6” competitive multi-party elections with universal suffrage. Spanning
1800-2013 and including 221 independent polities, the LIED has the most comprehensive coverage
among extant measures of political regimes. This allows us to move beyond the conventional post-1945
period, substantially widening the scope of our inferences.
We find little evidence to support the inverted-U hypothesis. Indeed, we show that all levels of
LIED above 0 negatively affect the risk of civil conflict. Moreover, there seems to be a threshold effect,
where the critical distinction is whether elections are genuinely competitive or not.
The paper proceeds as follows. Section One reviews previous research on the relationship
between political regime and outbreak of civil conflict, highlighting the above mentioned mismatch
between theoretical concepts and their empirical proxies; Section Two presents our theoretical
3
framework and specifies the hypotheses on the effects of electoral democracy on civil conflict onset;
Section three introduces the LIED and presents the empirical analysis. Section Four summarizes the
main results, discusses their implications, and offers suggestions for future research.
Regime type and civil conflict: Literature review
Building on the democratic peace theory put forward in international relations research, civil conflict
researchers have proposed the democratic civil peace hypothesis: democracy decreases the likelihood
of domestic armed conflict (Hegre et al. 2001; Krain and Myers 1997). A number of arguments have
been suggested in favor of this hypothesis.
First, it has been argued that societies in democratic states experience fewer political grievances,
as democratic regimes are generally less repressive, politically inclusive, and tolerant (e.g., Gurr 2000).
While it does not imply absence of societal discontent (as grievances may arise out of other sources, for
example, economy), fewer political grievances means fewer reasons for politically-motivated violence
against the state.
Second, even if political grievances arise, democratic regimes provide institutional means to
address these grievances in non-violent ways. Democratic institutions, most importantly, elections,
allow equal opportunities for political contenders to pursue their interests and channel discontent.
While such institutions are likely to invite contestation and may mobilize previously inactive groups for
political action, they constitute a major–and arguably more attractive–alternative to political violence.
The more accommodative the democratic institutions, the less likely that political contenders (and
citizens more generally) will pursue their interests using violent means.
Finally, leaders of democratic states are confronted with higher political costs than leaders of
autocratic states. Through the election of representatives, democratic settings ensure that interests of
4
society are taken into consideration. As domestic armed conflicts incur direct costs to population,
decisions to initiate an armed fight might be deemed unpopular. This will most likely be taken into
account by leaders willing to stay in power (Hegre 2014: 161).
A number of studies have argued, however, that full autocracies are potentially as peaceful as (or
even more peaceful) than full democracies. Because of restricted political competition, corruption, and
violations of human rights, societies in autocratic settings experience more political grievances than in
democratic ones. Yet, unlike in democracies, grievances in autocracies are often successfully contained
using repression, reducing the likelihood of civil conflict:
Both democracies and non-democracies use military force to counter illegitimate armed opposition,
but autocracies may make much more extensive use of repression without losing legitimacy–using
violence to silence opponents, censorship, arbitrary imprisonment without trial, etc. Autocracies
may indiscriminately target entire population groups to coerce influential individuals…Autocracies
also buy off other parts of the opposition by granting ministerial posts and by the selective
channeling of public funds…The combination of these two methods allows effective divide-andrule strategies. Autocracies also repress the formation of organizations before they can reach the
stage of armed insurgencies (Hegre 2014: 163).
While democracies provide institutional means to accommodate contenders in non-violent ways,
and autocracies possess effective means to repress contenders in violent ways, it is commonly argued
that anocracies usually lack both, which is why regimes falling between the poles of full autocracy and
full democracy are the most conflict-prone. The inconsistent nature of hybrid regimes is at the core of
this argument: mixing repression–not effective enough to quell political opposition–with the extent of
5
political openness–not sufficient to accommodate the opposition–anocracies provide both opportunities
and motivation for civil conflict (Gurr 2000; Hegre et al 2001).
Empirical research has provided considerable evidence to support these claims. A number of
studies have shown that fully democratic regimes are least likely to experience civil conflict (e.g.,
Gleditsch and Ruggeri 2010; Hegre et al. 2001) and, more generally, political instabilities (Gates et al.
2006). Though, Goldstone et al. (2010) and Buhaug (2006) have challenged these claims, showing that
full autocracies can be as peaceful as full democracies. Most studies, however, seem to have converged
on the finding that hybrid regimes are the most conflict-prone, suggesting that the relationship between
regime type and civil conflict indeed follows the pattern of an inverted-U (Gleditsch, Hegre, and Strand
2009; Golstone et al. 2010; Fearon and Laitin 2003; Hegre et al. 2001; Mueller and Weede 1990).
The inverted-U hypothesis, however, has been challenged in more recent research. Most
notably, Vreeland (2008) has demonstrated that results of the previous studies have been affected by
endogeneity problems: the middle values of the Polity index (Marshall, Gurr, and Jaggers 2013), which
previous studies almost exclusively relied on to proxy for anocracies, include regimes that are already
plagued by civil conflicts. When the element of civil conflict is excluded from the index, the middle
values seem to have no effect on civil conflict. Moreover, Vreeland has found that substituting the
adjusted Polity index with other measures of regime type makes little difference: anocracies have no
higher risk of civil conflict than other regimes. The inverted-U hypothesis has also been challenged in a
few other studies (Collier and Hoeffler 2004; Elbadawi and Sambanis 2002).
We argue that the inconsistencies in the findings of previous research stem from the mismatch
between the theorized causal mechanisms through which regime type is proposed to influence civil
conflict and the measures used to account for these. As suggested above, anocracies are often thought
to experience more civil conflict because of their inability to fully accommodate contenders and, at the
6
same time, to effectively repress them. The measures used to account for these qualities of regimes,
however, are usually very crude: they represent the “overall degree of democracy” without linking
specific characteristics of regimes to the different levels. A country with a value of “6” on the Polity
scale, for example, is considered more democratic than that with a value of “5.” This, however, tells us
little about the particular (qualitative) differences between the two regimes in terms of, for example, the
institutions regulating the access to political power.
Furthermore, the values of Polity index (but also other extant measures of political regimes) are
aggregates of many component variables; and different combinations of values on these component
variables may lead to the same values on the aggregate index (see Cheibub et al. 2010; Ravallion
2013). This implies that a country, for example, scoring low on repression and high on democratic
institutions may be assigned the same overall value as a country scoring low on democratic institutions
and high on repression. As a result, such indices generate rather heterogeneous pools of autocracies,
anocracies, and democracies, which hinders our ability to distinguish between the exact characteristics
of regimes that are associated with civil conflict.
Notably, a few recent studies have focused on more specific qualities of political regimes.
Goldstone et al. (2010), for example, have emphasized the sub-types of anocracies. They have found
that partial democracies with “factionalism” were the most conflict-prone. Further, Fjelde (2010) has
shown that electoral multi-party autocracies and military dictatorships have a higher risk of armed
conflict than one-party autocracies. We join this recent line of research and propose to specifically
focus on electoral qualities of regimes–qualities that are at the center of democracy concept, and that
are likely to underlie the regime type-civil conflict nexus.
7
Electoral democracy and civil peace: Theory and hypotheses [in progress]

Our overall theoretical argument is straightforward: democracy, when understood in an
electoral sense, has negative and generally monotone effect on the risk of civil conflict. We
explain this on the following grounds:
o First, electoral democracies offer representation as a substitute for political violence,
thereby increasing the chances that popular demands will be articulated and taken into
account by the relevant governing bodies.
o Second, electoral democracies refrain from repression (which often spurs civil
conflicts), as elected officials scrutinize governments’ actions and voters punish
governments for human rights violations.
o Finally, electoral democracies cultivate norms of non-violent political action, since
elections are recognized as the sole legitimate mean for access to political power
(Davenport 2007; Davenport & Armstrong 2004; Rummel 1997; Hegre 2014).

While we propose that there is a negative association between electoral democracy and the
onset of civil conflicts, we are less confident about whether the relationship is linear or is best
characterized as threshold effect. The main argument for the latter perspective would be that
some of the positive effects of democratic institutions only set in when a critical number of
them are in place and function more or less in accordance with democratic principles. For
example, it might be the case that only when elections are not a mere façade but actually
determine who holds legislative and executive power, the mechanisms will be operative.
Accordingly, civil war initiations would only become less likely when the political regime is
characterized by particular combination of institutions (cf. Davenport and Armstrong 2004).
8
Research Design
We have built our benchmark model following the principle of parsimony. Therefore, instead of relying
on complex statistical analysis techniques—based on multiple (and often untenable) assumptions—and
specifications involving dozens of potentially collinear controls (Schrodt 2014), we have enlisted a
standard country-year logistic regression and employed only a limited set of controls, selected based on
theoretical criteria. The overall pool of the country-years was taken from our new dataset on the LIED
(described below), which uses Correlates of War (2011), Gleditsch (2013), and various countryspecific sources to identify independent countries. The base of the dataset includes 221 unique polities
and spans 1800-2013.
Lexical Index of Electoral Democracy
The LIED focuses on the electoral aspects of democracy. The basic understanding is that “democracy is
achieved through competition among leadership groups that vie for the electorate’s approval during
periodic elections before a broad electorate” (Skaaning, Gerring, and Bartusevičius 2014: 6; see also
Przeworski et al. 2000; Schumpeter 1950; Møller and Skaaning 2013).
Similar minimalist conceptions underlie most extant indices of democracy, including the ordinal
Polity index (Marshall et al. 2013) or the dichotomous measure of political regime from the
Democracy-Dictatorship dataset (Cheibub et al. 2010). The LIED is constructed combining
dichotomous indicators of different regime qualities in a systematic fashion, where a series of a
necessary and jointly sufficient conditions are arranged in an ordinal scale. Based on six indicators we
construct an index with the following values:
0.
No elections.
1.
No-party or one-party elections.
9
2.
Multiparty elections for legislature.
3.
Multiparty elections for legislature and executive.
4.
Minimally competitive, multiparty elections for legislature and executive.
5.
Minimally competitive, multiparty elections with full male or female suffrage for legislature
and executive.
6.
Minimally competitive, multiparty elections with universal suffrage for legislature and
executive.
We aggregate the indicators using the cumulative logic of a lexical scale (see Gerring,
Skaaning, and Pemstein 2013), where ordering is determined by theoretical considerations over the
centrality of particular features to the concept of electoral democracy. Through this procedure our
index performs both a classificatory function, as each level is connected to a particular combination of
regime features, and a discriminating function, as it distinguishes between the seven levels. Thus,
unlike extant indices, the LIED provides an opportunity to investigate the effects of particular qualities
of regimes on civil conflict (as opposed to aggregate characteristics). For further details on the index
construction see Appendix A.
Outcome variable
The LIED expands the conventional post-1945 time-span threefold, thereby substantially widening the
temporal scope of our inferences. To match the temporal scope of LIED on the left hand side of the
equation we employed the data on civil wars from the Expanded War Data that spans 1816–2010
(Gleditsch 2004) (1 May 2013 update). The Expanded War Data is based on the COW Wars v4.0,
1816-2007 dataset (Sarkees and Wayman 2010) and thus employs the original COW definition of intrastate conflict: a sustained combat between a government of a state and non-state actor(s) that results in
10
at least 1000 battle-related deaths over a year, and takes place within a territory of one state (see the
original source for full definitions and details).
As we exclusively focus on civil wars, our analyses only included those conflicts that took place
within the borders of independent states, excluding any overseas colonies. While this criterion results in
a somewhat conservative list of civil conflicts, it allows us to effectively exclude all the cases of the socalled “extra-systematic” wars—colonial or anti-colonial conflicts that took place outside the core
territories (i.e., in colonial possessions) of SideA’s. We believe that both colonial and anti-colonial
wars and civil wars represent qualitative different phenomena.
Since this study focused on the outbreak of conflict, country-years after the year of civil war
onset were set to 0. In line with previous research (e.g., Buhaug 2006), we have also applied the twoyear intermittency rule. The final sample includes 293 separate civil war onsets, recorded between 1816
and 2008.
Control variables
Our aim was to isolate the effects of electoral democracy on civil conflict onset—and not to explain as
much as possible variation in the outcome variable. Therefore, we limited the set of control variables in
the main empirical model to the likely confounders, i.e., variables that were potentially linked to both
electoral democracy and civil war onset. Further, to retain as many observations as possible, and avoid
potential colinearity problems, we included only the most important confounders (as identified in the
literature). These include the absolute level of income, economic growth, and oil wealth.
Wealth and economic growth have been shown to be associated with democracy (e.g., Boix and
Stokes 2003; Lipset 1959; Przeworski et al. 2000). The absolute level of income and growth has also
been shown to vary with civil conflict onset (e.g., Hegre and Sambanis 2006; Miguel, Satyanath, and
11
Sergenti 2004). To account for potential confounding effects of the absolute level of income and
growth, we thus include control variables in the form of the natural log of GDP per capita (GDP per
capita) and annual GDP per capita growth (GDP per capita growth) (both lagged one year). Data on
GDP per capita was taken from Bolt and van Zanden (2013).
Further, researchers have shown that democracy levels are negatively related to energy resource
wealth, in particular, oil (e.g., Jensen and Wantchekon 2004; Ross 2012) and that oil wealth is also
positively associated with the risk of civil conflict onset (Ross 2004). To control for oil wealth effects,
we also include in the model the Total oil income per capita (Oil income) from Haber and Menaldo
(2011).
Following standard practice (e.g., Hegre and Sambanis 2006), we furthermore control for
population size (ln) with data from the National Material Capabilities dataset (Singer 1987) and time
dependency using peace years with a decay function e^(- peace years / x).1 In additional models we
introduce a number of other potential confounders, which we describe below. Table I provides
summary statistics for all the variables employed in the analyses below.
1
Where peace years stands for the number of years since the last civil war (or independence). X determines the rate of
decay. We followed Hegre et al. (2001) and set X to 4, which halved the effects of the peace years with every additional
three years in peace.
12
Table I. Summary statistics
Name
N
Years
Mean
S.D.
Min
Max
Civil War (COW) (Singer, 1987)
17 169
1800-2008
0.017
0.130
0
1
GDP per capita (ln) (Bolt and Zanden 2013)
10 502
1800-2008
7.856
0.979
5.313
10.667
GDP per capita growth (annual) (Bolt and Zanden 2013)
10 247
1801-2008
1.926
6.277
-61.473
86.900
Total oil income per capita (Haber & Menaldo 2011)
13 659
1800-2006
0.316
2.559
0
78.589
Population size (ln) (Singer 1987)
14 090
1800-2008
8.624
1.761
2.197
14.097
Peace years (with decay function) (authors’ coding)
17 169
1800-2008
0.153
0.292
0
1
Vertical economic inequality (Gini coefficient)
7 024
1946-2008
41.666
10.567
15.9
73.9
Ethno-political exclusion (max exclusion)
7 701
1947-2008
0.071
0.177
0
0.98
Ethno-political discrimination (max discrimination)
7 701
1947-2008
0.161
0.235
0
1
Horizontal economic inequality (max low ration)
7 701
1947-2008
1.197
0.543
1
6.774
MAIN MODEL
ADDITIONAL MODELS*
*All variables were imported from Cederman, Gleditsch, and Buhaug (2013)
Results
Table II presents the regression estimates. In line with our expectations, LIED has a negative and
highly significantly (p < 0.001) effect on the outbreak of civil war, when regressed separately (Model
1.1). The same is party true when we include GDP per capita and GDP per capita growth into the
same block (see Models 1.2 and 1.3): while the effect size and the p-value drops, the coefficient for
LIED remains negative and significant at 5% level. We believe the reduction in the p-value is party a
consequence of a substantial decrease in the sample size (due to which the analysis drops 76 cases of
civil wars). Model 1.4 shows that Oil income per capita has virtually no confounding effect on the
relationship between LIED and civil war.
13
Table II. Logistic regression estimates of civil war onset (full sample)
(1.1)
(1.2)
(1.3)
(1.4)
(1.5)
-.177***
(.029)
-.079*
(.038)
-.087*
(.040)
-.088*
(.041)
Lexical scale^2
-
-
-
-
-.050
(.138)
-.007
(.022)
i.Lexical scale:
-
-
-
-
-
L1
-
-
-
-
-
L2
-
-
-
-
-
L3
-
-
-
-
-
L4
-
-
-
-
-
L5
-
-
-
-
-
GDP per capita(ln)
-
-.559***
(.101)
GDP per capita growth
-
-
-.544***
(.104)
-.017
(.013)
Total oil income per capita
-
-
-
.257***
(.034)
1.425***
(.142)
-6.188***
(.332)
.201***
(.045)
1.161***
(.191)
-1.656*
(.796)
.196***
(.047)
1.084***
(.201)
-1.685*
(.834)
-.546***
(.110)
-.017
(.013)
-.029
(.034)
.193***
(.049)
1.068***
(.202)
-1.621†
(.850)
-.544***
(.110)
-.017
(.013)
-.027
(.034)
.196***
(.049)
1.072***
(.203)
-1.688†
(.873)
-.715†
(.427)
-1.255†
(.737)
-.684*
(.272)
-.514***
(.109)
-.0175
(.013)
-.039
(.041)
.197***
(.050)
.976***
(.207)
-1.764*
(.891)
Wald χ2
250.84
208.43
194.53
201.70
202.58
211.36
N
13936
9960
9745
9545
9545
9545
267
191
179
178
178
178
Lexical scale
L6
Population size(ln)
Peace years
Constant
N of civil wars
(1.6)
-.332
(.230)
-.863*
(.413)
-.027
(.219)
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
14
To test the claim that regime type affects the outbreak of civil war in a non-linear fashion, we
regressed a quadratic term LIED^2 in the same block. As shown in Model 1.5, the coefficients for both
LIED and LIED^2 are negative and insignificant, lending little support for the inverted-U hypothesis.
This is largely corroborated in Model 1.6 that includes a “dummy-coded” version of LIED (with Level
0 [no elections] serving as a reference category). The results of this model demonstrate that each of the
six levels—compared to Level 0—have negative effects on the outbreak of civil war, though the
coefficients for levels L1 and L3 are insignificant. While the coefficient for L1 is close to significance
at 10% level (p = .148) the coefficient for L3 is far from achieving statistical significance (p = .901),
which, we believe, hints to the potential source of the inconsistencies in the findings of previous studies
(an issue we return to below).
Additional tests
To assess the robustness of these results we implemented a number of additional tests. A time-span of
over 200 years contains countries from many different periods and therefore has the potential to
introduce heterogeneity problems. One could argue, for example, that findings based on a sample that
includes countries from the early 19th century, can hardly be applicable to the countries of
contemporary era. To address this consideration, we regressed the same set of variables in a sample
limited to 1946–2008 (see Table III), which is the typical time-span covered in the studies of civil
conflict onset. As indicated in Model 2.1 the effect of LIED remains negative and significant at 5%
level, despite the fact that the sample size drops to just 7036 observations (and the number of civil war
onsets to just 139).
15
Table III. Logistic regression estimates of civil war onset (post-1945 sample with additional controls)
(2.1)
(2.2)
(2.3)
(2.4)
(2.5)
(2.6)
-.098*
(.044)
-.106*
(.046)
-.088†
(.049)
-.086†
(.049)
-.086†
(-.086)
Lexical scale^2
-
-
-
-
-
-.095
(.179)
.002
(.028)
i.Lexical scale:
-
-
-
-
-
-
L1
-
-
-
-
-
-
L2
-
-
-
-
-
-
L3
-
-
-
-
-
-
L4
-
-
-
-
-
-
-.462***
(.107)
-.020
(.015)
-.054
(.045)
-.469***
(.117)
-.022
(.016)
.208**
(.068)
-.006
(.009)
-.445***
(.117)
-.020
(.016)
.207**
(.070)
-.005
(.009)
1.368***
(.330)
-.441***
(.117)
-.020
(.016)
.205**
(.071)
-.005
(.009)
1.227**
(.477)
.184
(.455)
-.439***
(.118)
-.020
(.016)
.205**
(.070)
-.005
(.009)
1.226**
(.477)
.187
(.454)
-.010
(.125)
.203**
(.065)
.742**
(.238)
-2.338†
(1.211)
-.467†
(.271)
-1.042
(.737)
-.117
(.274)
-.741*
(.307)
-.403***
(.121)
-.020
(.016)
.207**
(.068)
-.006
(.009)
1.183*
(.475)
.275
(.459)
-.004
(.128)
.200**
(.068)
.677**
(.244)
-2.394†
(1.276)
Lexical scale
-
Gini coefficient
-
Max exclusion
-
-
Max discrimination
-
-
-
Max low ration
-
-
-
-
.210***
(.053)
.890***
(.221)
-2.280**
(.849)
.179**
(.059)
.833***
(.231)
-1.646
(1.182)
.199***
(.061)
.751***
(.236)
-2.260†
(1.179)
.202***
(.062)
.743**
(.238)
-2.333†
(1.204)
-.439***
(.118)
-.020
(.016)
.205**
(.070)
-.005
(.009)
1.225**
(.475)
.187
(.454)
-.012
(.123)
.204**
(.065)
.742**
(.238)
-2.347†
(1.215)
144.77
111.32
142.30
142.25
143.12
147.24
161.62
N
7036
6317
6275
6275
6275
6275
6275
N of civil wars
139
127
127
127
127
127
127
GDP per capita(ln)
GDP per capita growth
Total oil income per capita
Population size(ln)
Peace years
Constant
Wald χ2
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
16
Limiting the time-span to the post-1945 period, allows for exploring potential confounding
effects of additional variables. A large body of recent scholarship has found a robust relationship
between horizontal inequalities (i.e., inequalities between groups) and outbreak of civil conflict (e.g.,
Cederman, Weidmann, and Gleditsch 2011; Buhaug, Cederman, and Gleditsch 2013; Cederman,
Gleditsch, and Buhaug 2013). While the relationship between horizontal inequalities and democracy
has been less explored, it is likely that economic discrimination or political exclusion on a group level
also correlate with electoral qualities of regimes. We therefore introduced into the model two additional
variables: horizontal ethno-political inequality—captured by max exclusion (the size of the largest
politically excluded group relative to the combined size of the excluded group and the ethnic group(s)
in power) and max discrimination (the relative demographic power of the largest ethnic group subject
to active discrimination)—and horizontal economic inequality, proxied by max low ration (the relative
income gap between the poorest group and the national average). To account for the level of inequality
in the total population we also introduced vertical economic inequality proxied by Gini index. The four
indices were taken from Cederman, Gleditsch, and Buhaug (2013).
As shown in Table III (Models 2.2–2.5), the four indices have little confounding effect on the
relationship between LIED and civil war. To tests for the non-linear effects of electoral democracy in
the post-1945 period (while also controlling for political and economic inequalities) we similarly
regressed LIED and LIED^2 in the same block (Model 2.6) and a dummy-coded version of LIED
(Model 2.7) (as there is little variation in suffrage extensions in the post-1945 period we collapsed
levels 4 through 6 into a single category, L4). Once again, the estimates provide little support for the
inverted-U hypothesis—all four levels have negative coefficients, with L1 and L4 being significant and
L2 and L3 insignificant. Once again, while L2 is close to significance at 10% level (p = .136), the
coefficient for L3 is far from any conventional level of statistical significance (p = .394).
17
Finally we have tested the sensitivity of the results to alternative measures of the dependent
variable. First, we substituted the COW category of civil war with intrastate war form the UCDP/PRIO
Armed Conflict Dataset v.4-2011, 1946–2010 (Gleditsch et al. 2002; Themnér and Wallensteen 2011)
(further UCDP/PRIO). As shown in the Table IV (Model 3.1), the effect of LIED remains negative and
significant at 5% level. In line with previous estimates, we find little evidence to support the inverted-U
hypothesis (though note that the coefficient for L2 is now positive, p = .555). Second, we substituted
the COW category of civil war with the UCDP/PRIO minor armed conflict, which employs
considerably lower threshold of violence (25 battle-related deaths). As shown in Model 4.4, the
coefficient for LIED remains negative; yet, it drops below the level of significance (p = .400). Also,
unlike in previous models, the L2 and L3 has now positive effect on minor armed conflict onset
(though far from statistical significance), which suggests that one should be hesitant to generalize our
main findings to low intensity conflict.
18
Table IV. Logistic regression estimates of civil war onset (alternative coding of the dependent variable)
UCDP/PRIO (war)
Lexical scale
(3.1)
(3.2)
-.112*
(.045)
.074
(.177)
-.030
(.028)
Lexical scale^2
-
i.Lexical scale:
-
-
L1
-
-
L2
-
-
L3
-
-
L4
-
-
-.321**
(.122)
.0123
(.017)
.239***
(.060)
.006
(.009)
1.039†
(.557)
-.284
(.509)
.204†
(.113)
.341***
(.065)
1.019***
(.208)
-5.426***
(1.341)
UCDP/PRIO (minor armed conflict)
(3.3)
(4.4)
(4.5)
-
-.028
(.033)
-
-
.142
(.132)
-.027
(.020)
-
-
-
-
-
-
-
-
-
(4.6)
-
-
-
-.323**
(-.323)
.012
(.017)
.243***
(.062)
.005
(.009)
.992†
(.553)
-.269
(.507)
.186
(.117)
.353***
(.068)
1.027***
(.208)
-5.573***
(1.345)
-.405
(.293)
.247
(.418)
-.030
(.281)
-.793**
(.291)
-.319**
(.123)
.012
(.017)
.240***
(.062)
.004
(.009)
1.072†
(.577)
-.247
(.526)
.228†
(.117)
.335***
(.070)
.969***
(.201)
-5.284***
(1.389)
-.477***
(.091)
-.006
(.013)
.230***
(.055)
.003
(.007)
.944*
(.436)
-.418
(.384)
.167†
(.096)
.280***
(.049)
.654***
(.160)
-2.789**
(.946)
-.476***
(.091)
-.006
(.013)
.233***
(.056)
.002
(.007)
.905*
(.436)
-.406
(.384)
.152
(.097)
.290***
(.050)
.661***
(.160)
-2.941**
(.957)
-.295
(.220)
.289
(.362)
.170
(.213)
-.280
(.219)
-.473***
(.093)
-.006
(.013)
.232***
(.056)
.001
(.007)
.933*
(.446)
-.404
(.399)
.184†
(.098)
.275***
(.052)
.615***
(.157)
-2.643**
(1.002)
138.31
133.31
138.58
157.03
154.98
161.18
N
6243
6243
6243
6243
6243
6243
N of civil wars
127
127
127
231
231
231
GDP per capita(ln)
GDP per capita growth
Total oil income per capita
Vertical eco. inequality
Ethno-political excl.
Ethno-political discrim.
Horizontal eco. inequality
Population size(ln)
Peace years
Constant
Wald χ2
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
19
Discussion [in progress]
What are the specific implications of these findings?

First and foremost, our analysis provides considerable support to the democratic civil peace
hypothesis. We thus concur with Hegre et al. that “There is a democratic civil peace” and that:
The most reliable path to stable domestic peace in the long run is to democratize as much as possible. A
change in that direction ensures the strongest ratchet effect in terms of consolidating political institutions
and makes it less likely that the country will slide back into a state in which it is more prone to civil war
(2001: 44)

Subsequently, in contrast to a number of previous studies, we find little evidence to support the
inverted-U hypothesis. Electoral regimes in the middle range of the political regime spectrum
covered by LIED do not have higher risk of civil conflict than non-electoral autocracies. Indeed,
the results indicate that all levels of the LIED characterized by national elections have lower
probabilities of civil conflict. Further, we find that the most pacifying effect of electoral
democracy comes from the values of the LIED representing competitive elections.

The fact that our results stand in contrast to multiple previous studies can be explained on
several grounds. One explanation is that our conceptualization of electoral democracy does not
take into account “non-electoral repression,” which is central to the arguments put forward in
support of the inverted-U hypothesis. Moreover, some of the differences are probably also due
to the problems with the middle scores of the Polity index emphasized by Vreeland (2008).

More generally, our analysis provides a glimpse to the mechanisms underlying the democratic
civil peace hypothesis. In contrast to previous research, we have focused on particular qualities
20
of regimes, which has enabled us to shed new light on exact mechanisms through which
democracy is thought to influence conflict.
21
Appendix A: Details on Index Construction (extracted from Skaaning, Gerring, and
Bartusevičius 2014)
To operationalize the levels of the index we make use of four variables from the Political Institutions
and Events (PIPE) dataset (Przeworski et al. 2013), defined as follows:
LEGSELEC: If there is a legislative body that issues at least some laws and does not
perform executive functions and (the lower house of) the legislature is at least partly
elected and has not been closed, the variable is scored 1; otherwise 0.
EXSELEC: If the chief executive is either directly elected or indirectly elected (i.e.,
chosen by people who have been elected), the variable is scored 1; otherwise 0.
OPPOSITION: If there is an legislature that is at least in part elected by voters facing
more than one choice, meaning that all candidates at elections are not presented on the
same, single list, there is not only one party while some candidates run as independents,
parties are not generally banned and everyone run without party labels, and the
legislature has not been dissolved, this variable is scored 1; otherwise 0.
FRANCHISE: This variable is divided into two components, which we refer to as
MALE SUFFRAGE and FEMALE SUFFRAGE. If there is virtually universal suffrage
for male and female citizens, respectively, for national elections, these variables are
scored 1; otherwise 0. Restrictions referring to age, criminal convictions, legal
incompetence, and local residency are not considered as violations of the suffrage
criterion.
Since the quality of elections is not taken into account in the foregoing variables, we generate a new
variable, as follows:
COMPETITION: The chief executive offices and the seats in the effective legislative
body are filled by elections characterized by uncertainty, meaning that the elections are,
in principle, sufficiently free to enable the opposition to gain power. If control over the
key executive and legislative offices is determined, in practice, by the electorate by
means of contested elections, the executive and members of the legislature have not
been unconstitutionally removed, the legislature has not been dissolved, (non-extremist)
22
parties have not been banned, and the constitutional timing of elections is not violated in
a more than marginal fashion, the variable is scored 1; otherwise 0.
This indicator captures whether or not elections are contested, that is, whether there is a
positive probability that the opposition can win government power (see Przeworski et al.
2000: 15–18; Møller and Skaaning 2013). In common with the DD and the BMR
indices, we consider instances of incumbent turnover (as a result of elections) as a
strong indicator of contested elections. However, we do not consider executive turnover
to be either necessary or sufficient for genuinely contested elections.2 It should be clear
that in classifying an election as competitive we are establishing a modest threshold, not
insisting on an entirely level playing field or a high level of respect for civil liberties.
Specifically, elections are considered competitive if the winner of executive and
legislative elections reflects the votes cast by the electorate, as near as can be
determined (from extant sources). This state of affairs is reached if opposition
candidates are generally free to run for these key offices, voters experience little
systematic coercion in exercising their electoral choice, and electoral fraud does not
determine who wins.
A country-year is assigned the highest score (0–6) for which it fulfills all the requisite criteria,
as follows:
0. LEGSELEC=0 & EXSELEC=0.
1. LEGSELEC=1 or EXSELEC=1.
2. LEGSELEC=1 & OPPOSITION=1.
3. LEGSELEC=1 & OPPOSITION=1 & EXSELEC=1.
4. LEGSELEC=1 & OPPOSITION=1 & EXSELEC=1 & COMPETITION=1.
2
It is not necessary since an incumbent party can be sufficiently popular to win a long sequence of genuinely contested
elections, as happened for decades in, e.g., Botswana, Japan, and Sweden. It is not sufficient because opposition power can
gain power through a flawed election if the incumbents have only weak control on power or have stepped down. That said,
in all but a very few cases executive turnover in conjunction with elections is associated with a coding of 1 for
COMPETITION.
23
5. LEGSELEC=1 & OPPOSITION=1 & EXSELEC=1 & COMPETITION=1 &
(MALE SUFFRAGE=1 or FEMALE SUFFRAGE=1).3
6. LEGSELEC=1 & OPPOSITION=1 & EXSELEC=1 & COMPETITION=1 &
MALE SUFFRAGE=1 & FEMALE SUFFRAGE=1.
Countries are coded for the length of their sovereign existence within the 1800–2008 timespan,
generating a dataset with 220 countries and 17,169 country-years. To identify independent countries we
rely on Gleditsch (2013) and Correlates of War (2011), supplemented from 1800 to 1815 by various
country-specific sources. Importantly, electoral democracy does not presume complete sovereignty. A
polity may be constrained in its actions by other states, by imperial control (as over a colony), by
international treaties, or by world markets. Thus, to say that a polity is an electoral democracy is to say
that it functions as such for policies over which it enjoys decision-making power.
Scores for each indicator reflect the status of a country on the last day of the calendar year (31
December) and are not intended to reflect the mean value of an indicator across the previous 364 days.
To qualify as an election the electorate may be quite small—though it must be separable from,
and much larger than, the group of officials it is charged with selecting. Examples would be South
Africa under Apartheid and virtually all national elections in Europe during the nineteenth century. In
measuring universal male and female suffrage we take a juridical approach. Suffrage is achieved when
constitutionally prescribed, even though local or informal practices may impede the achievement of this
right (as in the American South prior to the Civil Rights movement). This is consistent with the usage
of the concept by Schumpeter and Przeworski and also with many extant indices (e.g., Polity2 and
DD). Indirect elections do not qualify as “elections” unless the electors endorse specific candidates or
parties, as in US presidential elections.
Although we employ PIPE as an initial source for coding LEGSELEC, EXSELEC,
OPPOSITION, and FRANCHISE (MALE SUFFRAGE and FEMALE SUFFRAGE), we deviate from
PIPE codings—based on our reading of country-specific sources—in several ways. First, with respect
to executive elections, in the PIPE dataset “Prime ministers are always coded as elected if the
legislature is open.” However, for our purposes we need an indicator that also takes into account
3
In no extant cases was universal female suffrage introduced before universal male suffrage, so in practice this level is
reserved for countries with male (only) suffrage.
24
whether the government is responsible to an elected parliament if the executive is not directly elected—
a situation generated by a number of European monarchies prior to World War I, by episodes of
international supervision such as Bosnia-Herzegovina in the first years following the civil war, and by
some monarchies in the Middle East and elsewhere (e.g., Liechtenstein, Monaco, and Tonga) in the
contemporary era. To illustrate, PIPE codes Denmark as having executive elections from 1849 to 1900
although the parliamentary principle was not established until 1901. Before then, the government was
accountable to the king. Among the current cases with elected multiparty legislatures not fulfilling this
condition, we find Jordan and Morocco. In order to achieve a higher level of concept-measure
consistency, we have thus recoded all country-years (based on country-specific accounts) for this
variable where our sources suggested doing so.
Moreover, we complete all missing values (and missing countries, e.g., the German
principalities of the nineteenth century) in the PIPE dataset, generating a complete dataset for all
conditions for all independent countries of the world in the period under study (1800–2008). Whereas
the numbers of observations for the employed PIPE indicators range between 14,465 and 15,302, our
dataset provides 17,179 observations for all indicators. Except for minor adjustments regarding
executive elections (mentioned above), this additional coding follows the rules laid out in the PIPE
codebook. Coding decisions are based on country-specific sources that are too numerous to specify. In
rare instances we stumbled upon information that required a recoding of PIPE variables, so the two
datasets do not correspond exactly.
25
References [not fully completed]
Auvinen, Juha. 1997. Political conflict in less developed countries. Journal of Peace Research 34 (2):
177-95.
Boix, Carles, and Susan C. Stokes. 2003. Endogenous democratization. World Politics 55 (04): 517-49.
Jutta Bolt and Jan Luiten van Zanden, "The First Update of the Maddison Project: Re-Estimating
Growht before 1820" Maddison-Project Working Paper WP-4, 2013),
http://www.ggdc.net/maddison/maddison-project/publications/wp4.pdf (accessed 27 May 2014).
Buhaug, Halvard. 2006. Relative capability and rebel objective in civil war. Journal of Peace Research
43 (6): 691-708.
Buhaug, Halvard, Lars-Erik Cederman, and Kristian Skrede Gleditsch. 2013. Square pegs in round
holes: Inequalities, grievances, and civil war. International Studies Quarterly.
Cederman, Lars-Erik, Gleditsch, Kristian S., and Buhaug, Halvard. Inequality, grievances, and civil
war. New York, NY: Cambridge University Press, 2013.
Cederman, Lars-Erik, Kristian Skrede Gleditsch, and Simon Hug. 2013. Elections and ethnic civil war.
Comparative Political Studies 46 (3): 387-417.
Cederman, Lars-Erik, Nils B. Weidmann, and Kristian Skrede Gleditsch. 2011. Horizontal inequalities
and ethnonationalist civil war: A global comparison. American Political Science Review 105 (3):
478-95.
Collier, Paul, and Anke Hoeffler. 2004. Greed and grievance in civil war. Oxford Economic Papers 56
(4): 563-95.
Davenport, Christian and David Armstrong. 2004. Democracy and the violation of human rights: A
statistical analysis from 1976–1996.” American Journal of Political Science 48:3, 538–54.
Elbadawi, Ibrahim, and Nicholas Sambanis. 2002. How much war will we see?: Explaining the
prevalence of civil war. Journal of Conflict Resolution 46 (3): 307-34.
Fearon, James D., and David D. Laitin. 2003. Ethnicity, insurgency, and civil war. American Political
Science Review 97 (1): 75-90.
Fjelde, Hanne. 2010. Generals, dictators and kings: Authoritarian regimes and civil conflict 1973 –
2004, Conflict Management and Peace Science, 27 (3): 195-218.
Gleditsch, Kristian Skrede. 2004. A revised list of wars between and within independent states, 18162002. International Interactions 30 (3): 231-62.
26
Gleditsch, Kristian Skrede and Andrea Ruggeri. 2010. Political opportunity structures, democracy, and
civil war. Journal of Peace Research 47(3): 299-310.
Goldstone, Jack A., Robert H. Bates, David L. Epstein, Ted Robert Gurr, Michael B. Lustik, Monty G.
Marshall, Jay Ulfelder, and Mark Woodward. 2010. A global model for forecasting political
instability. American Journal of Political Science 54 (1): 190-208.
Gurr, Ted R. Peoples versus states: Minorities at risk in the new century. Washington, DC: United
States Institute of Peace Press, 2000.
Haber, Stephen, and Victor Menaldo. 2011. Do natural resources fuel authoritarianism? A reappraisal
of the resource curse. American Political Science Review 105 (01): 1-26.
Hegre, Håvard. 2014. Democracy and armed conflict. Journal of Peace Research 51 (2): 159-72.
Hegre, Håvard, Tanja Ellingsen, Scott Gates, and Nils Petter Gleditsch. 2001. Toward a democratic
civil peace? democracy, political change, and civil war, 1816–1992. American Political Science
Review 95 (1): 33-48.
Hegre, Havard, and Nicholas Sambanis. 2006. Sensitivity analysis of empirical results on civil war
onset. Journal of Conflict Resolution 50 (4): 508-35.
Jensen, Nathan, and Leonard Wantchekon. 2004. Resource wealth and political regimes in Africa.
Comparative Political Studies 37 (7): 816-41.
Krain, Matthew, and Marissa E. Myers. 1997. Democracy and civil war: A note on the democratic
peace proposition. International Interactions 23 (1): 109-18.
Lipset, Seymour Martin. 1959. Some social requisites of democracy: Economic development and
political legitimacy. The American Political Science Review 53 (1): 69-105.
Marshall, Monty G., Gurr, Ted R. and Jaggers, Keith. Polity IV project: Dataset user's manual. 2013
[cited 5/27 2014]. Available from http://www.systemicpeace.org/inscr/p4manualv2012.pdf.
Miguel, Edward, Shanker Satyanath, and Ernest Sergenti. 2004. Economic shocks and civil conflict:
An instrumental variables approach. Journal of Political Economy 112 (4): 725-53.
Møller, Jørgen, and Svend-Erik Skaaning. 2013. Regime Types and Democratic Sequencing. Journal
of Democracy 24 1): 142–56.
Møller, J. and S. Skaaning. 2012. Systematizing Thin and Thick Conceptions of the Rule of Law,
Justice Systems Journal, 33 (2): 136-153.
27
Møller, J. and S. Skaaning. 2014. The rule of law: Definitions, measures, patterns and causes.
Houndmills: Palgrave.
Muller, Edward N., and Erich Weede. 1990. Cross-national variation in political violence: A rational
action approach. Journal of Conflict Resolution 34 (4): 624-51.
Przeworski, Adam, et al. 2013. Political Institutions and Political Events (PIPE) Data Set. Department
of Politics, New York University.
Przeworski, Adam, Alvarez, Michael E., Cheibub, Jose A., and Limongi, Fernando. Democracy and
development. Cambridge: Cambridge University Press, 2000.
Ravallion, Martin. 2011. Mashup indices of development. World Bank Research Observer 27: 1–32.
Reynal-Querol, M., 2002. Political systems, stability and civil wars. Defense and Peace Economics 13,
465– 483.
Ross, Michael L. 2012. The oil curse: How petroleum wealth shapes the development of nations.
Princeton: Princeton University Press.
Ross, Michael L. 2004. What do we know about natural resources and civil war? Journal of Peace
Research 41 (3): 337-56.
Ross, Michael L. 2001. Does oil hinder democracy? World Politics 53 (03): 325-61.
Rummel, Rudolph J. 1997. Power Kills. New Brunswick: Transaction
Publishers. Sarkees, Meredith R., and Wayman, Frank. Resort to war: 1816-2007. Washington, DC:
CQ Press, 2010.
Schrodt, Philip A. 2014. Seven deadly sins of contemporary quantitative political analysis. Journal of
Peace Research 51 (2): 287-300.
Schumpeter, Joseph A. 1950. Capitalism, Socialism and Democracy. New York: Harper & Bros.
Singer, J. D. 1987. Reconstructing the correlates of war dataset on material capabilities of states, 1816–
1985. International Interactions 14 (2): 115-32.
Strand, Håvard. 2007. Retreating from a civil democratic peace? Revisiting the relationship between
political institutions and civil war. Oslo: Political Science Department, University of Oslo.
28