Content uploaded by Dan Reiter
Author content
All content in this area was uploaded by Dan Reiter on Mar 06, 2015
Content may be subject to copyright.
“State Power, Paramilitary Forces, and Internal Violence”
Jun Koga and Dan Reiter
Department of Political Science
Emory University
Atlanta, Georgia 30322
(404) 727-0111
dreiter@emory.edu
February 23, 2011
Abstract: This paper explores the relationship between state power and internal violence.
Conventional wisdom predicts that more powerful states should be less likely to experience
internal violence. However, existing scholarship has failed to uncover robust empirical
relationships between various measures of state power, such as geography and roads, and
internal conflict. This paper presents a new empirical test, exploring the relationship between
paramilitary forces and internal violence. The paper presents analysis of a new data set on the
national paramilitary forces for all states from 1969-2003. The analysis finds that larger
paramilitary forces significantly decrease internal violence. This relationship holds for some but
not all measures of internal violence.
Draft. Comments welcome. Do not circulate.
What factors make it more likely that a state will experience internal violence? The most
straightforward answer to this question focuses on state power, the state’s ability to exercise
control and autonomy over its society and territory, and to carry out its policies. Indeed, some
view the very function of the state to be the prevention of violence. This was certainly Thomas
Hobbes’ central point, that society needed a Leviathan to prevent people from warring against
each other. More recently, the examination of “failed states” has focused on the synergistic
collapse of state power and authority coupled with the rise of internal violence. Beyond the
failed states concept, modern social scientists have endeavored to explore the relationship
between state power and internal violence, whether variations in different measures of state
power correlated with variations in internal violence.
Surprisingly, empirical scholarship has generally not borne out the theoretical expectation
that more state power translates into lower levels of internal violence. Namely, very few
consistent relationships have emerged between nearly any measure of state power and the
likelihood of internal conflict. Studies have either found no relationship between state power
and internal violence, the opposite of the predicted relationship (more state power is associated
with higher levels of internal violence), or perhaps scattered evidence of the predicted
relationship. The absence of consistent, empirical patterns that stronger states experience greater
internal political stability is a real puzzle.
That being said, previous research on the causes of internal conflict has not examined the
full range of measures of state power, generally focusing on measures of state power that are
perhaps only indirectly related to a state’s ability to deter and combat violence. These measures
include factors such as road networks, terrain, levels of economic development, extractive
capacity of the government, and others. Little scholarship has yet examined elements of state
1
power which are more directly useful and immediately available for deterring and combating
internal violence.
This paper presents an empirical test of a measure of state power which gets directly at a
state’s ability to deter and combat violence: a state’s level of paramilitary forces. States use
paramilitary forces more than regular military forces to combat insurgencies and address other
forms of internal violence. The central question this paper asks is: Do higher levels of
paramilitary forces reduce the incidence of internal violence?
To answer this question, this paper analyzes new data on the paramilitary forces of all
states from 1969-2003. In statistical analyses of internal violence for all states across this time
period, it finds evidence that higher levels of paramilitary forces reduce the incidence of internal
violence. Across four dependent variables, the analysis finds there is a negative and clearly
statistically significant relationship for two measures, a negative and marginally statistically
significant relationship for one measure, and no statistically significant relationship for the last
measure. In a separate analysis, the paper finds that higher levels of internal violence do not
affect a state’s level of paramilitary forces, addressing endogeneity concerns.
The paper proceeds in four parts. First, it discusses existing scholarship on the
relationship between state power and internal conflict. Second, it discusses the research design
used to test the hypothesis that higher levels of paramilitary forces are associated with lower
levels of internal conflict. Third, it presents the results of statistical analysis. Fourth, it
concludes.
State Power and Internal Conflict
2
The sociologist Michael Mann distinguished between two different forms of state power.
The first is despotic power, “the range of actions which the elite is empowered to undertake
without routine, institutionalized negotiation with civil society groups.” The second is
infrastructural power, the “capacity of the state to actually penetrate civil society, and implement
logistically political decisions throughout” the country (Mann 1988, 5).1
Scholars have offered a variety of hypotheses connecting state power, and state
infrastructural power in particular, with various behavioral or political outcomes. Dan Slater
(2010) argued that regime type and in particular party structures provide a state with
infrastructural power, and that state infrastructural power is an important tool which can extend
the duration of authoritarian regimes. Brian Lai and Slater (2006) used the same concept,
proposing that because single party regimes can use their party structures to maintain control of
society, they are unlikely to need to initiate the use of military force as a means of diverting
public attention away from declining domestic political and economic conditions. However,
military juntas have less infrastructural power because they do not have such party structures,
and hence are more likely to initiate military force to divert public attention away from declining
domestic conditions. Daniel Schensul (2008) found that variations in state infrastructural power
explain variations in the success of attempts in post-apartheid Durban, South Africa to reduce
economic and other forms of inequality. Matthias Vom Hau (2008) posited that state
1 Mann’s notion of state infrastructural power has attracted substantial scholarly attention, and spirited debate.
Scholars have differed on whether the focus should be on resources or on outcomes. Some have been satisfied by
focusing on indices of capability, such as revenue collection or numbers of armed troops. Others have challenged
the assumption that capabilities necessary translate into power, arguing that true power can only be judged by the
degree to which the behavior and identities of society and societal actors have actually been changed (Soifer 2008).
“State power” is similar to the concept of “state capacity.” In the context of the civil war scholarship, definitions of
state capacity have paralleled Mann’s notion of state power, and state infrastructural power in particular. State
capacity has been defined as the ability to “deter or repel challenges to its authority with force,” “the repressive
capacity of the state during the early stages of conflict,” its “ability to address the demands of their citizens in ways
that reduce the incentive for political violence,” and the “ability to accomplish those goals it pursues” (Hendrix
2010; Sobek 2010; DeRouen et al 2010).
3
infrastructural power affected the development of legitimacy in Mexico and Argentina, and
especially the evolution of popular nationalism. Hillel Soifer (2009) demonstrated that higher
state infrastructural power mediates the relationship between income inequality and regime
outcomes. Daniel Ziblatt (2008) connected state infrastructural power to the provision of public
goods.
Though there is a growing body of scholarship on the effects of state infrastructural
power, there is perhaps surprisingly relatively little scholarship on the sources of state
infrastructural power. One exception is Slater (2008), who proposed that postcolonial
democracies are more likely to build infrastructural power in the context of competitive elections
amid widespread mass mobilization. Cullen Hendrix (2010) explored some fifteen different
measures of state power to explore for possible dimensionality underlying these different
measures. Using factor analysis, he discovered three different conceptual sets, rationality
legality, rentier-autocraticness, and neopatrimoniality.
There are two different ways to think about the manifestations of state infrastructural
power. The first is to think of state infrastructural power as similar to factor endowments, assets
controlled by a state which are available merely because of their existence. This factor
endowment approach imagines elements like natural resource wealth that can provide revenue
that can then be converted into more specific manifestations of state power, like roads or well-
armed troops. The second way is to think about the specific manifestations of state power
themselves, like roads or troops. Note that if one focuses on the specific manifestations, then the
level of these manifestations is due not only to factor endowments, but also to explicit decisions
made by governments to build state infrastructural power, and they are often the product of
budgetary choices.
4
In the following subsections, we examine a variety of measures of state power that past
scholarship has considered as possibly affecting the likelihood of internal violence. We explore
the theoretical logic as to why it might be related to internal violence and discuss existing
empirical findings connecting it to internal violence. In general, the empirical evidence
connecting state power and internal violence is surprisingly weak.
Terrain
Some have proposed that a country’s terrain affects the state’s infrastructural power as it
relates to the likelihood of internal violence. More mountainous or heavily forested terrain gives
insurgents more places to hide, and makes it difficult for states to extend their reach throughout
their national territories through the deployment of troops, tax collectors, government
bureaucrats, or other manifestations of the state. Rough terrain also reduces the advantage of
mechanization the government might have over insurgents, as it is more difficult for vehicles to
traverse mountainous and forested terrain, and forests provide insurgents cover from government
aircraft. That being said, some have proposed that mechanization may actually serve to hinder
rather than aid counterinsurgent forces, because it discourages efforts to gather information from
the population (Lyall and Wilson 2009).
The empirical findings of the effects of terrain on insurgency are mixed. For their cross-
national study of civil war onset in the post-World War II period, James Fearon and David Laitin
(2003) hired a geographer, A. J. Gerard, to develop a single score for each country in the world
measuring the extent to which that country is mountainous. They found that civil war onset is
more likely in countries with more mountainous terrain.
5
This is a contentious finding. Matthew Lange and Hrag Balian (2008), Stathis Kalyvas
and Laia Balcells (2010), and Clayton Thyne (2006) also conducted cross national statistical
analyses of civil war onset, used the Gerard data, and found no relationship between mountains
and civil war onset. Beyond reanalysis of the Gerard data, some scholars have critiqued the
approach of using a single mountains score for an entire country. They posit that countries are
not uniformly mountainous or flat, and to assume otherwise risks aggregation error. They
suggest breaking countries into subnational zones, exploring whether terrain is correlated with
the onset of violence within each smaller zone. This enables the distinction between subnational
areas which are highly mountainous (such as the Himalayan regions of China) and areas which
are flatter (such as the coastal plains in southeast China), and avoids the problem of having to
provide a single score for the variable terrain of a single country.
Halvard Buhaug and Jan Ketil Rød (2006) took this subnational approach, using
Geographical Information Systems (GIS) software to look at sub-state regions, dividing African
countries up into 100 km x 100 km squares. They found results opposite to those of Fearon and
Laitin, that more mountainous regions were less likely to experience the onset of violence than
other regions. They also found that forested regions were not more or less likely to experience
violence.2 Lars-Erik Cederman et al (2009) also used GIS data on sub-state regions, and they
found only mixed evidence that mountains made civil war onset more likely. Steven Pickering
(2009) developed a new measure of terrain using NASA data, finding suggestive evidence that,
again using sub-state regions, there is a positive relationship between rugged terrain and internal
conflict. Clionadh Raleigh (2010) also used subnational data, but found in multivariate models
that elevation and forests were uncorrelated with conflict, for six central African states. Alok
2 Buhaug and Lujala (2005) found that civil wars in mountainous regions lasted longer, but civil wars in forested
regions were shorter.
6
Bohara et al (2006) found that terrain mattered in comparing violence levels across subnational
regions within Nepal, though Quy-Toan Do and Lakshmi Iyer (2010) found that in Nepal
geography is related to conflict intensity, but not onset.
Terrain is seen as the least manipulable of the factors affecting state power, though it is
not totally immutable. Mountains are generally unmanipulable, “mountain top removal”
schemes aside, but states can build roads and rail networks which reduce the ability of mountains
to restrain the extension of state power (see below). Forest cover is difficult to manipulate,
though some counterinsurgents might destroy natural ecosystems to make insurgents more
vulnerable, in some ways literally “draining the swamp.” For example, the United States
employed herbicides such as Agent Orange in the Vietnam War to defoliate Vietnamese forests
and eliminate potential cover for the Viet Cong (Buckingham 1983). In 1992, Iraqi government
forces drained the marshes in southern Iraq to facilitate counterinsurgency against the Shiite
rebels (Cordesman and Hashim 1997, 103).
Centrality of National Capital
Jeffrey Herbst (2000) examined the effects of geography on political stability in African
states. He proposed that one reason that African countries have been more susceptible to civil
war onset is because of the locations of their national capitals. Many African countries were
European colonies, and the cities which now serve as national capitals were built by Europeans.
However, the Europeans often built these cities on the coasts, to facilitate the export of the raw
materials produced by the colony. Herbst argued that this path-dependent process causing
African capitals to be located far from the geographic center of the country undermined state
7
power. The capital is the hub of state power, and the ability of the state to extend and apply its
power within a geographical district decreases as the distance between the capital and the district
increases. Locating the national capital on a national border increases the distance to many
geographic districts, making civil war outbreak more likely.
Empirical scholarship on this proposition has found at best mixed support. Weidmann et
al (2010) reanalyzed the Fearon and Laitin (2003) data by adding a variable measuring the
distance between the national capital and the geographic center of a country, and found no
relationship between this distance and the onset of civil war. Buhaug and Rød (2006), on the
other hand, found empirical support for the capital distance proposition, discovering that in
Africa the farther a geographic district is from the national capital, the more likely it is that civil
war will break out there. Cederman et al (2009) found mixed support for the proposition. In a
study of subnational regions within Liberia, however, Hegre et al (2009) found that regions
closer to the national capital were more likely to experience violence. More recently, Buhaug
(2010) found that when the state is strong in relation to the rebels, conflict tends to occur farther
from the capital, and when the state is weak in relation to the rebels, conflict tends to occur
closer to the capital.
The location of the national capital is relatively but not completely unmanipulable.
Sometimes governments elect to move their national capitals. The Soviet government moved its
national capital in 1918 from St. Petersburg to Moscow as German forces advanced, and
prepared to evacuate Moscow in October 1941 as German forces again threatened. Soon after its
revolution the United States moved its capital from Philadelphia to New York to the newly
created District of Columbia, placing the last on the banks of the Potomac River in part because
such a location was close to the geographic center of the new republic (Crew et al 1892, esp. 75).
8
Brazil moved its capital from Rio de Janeiro to the newly created, more centrally located city of
Brasilia in the 1950s and 1960s. The Federal Republic of Germany moved its capital from Bonn
to Berlin soon after German reunification in 1990. Sometimes countries decide to move their
national capital in reaction to threats of internal instability. Burma is moving its national capital
from Rangoon to a more central location near the town of Pyinmana, perhaps to make it easier to
control potentially restive ethnic minority groups in the Shan, Chin, and Karen border states
(McGeown 2005).
Economic Development
Scholars have posited a relationship between economic development and civil war onset
for decades, as far back at least as Ted Gurr’s (1970) relative deprivation hypothesis. A more
contemporary proposition has been that lower levels of economic development generate higher
levels of grievance, which in turn makes the population more willing to support insurgency
(Collier and Hoeffler 2004). Fearon and Laitin (2003) drew a new connection between
development and internal conflict, positing that higher levels of development mean that states
have greater “financial, administrative, police, and military capabilities” and enable the state to
better penetrate the entire country with more extensive roads. They also posit that higher levels
of development mean that individuals are less likely to join insurgencies. They found that gross
domestic product (GDP) per capita, a development proxy, is negatively associated with the
outbreak of civil war in the post-1945 period.
Three caveats to the Fearon/Laitin proposition and finding are in order. First, economic
development can account for several different causal forces affecting civil war onset. Some of
9
these causal processes do relate to state infrastructural power, such as the bureaucratic
effectiveness of the state, the military and police assets of the state, and the physical
infrastructure of the economy. However, GDP/capita also reflects economic privation, and
higher levels of economic privation may make insurgency more likely (the grievance hypothesis
(Collier and Hoeffler 2004)), state infrastructural power aside. Second, the empirical evidence
on the relationship between GDP/capita and internal conflict is mixed. Cederman et al (2009)
found that GDP/capita was not correlated with intrastate violence over roughly the same time
period, and Buhaug et al (2008) found significant relationships in only some of their models. In
a careful study of Liberia, Hegre et al (2009) found that richer districts within Liberia were
actually more prone to violence than poorer districts. Third, there is evidence that the occurrence
of civil war undermines development, raising the question of the direction of the causal arrow.
People do not engage in economically productive activity during civil war, because they are
involved in fighting, because they are deterred from going to work, and/or because their factories
or offices have been destroyed. Civil war has other negative effects on development, such as
undermining public health and destroying physical infrastructure (Iqbal 2010). Paul Collier
(2007) refers to the mutual interaction between civil war and poverty as the “conflict trap.”
Transportation Infrastructure
A nation’s transportation infrastructure is an important component of state infrastructural
power. Roads, rail, and airports are the principal means by which the state can distribute goods
and services to the rest of the country. More extensive and higher quality roads, rail, airports,
and seaports mean that governments can more cheaply and more quickly dispatch teachers,
10
construction equipment, health care supplies, food, and administrators to every part of the
country. The 2006 United States Army/Marine Counterinsurgency Manual recommends
investing in infrastructure projects like building roads as a means of supplying goods of direct
use to the population, thereby reducing popular grievances against the government and
decreasing likely popular support for insurgents (Counterinsurgency 2006, 5-21). Further, the
state uses its physical infrastructure to project effectively its military and policing power across
the breadth of the country. Roads, rail, airports, and seaports of higher quality and quantity mean
that the government can more quickly and cheaply dispatch larger numbers of troops and masses
of material through the entire breadth of the country to show force or put down insurgents.
Few studies have explored the relationship between physical infrastructure and civil war
onset, and those studies that have examined the connectiondo not provide evidence clearly
supporting the hypothesis that more extensive infrastructure means a lower chance of internal
violence. Perhaps surprisingly, Buhaug and Rød (2006) found only very limited empirical
support for the proposition that higher road density translates into a lower probability of civil
war. Raleigh (2010) showed that in six African countries subnational areas with more roads,
cities, and airports were more likely to experience conflict. She hypothesized that these areas
were of higher value, and hence more likely to see rebels and government forces clashing over
them. Relatedly, Jason Lyall and IsaiahWilson (2009) found that less mechanized
counterinsurgency efforts are more successful than more mechanized efforts, as
counterinsurgency forces will be more effective when dismounted. However, more extensive
road networks do not necessarily mean that counterinsurgents must stay in their trucks and
helicopters.
11
The causal arrow may run in the opposite direction, as the threat of internal violence may
affect the incentives of states to invest in transportation networks. States will expand and
improve their transportation infrastructures to increase their power and deter and combat
insurgents. The Roman Empire built the world’s first road network to maintain internal order
(Starr 1982, 116-120). The urban landscape (including the streets) of Paris was substantially
redesigned after the revolutions of 1848, partly to make it easier to put down insurrections
(Jordan 1995, 188-189). Iraq built roads in the southern portion of the country in 1992 to help
combat the Shiite insurgency (Cordesman and Hashim 1997, 103). In recent years China
completed a high altitude railway to Tibet as a means of tightening its control over the restive
population there (Ni 2006). India has built roads to sap support for the Maoist insurgency, while
Maoists have sabotaged roads in India in an attempt to undermine state power (Yardley 2009,
16). Road construction is seen by some as a “cornerstone” of the counterinsurgency effort in
Afghanistan (Mockenhaupt 2009, 136). Conversely, some leaders may deliberately neglect their
state’s physical infrastructure so as to hamper potential rebels. The Zairean dictator Joseph
Mobutu Sese Seko resisted investing in roads and infrastructure in his country, as he believed
that better roads would facilitate collective opposition to his regime (Dunning 2005, 465-6).
There are difficulties in executing cross-national analysis of the effects of road density
(perhaps the most important element of physical infrastructure) on civil war onset. Among all
measures of physical infrastructure (non-transportation measures of infrastructure include items
such as access to water and energy grids), the data on transportation infrastructures may be the
worst. Existing cross-national data does not describe road quality, does not distinguish between
rural roads and multilane superhighways, and does not account for maintenance levels of roads
(Estache and Fay 2007; Canning 1998). There are also signs that codings in the data may have
12
changed because of undescribed, state-by-state changes in coding rules (that is, the way that
nations count roads) rather than changes in the actual transport networks. For example, in the
major cross-national data set on national road networks, there are a number of instances in which
the kilometers of recorded road network went down substantially from one year to the next,
including Congo from 1955 to 1957, Madagascar from 1969 to 1970, the US from 1980 to 1981,
Japan from 1986 to 1987, Czechoslovakia from 1982 to 1985 (a decrease of 49%), and others.
Short of major interstate war, it is hard to imagine circumstances in which a consequential length
of roads would disappear from one year to the next. This is not a trivial problem. In the
International Roads Federation data set, which includes annual roads data for all countries from
1963-1989,3 the IRF codes a decline in a country’s number of kilometers of road from one year
to the next in 14% of the country-years. Buhaug and Rod (2006) use the Digital Chart of the
World to produce roads data.4 This data set was originally produced by the Environmental
Systems Research Institute for the Defense Mapping Agency, in 1993. One shortcoming of the
data set is that it provides a one-time snapshot of road network data, that is, there is not temporal
variation.
Foreign-Imposed Regime Change
Occasionally, one state will attack another state and overthrow its leadership. Peic and
Reiter (forthcoming; see also Enterline and Greig 2008) hypothesized that foreign imposed
regime change (FIRC) following war is likely to eviscerate the infrastructural power of the
targeted state. A state suffering a FIRC following a war may see its armed and police forces
3 Available at http://www.hsph.harvard.edu:80/faculty/david-canning/data-sets/ <downloaded October 20, 2009>.
4 For more information on the Digital Chart of the World, see http://www.maproom.psu.edu/dcw/ <downloaded
October 15, 2009>.
13
destroyed, its physical infrastructure wrecked, its national capital leveled, and the government
administrators who enable the day to day operation of state and society by performing mundane
tasks such as operating utilities and collecting taxes removed from power. Because FIRCs
following wars destroy state infrastructural power, they make the outbreak of civil war more
likely. Examining all civil wars from 1920-2004, Peic and Reiter found that foreign imposed
regime changes which followed war and imposed changes on political institutions increased the
likelihood of civil war onset greatly. In a related paper, Krisitian Gleditsch and Andrea Ruggeri
(2010) focused on the more general phenomenon of irregular entry into and exit from power,
FIRC being one example of irregular leadership transition. They proposed that irregular
leadership transitions can indicate state weakness, and found that irregular leadership transitions
make the onset of internal violence more likely.
National Census
Hillel Soifer (2009) has argued that conducting a national census both reflects and
contributes to state infrastructural power. It reflects state infrastructural power because a state
needs to have a certain degree of control over its territory and population in order to implement a
census. It contributes to state infrastructural power because the information provided by a
national census can be used to effectively distribute resources. He explored the relationship
between regime type and economic redistribution, and found higher state infrastructural power as
indicated and boosted by the presence of a national census to be a key intervening variable.
However, Jun Koga and Dan Reiter (2009) found that taking a national census did not make civil
war onset significantly less likely.
14
The decision to take a national census may be endogenous to the risk of internal violence.
A census can be used to reapportion economic and political power, and in a region or nation in
which there are deep ethnic conflicts (and hence a high structural risk of civil war) states may be
unwilling to carry out a census. Even though the new Iraqi constitution calls for a census, the
national government has delayed holding it, as the political implications of any census results
(such as indicating that the Shiites are not as numerous as they claim to be) might be sufficient to
reignite civil war in that country. Lebanon has not held a census since 1932, and any new census
there that reapportioned political power might cause the groups losing representation to resort to
civil war (Nordland 2009; Myers 2010).
Size of Public Sector
Hanne Fjelde and Indra de Soysa (2009) made two arguments as to why the size of the
public sector, roughly speaking, would affect state infrastructural power and the likelihood of
civil war.5 They looked at two measures, the state’s capacity for revenue extraction from the
economy and the size of government expenditure as a percentage of GDP. They proposed that
the first reflects the government’s ability to deter and crush insurgency, and the latter gives the
state the power to distribute goods to reduce societal grievance. In quantitative tests for the years
1961-2004, they found that revenue extraction was not significantly related to civil war onset,
but government expenditures was.6 Alex Braithewaite (2010) found that internal violence was
5 They use the phrase “state capacity,” though their definition (7), “the extent to which governmental agents control
state activities and resources within the government’s territory,” generally equates to Mann’s definition of state
infrastructural power.
6 They also hypothesized protection of property rights generates public trust in government, and such legitimacy can
in turn be used by the government as a means of social control. Their measure of property rights was negatively and
significantly related with the onset of civil war.
15
less likely to spread to a state with higher levels of relative political capacity, though relative
political capacity itself was uncorrelated with civil war onset. Cameron Thies (2010) also found
that levels of government revenue, taxation, and relative political capacity did not affect the
likelihood of civil war onset. Thies did find, however, that civil war onset undermined state
capacity, specifically government revenue, taxation, and political extraction.
Summary
Many factors which ought to be critical aspects of state infrastructural power affecting
the outbreak of civil war are not consistently correlated with the outbreak of civil war across
scholarship and research designs. Many variables do not acquire consistently supportive
empirical results. Economic development does attract consistent empirical results, but that
positive relationship may be due to factors beyond state power. There is good evidence that
foreign imposed regime change following interstate war makes civil war more likely, but those
events are relatively rare, and so can not constitute a comprehensive account of the outbreak of
civil war.
There are a few reasons why past research has failed to find strong connections between
state power and internal violence. One possibility is that greater state power might both increase
and decrease internal violence. Lange and Balian (2008) speculated that greater state power
might make civil war more likely if the reach of the state served to provoke the population to
support insurgents (see also Peic and Reiter (forthcoming)). Larger military and paramilitary
forces in particular might be especially likely to provoke a population to support insurgents. The
presence of occupation forces following a foreign imposed regime change might also provoke a
16
population to support insurgency, though Peic and Reiter (forthcoming) found that there was not
a significantly different risk of civil war onset following FIRCs with occupation forces as
compared to FIRCs without occupation forces. Though the common view from Iraq and
Afghanistan is that an occupation force can inspire insurgency, studies have found that at least in
some instances an occupation force can deter insurgent activities (Willard-Foster 2009; Lyall
2009). Greater extractive capacity might be more provocative, as the population might resent the
presence of tax collectors. A national census might be especially likely to provoke internal
violence in nations which could be characterized as having ethnic divisions. Hence, when
aggregating together both kinds of cases, dampening violence and provoking violence, into a
single data set, the net observed effect may wash out.
A second possible explanation of the null results between measures of state
infrastructural power and civil war onset regards the possible endogeneity of state infrastructural
power. One of the very core functions of state infrastructural power and indeed of the state itself
is to maintain internal order. A state may be motivated to spend more to expand its
infrastructural power if it perceives a growing threat of internal disorder. This dual causality
may introduce unrecognized nuance into empirical results.
A third point is that physical assets of state power may encourage internal violence, either
because insurgents want to destroy things the state values, or the insurgents want to capture items
of value. This is generally true of commodities like secondary diamonds and oil, which can
encourage rebel activity. It can also be true of other items which directly contribute to state
power. Raleigh (2010) found that areas which contain several items which are both valued and
contribute to state power, such as airports, cities, and roads, experience higher levels of internal
violence. Historically, the construction of railroads and more broadly industrialization in Russia
17
helped engender political and social change which eventually led to the 1905 revolution
(Acemoglu and Robinson 2006, esp. 116).
Fourth, there may also be unrecognized interactive factors which distort the relationship
between state power and internal violence. For example, studies examining the relationship
between terrain (specifically, variance in levels of elevation) and internal violence have as yet
failed to examine the interactive effects of urbanization, as cities tend to be as flat (in terms of
elevation differentials) as deserts, and yet state power, especially airpower, is much more easily
wielded against rebels in deserts than in cities, as buildings provide man-made terrain which can
make it easier for insurgents to organize and hide. Further, studies have not yet attempted to
understand how transportation networks might ameliorate the effects of rugged terrain.
Paramilitary Forces and Internal Conflict
The factor most immediately useful in deterring and combating insurgents is also the
factor which has ironically been studied the least: paramilitary forces. At least one element of
counterinsurgency strategy will be the application of violent force, ranging from putting down
political demonstrations with force to small operations to capture or assassinate insurgent leaders
to major military offensives. No quantitative studies have yet assessed the relationships between
size of paramilitary forces and likelihood of internal violence.7 David Cunningham et al (2009)
coded a dichotomous variable on the fighting capacity of rebels in relation to the government, for
insurgencies which have occurred. They found that greater rebel strength was not significantly
7 Some studies have explored the relationship between police forces and crime rates. They have found that once
endogeneity is controlled for, there is no relationship between larger police forces and crime rates, though some
studies suggest that variation in policing strategies can affect the crime rate (see Eck and McGuire 2006; Leavitt
1997; Fajnzylber et al 2002).
18
related to conflict duration, but that greater rebel strength did make a rebel victory more likely in
relation to other outcomes such as government victory or compromise through formal agreement.
Lyall and Wilson (2009) included a variable measuring the government’s average share of power
among all states in the system across the six Correlates of War (COW) indicators of population,
urban population, number of troops, military spending, energy consumption, and iron and steel
production. They found that this measure was positively correlated with counterinsurgent
victory in the pre-1918 “foraging” era, but was not correlated in the post-1917 “mechanized” era.
Kalyvas and Balcells (2010) found that the number of a state’s military personnel was negatively
correlated with the onset of symmetric non-conventional civil wars, but not with conventional
civil wars. They used the COW measure of regular military personnel.
One possible reason for the shortage of analysis on paramilitary forces is the paucity of
useful data. There is systematic data on regular military troops, and has been used in some
analyses (see above). However, not all regular military troops are trained or equipped for
counterinsurgent activity, and focusing one’s regular military on internal dissent may make it
less able to confront external, conventional threats (Andreski 1980). Hence, aggregate numbers
of regular military forces may not convey an accurate sense of the state’s ability to combat
insurgents. This is perhaps a particular problem with the COW data on troop levels, which
aggregates air, naval, and ground forces into a single measure, as air and naval forces are likely
less helpful in combating insurgencies than ground forces (Quinliven 1999).
Paramilitary forces may be a more appropriate measure of a state’s ability to exert
military power to deter or combat insurgents. There is no consensual definition of paramilitary
forces, though such forces are generally understood to be armed agents of the state other than
regular military forces. The Military Balance (MB) is one possible source of information on a
19
state’s paramilitary forces and its armed forces more generally.8 This annual publication lists all
subcomponents of a state’s armed forces, providing troop counts for each sub-category. There
are at least two problems with these data, however.9 First, there are substantial gaps in the data,
both across years within nations and across nations. Second, there are inconsistencies in the
coding rules in terms of what factors get included in which subcategories, both across years
within nations and across nations.
We built a new data set on the size of the paramilitary forces for all states for the time
period 1969-2003. We began with the MB data set, and cross-checked with another
comprehensive source of paramilitary data, the annual publication Statesmen’s Yearbook. We
used the two sources to identify and correct possible gaps and inconsistencies in the data. We
also in some instances employed supplemental sources. These efforts allowed us to reduce
substantially the amount of data missing in MB. With MBdata alone, there is paramilitary data
only for about 27% of the 7418 country-years across this time period. With our combined data
set, we have paramilitary data for about 64% of those country years. The value ranges from 0 to
12,000,000.
When collecting data, we observed that in these two data sets the term “paramilitary
forces” often include a variety of different kinds of forces, such as militia, gendarmerie, border
guards, and others. We included all of the categories of paramilitary forces listed by MB and
Statesmen’s Yearbook, as different governments use different kinds of paramilitary forces to
deter or combat internal violence. Some governments use Interior Ministry troops, such as Peru
8 Note that the rebel capability data set from Buhaug et al (2009) is not helpful here, because those data only contain
information on the relative power of a rebel group once a rebellion has broken out. A group of scholars is currently
in the process of collecting data on “pro-government armed groups,” which include paramilitary, militia, vigilante,
and other groups often intended to combat violently anti-government groups and insurgents (Carey et al 2009). This
data set will include a variety of descriptive information about such groups over the time period 1981-2004, though
it apparently will not include data on the size of such groups. This data is not yet available.
9 The editors at MB have to this date not replied to inquiries about their data methodologies. We note that some
scholars have used the MBdata as is, eg Colaresi and Carey (2008) and Boehmelt and Pilster (forthcoming).
20
in its campaign against the Shining Path or Russia in its campaign in Chechnya. Some
governments commission the military to train civilians to serve as militia, as Colombia has done
to combat narcoterrorist and other groups. One critique of our decision to include all categories
of paramilitary forces might be that it is a crude approach, including troops which are unlikely to
be used to defeat insurgents. This is a concern, but we note that our approach is similar to the
common approach in studies of interstate war of including the entirety of a nation’s regular
armed forces to understand its ability to deter and defeat the range of military threats it might
face, even though not all elements of a nation’s regular armed forces would be used against all
possible military threats. The United States Navy, for example, played little role in the 2001
interstate war against Afghanistan, though American sailors get included in COW troop counts
for the US, and American spending on its navy gets included in the COW measure of spending.
We might also note that accurately differentiating categories would be difficult, as states have
many different names for their military formations (militia, gendarmerie, volunteer defense
corps, people’s militia, security police, etc.), though the functions may be similar across states.
Our central independent variable is the number of paramilitary forces. Depending on the
dependent variable, we use either the number of paramilitary forces, or the number of
paramilitary forces per capita (using COW population data).10
We use four different dependent variables.11 First, we used Uppsala data on internal
violence, creating a dichotomous variable coded 1 if there was internal violence (at least 25
dead) in a nation-year, and 0 otherwise.12 Second, we used the Peic and Reiter (forthcoming)
10 Studies on post-conflict nation-building have focused on the number of occupation troops per capita as a key
factor determining peace and stability (Dobbins 2003).
11 We did not include Polity, following James Vreeland’s (2008) discovery that the standard Polity measure of
regime type includes a subcomponent which measures internal violence. When this problem is corrected, Polity is
not correlated with internal violence onset. Peic and Reiter (forthcoming) reached a similar conclusion.
12 Available at http://www.pcr.uu.se/research/ucdp/datasets/.
21
dichotomous variable on the outbreak of civil war. They expanded the Fearon and Laitin (2003)
data on civil war onset, who coded violence as constituting civil war if there were at least 1000
casualties. Third, we used the Bethany Lacina and Nils Petter Gleditsch (2005) data on battle
deaths from internal violence in a country in a given year. If a country suffers fewer than 25
such deaths in a year, they code it as zero. We use their “best estimate” measure. That variable
ranges from 0 to 350,000. Fourth, we used a dependent variable measuring the presence of
guerrilla war, as coded by the Cross National Time Series data.13 That data set defines guerrilla
war as, “Any armed activity, sabotage, or bombings carried on by independent bands of citizens
or irregular forces and aimed at the overthrow of the present regime.” The variable is a count of
the number of times the media (usually, the New York Times) mentions guerrilla activity in the
country during the year in question, and ranges from 0 to 34.
We also included a number of control variables, including whether the state was
experiencing an interstate war (COW) during the year in question, population, logged
energy/capita as a proxy for logged GDP/capita (COW), and the number of troops in the army
(MB).
Table 1 presents our findings. In all models, we used robust standard errors clustered on
country. Model 1 presents results from analyzing the battle deaths dependent variable. This
variable is a count, so we used negative binomial regression. The results indicate that larger
paramilitary forces are negatively and significantly correlated with the number of annual battle
deaths, providing evidence in favor of the hypothesis that paramilitary forces boost state power
and prevent violence. Among the controls, only logged GDP/capita is significant.14 The results
13 Available at http://www.databanksinternational.com/.
14 Given that there are several zeroes in the dependent variable, an alternative approach would be zero-inflated
negative binomial (ZINB). We did not have strong theoretical expectations as to which variables should be in the
inflation equation and which variables should be in the count equation. In exploring a variety of different ZINB
22
do not change if we instead use lagged values of paramilitary forces, or drop the Army variable.
Model 2 presents results from analyzing the guerrilla war dependent variable. This variable is
also a count, so we used negative binomial regression. The results were similar to Model 1,
though in Model 2 Interstate War and Population became significant, as well. Model 3 presents
results from analyzing the Uppsala dichotomous dependent variable. We used logit. Here, we
used paramilitary forces/population, recognizing that if the measure is simply whether or not any
violence breaks out (above the 25 dead threshold), then the appropriate measure is the
demographic density of paramilitary forces, not the overall number. The paramilitary/capita
variable is in the predicted direction, but only marginally significant (p<.10, one-tailed test). The
result does not change if we drop the Population control variable. If we instead use the aggregate
count of paramilitary forces (as in Models 1 and 2), then the paramilitary variable becomes
statistically significant (p<.01). Model 4 presents results from analyzing the Peic and Reiter
(forthcoming) civil war onset data. We use rare events logit because the frequency of ones is
quite low, again using paramilitary forces per capita. The paramilitary forces variable is not
statistically significant, and remains so if we drop the population variable, or use aggregate
paramilitary forces.
The substantive significance of the effect is difficult to assess because of the odd
distributions of the dependent and principal independent variable. However, using Model 1, if
we set the control variables to their means, the estimated battle deaths decrease from 398 to 260
if the paramilitary forces increase from 0 to the mean of 213,000.
functional forms, we found that paramilitary forces were negatively and significantly correlated with battle deaths in
some but not all ZINB functional forms.
23
Table 1: Paramilitary Forces and Internal Violence, 1969-2003
Model 1 Model 2 Model 3 Model 4
Battle deaths Guerrilla Warfare Uppsala
Internal
Violence
Measure
Peic/Reiter,
Civil War
Onset
Negative Binomial Negative Binomial Logit Rare Events
Logit
Paramilitary -.00000196***
(.000000406)
-.000000725***
(.000000132)
--- ---
Paramilitary/pop --- --- -.0186●
(.0140)
-.00371
(.0141)
Interstate War .439
(.909)
1.52***
(.270)
1.29**
(.541)
1.24**
(.470)
Population .00000386
(.00000675)
.00000204**
(.000000655)
.00000258●
(.00000162)
.00000141***
(.000000378)
GDP/capita
(logged)
-.868**
(.290)
-.367***
(.0602)
-.386***
(.104)
-.193***
(.0542)
Army .00000258
(.00000172)
.000000573
(.000000603)
.0141
(.0206)
-.0139
(.0258)
Ln(alpha) 4.14
(.207)
.356
(.385)
--- ---
Constant --- --- -1.39***
(.209)
---
Pseudo r squared --- --- .106 ---
Log likelihood -7146.9922 -2445.8513 -2133.1367 ---
n 4413 4067 4413 4413
***p<.001. **p<.01. All significance tests are one-tailed.
Some might be concerned about possible endogeneity. Specifically, it may be that the
true relationship is that the rise of internal violence causes a state to expand its paramilitary
forces, and then the expansion of a state’s paramilitary forces causes internal violence to
decrease. This kind of relationship is demonstrated graphically in Figure 1.
[Figure 1 about here]
24
If this relationship is occurring, we would expect there to be a lagged, positive effect
between internal violence as an exogenous variable and size of paramilitary forces as a
dependent variable. We explored for this possibility, using a regression model with robust
standard errors clustered on country code to analyze the determinants of per capita paramilitary
forces (note that because the measure is per capita providing many fractional values, count
models are inappropriate). As an independent variable, we used the Lacina and Gleditsch (2005)
data on annual battle deaths, as count data such as battle deaths will provide a better measure of
the ebb and flow of internal violence than will categorical variables like the onset or ongoing
occurrence of internal violence. We used a variety of lags of battle deaths, of one to four years,
and included the control variables included in Models 1-4, as well as a Polity measure. In no
case was lagged battle death significantly related to size of paramilitary forces, providing some
confidence that the results in Table 1 are not contaminated by endogeneity.
Conclusions
Existing empirical research has generally failed to find systematic connections between
measures of state power and internal violence. This paper explores the possible relation between
perhaps the single most relevant measure of state power, paramilitary forces, and internal
violence. It found that greater paramilitary forces were significantly and negatively correlated
with internal violence, though not for all measures of internal violence.
These findings encourage continued work on the relationships between state power and
internal violence. Coupled with past results, the general picture may be that for factors which
only generally or indirectly predict to the kinds of power a state would need to deter or combat
25
internal violence (such as terrain or censuses), there is not, at least not cross-nationally, a
significant relationship with internal violence. But, for factors which are directly related to a
state’s ability to deter or combat internal violence, such as paramilitary forces or FIRC following
war, there does appear to be a significant relationship.15
That being said, there remains significant work to be done to flesh out our understanding
of the relationship between paramilitary forces and war, and between state power and war.
Future work might more precisely explore exactly what kinds of paramilitary forces are
associated with deterring and combating violence, and what kinds of paramilitary forces are not.
This kind of separation would allow the refinement of the measure of paramilitary forces.
Relatedly, future work might separate out more effective from less effective kinds of
paramilitary forces, perhaps based on existing research that has found that civilian defense forces
like local militias can be relatively effective counterinsurgents, (Lyall 2010; Peic 2011), but that
mechanized counterinsurgents may be ineffective (Lyall and Wilson 2009).
More generally, the analysis here suggests deepening further our theoretical
understanding of the role state power plays in affecting internal violence, towards building a
theory that differentiates the variety of different types and manifestations of state power.
Developing an extensive theory connecting state power and internal violence needs to
accomplish several tasks. It should list all elements of state power which might affect the onset
of internal violence, such as geography, economic wealth, paramilitary forces, roads, etc. It
should distinguish between elements that both contribute to state power and constitute a prize
that, if captured, would boost insurgent power (such as, perhaps, airports), and elements that
contribute to state power but do not constitute a valued prize for insurgents (such as government
15 Notably, even uncovering robust, statistically significant relationships may not be enough to properly inform
policy seeking to predict civil war onset (Ward et al 2010).
26
legitimacy). It should classify elements on the degree to which they can be manipulated, ranging
from relatively difficult to manipulate (terrain) to moderately difficult to manipulate
(transportation infrastructure) to less difficult to manipulate (paramilitary forces). Varying
degrees of manipulation will indicate the degree to which endogeneity may pose a threat to
inference. It should also allow for elements of state power which might both reduce insurgent
activity, through deterrent or redistributive means, or provoke insurgent activity (such as
paramilitary or military forces, or tax collection infrastructures). It should develop theoretical
expectations as to when a certain aspect of state power is likely to increase violence, and when it
should decrease violence. Lastly, it should explore for possible interactive effects among
different aspects of state power. Roads can ameliorate the degree to which mountains restrain
state power, for example.
Solving this puzzle is not purely of academic interest. American foreign policy is
centrally concerned with helping the Afghan and Iraqi governments avoid internal violence into
the future. In Afghanistan in particular, important decisions need to be made about what
directions to take, including focusing on meeting basic societal needs of education and health
care, building an adequate transportation network, expanding an effective military and
paramilitary force sufficient to deal with the Taliban threat, and so forth. Each approach is
expensive, and as noted there is reason to believe that some of these policy initiatives if
mishandled could serve to exacerbate rather than mitigate violence. Building an empirically and
theoretically grounded understanding of what kinds of state power can reduce violence will help
both the Afghan and Iraqi regimes survive and thrive in peace in the years to come.
27
Figure 1: Mutual Causality Between Paramilitary Forces and Internal Violence
Internal Violence Paramilitary Forces
Time
28
References
Acemoglu, Daron and James A. Robinson. 2006. Economic Backwardness in Political
Perspective. American Political Science Review 100 (February): 115-131.
Andreski, Stanislav. On the Peaceful Disposition of Military Dictatorships. Journal of Strategic
Studies 3 (December 1980): 3-10.
Boehmelt, Tobias and Ulrich Pilster. Forthcoming. Do Democracies Engage Less in Coup-
Proofing? On the Relationship between Regime Type and Civil-Military Relations.
Foreign Policy Analysis.
Bohara, Alok K., Neil J. Mitchell, and Mani Nepal. 2006. Opportunity, Democracy, and the
Exchange of Political Violence. Journal of Conflict Resolution50 (February): 108-128.
Braithewaite, Alex. 2010. Resisting Infection: How State Capacity Conditions Conflict
Contagion. Journal of Peace Research 47(3): 311-319.
Buckingham, William A., Jr. 1983. Operation Ranch Hand: Herbicides in Southeast Asia. Air
University Review (July-August).
Buhaug, Halvard. 2010. Dude, Where’s My Conflict? LSG, Relative Strength, and the Location
of Civil War. Conflict Management and Peace Science 27 (April): 107-128.
Buhaug, Halvard, Lars-Erik Cederman, and Jan Ketil Rød. 2008. Disaggregating Ethno-
Nationalist Civil Wars: A Dyadic Test of Exclusion Theory. International Organization
62 (Summer): 531-551.
Buhaug, Halvard, Scott Gates, and Päivi Lujala. 2009. Geography, Rebel Capability, and the
Duration of Civil Conflict. Journal of Conflict Resolution 53 (August): 544-569.
Buhaug, Halvard and Päivi Lujala. 2005. Accounting for Scale: Measuring Geography in
Quantitative Studies of Civil War. Political Geography 24: 399-418.
Buhaug, Halvard and Jan Ketil Rød. 2006. Local Determinants of African Civil Wars, 1970-
2001. Political Geography 25: 315-335.
Canning, David. 1998. A Database of World Stocks of Infrastructure. World Bank Economic
Review vol. 12 no. 3: 329-347.
Carey, Sabine, Neil J. Mitchell, and Will Lowe. 2009. A New Database on Pro-Government
Armed Groups. Presented at the annual meeting of the International Studies Association,
February 15-18, New York.
Cederman, Lars-Eric, Halvard Buhaug, and Jan Ketil Rød. 2009. Ethno-Nationalist Dyads and
Civil War. Journal of Conflict Resolution 53 (August): 496-525.
Colaresi, Michael and Sabine Carey. 2008. To Kill Or To Protect: Security Forces, Domestic
Institutions, and Genocide. Journal of Conflict Resolution 52 (February): 39-67.
Collier, Paul. 2007. The Bottom Billion: Why the Poorest Countries Are Failing and What Can
Be Done about It. Oxford: Oxford University Press.
Collier, Paul and Anke Hoeffler. 2004. Greed and Grievance in Civil War. Oxford Economic
Papers 56 (October): 563-595.
Cordesman, Anthony H. and Ahmed S. Hashim. 1997. Iraq: Sanctions and Beyond. Boulder:
Westview.
Counterinsurgency. 2006. Washington: Department of the Army. Available at
http://www.fas.org/irp/doddir/army/fm3-24.pdf.
Crew, Harvey W., William Bensing Webb, and John Wooldridge. 1892. Centennial History of
the City of Washington, D.C. Dayton, Ohio: United Brethren Publishing House.
29
Cunningham, David E., Kristian Skrede Gleditsch, and Idean Saleyhan. 2009. It Takes Two: A
Dyadic Analysis of Civil War Duration and Outcome. Journal of Conflict Resolution 53
(August): 570-597.
DeRouen, Karl, Jr., et al. 2010. Civil War Peace Agreement Implementation and State
Capacity. Journal of Peace Research 47(3): 333-346.
Do, Quy-Toan and Lakshmi Iyer. 2010. Geography, Poverty, and Conflict in Nepal. Journal of
Peace Research 47 (November): 735-748.
Dobbins, James. 2003. Nation-Building: The Inescapable Responsibility of the World’s Only
Superpower. RAND Review 27 (Summer): 17-27
Dunning, Thad. 2005. Resource Dependence, Economic Performance, and Political Stability.
Journal of Conflict Resolution 49 (August 2005): 451-482.
Eck, John E. and Edward R. Maguire. 2006. Have Changes in Policing Reduced Violent
Crime? In Alfred Blumstein and Joel Wallman, eds., The Crime Drop in America,
revised edition. Cambridge: Cambridge University Press, 207-265.
Enterline, Andrew J. and Michael Greig. 2008. Perfect Storms? Political Instability in Imposed
Polities and the Futures of Iraq and Afghanistan. Journal of Conflict Resolution 52
(December): 880-915.
Estache, Antonio and Marianne Fay. 2007. Current Debates on Infrastructure Policy. World
Bank Policy Research Working Paper #4410, November.
Fajnzylber, Pablo, Daniel Lederman, and Norman Loayza. 2002. Inequality and Violent Crime.
Journal of Law and Economics 45 (April): 1-40.
Fearon, James D., Kimuli Kasara, and David D. Laitin. 2007. Ethnic Minority Rule and Civil
War Onset. American Political Science Review 101 (February): 187-193.
Fjelde, Hanne and Indra de Soysa. 2009. Coercion, Co-optation, or Cooperation? State Capacity
and the Risk of Civil War, 1961-2004. Conflict Management and Peace Science 26
(February): 5-25.
Gleditsch, Kristian Skrede and Andrea Ruggeri. 2010. Political Opportunity Structures,
Democracy, and Civil War. Journal of Peace Research 47(3): 299-310.
Gurr, Ted Robert. 1970. Why Men Rebel. Princeton: Princeton University Press.
Hegre, Håvard, Gudrun Østby, and Clionadh Raleigh. 2009. Poverty and Civil War Events: A
Disaggregated Study of Liberia. Journal of Conflict Resolution 53 (August): 598-623.
Hendrix, Cullen S. 2010. Measuring State Capacity: Theoretical and Empirical Implications for
the Study of Civil Conflict. Journal of Peace Research 47(3): 273-285.
Herbst, Jeffrey. 2000. States and Power in Africa: Comparative Lessons in Authority and
Control. Princeton: Princeton University Press.
Iqbal, Zaryab. 2010. War and the Health of Nations. Stanford: Stanford University Press.
Jordan, David P. 1995. Transforming Paris: The Life and Labors of Baron Haussmann. New
York: Free Press.
Kalyvas, Stathis and Laia Balcells. 2010. International System and Technologies of Rebellion:
How the End of the Cold War Shaped International Conflict. American Political Science
Review 104 (August): 415-429.
Koga, Jun and Dan Reiter. 2009. State Infrastructural Power and Civil War Onset. Presented at
the annual meeting of the Peace Science Society (International), Chapel Hill, NC,
November 21.
Lacina, Bethany and Nils Petter Gleditsch. 2005. Monitoring Trends in Global Combat: A New
Dataset of Battle Deaths. European Journal of Population 21 (2-3): 145-166.
30
Lai, Brian and Dan Slater. 2006. Institutions of the Offensive: Domestic Sources of Dispute
Initiation in Authoritarian Regimes. American Journal of Political Science 50 (January):
113-126.
Lange, Matthew and Hrag Balian. 2008. Containing Conflict or Instigating Unrest? A Test of
the Effects of State Infrastructural Power on Civil Violence. Studies in Comparative
International Development 43 (Fall/Winter): 314-333.
Levitt, Steven D. 1997. Using Electoral Cycles in Police Hiring to Estimate the Effect of Police
on Crime. American Economic Review 87 (3): 270-91.
Lyall, Jason. 2009. Does Indiscriminant Violence Incite Insurgent Attacks? Evidence from
Chechnya. Journal of Conflict Resolution 53 (June): 331-362.
Lyall, Jason. 2010. Are Coethnics More Effective Counterinsurgents? Evidence from the
Second Chechen War. American Political Science Review 104 (February): 1-20.
Lyall, Jason and Isaiah Wilson III. 2009. Rage Against the Machines: Explaining Outcomes in
Counterinsurgency Wars. International Organization 63 (Winter): 67-106.
Mann, Michael. 1988. States, War and Capitalism. Oxford: Basil Blackwell.
McGeown, Kate, 2005. Burma’s Confusing Capital Move. BBC News, November 8. Available
at http://news.bbc.co.uk/2/hi/4416960.stm <downloaded March 9, 2010>.
Military Balance. Various years. London: International Institute for Strategic Studies.
Mockenhaupt, Brian. 2009. Fire on the Mountain. Outside (November): 95-100, 136-139.
Myers, Steven Lee. 2010. Delays in a Head Count Keep Crucial Numbers a Matter of
Guesswork. New York Times, December 7, A10.
Ni, Ching-Ching. 2006. Rail Debuts in Icy Mountains of Tibet. Houston Chronicle, Jul2, p.
A34.
Nordland, Rod. 2009. Now It’s a Census That Could Rip Iraq Apart. New York Times, July 29,
p. WK-4.
Peic, Goran. 2011. Civilian Defense Forces and Counterinsurgency Outcomes. Unpublished
manuscript, Emory University, Atlanta, Georgia.
Peic, Goran and Dan Reiter. Forthcoming. Foreign Imposed Regime Change, State Power, and
Civil War Onset, 1920-2004. British Journal of Political Science.
Pickering, Steve. 2009. Rugged Terrain and War: Disaggregating From the State and Moving
Beyond the Mountain Binary. Presented at the annual meeting of the Peace Science
Society (International), Chapel Hill, NC, November 19-22.
Quinlivan, James T. 1999. Coup-Proofing: Its Practice and Consequences in the Middle East.
International Security 24 (Fall): 131-165.
Raleigh, Clionadh. 2010. Seeing the Forest for the Trees: Does Physical Geography Affect a
State’s Conflict Risk? International Interactions 36 (October-December): 384-410.
Schensul, Daniel. 2008. From Resources to Power: The State and Spatial Change in Post-
apartheid Durban South Africa. Studies in Comparative International Development 43
(December): 290-313.
Slater, Dan. 2008. Can Leviathan Be Demoratic? Competitive Elections, Robust Mass Politics,
and State Infrastructural Power. Studies in Comparative International Development 43
(December): 252-272.
Slater, Dan. 2010. Ordering Power: Contentious Politics and Authoritarian Leviathans in
Southeast Asia. Cambridge: Cambridge University Press.
Smith, Michael. 1994. “Taking the High Ground: Shining Path and the Andes”, in Palmer, ed.
Shining Path of Peru.
31
32
Sobek, David. 2010. Masters of Their Domains: The Role of State Capacity in Civil Wars.
Journal of Peace Research 47(3): 267-271.
Soifer, Hillel. 2008. State Infrastructural Power: Approaches to Conceptualization and
Measurement. Studies in Comparative International Development 43 (Fall/Winter): 231-
251.
Soifer, Hillel David. 2009. The Redistributive Threat: State Power and the Effect of Inequality
on Democracy. Brooks World Poverty Institute, University of Manchester, Working
Paper #93.
Starr, Chester G. 1982. The Roman Empire 27 B.C.-A.D. 476: A Study in Survival. New York:
Oxford University Press.
Statesmen’s Yearbook. Various years. London: Palgrave.
Thies, Cameron G. 2010. Of Rulers, Rebels, and Revenue: State Capacity, Civil War Onset,
and Primary Commodities. Journal of Peace Research 47(3): 321-332.
Thyne, Clayton L. 2006. ABC’s, 123’s, and the Golden Rule: The Pacifying Effect of
Education on Civil War, 1980-1999. International Studies Quarterly 50 (December):
733-754.
Vom Hau, Matthias. 2008. State Infrastructual Power and Nationalism: Comparative Lessons
From Mexico and Argentina. Studies in Comparative International Development 43
(December): 334-354.
Vreeland, James Raymond. 2008. The Effect of Political Regime on Civil War: Unpacking
Anocracy. Journal of Conflict Resolution 52 (June): 401-425.
Ward, Michael D., Brian D. Greenhill, and Kristin M. Bakke. 2010. The Perils of Policy By P
Value: Predicting Civil Conflicts. Journal of Peace Research 47(4): 363-375.
Weidmann, Nils B., Doreen Kuse, and Kristian Skrede Gleditsch. 2010. The Geography of the
International System: The CShapes Dataset. International Interactions 36: 86-106.
Willard-Foster, Melissa. 2009. Planning the Peace and Enforcing the Surrender: Deterrence in
the Allied Occupations of Germany and Japan. Journal of Interdisciplinary History 40
(Summer): 33-56.
Yardley, Jim. 2009. Rebels Widen Deadly Reach Across India. New York Times. November 1,
1, 16.
Ziblatt, Daniel. 2008. Why Some Cities Provide More Public Goods than Others: A Subnational
Comparison of the Provision of Public Goods in German Cities in 1912. Studies in
Comparative International Development 43 (December): 273-289.
