Decentralization and Conflict: The Missing Link in State Capacity Literature


Do greater levels of decentralization deter conflict onset? Drawing from the decentralization and state capacity literature, I attempt to demonstrate that higher levels of political and fiscal decentralization deter the onset of intrastate-armed conflict. Using a rare events logistical regression model, I fail to find evidence of the theorized relationship between decentralization and conflict onset. This study, however, draws attention to the many data problems that need to be solved before we can rule out the possibility that decentralization deters civil war.

Table of Contents: 


    One of the broadest connections that conflict onset scholars identify is the thought process that increased levels of state capacity facilitate the conditions necessary to reduce the likelihood of conflict. Despite the acceptance of this theme, state capacity has failed to explain some particular anomalies. Two particular cases stand out when evaluating the link between capacity and conflict onset.

    One of the cases that deviate outside normal expectations is post Bosnian War Bosnia-Herzegovina. The former Republic of Yugoslavia created a variety of grievances for various parties, accelerated by economic failure and ethnic tensions. The fall of the state coincided with a power grab of groups with competing claims. Cultural and political divisions facilitated the conflicts that arose, with different parties competing for control of the state. Although the conflict ultimately ended, the divisions were not only left unresolved, but in many cases, exacerbated. Thus, the signing of the Dayton Agreement led to a state that puts little faith in the central government. Instead, Bosnia-Herzegovina was established as a state that was highly decentralized along ethnic lines, split between ten cantons and a weak central government at Sarajevo. Despite a low level of state capacity, Bosnia has not only been able to deter conflict from arising, but build institutions that serve the population.

    Pakistan demonstrates the opposite case. Despite arguments that point out Pakistan’s high level of state capacity, incidences of armed conflict in the autonomous territories demonstrate consistency issues. The inability to prevent conflicts, however, is not just isolated in the autonomous territories; instead, the threat of conflict has been seen even in Islamabad. While these two cases highlight many of the same characteristics, state capacity has failed to provide a universal explanation on how a country should deter conflict. Because of the conditions that should show conflict in Bosnia-Herzegovina and effective conflict deterrence in Pakistan, this leads to the puzzle: Does decentralization play a role in the ability to quell conflict?

    These two cases demonstrate the wide variation of state capacity observed in reality and the problems that arise when overgeneralizing the literature. What is missing is a proper analysis of what constitutes state capacity. The current conceptualization of state capacity often focuses on the national government. This focus in academic discussion of state capacity thus ignores the building blocks of the determinants of state capacity. While this standard is commonly used, the notion that state capacity can only be observed and measured through the central government is problematic. The complication in the literature is that the role of the central government’s capacity is overstated, and the role of local government is downplayed. If scholars continue to identify both conflict and capacity at the state level, we will overgeneralize issues that are normally contained within a specific section of the country.

    This oversight by existing research is the missing area that my paper seeks to explore. State capacity is determined by observing the central government’s capabilities, which reduces local capacity as an insignificant factor. A federal structure necessitates an examination of capacity at the local capacity of a government.  If scholars are able to understand how a country governs at the local level, the scholars should also be able to determine how the central government reacts to specific situations. Thus, decentralized governments should have greater capacity to prevent escalation of violence outside the national capital.  

    This paper seeks to pinpoint factors that individuals have identified as influences on the link between state capacity and conflict. I will first establish the arguments that scholars have made to explain conflict onset. Then, I will explain how decentralization is a factor that one must consider when discussing all of these explanations. Next, I develop a theory on how decentralization and state capacity interact in order to show how it reduces the likelihood of conflict onset, after which I will provide my operationalization of the factors that are identified in the above portion. I will then discuss these results and what it means for future research in the area of conflict studies.

    Literature Review

    Existing literature has offered a variety of explanations for conflict onset (Bakke and Wibbels 2006; Buhaug, Gates, and Lujala 2009; Collier and Hoeffler 1998; Fearon and Laitin 2003; Garcia-Montalvo and Reynal-Querol 2005; Hendrix 2011; Kaufmann 2010; Ross 2001, 2006; De Soysa and Neumayer 2007; Thies 2010). I examine the incidences of conflict onset that may have been prevented by the build-up of state capacity. Four key explanations articulate the link between state capacity and conflict onset: (1) countries that have high resource wealth are often unable to control their citizens (Ross 2001); (2) ethnic fractionalization creates conditions for ethnic minorities to rebel (Kaufmann 2010); (3) geography allows for guerillas to mobilize, dampening state capacity (Hendrix 2011); and (4) decentralized governance allows for states to be responsive to the threat of conflict onset (Bakke and Wibbels 2006).

    Various scholars have articulated the potential for conflict with roots in resources (Collier and Hoeffler 1998; De Soysa and Neumayer 2007). Collier and Hoeffler (1998) provide a foundation for future development in regards to conflict onset. As these authors point out, a high-income population is more likely to prefer the status quo, while those in low income populations have less to lose when attempting to begin a conflict. The authors argue that increased wealth decreases the likelihood of conflict onset. The debate over wealth, however, has been filtered with the distinction of how a country comes into its wealth; in particular, states that rely on mineral wealth have a direct relationship with increased conflict onset. This point, however, has been disputed. Collier and Hoeffler (1998) add that resources have a curvilinear relationship with conflict onset. They argue that while resources might be a gateway to conflict, the government should be able to use funds from the resource wealth to repress dissidents. While the findings are statistically significant, the authors’ use of GDP per capita as a proxy for wealth may be problematic, often overstating wealth when money is concentrated at the top (Fleurbaey 2009).

    Regardless of these criticisms, the resource curse literature has been valuable for linking wealth with conflict. For most contemporary scholars, the focus has been an issue of export goods and conflict. Building on the findings of Collier and Hoeffler (1998), Michael Ross (2001, 2006) identifies the resources that correlate to higher amounts of violence. While Collier and Hoeffler’s analysis lends credence to the idea that all minerals may lead to the conflict, Ross identifies oil and non-fuel mineral wealth as determinants of conflict. What makes these states so different from others is that authoritarian governments often have ready access to looting these resources.

    Ross’s argument synthesizes the rentier effect, repression effect, and modernization effect, creating conditions necessary for the inability to control conflict. As Ross explains, the rentier effect is when the central government puts out a low tax rate in order to satisfy the population. Rentier states do this in order to prevent any transitions into democracy. Next, the repression effect is where states depend on oil revenues to create military capacity to repress their local population. This creates the conditions that lessen the likelihood for a democratic transition. Lastly, the modernization effect is where states with high levels of mineral exports are unable to create social changes for democracy. Ross indicates that despite these setbacks, nations are able to transition from their resource curses to countries with good governance.  

    Despite validity for Ross’s arguments, scholars are questioning if there is a deeper connection between resources and conflict. Ron (2005) notes that the link between resources and civil conflict is at best just a mechanism, in which authoritarian regimes build up their militaries in order to stifle conflict. Political opportunities, not economic, have been critical to determining whether or not there will be an ensuing conflict. In fact, large issues such as resources can potentially be one of the gateways to prevent, diffuse, or end a conflict. As such, research that connects economics and politics is necessary to fully understand these relationships. As Ron concludes, the literature regarding conflict onset and minerals is becoming more tenuous; the connection between resources and how the state collects revenue, however, become more persuasive for understanding conflict onset.

    Others have argued that the root of conflict onset is tied to levels of ethnic fractionalization (Garcia-Montalvo and Reynal-Querol 2005; Kaufmann 2010). Montalvo (2010) elaborates on the economic stances that facilitate conflicts. As he notes, high levels of ethnic fractionalization might deter investments in a country. There are three reasons for this. First, as Garcia-Montalvo (2005) indicates, highly heterogeneous societies lack connections for diffusion of technologies, and lack of investment in turn creates social deprivation that mobilizes individuals into conflict. Secondly, instability from strife between an ethnic group and a central authority increases the probability of disinterest in investment, which leads to a drop in trust of the central government. Last, ethnic diversity creates conditions for high levels of corruption amongst the ruling parties’ interest.

    When ethnic fractionalization is high, the dominant ethnicity creates the conditions that allow for corrupt regimes. The ruling party may create conditions that ensure their success, and thus creates conditions favorable for a protest. Kaufman (2010) furthers the arguments presented by Montalvo, pinpointing that while ethnic fractionalization may not be a good determinant of conflict onset, it is helpful in predicting the incidence of a conflict. The ambiguity between onset and incident has become an issue for individuals trying to isolate ethnic fractionalization as a determinant of conflict. These findings have been problematized with Fearon and Laitin (2003) finding that there is no relationship between the levels of ethnic fractionalization and conflict onset. Instead, ethnic fractionalization must be conditioned with factors such as low levels of wealth and dire living situations to justify the rise of a conflict.

    Certain geographic characteristics have been identified as conditions that encourage insurgent groups to operate (Buhaug, Gates, and Lujala 2009; Fearon and Laitin 2003; Hendrix 2011). While the literature has discussed the necessity of insurgents to find a location for operation, this is an understudied relationship. While both Collier and Hoeffler (1998) and Fearon and Laitin (2003) add the control in their dataset, they have done little to explain how geography interacts with the capacity of the state.

    Buhaug, Gates, and Lujala (2009) pick up on the geography literature. They find that geographic markers play a direct role in the ability of conflict to sustain itself. In particular, terrain and location create the conditions that nullify the power that the military traditionally enjoys. As they find, controlling strategic points often empower rebels to fight the effects of the government.

    Hendrix (2011) picks up on the literature that scholars had left unexplained, pointing out that certain geographic features provide the right conditions for conflicts to occur while leaving the government forces immobilized. In particular, guerilla groups are more likely to find themselves in mountainous regions as they diminish both the tax and military capacity of the state. These mountainous terrains negate the traditional military advantage that states enjoy. Additionally, mountainous territories create conditions that lead to dampened governance. These areas provide sanctuary for insurgents and affect the state’s ability to deter challengers. As witnessed by operatives in the Hindu-Kush region of the Afghan-Pakistan border, the ability to operate in these locations has been detrimental to military operations.

    Despite the substantive findings that Buhaug, Gates, and Lujala (2009) and Hendrix (2011) show, they hint at the necessity of disaggregating capacity and geographic features. In particular, the literature would benefit were scholars able to point out what geographic features enabled insurgents to organize to initiate a conflict. As Buhaug, Gates, and Lujala indicate, disaggregating a large nation such as Russia would be beneficial for the literature, as it is unlikely that a conflict that occurs in Chechnya is related to extracting resources from the Urals. Additionally, the literature fails to identify the government structures that govern in these mountainous areas.

    Contemporary literature has changed the focus of conflict onset. Thies (2010) predicted that state capacity would have an effect on civil war; he instead found that civil war decreased the capacity of the state. As he points out, taxation, one of the main ways of determining state capacity, may be endogenous to the risk of rebellion. Thies adds to the works of state capacity by indicating that weak central governments might lead to the increased likelihood of civil war. In order to accurately study the phenomena, however, Thies indicates that it is necessary to investigate state capacity in terms of subnational governments.

    Contemporary literature has emerged outside the opportunity framework. For example decentralization is a potential gateway to lessen the amount of violence a country may face (Bakke and Wibbels 2006). While the previous literature sought to find the causal link between conflict onsets, decentralization literature has focused primarily on what must be done to prevent a conflict from escalating. As Bakke and Wibbels (2006) point out, decentralized structures allow the government to tailor specific responses to their ethnic populations. The literature has linked decentralization with economic growth, and this literature synthesizes the arguments made by Collier and Hoeffler in 1998; the ability to promote growth is highly correlated with the decreased incidents of conflict onset. Additionally, Bakke and Wibbels find that decentralization has prevented protests from escalating because it provides a foundation in which individuals can still be in the constraints of the legal system without necessitating an overthrow of government.  The authors warn, however, that fiscal decentralization is not the answer in cases where there is a large degree of inequality. Instead, the existence of the national government and copartisans help include groups that might be willing to engage in conflict by bringing them into the political arena. The government then is able to find outlets to handle a conflict, and can address the grievances that may cause some individuals to desire conflict. This patchwork of decentralized governance allows for governments to collect resources, a determinant of state capacity.

    State capacity as a concept has continuously been linked with the potential to remedy conflict onset (Besley and Persson 2008, 2010; Braithwaite 2010; Thies 2010).  Building on a number of literature bases, authors have attempted to draw the connection between these two ideas. This portion of the text will look at the different conceptualizations of state capacity.

    Despite findings linking state capacity with conflict onset, scholars still have difficulty identifying the meaning of state capacity. As Englehart (2009) explains, state capacity has never been pinpointed to a single concept, but rather has been seen as a multidimensional idea. The lack of clarity from any source of state capacity has become an issue for those who are attempting to study the subject matter. Scholars have operationalized the concepts, with two notable works by Hanson and Sigman (2013) and Hendrix (2010). Both of these works have created idealized frameworks to evaluate what constitutes state capacity.

    According to Hendrix (2010), the literature focuses on various ideas of state capacity, including military capability, bureaucratic capacity, or quality and coherence of political institutions. The issues associated with difficulty in identifying construct validity and causal mechanism plagues the literature, with various tenets of state capacity hampering the overall design of the work. Deciding which metrics are a measure of state capacity has been an issue for scholars studying the matter. Thus, Hendrix conducts factor analysis on the existing work, finding that scholars are unwilling to measure state capacity in a quantitative fashion. Instead, Hendrix indicates that the best way of measuring state capacity is in a qualitative manner. In his view, the definitions of state capacity must work parallel with one another for an accurate framing of the literature, synthesizing opinion and tax information to create a picture of what state capacity is. As he concludes, it is necessary to look at state capacity as either bureaucratic quality or total taxes/GDP as these two concepts have ties with the themes that surround capacity literature.

    Hanson and Sigman (2013) present their case of what they see as state capacity. The authors point out that the current literature fuses multiple models of state capacity, with each addressing a different subject matter. For clarity, Hansom and Sigman break down state capacity in different conceptual models.  As they point out, there has been a heavy use of GDP per capita as an indicator of state capacity. By using Fearon and Laitin’s original dataset, Hansom and Sigman apply the logic that Fearon and Laitin use with more recent data on the topic, running the same regression tests. However, they find that GDP per capita has less explanatory power than research has suggested. As they conclude, our broader understanding of state capacity is flawed by the cross-applications in various fields. 

    The literature now finds difficulty in identifying the appropriate link between conflict onset and state capacity. While the literature has placed heavy emphasis on the national government as the agent in state capacity, there is little discussion of the role decentralization plays. The previous literature has been accurate in discussing state capacity in many cases, but there are some cases that cannot be explained by state capacity alone. Building on the foundations of Hendrix, I indicate that there is a necessity to use the metrics used in describing state capacity and connecting it to local capacity. While the previous research has been helpful in explaining various facets of conflict onset, the lack of emphasis on decentralized governance seems to be at odds with the findings of Bakke and Wibbels’s position on decentralization. This paper will attempt to synthesize state capacity in the context of conflict onset.


    I contend that the ability of the state to prevent the onset of conflict is just a symptom of decentralized state capacity. My explanation assumes that the state is sovereign over its territory. Additionally, it assumes that individuals have no incentive to ignite civil conflict unless a negative action is committed. A negative state action entails actions by the state that violate the rights of constituents. The theory presupposes that individuals are willing to maintain the status quo as long as there are no infringements on basic necessities. Rational actor theory is a necessary condition to understand the origins of conflict, as individuals only have incentives to fight when basic living conditions cannot be met. The explanation also assumes that the government is able to provide the basic services of the state. This does not preclude, however, the possibility that external actors may ignite civil conflict. Last, it assumes that we are able to distinguish which states have a decentralized framework and which states do not.

    One of the basic components of intrastate conflict onset is the demonstration of a government who cannot prevent or deter conflicts from escalating. The ability for the state to deter or stop a conflict from escalating is a function of state capacity. This conceptualization of state capacity in this instance is drawn from the literature in terms of bureaucratic quality. The issue that is being addressed is how the government is able to address conflict onset. State capacity is classified as either centralized or decentralized, two different forms of government with two different outcomes for conflict.

    Centralized governments tend to have power located at one strategic point, typically the capital. Because of this framing, conflicts that arise will either be (1) at the center of the state where the government has the most power or (2) in the areas outside of the government’s control. I theorize that a state does not have conflict throughout the country at the same time. Rather, a country has an isolated incidence of conflict and the lack of ability to control that conflict leads to its spread. Rather, conflict onset spreads from one point and expands from that location. There are a couple of scenarios in which centralized government’s ability to control the state might be hindered. First, individuals who protest at strategic points will typically have their demands met. This is because once protests come closer to home, there are incentives for the central government to give concessions to those protesting. If the government does not give into protesters, they will have to put down the protesters, which create the conditions for individuals to desire a larger conflict. This is a function of the state’s military capacity.

    Second, individuals who live in the periphery outside the capital have less of a connection to a governmental structure. We witnessed this in the cases of Islamabad, Kabul, and Freetown, as the government was unable to exert control. This lack of governance opens up the possibility for external actors to come in and create a separate entity outside of the central government that dictates the law in the region. These external groups often proliferate their influence onto these regions that lack central government support. The central government will not cede control of the periphery to the external group, opening the gateway for a civil war.

    Third, centralization creates one focal point of failure. Failures can range from corruption to inability to appease a local population, and inability to provide basic necessities. This creates cleavages in which the government will have issues maintaining control, and facilitates mobilization. The central government gets the direct blame for all these failures, and individuals who see these issues will rebel against the central authority.

    Last, states with high levels of centralization will also be the states that are unable to address the different needs of the population. When a state has one ethnic group at the core of its government, the state will also give preferential treatment to this group. This preferential treatment creates the conditions necessary for individuals to want to rebel if the government is unable to provide basic services.

    Decentralized governments on the other hand have various facets that prevent a complete failure of government. First, decentralized states are more locally responsive to their constituents. When a state goes into the process of decentralizing, the state in effect creates more avenues for their population to voice their grievances. Instead of the one focal point of contention, decentralized governments enlist more bureaucrats in order to have their voices heard.

    Second, decentralized states are able to stop the contagion of conflict. When a government decentralizes, they create focal points that stop the contagion of conflicts. While locations of protests increase, the relative power of each of these sites of protests can now be dampened. In this practice, they also create the conditions of dividing overarching themes of government so that they are now isolated to specific instances.

    Third, decentralized states are able to avoid a collapse of the central government. When a government is highly decentralized, they do not have to risk the loss of the entire government. Instead, decentralized states are able to make concessions if there is a necessity to do so. This is not a one-way direction: if a decentralized government is on the brink of failing, the central authority may have the power to reincorporate the failing states.

    Last, the decentralized government internally checks one another to provide a better framework of governance. As such, if a state/province is deficient in an area, decentralization allows for an oversight mechanism outside the central government. This creates multiple incubators for political and fiscal change in order to maintain a strong government.


    In this section, I broaden the conceptual link between state capacity and conflict to include the element of decentralization. The hypothesis outlined here is that a high level of decentralization has a positive effect on relative state capacity and the way that the state copes with civil conflict onset. In order to study the phenomena, I will look at relative state capacity and its relation to civil war onset. For the purposes of this study, the unit of analysis will be country-year. While some scholars find country-year problematic because it inflates small conflicts, this experiment does not attempt to explain the intensity of a conflict. The mere occurrence of a conflict demonstrates a failure at one part of government, whether it is a large-scale assault at the capital or a small armed rebellion that failed.

    Dependent Variable

    In order to test the hypothesized relationship, I use the dependent variable “onsetcv1” from the Onset of Intrastate Armed Conflict dataset as specified by Themnér, Lotta, and Wallensteen. The measurement of conflict onset is a binary variable (1,0), with 1 specifying the occurrence of a conflict. In order to meet the criteria to be coded as a 1, a state must not have conflict occurring within the last year. Conflicts that happen sooner than one year are often conflicts that are based off of previous grievances. The one-year measure demonstrates that conflict that should have occurred be filtered out. The dataset examines all incidences of intrastate-armed conflict from the year 1946 to 2011. The dataset is coded in accordance with the definition set by Gleditsch and Ward (1999). As they indicate, “onsetcv1” refers to an onset of intrastate-armed conflict that led to more than 25 individuals dying. Additionally, there must have been more than one year since the previous conflict occurred. Gleditsch and Ward (1999) explain that while governments have incentives to underreport incidences of conflict, it still remains as the most reliable data source that we have.

    Independent Variables

    The independent variables that are identified in this paper are the elements of centralization. Currently, the literature does not try to synthesize the elements of state capacity and decentralization. This lack of synthesis leads to my belief that we have to evaluate these concepts in unison with one another. 

    In order to gauge for political decentralization, I use the World Bank’s Political Decentralization Indicators. The World Bank splits its data by the divisions of either state or municipal decentralization. The data is then coded country-year, in which 0 denotes a country that has no elections at either the municipal or state level, 1 in which there are elections for one branch of government, and 2 in which both executive and legislative branches of government are elected. The World Bank is one of the few organizations that collect data on political decentralization, and while it is one of the most comprehensive, there are certain limitations to the data used. First, the World Bank data is limited in both scope and focus, with western democracies being a majority of the dataset. This inflates the number of states that are likely to see conflict onset and skews the data to show an insignificant result. Second, countries with low levels of governance will underreport the number of cases that underperform. Last, the current measure of political decentralization only describes instances in which specific circumstances occurred. In order to properly measure the level of political decentralization, data must be collected on the amount of rule that exists at the local/state level. Despite the many flaws that exist within the data, the World Bank data is the only one that attempts to map the configuration of power within a state.

    In order to test the effects of tax decentralization on conflict onset, I use the World Bank’s Fiscal Decentralization Indicators. The World Bank dataset is preferable to that of both the International Monetary Fund and the Organization of Economic Cooperation and Development for three reasons: First, the data that the World Bank compiles incorporates the measures found in both of these other datasets. Additionally, the data that the World Bank gathers provides more countries to analyze. Last, the World Bank data goes further back than either the International Monetary Fund and Organization of Economic Cooperation and Development. The World Bank operationalizes tax decentralization in seven different ways. For the purposes of the discussion of fiscal decentralization, all the indicators point to different themes found within the decentralization literature.

    Currently in the literature, fiscal decentralization has been measured through tax transfers from the central government to the provincial/local government. I will continue this practice used by Bakke and Wibbels. However, I add on two other variables, the first being vertical imbalance. Vertical imbalance is constructed to show the actual power of local governments in terms of collecting revenues and the levels of subsidization from the central government. Thus, a lower level of vertical imbalance represents independence from the central government. Next, I add subnational revenue as the total part of gross domestic product. Subnational revenue as the total percent of gross domestic product shows the extractive capability of the local governments. If local governments are the ones that make up the majority of the revenue stream, it will show how decentralized and how independent these subunits are from the central government.

    Control Variables

    Multiple facets of state capacity have to be considered when testing for its effect on conflict onset. These various facets will be grouped into the parameters that Hendrix (2010) identifies. The first facet of state capacity is the state’s tax capacity. To test for the effects of tax capacity, I follow Hendrix’s model and use Kugler & Arberman (1997) to measure Taxes/GDP.

    The next facet of state capacity that I include is the bureaucratic quality measure. This is measured with the efforts of the International Country Risk Guide (ICRG). The ICRG data is built on a scholarly opinion of the quality of the government because there is no real framework in which we are able to calculate the effectiveness of a bureaucracy.

    Another facet of state capacity that I include is relative political capacity. Relative political capacity demonstrates the strength of the state in regards to other states with the same amount of capacity. Relative political capacity does this by comparing the total taxes collected by the state as opposed to the expected amount that it earned. The number 1 denotes a standard measure when comparing states with one another, with scores higher than 1 indicating a state with greater amounts of capacity and scores lower than one showing states less capable.

    The last facet that I add is military capacity. While this has been shown as one of the poorest indicators of state capacity, the suppression qualities of the state can only be measured through the effects of the military and their relative strength against the local population. In order to measure for this effect, I log for the military expenditure of the country being analyzed.

    The depth of the conflict onset literature points to a myriad of alternative explanations for what are the elements that lead to conflict onset. The broadness of the literature points me to constrain my controls to indicators that directly affect the ability of the state to govern. This helps facilitate better discussion of the works in the literature while ensuring that research being done in the area is still relevant.

    Economic standing has been negatively correlated with the likelihood of conflict (Collier and Hoeffler 2004). As scholars contend, wealth facilitates the conditions that ensure individuals have no incentives to get into a conflict. Because of this consensus, GDP per capita is used in this project to help determine the wealth of a nation and whether or not that leads to conflict. The data that I use is compiled from the World Bank (2011).

    Population is discussed in the literature as having an impact on state capacity (Fearon and Laitin, 2003). As the population of a state grows, it has to increase its level of governance. The increase in population size brings about the conditions that can lead to the failure of the state if it is unable to cope. In order to test for the effects of population, I use data from the World Bank (2011).

    As indicated by Hendrix (2011), terrain has a relationship with lowering state capacity. In order to test for the effects of terrain, I use the data found in Fearon and Laitin’s 2003 dataset. Because the analysis ends in the year 1999, I impute the values for percentage of mountains that are found in the country. This is justified, as shifts in mountain size are slow and should not create large variations as a control.

    Previous literature  has shown that democratic states are often the states that have the fewest conflicts. In order to gauge the effects of a regime type, I use polity2 as coded by Marshall and Jagger (2012).

    Greater amounts of resources in a country are correlated with the increased likelihood of conflict onset (Ross 2001, 2006). In order to control for the effect of resources and civil conflict onset, I use Ross’s (2011) data on oil and natural gas exports. In particular, I use Ross’s measure of resource revenue.

    Ethnic fractionalization has been shown to have an effect with other factors to create the conditions of conflict (Kaufmann 2010). In order to control for the effects of ethnic fractionalization, I use Fearon and Laitin (2003) and their data on ethnic fractionalization. While their data goes up only to 1999, all their previous values were imputed. Thus, I will follow their practice.


    Running the rare events logistics test produced results that seem to conflict with the findings that are already made in the literature. Table 1 and Table 2 present the results of the tests in question. Table 1 reports my results prior to control variables being added, while Table 2 adds all the controls that are indicated from above.

    Table 1 reports my findings before adding the control variables that are found in the literature. I begin by doing an investigation between the two at the one-year level. The results indicate that political decentralization has very little explanatory power in terms of the decreased amount of conflict onset. While the direction of the link is consistent with the direction as found in the literature, the results should raise some questions. The lack of significance contradicts the findings of Bakke and Wibbels (2006) that political decentralization has the ability to deter conflict onset. These findings suggest that scholarly understanding of political decentralization may not be accounting for other literature bases that are intertwined.

    Table 2 reports my findings after adding the control variables that are found in the literature. While the direction of the link remains the same, political decentralization remains as an indicator that has little explanatory power. Each indicator of state capacity comes up with different results. Relative political capacity remains as a quality indicator of state capacity, as the risk of conflict onset decreases with a more capable government. As my theory suggests, states with a higher level of decentralization are more responsive to their constituents. Additionally, the bureaucratic quality indicator of state capacity continues to be consistent with the decreased risk of conflict onset. However, the other traditional indicator of state capacity, total taxes/GDP shows a positive relationship with the onset of conflict. While scholars contend that tax capacity alone has no effect on conflict, the addition of political decentralization shows issues with institutions and how money is collected. When there is a large level of decentralization and a majority of gross domestic product comes from tax revenue, that means that more money is being extracted from the pockets of the constituents. This creates conditions that can lead to a conflict onset. Military expenditures, while typically considered a poor indicator of state capacity, has a negative relationship with conflict onset. Traditionally, military expenditures are indicative of the suppressive nature of the state; however, once a state decentralizes, the necessity to suppress is diminished.

    The effects on the controls lead to some intriguing results. Population size is still significant when accounting for the various controls that are found in the literature. In order to have proper governance with a large population, states have to decentralize. Effective politically decentralized states will thus be able to respond to the needs of their population.

    The other noteworthy control is a country’s polity score. As observed, while polity2 isn’t statistically significant when explaining the onset of conflict, the tests suggest that regime type still shapes the potential for conflict. As the tests suggest, there is a missing component about the structure of a government that leads to the risk of conflict.

    The other traditional alternative explanations of the onset of conflict have their explanatory power diminished. Continuing with the literature that Ross (2001, 2006) builds on, I find oil resource revenue as a potential explanation for the onset of conflict. However, the results may be erroneous because of the lack of an indicator of how the revenue is distributed.

    The last three controls that are discussed don’t allude to a direction of a link. In particular, ethnic fractionalization remains statistically insignificant, as suggested by Fearon and Laitin (2003). Ethnic fractionalization in the realm of decentralization alludes to the idea that different ethnic groups have the potential to let their voices be heard. Additionally, the wealth explanation of the deterrence of conflict is tenuous, as the level of significance drops. This seems to say that even weak states with little wealth can still potentially deter conflict.

    Last, the effects of mountains in the realm of conflict drops significantly as well. I reason that a large level of mountain coverage prevents there being a central state to control. Decentralization, however, provides a proper outlet and an oversight component to deter individuals from engaging in guerilla conflicts that the literature has previously suggested.


    The purpose of this piece was to evaluate the claims that decentralization has an effect on intrastate-armed conflict. In particular, I postulated that political and fiscal decentralization were a function of state capacity. However, upon evaluation of the World Bank data on political and fiscal decentralization, the data seems to be inconclusive. This portion of the text will evaluate the data and the flaws that led to the conclusions that I got.

    First, the World Bank fiscal decentralization that is used in this dataset created some epistemic flaws with the arguments that are being advanced by this paper. When the World Bank compiles their data in terms of decentralization, they rely on the data that governments disclose. From here, there is a tacit agreement that the numbers that are disclosed are accurate to the overall economic conditions that a country is facing. The numbers that the World Bank gets often lacks the administrative oversight to show figures and shut off the possibility of external oversight. 

    Second, the dataset that I was able to compile is largely incomplete. Because of this, upon testing for my hypothesis, I ended up with a total of 27 countries to be analyzed. Of these 27 countries that were being examined, however, 13 of those nations that were examined were Western democracies or Australia that haven’t seen intrastate armed conflict for a long time. Additionally, the data that was compiled also showed instances of countries with limited disclosure of their governments. For example, upon investigation of countries such as the Philippines or Thailand, I only have one year to base our study on. This prevents analysis for multiple years and the trends that might come from running these tests.

    Third, the World Bank Political Decentralization Indicators do not measure political decentralization; rather, the political decentralization indicators only indicate the power of the constituents and their power. Although this was the only dataset that was accessible, the dataset that the World Bank compiled only describe the setup of the governmental system. However, because this was the only setup of the evaluation of power of the local level, I used it. In order to better describe the effects of political decentralization in terms of state capacity, social scientists must find a better indicator. 

    Fourth, the absence of replication data from various authors cited in this piece raises  the question of how to verify some of the studies that were done. Much research done in social sciences is peer reviewed; however, upon reviewing scholastic pieces, we have been negligent about the data driving our research. If the data that we base our claims on are questionable, we should cast doubts on our findings. This is especially the case with social science research that reaches the realm of policymakers. One of the best examples of bad social science research driving policymaking was Reinhart and Rogoff’s finding and the subsequent justification of austerity. When Reinhart and Rogoff began their study in 2008, they purposefully omitted a row of data that altered their findings. As they found, austerity was necessary for the growth of an economy. When a graduate student finally requested their dataset, he found the omission of a column of data to make the difference in the research. Absent the ability to replicate one’s results, we should cast doubt on other’s findings.

    In particular, Bakke and Wibbels’ (2006) publication in World Politics seems to have created conditions that have broader implications. This piece cites much of the research that the pair conducted, and was also the inspiration of this piece. Upon trying to replicate the findings of Bakke and Wibbels, however, my efforts were fruitless. Upon revisiting their piece, I find that much of the data that the pair uses are predicated off of another piece from Wibbels, in which Rodden and Wibbels imputed various values because of the large number of missing cases. If the practice of imputation is employed in a paper, there should be a transparent disclosure of the methods that the argument employed.

    Because the synthesis of political decentralization and state capacity is a novel idea, there isn’t much literature on how to connect the two concepts. Future avenues of research should attempt to find the connection between these two concepts. This would entail the breakdown of the studies of state capacity to match up with the literature that show the potential that political decentralization has in deterring conflict.


    • Bakke, Kristin M., and Erik Wibbels. 2006. “Diversity, Disparity, and Civil Conflict in Federal States.” World Politics 59(01): 1–50. (June 8, 2013).
    • Buhaug, Halvard, Scott Gates, and Paiva Lujala. 2009. “Geography, Rebel Capability, and the Duration of Civil Conflict.” Journal of Conflict Resolution 53(4): 544–69. (June 20, 2013).
    • Collier, Paul, and Anke Hoeffler. 1998. “On Economic Causes of Civil War.” Oxford Economic Papers 50: 563–73. (June 8, 2013).
    • Collier, Paul, and Anke Hoeffler. 2004. “Greed and Grievance in Civil War.” Oxford Economic Papers 56(4): 563–95. (May 21, 2013).
    • Coplin, William; Michael O’Leary & Tom Sealy (2007) A Business Guide to Political Risk for International Decisions. Syracuse, NY: PRS Group.
    • De Soysa, Indra, and Eric Neumayer. 2007. “Resource Wealth and the Risk of Civil War Onset: Results from a New Dataset of Natural Resource Rents, 1970-1999.” Conflict Management and Peace Science 24(3): 201–18.||D404A21C5BB053405B1A640AFFD44AE3 (May 21, 2013).
    • Fearon, James D., and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American Political Science Review 97(01): 75. (June 20, 2013).
    • Fleurbaey, Marc. 2009. “Beyond GDP: The Quest for Beyond Measure of Social Welfare.” Journal of Economic Literature 47(4): 1029–75. (June 20, 2013).
    • Garcia-Montalvo, Jose, and Marta Reynal-Querol. 2005. “Ethnic Polarization, Potential Conflict,and Civil Wars.” UPF Economics and Business Working Paper 770. (June 20, 2013).
    • Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Håvard Strand. 2002. “Armed Conflict 1946-2001: A New Dataset.” Journal of Peace Research 39(5).
    • Hanson, Jonathan K., and Rachel Sigman. 2013. “Leviathan’s Latent Dimensions: Measuring State Capacity for Comparative Political Research.” Presented at the World Bank Political Economy Brown Bag Lunch Series, Washington, DC, March 21, 2013.
    • Hendrix, Cullen S. 2010. “Measuring State Capacity: Theoretical and Empirical Implications for the Study of Civil Conflict.” Journal of Peace Research 47(3): 273–85. (May 28, 2013).
    • Hendrix, Cullen S. 2011. “Head for the Hills? Rough Terrain, State Capacity, and Civil War Onset.” Civil Wars 13(4): 345–70.
    • Jones, Daniel M; Stuart A Bremer & J David Singer (1996). Militarized interstate disputes, 1816–1992: Rationale, coding rules, and empirical patterns. Conflict Management and Peace Science 15(2): 163–215.
    • Kaufmann, Eric. 2010. “Why Ethnicity Is Linked to Civil War: The Case for, and Limits of, Perennialism.” Perennialism (August 16, 2010): 1–27. (June 20, 2013).
    • Kugler, Jacek & Marina Arbetman (1997). Relative political capacity: Political extraction and political research. In M. Arbetman and J. Kugler, (eds) Political Capacity and Economic Behavior. Boulder, CO: Westview, 11–46.
    • Marshall, Monty G. and Ted Robert Gurr. 2013. “Polity IV Project: Political Regime Characteristics and Transitions, 1800–2012.” URL:
    • Ron, James. 2005. “Paradigm in Distress?: Primary Commodities and Civil War.” Journal of Conflict Resolution 49(4): 443–50. (May 21, 2013).
    • Ross, Michael L. 2001. “Does Oil Hinder Democracy?” World politics 53(April): 325–61. (June 8, 2013).
    • Ross, Michael L. 2006. “A Closer Look at Oil, Diamonds, and Civil War.” Annual Review of Political Science 9(1): 265–300. (June 4, 2013).
    • Ross, Michael L. 2013-02, "Oil and Gas Data, 1932-2011", UNF:5:dc22RlDasveOTAJvwIjBTA== V2
    • Thies, Cameron G. 2010. “Of Rulers, Rebels, and Revenue: State Capacity, Civil War Onset, and Primary Commodities.” Journal of Peace Research 47(3): 321–32. (June 13, 2013).
    • World Development Indicators. 2013. The World Bank, Washington, DC: World Bank Group. URL:

    Table 1: Rare Events Logistic Model without Control Variables

    Onset (1 Year)


    Standard Error

    Z Score

    P Value

    Political Decentralization





    Table 2: Rare Events Logistic Model with Control Variables



    Standard Error



    Municipal Decentralization






    Relative Political Capacity






    Military Expenditures (Logged)











    Tax Ratio





    Bureaucratic Quality





    Fiscal Decentralization (Transfers)










    Population (Logged)





    Mountains Percentage










    Ethnic Fractionalization