State Capacity and the Capability for
Comprehensive Peace Accord Implementation
David Edberg Landeström HT 2021
Peace and Conflict C
Department of Peace and Conflict Studies Uppsala University
Supervisor: Jonathan Hall Word Count: 10244
Table of Contents
ABSTRACT: ... 1
1. INTRODUCTION ... 2
1.1.RESEARCH QUESTION ... 4
2. THEORETICAL CONSIDERATIONS ... 5
2.1.LITERATURE REVIEW ... 5
2.2.PHENOMENON OF INTEREST ... 8
2.4.HYPOTHESIS ... 11
3. RESEARCH DESIGN ... 12
3.1.CHOICE OF METHOD ... 12
3.2.DATASET ... 13
3.3.OPERATIONALIZATION ... 15
3.4.DESIGN CONSIDERATIONS ... 23
4. FINDINGS AND ANALYSIS ... 25
4.1.DESCRIPTIVE STATISTICS ... 25
4.2.REGRESSION STATISTICS ... 26
4.3.ANALYSIS AND DISCUSSION ... 28
4.4.ALTERNATIVE EXPLANATIONS ... 30
5. CONCLUSION ... 32
6. BIBLIOGRAPHY: ... 34
Abstract:
Recent empirical studies have suggested that the implementation of a comprehensive peace agreement is the primary predictor of whether or not peace will last after a civil war.
However, it is less certain what factors that lead to high implementation rates of peace
agreements. Qualitative research has suggested that state capacity is a necessary condition for peace agreement implementation. Quantitively the relationship between state capacity and peace agreement implementation has only been controlled for in two studies. In this paper it is argued that this relationship has not been studied in a sufficient manner in either of them.
Consequently, this study measures the relationship between state capacity and peace agreement implementation rate, operationalizing state capacity as the extraction rate and political reach of the state. This relationship is tested on 34 comprehensive peace agreements during the years of 1989 to 2015. However, the hypothesis did not find support as extraction rate has a negative correlation while political reach has a positive correlation. These findings are significant as they further the study on peace agreement implementation rate; how best to measure state capacity and moreover these findings can become important for what policies to prioritize in order to increase the implementation rate concerning peace agreements.
1. Introduction
When a comprehensive peace agreement (CPA) is signed between two warring parties in a country it is often viewed as a momentous hurdle that has been overcome on the way from a conflictual relationship towards a peaceful one. But the threat of civil war does not end there due to the fact that civil wars ended through a negotiated settlement often risk leading to a recurrence of conflict (DeRouen & Bercovitch, 2008: 69–70). This problem has led many researchers to study how a durable peace is created after prolonged civil war. Many have analyzed, compared and evaluated CPAs based on their content such as third-party intervention in the form of peacekeeping (Fortna, 2008); power sharing mechanisms
(DeRouen et al., 2009: 383–384) and security sector reform (Brzoska 2003: 1–49). It has here been expected that the inclusion of certain provisions in itself could affect the durability of the post-accord peace.
A growing body of literature have taken another approach and instead focused on the relationship between the degree to which provisions in the peace agreement have been
implemented and the durability of the peace. This perspective has manifested a lot of promise, as it has been able to show a close relationship between the implementation of a peace
agreement and the durability of peace (Joshi & Quinn, 2015a: 869–892; Hartzell & Hoodie, 2003: 330–331; Jarstad & Nilsson, 2008: 220–222).
Under which circumstances are the parties in a CPA able to succeed in the implementation of the provisions in the peace agreement? One compelling theory that has been presented by DeRouen et al. (2010: 333–346) is that the state capacity of the government is important for the implementation of the CPA. This relationship was shown through a number of case studies where high state capacity was correlated to a higher degree of peace agreement implementation rate.
So far, the quantitative study of CPA implementation has not provided satisfactory results on the subject. First of all, of the two studies that have controlled for this mechanism a significant relationship was found in only one study, while the other did not have any association between the variables. Secondly, in this paper it is argued that neither of these studies measured state capacity in a valid way as they either used measures that are too closely related to other concepts or because they failed to measure the governments innate ability to make and implement their own choices (Arbetman-Rabinowitz, 2012: 18–54;
Hendrix, 2010: 273–285). As a consequence, it is uncertain if the relationship between these variables is generalizable or if they can be said to measure state capacity at all.
In this thesis the aim is to contribute to the literature of conflict resolution by covering the gap on the relationship between state capacity and the implementation of a CPA. This will be done through a quantitative analysis. The paper hypothesizes that as state capacity increases so will the overall implementation rate of the CPA. Further, by measuring the concept of state capacity as the extraction rate (ability of the state to appropriate portions of the national output) and the political reach (government’s involvement and influences over people’s daily life) of the state it is here proposed that state capacity will be measured in a more accurate way and through different dimensions as opposed to previous literature on the subject (Arbetman-Rabinowitz et al., 2012: 26–32).
Using data on 34 CPAs between the years of 1989 to 2015 the presented hypothesis does not find support in the findings of the study. While both variables had a significant
relationship with CPA implementation rate, political reach had a positive correlation meanwhile extraction rate had a negative correlation with CPA implementation rate. These findings lead to the conclusion that the political reach of the state might not be related to state capacity, since in that case it should show a similar correlation to the extraction rate of the state. This study becomes an important springboard for further theorization on how these
variables presented are related to CPA implementation rate as the hypothesis was not supported.
The study will continue as follows. In section two previous research, phenomenon of interest and the hypothesis will be discussed. Section three will propose the research design of the study. In section four the findings of the study will be presented and analyzed. Lastly, in section five the conclusion to the study will be presented.
1.1. Research Question
How does state capacity affect peace agreement implementation?
2. Theoretical Considerations
The theoretical considerations will start with a literature review where previous research on the subject of peace agreements will be discussed. Then the phenomena of interest will be discussed and how these are conceptualized in this study. Lastly, the hypothesis of the study will be presented.
2.1. Literature Review
Much of the focus of the conflict resolution literature has so far largely been concerned with the contents of peace agreements and their effect on durable peace. Fortna (2004: 172–179) focused on the effects of peacekeeping. She finds that even when controlling for the fact that peacekeepers are deployed in the most difficult cases, the risk of war resuming is much lower when peacekeepers are present in a country than when the parties are left to their own
devices. Power sharing has been discussed as it may have positive effects for short term peace as it creates incentives for the parties to lay down their arms. Though it has also been
criticized for its effect on long term peace as it might harm and undermine the
democratization process of a country (Jarstad, 2008: 130). Others have focused on who took part in the negotiation process. Krause et al. (2018: 985–1016) have shown that peace agreements signed by women often lead to a more durable peace as these signatories often have a linkage to women civil society groups.
A growing body of literature within the field have taken another approach and focused on the degree to which different provisions of interest within a peace agreement were later implemented. Jarstad and Nilsson (2008: 206–223) showed that the implementation of political, military and territorial power sharing agreements helped in contributing to a more durable peace. This finding was used by Joshi & Quinn (2015: 869–892) who argues that the implementation of the CPA has significant effect on how long the peace will last. Thus, the
implementation is viewed as a peacebuilding process in itself and an outcome that can overcome commitment problems, normalize political relations between the parties and address root causes of conflict as the parties continue to show their willingness to uphold the agreement. From these findings Joshi et al. (2017: 994–1018) argued that the implementation process of a peace agreement is a highly contested period where tensions might rise up again and lead to renewed fighting due to credible commitment problems where the parties might not trust that their opponent will abide to the negotiated settlement. Because of this,
safeguards need to be put in place that can resolve these issues while the implementation process moves on. These safeguards need to be put in place swiftly to address issues of mutual suspicion. They need to be facilitative so as to enhance the implementation of the accords, the longevity of the peace process and to address issues that arise during the execution process of the CPA. Lastly, the safeguards need to be temporary in nature and deliberately designed to cover the post-accord transition period. Once these goals are achieved the safeguards become obsolete and should be decommissioned. Based on these selection criteria the authors present three safeguards: transitional power sharing, dispute resolution and verification mechanisms. In the study it was found that together these three safeguards were able to increase the peace accord implementation rate by over 47% (ibid.:
994–1018).
Some have meanwhile pointed out that failed agreements cannot always be attributed to voluntary defection by one of the signing parties. Instead, the implementation of peace
agreements also includes other actors than the signatories themselves, which can lead to other types of commitment problems namely involuntary defections. These are especially prevalent in countries with strong polarized voting and a limited state capacity. Here, the political space for punishing politicians who do not comply with the results can be non-existent and thus they might not implement the peace agreement. As a solution to this problem the authors argue for
third party mediation through UN missions. Such a solution can expand the political space for peace building by either replacing weak state institutions or by strengthening local capabilities as part of a more permanent solution which leads to a higher implementation rate. By
comparing data on CPAs and personnel commitments from the UN the authors find support for their argument that UN missions aid in the implementation process (Maekawa et al., 2018:
397–416).
One factor that has been shown to reduce implementation rate of peace agreements is one- sided violence. One sided violence during the implementation process can be a consequence of actors who perceive that they benefit more from maintaining the status quo and thus attempt to spoil the implementation process (Joshi, 2020: 1–19).
Through a qualitative lens, researchers have centered on when and when not, an implemented peace agreement will hold. Some have argued that the peace agreement implementation process is more likely to proceed when there are mutual vulnerabilities between the parties (Bekoe, 2003: 256–294). Others argue that ethnocentrism and top-down bias of the liberal peace agenda are what lead to a failure of peace agreement implementation.
DeRouen et al. (2010: 333–346) shows through a number of case studies on Northern Ireland, Indonesia, Mali, Burundi and Somalia that state capacity seems to be a key factor in the implementation of CPAs. Based on these case studies they argue that state capacity is central to which functions a government is able to perform. Where state capacity is high, the cost of taking action in accordance with the agreement will be low, as the state will have both the resources and the capability to do so. This is important as the state is the main executor of the agreement and the implementation process can serve as a costly signal towards the rebels that the state will uphold the agreement which would give the rebels enough confidence in the process in order to disarm and return to society. Through these case studies, the authors found that state capacity was a strong predictor of implementation success.
State capacity has been controlled for in two different quantitative studies on the
implementation of peace agreements. Both Joshi et al. (2017: 994–1018) and Maekawa et al.
(2018: 397–416) controlled for state capacity and its effect on implementation rate success in their studies. However, their results were inconclusive. Using different operationalizations for state capacity Joshi et al. found a significant relationship while Maekawa et al. did not.
Maekawa et al. did argue that disaggregating the effects of bureaucratic state capacity more precisely could help clarify the relationship between bureaucratic capacity and peace accord implementation. Furthermore, this study argues that both of these papers used insufficient measures of state capacity as they could both be attributed to other concepts apart from state capacity. This will be further discussed in the operationalization of state capacity.
The aim of this paper is to contribute to the growing literature on the implementation of CPAs by further disaggregating the concept of bureaucratic state capacity. This is done through a quantitative analysis of the relationship between state capacity and peace accord implementation. This is in line with the conclusions drawn by Maekawa et al. (2018: 397–
416) as they argued for further disaggregation of the concept of bureaucratic state capacity. It is expected that the result of this study will be in line with the conclusions drawn by DeRouen et al. (2010: 333–346) that state capacity reduces the cost of implementation for the state, which leads to a higher implementation rate of the peace accord.
2.2. Phenomenon of Interest
2.2.1. Independent Variable
The independent variable that will be used for this study is state capacity. This is a variable that has been used by numerous researchers to explain the occurrence of different phenomena in the social sciences and the study of civil conflicts in particular. According to Hendrix (2010: 273–274) the conceptualizations of state capacity mainly falls into one of three
categories: military capacity, bureaucratic/administrative capacity and political institutional coherence and quality. Military capacity can be understood as the ability of the state to deter or repel challenges to its authority with force. This conceptualization is used by Fearon and Laitin (2003: 75–90) when they argue for weak state presence in rural areas as a key explanatory factor for civil wars. Second, bureaucratic/administrative capacity is
characterized by professionalization of the state bureaucracy. The ability of the state to uphold its own rule of law comes from its ability to collect and disseminate information. The last branch of state capacity is related to the amount of democratic or non-democratic features that exist in the political system. The argumentation here is that states with mixed regimes that are neither fully democratic nor autocratic are more likely to experience civil conflict. The reason for this is that the state is neither fully capable of repressing nor accommodating other
political viewpoints (Hendrix, 2010: 274–279). All of these conceptualizations carry with them both advantages and disadvantages that can be relevant for different studies.
The bureaucratic/administrative state capacity branch provides compelling reasons for its relationship to the implementation of CPAs. The goals that the state sets out to achieve are mainly accomplished through the bureaucratic systems it creates. Neither military capacity nor the coherence of political institutions is of the same importance for the implementation of a CPA as the bureaucratic/administrative capacity, since it does not revolve around a military victory or the democratic or autocratic nature of the political system.
In their paper, DeRouen et al. (2010: 334–335) uses a definition of state capacity that is closely related to bureaucratic/administrative capacity when they define state capacity as “the state's ability to accomplish those goals it pursues, possibly in the face of resistance by actors within the state”. This definition is adequate in many ways; it is clear in the sense that there is no confusion as to what it represents; it is well delineated in the way the different components of the concept are easy to comprehend; the concept has a moderate scope in the sense that it
includes the defining characteristics of state capacity and nothing more; the concept is coherent in the sense that the properties within it (the states attempt to pursue its goals and possible resistance) are logically connected to each other; and lastly, it has discriminatory power as it can be distinguished from other concepts. But the definition has limitations concerning its ability to compare states with each other. For example, if state A achieves seven out of ten goals, does that state have more or less state capacity than state B, which has four out of five accomplished goals. Nor does the definition include whether the size of the goals matter. Consequently, this conceptualization, although it is adequate in many ways, it makes it difficult to compare different cases, which is the point of a comparative analysis. As a result, in order to have a conceptualization that is comparable across cases. This paper will define state capacity as; the ability of the state to accomplish goals through bureaucratic means, possibly in the face of resistance by actors within the state. This conceptualization should allow for better comparisons between states as each state is not limited to the goals that it set out for itself.
2.2.2. Dependent Variable
The dependent variable in this study is the implementation of a CPA. Joshi et al. (2015: 552–
553) argues that peace agreements are comprehensive in two ways. First, the major parties of the conflict were involved in the negotiations that produced the agreement. Secondly, the substantive issues underlying the dispute were included in the negotiations. A major party is one that has sufficient mobilizational capacity and influence to alter the outcome of the peace process. Substantive issues refer to incompatibilities underlying the dispute, and which represent the main areas of contention between the warring parties.
Furthermore, a CPA contains a number of provisions that are supposed to be implemented.
A provision is a “goal-oriented reform or stipulation that is costly to one or both conflict actors and that falls under a relatively discrete policy domain” (ibid.: 554).
In this paper it is argued that implementation of a CPA is an adequate concept. It is clear and easy to understand what it entails; it is well delineated as components of the concept are identified and defined; it only captures what it is meant to capture; it is coherent; and it is possible to discriminate this concept from other similar concepts.
2.4. Hypothesis
The higher the state capacity within a state, the higher the rate of overall peace accord implementation will be.
3. Research Design
In research design, the choice of method will be discussed. The datasets used for the study will be presented and analyzed. The independent and dependent variables will be
operationalized, and design considerations will be made.
3.1. Choice of Method
The study will use a quantitative and comparative approach as a way to examine the
relationship between state capacity and the implementation of peace agreements to find out if the relationship is significant or not.
In order to test the hypothesis - that state capacity increases the implementation rate of a CPA - there is a need to carefully consider the criteria for causality. Namely, is there a credible causal mechanism that connects the variables? Is the time-order correct? Does the independent variable co-vary with the dependent variable and is the relationship isolated from other alternative explanations (Kellstedt & Whitten, 2013: 55)?
First of all, a credible causal mechanism has already been presented between the two
variables. That state capacity reduces the cost of implementation for the state which leads to a higher implementation rate. Second, there could be a time-order issue with CPA
implementation leading to more state capacity. This will be further discussed in the
operationalization of the independent variable. Lastly, in order to analyze the last two criteria of causality this study will use the Ordinary Least Squared (OLS) regression model. This statistical model is used to analyze the relationship between one or more variables (including the independent variable) and the dependent variable. The model estimates the relationship between the variables by minimizing the sum of the squares in the observed and the predicted values of the dependent variable organized as a straight line (ibid.: 171–246). The model can thus estimate what the relationship is between the independent and dependent variable as well
as if there are other variables that could explain the relationship in a better way than the independent variable.
Of note here is that there is always a risk that a variable that has been omitted from the equation might explain the relationship better than any of those that are included in the equation. This study will use similar control variables as has been done in previous studies in order to reduce such a risk of omitting variables that other researchers have deemed
important.
Moving on, as with to Joshi et al. (2017: 1002) the unit of analysis in this study is a post- CPA year that is observed once the CPA implementation process begins. And the scope conditions are all cases of civil war that were ended through a CPA.
3.2. Dataset
For the purpose of researching the relationship between state capacity and the implementation of a CPA, two different datasets have been merged. The first is the Relative Political Capacity Dataset (RPCD). The dataset covers countries with a population of at least 350 000 during the year of 2018. Furthermore, the dataset covers data on these countries in between the years of 1960 to 2018. The data is gathered from a multitude of data sources including: World
Development Indicators by the World Bank, National Account Estimates of Main Aggregates and Global Indicator Database by United Nations Statistics Division and Government Finance Statistics by the International Monetary Fund (Fisunoglu et al., 2020; Kugler & Tammen, 2012). What this indicates is that the information in the dataset has been collected
transparently, systematically and has been triangulated when possible. All of this should lead to a more trustworthy dataset. However, there is one problem concerning this data collection.
The International Monetary Fund relies on national reporting. This can lead to situations where countries for some reason or another might want to report inaccurate data as they
possibly have other incentives to do so. According to the codebook, these issues are often adjusted with the passage of time. National Central Bank reports have also been used
sometimes to supplement the International Monetary Fund (Fisunoglu et al., 2020; Kugler &
Tammen, 2012).
The second dataset to be included is the replication dataset by Joshi et al. (2017). This dataset is mainly built on the Peace Accord Matrix Implementation Dataset (PAM_ID) (Joshi et al. 2015; Joshi & Darby, 2013). PAM_ID contains the degree to which the provisions found in CPAs are implemented in the decade following the signing of the agreement. Included in the dataset are all CPA implementation processes in between the years of 1989 to 2015.
The replication dataset has also included one variable from the Polity IV dataset (Marshall et al.: 2013), two variables from the World Development Indicators (World Bank, 2013) and lastly one variable from the UCDP dyadic dataset (Harbom et al. 2008; Pettersson & Öberg, 2020).
Concerning PAM_ID it can be noted that although they mention many common sources that are used to gather information on the implementation status of different conflicts, it is still not entirely clear which sources provided the information for the different coding decisions.
The codebook mentions 14 different websites (such as UCDP, Amnesty International, UN Peacekeeping) and it also mentions media reports and conflict specific books and articles (Joshi & Quinn, 2015b: 1; Joshi & Darby, 2013). All of this does point to the data being triangulated when possible. This decision could have been made due to the complexity of the dataset as it contains many different provisions, and it is gathered over multiple years. But it can still result in a transparency issue as it is unclear what coding decisions went into the creation of the dataset. This is not an issue that can be handled in any easy way as this dataset is unique in how well is disaggregates the implementation process of CPAs, which makes it a vital dataset for this study. Concerning the other sources used for the replication dataset
(UCDP, Polity and World Development Indicators) all of them are widely used in the literature of peace and conflict studies, which one could argue makes them trustworthy sources (Marshall et al.: 2013; World Bank, 2013; Harbom et al. 2008; Pettersson & Öberg, 2020).
The replication dataset and RPCD are easily compatible as they both provide yearly data on the countries included in the datasets. In total, the merged dataset contains data on 34 CPAs negotiated between the years of 1989–2015. This results in 343 units of analysis with 15 variables.
3.3. Operationalization
3.2.1. Independent Variable
According to Hendrix (2010: 273–285), one of the most common ways to operationalize state capacity is as GDP per capita. This operationalization does come with a few advantages.
Firstly, GDP per capita is highly correlated with a variety of measures for bureaucratic state capacity and may also plausibly be viewed as a cause and effect of high bureaucratic quality.
It is also widely available for a large quantity of countries over a long timeframe.
However, GDP per capita is also a widely used operationalization for other variables. For example, Collier and Hoeffler (2002: 13–28; 2004: 563–595) demonstrates a relationship between GDP per capita and the onset of civil war as it affects the opportunity cost to wages in the legal economy versus participation in the illegal economy. As a result, there is a need to specify further how state capacity should be observed in reality so as not to accidently
become a proxy for some other mechanism as well. This line of reasoning is supported by Maekawa et al. (2018: 397–416) as they concluded that GDP per capita had no significant relationship with CPA implementation rate and that in order to properly analyze the role of
state capacity in CPA implementation the concept would be needed to be disaggregated further.
Meanwhile, Joshi et al. (2017: 994–1018) uses both GDP per capita and infant mortality rate as operationalizations of state capacity since the variables together can indicate individual well-being and the capacity of the state to deliver basic public goods. Furthermore, they did not find a relationship between GDP per capita and the implementation rate of CPAs. But they did find that lower infant mortality rate had a negative and significant relationship with CPA implementation.
Using infant mortality rate as a proxy seems to be a relatively unique way of measuring state capacity as it does not appear to be a widely used operationalization of state capacity in other studies. It is possible that state capacity and infant mortality rate are highly correlated which makes infant mortality a functional proxy for state capacity. But, at the same time it does not actually apprehend state capacity and there is a risk that it could capture something else instead. For example, infant mortality rate has been shown to have a significant
relationship with gender inequality. Infant mortality rate could therefore become a proxy for that instead (Brinda et al., 2015: 1–6). Further, proxies such as infant mortality rate relies on the assumption that states share and enact similar social, economic and political goals.
However, this presumption has been criticized as other researchers have argued that states do not in fact share similar goals. Instead, some states might prioritize social service provisions while others might act to facilitate economic growth etc. which would lead to different goals being prioritized (Arbetman-Rabinowitz, 2012: 26). From this perspective, the assumption behind using infant mortality rate as a proxy for state capacity is false.
In this paper it is argued that neither study has so far examined state capacity in a sufficient manner regarding the fact that both operationalizations can plausibly serve as proxies for other variables than state capacity and infant mortality rate may even rely on false premises
on how state prioritize goals. There is thus still a need for a study on the relationship between state capacity and CPA implementation rate using another operationalization for state
capacity.
One attractive way of measuring state capacity is to employ an extractive capacity with total taxes per GDP. This operationalization follows a sound logic since, in order for states to achieve anything they must first satisfy their resource demands. Further, taxation requires the state to invest in its capacity to monitor its population and effectively threaten them with consequences for non-compliance. This is also an indicator that has been widely used in the literature. Even so, this indicator carries with it some shortcomings. Basic measures of tax capacity do not in a sufficient manner distinguish between states that depend on
administratively sophisticated revenue instruments and those that do not. For any given level of taxation, a state that collects taxes on income, property, and capital gains would be
expected to have a much more developed bureaucratic capacity than a state that collects the same amount of taxes from international trade. However, this is not captured in this
measurement (Hendrix, 2010: 273–285).
The limitation of not being able to distinguish between sophisticated revenue instrument and those that are not sophisticated is addressed by the RPCD (Fisunoglu et al. 2020). Here the variable Relative Political Extraction (RPE) approximates the ability of the state to appropriate portions of the national output to advance public goals. They solve the
aforementioned issue by measuring the variable as the ratio of actual tax yield relative to the predicted extraction given a countries GDP per capita, mineral production, exports and other factors. States that collect tax revenues close to the predicted values have an RPE value of 1, while a state that collects twice as much revenue as predicted will have an RPE value of 2.
The variables range from 0 in Botswana in 1960 to 6.41 in Uganda in 1978. This variable carries with it two potential strengths. First, it controls for the effects of wealth which
diminishes concerns of collinearity. Second, the measure attempts to account explicitly for factors that might affect tax revenue by modeling the proportion of production coming from convenient tax handles and the demand for public expenditures; two sources of variation not previously captured by more aggregate measures of taxation. One potential drawback with this variable is that it has a consistent downward estimation of states that rely heavily on other sources of income than taxation. The reason for this is that these states do not have the same incentives to establish monitoring and coercive capabilities for their sources of income.
Instead, they can circumvent their populations. Nevertheless, the potential downward estimation of states that rely in other sources of income than taxes is inherent to every conventional measure of taxation as well. Furthermore, the RPE index has been able to heavily improve on the already existing indicators on tax revenue (Hendrix, 2010: 273–285;
Fisunoglu et al. 2020; Arbetman-Rabinowitz et al., 2012: 26–29).
The indicators of state capacity that have been mentioned so far have mostly focused on the material resources of the state and their indication for the state capacity of the country.
The RPCD provides another valuable variable to expand on the dimensions of state capacity.
The authors argue that for a state to function effectively and to achieve the objectives it is set out to achieve it needs both the material resources and the ability to mobilize human
resources. If a state has low capabilities to mobilize the population then the population might not accept the provisions that the state aims to implement. This is encompassed in the variable Relative Political Reach (RPR). RPR establishes the degree to which the government is involved in, and influences, people’s daily life. It reflects how well the state can mobilize human resources in order to achieve its goals. Societies that are characterized by little trust in the government elites are more likely to avoid the government and thus reduce any type of interaction with it. Therefore, if the informal sector of a country is calculated the political reach of the state is also calculated. RPR estimates the degree to which the government is
involved in economic activities of the population relative to the expected degree given the age, level of urbanization, education and employment of the population, among other factors.
If the output is lower than the average activity rate at any given interval, it will represent undeclared labor. The lower the ratio of the variable, the lower the capability of the state will be to reach the population as more people work outside the reach of the government and accordingly defy the authority of the government. RPR is calculated similarly to the RPE variable in such a way that a that at every level of economic development a certain activity rate is required to meet the output. The variable ranges from 0.26 in Algeria in 1995 to 1.88 in Japan in 1961. Together, RPE and RPR represents the available resources for the state to make policy choices (Fisunoglu et al. 2020; Arbetman-Rabinowitz et al., 2012: 29–32).
For the purpose of measuring different dimensions of state capacity this study will
operationalize state capacity as the annual RPE rate and RPR rate of the units of analysis. It is expected that both RPE rate and RPR rate will be important for the implementation of CPA provisions as the state needs the capacity to carry out these provisions, but also the support from the population of not hindering the implementation of the agreements. For example, a provision such as amnesty for rebel soldiers can be important to show the other party that the state truly believes in a joint society. But it is also a provision that can be immensely
unpopular within the population as the rebels may have done much harm to the civilians in the country. For this reason, if the state is lacking in sufficient reach the population might not accept the implementation of such a provision.
At the same time, it could also be the case that RPE will be more important than RPR.
Where RPR will make implementation easier as the government might face less resistance from the population. RPE is viewed as essential for implementation since extraction rate is vital for what services the government is able to provide. A high RPE value is critical for a high state capacity while a high RPR value is not.
Concerning the validity and reliability of RPE and RPR, these measures of state capacity have been chosen, as it is argued in the study because they have a higher internal validity than other measures of state capacity.
There are a few shortcomings regarding validity though. As mentioned before, the RPE variable consistently contains downward estimations of states that rely heavily on other sources of revenue than taxation. Another issue concerning the RPE variable is that the RPCD contains two models for this variable: one for developed countries and one for developing countries. This is due to the fact that the two groups often rely on different revenue sources.
For the sake of comparability across cases, this study has decided to only use the model for developing states as most of the states in the study are in this category. Such a decision reduces the internal validity for developed states. At the same time, it is the only way to move forward in order to compare between the groups. Continuing, as this operationalization focuses on concrete measurements in each country, it is here argued that these measurements are captured in a reliable way.
As previously mentioned, there could be a reverse time-order issue in the relationship between state capacity and implementation of CPA. This situation could be due to the implementation process of a CPA leading to more state capacity since the agreement written between the two warring parties might include provisions that make the government more effective and responsive to the population. This state of affairs could also reduce the amount of backlash that the government might face when attempting to implement its goals. Lastly, peace time itself might have a positive effect on state capacity as armed conflict has been shown to reduce state capacity (DeRouen et al. 2010: 335). To make sure that it is state capacity that leads to CPA implementation and not the other way around the independent variable will be lagged one year.
3.2.2. Dependent Variable
Implementation of a CPA will be operationalized as the annual CPA implementation rate.
This operationalization is used to identify the impact of the independent variable rather than a binary variable approach. Joshi et al. (2017: 1002) states that implementation is an
incremental process and those responsible for implementation do not see it in a binary way.
Further, using a binary variable would also minimize the variation to be explained. In
PAM_ID, for any provision year, the actual implementation value of any specific provision is coded on a scale of 0 for no implementation, 1 for minimum implementation, 2 for
intermediate implementation and 3 for full implementation. The annual CPA implementation rate is then calculated by summing up the value from each provision in the CPA, divided by the expected value of fully implementing those provisions and then multiplying by 100.
PAM_ID contains 40 provisions after external actor-related provisions and provisions that will be used as control variables have been excluded (Joshi et al., 2017: 551–562). For example, if there are 15 provisions in a CPA, 8 provisions achieve full implementation, 2 achieve intermediate implementation, 3 achieve minimum implementation and 2 achieve no implementation in a specific year. That year would attain a score of 67 (Joshi et al. 2015;
Joshi & Darby, 2013).
3.2.3. Control variables
Alongside the independent and dependent variables, the study will also use a number of control variables. These are variables that potentially could explain the CPA implementation rate better than could be the case with state capacity.
The implementation of a peace agreement review mechanism will be controlled for. The peace agreement implementation rate could be influenced when the CPA stipulates and implements a review of the agreement. Donor support will be controlled for as external
financial support can be critical for the success of many policies and programs (DDR, reparations etc.). The type of conflict, whether it was over governmental or territorial incompatibility will be controlled for as it could affect the difficulties in implementing the provisions. Further, the analysis will control for UN peacekeeping as previous literature has argued that it has an effect on the implementation rate (Joshi et al., 2017: 1003–1005;
Maekawa et al., 2018: 397–416). As previous research has argued that the built-in safeguards:
transitional power sharing, dispute resolution mechanisms and verification mechanisms, lead to a higher implementation rate, this will be controlled for as well. Each of these control variables are coded on a scale from 0 (no implementation) to 3 (full implementation) by utilizing data from PAM_ID (Joshi et al., 2017: 1003–1005; Joshi et al., 2015: 551–562).
Moving on, it could also be the case that the peace implementation rate gradually moves upwards as time moves on. Therefore, peace implementation year will be controlled for. This variable ranges from 1 (first year implementation) to 10 (the last year of implementation) in the PAM_ID dataset (Joshi et al., 2017: 1003–1005).
The implementation of a CPA could also be affected by government decisions that are made without oversight. Conditions that lack oversight can lead chief executives to be able to ignore the reforms stipulated in the CPA. Therefore, oversight such as a legislator or
independent judiciary on the decisions made by the chief executives will be controlled for with the constraint’s variable. When constraint is high (5–7) the executive constraints variable is coded as 1. While if it is lower, it is coded as 0 (ibid.: 1003–1005).
Number of provisions in the CPA will be controlled for as an agreement with many provisions could indicate that the parties had to make multiple compromises which signal a contentious implementation process that struggles to keep the actors satisfied. Further, CPAs with fewer provisions might be easier to implement. For this reason, in this study an above- average number of provisions will be controlled for. This variable is the difference between
the number of actual provisions in a CPA and the average number of provisions for all CPAs.
The difference ranges from negative to positive values. Finally, both GDP per capita and infant mortality rate will be included as control variables so as to compare with the use of RPE and RPR as operationalization’s of state capacity (ibid.: 1004).
3.4. Design Considerations
A consideration that needs to be done is that in the article by Joshi et al. (2017: 1007), they discuss that time series panel datasets such as PAM_ID have autocorrelation issues.
Autocorrelation issues refer to when two or more variables are systematically related to each other. The solution depicted by Joshi et al. to this problem in their analysis is by fitting feasible generalized least square (FGLS) panel data linear models with a first order
autoregressive process. As this study uses an OLS model and calculates the models using R Studio, while the original authors used STATA, some variations in the data are expected due to slightly different models and programs.
An attempt was made to replicate the data from Joshi et al. (2017) using a FGLS model.
While the data was closer to the original, it was still not able to fully replicate the results (see Appendix). Furthermore, using this model would have required an extensive time investment both to understand the model properly and to be able to accurately visualize the data which would have taken time from the analysis of the data itself. Therefore, the decision was made to continue with an OLS model. This will result in slightly different results than those from Joshi et al. (2017) but it should still be possible to make the relevant conclusions for this study. What this study will not be able to achieve though, is to discuss possible changes in other control variables as there is nothing to compare them with. The change in significance level and coefficient might be because of new variables introduced in the model, because OLS is used instead of FGLS or, due to the fact that another program is used.
Moving on, using OLS regression model for time-series data comes with a risk of spurious regressions due to trends in time-series data. For every year that is measured, it is uncertain if it is the state capacity of that particular that had led to the implementation of the provisions.
Some of these processes take a long time and might not actually be affected by the current state capacity. It is argued by other researchers that the use of lagged variables has been a way to solve this problem. It is likely that a one-year lag should matter as the state capacity still needs to be close in time to affect the implementation rate (Kellstedt & Whitten, 2013: 256–
269).
4. Findings and Analysis
Now the empirical findings made in the study will be discussed. It will begin with a
descriptive analysis of the variables in the dataset. Thereafter the regression statistics will be presented and discussed alongside a discussion of alternative explanations.
4.1. Descriptive Statistics
As can be seen in Table 1, the dataset contain data on 34 conflict dyads between the years of 1989 to 2015. These dyads result in 343 units of analysis in total. As shown in Table 1, the dataset includes 15 variables. RPE lagged and RPR lagged measure the independent variable in the study. RPE lagged has 81 missing values while RPR lagged has 98 missing values. This is mainly due to the fact that Macedonia, Bosnia, Cote d’Ivoire, Congo-Brazzaville and Timor-Leste have completely missing values on the IVs while El Salvador and Cambodia miss values on RPR lagged during this time span. The variables also miss additional data due to the fact that the first year of each dyad lacking a lagged value. Every bit of missing data, especially on the independent variable, is detrimental to a study as it risks providing less accurate results. However, in this paper it is argued that there are enough units left to be able to discern if there is a significant relationship between the independent and dependent variables and if this relationship is positive or not. Apart from these missing data, only the variables year count, infant mortality rate and GDP per capita contains missing values.
Though these missing values are fewer than the ones for the independent variable and should therefore not be a problem for the results of the study either.
An important thing to note concerning the distribution of values in the implementation of different provisions and the aggregated implementation rate is that there is an overweight to the top of the curve, which is due to time-series data. That is to say as time moves forward
more and more provisions are fully completed, which means that they are counted multiple times. This result is a distribution that heavily favors more completion.
4.2. Regression Statistics
The aim of this paper is to analyze the relationship between state capacity and CPA
implementation and to test whether or not the theory presented by DeRouen et al. (2010) is generalizable or not on a bigger sample. For this purpose, the study has conducted a multiple OLS regression in order to compare the data between different models. The regression table includes four models. The first is a replication of the model presented by Joshi et al. (2017:
1007), the second has removed their proxies for state capacity and included RPE lagged, while the third model instead has included RPR lagged. The last model includes both the lagged variables in order to estimate their joint effect on CPA implementation rate.
A replication of the results from the article by Joshi et al. (2017: 1007) was done in order to compare the results from previous operationalizations of state capacity to the
operationalization presented in this study. As mentioned previously, the replication model will look different than in the paper by Joshi et al. (2017). This is due to differences in the model itself and the use of a different program to construct the model.
When comparing the model used by Joshi et al. (2017: 1007) and the replication of their model it becomes clear that most variables have similar enough values in both models. GDP per capita still does not show any strong significance either. However, a bigger difference is seen in transitional power sharing, donor support, conflict type and infant mortality rate. Both transitional power sharing and donor support lose their significance level and have their coefficients reduced while conflict type increases in both significance level and coefficient.
Infant mortality rate has a reduction in its significance level while the coefficient increases somewhat.
Next, the second model where RPE lagged is used instead of infant mortality rate and GDP per capita. The model indicates a negative correlation for RPE lagged with a high statistical significance. Further, according to the second model dispute resolution increases its
coefficient by 1.3 as compared to the first model. When looking at the third model in the regression table RPR lagged shows a positive coefficient while also remaining statistically significant at the 95% level. Further, both UN peacekeeping and executive constraints both indicate higher coefficients when RPR lagged is included. Lastly, in the fourth model both of the indicators for state capacity still have high degrees of statistical significance. The
coefficient for RPE lagged has become smaller while the coefficient for RPR lagged has grown bigger. These results are consistent with model two and three. Dispute resolution shows similar values as in model two, while donor support and executive constraints both show values similar to model 1. The variable verification mechanism has remained statistically significant and with a large coefficient throughout all models. Continuing, the variable above average provisions has reduced its coefficient from model 1 and lastly, year count has had a similar effect in all models.
4.3. Analysis and Discussion
While it is not ideal that the data could not be replicated entirely it is not detrimental to the analysis. Within the scope of the study, it will not be possible to discuss why these changes occurred and if they are important in themselves, but a comparison can still be made between the old indicators of state capacity with the new indicators, since a difference in coefficient and significance level might still be a relevant comparison. The difference between these levels in this study could indicate a difference between them in other studies and in the real world as well. Furthermore, the change in the variable infant mortality rate is not as big as the other three variables mentioned that had different results from the Joshi et al. (2017) article
which give more reason to believe that it can still be fruitful to include infant mortality rate in the analysis.
All in all, these findings do not indicate any support for the hypothesis that a high state capacity should increase the overall implementation of a CPA. The reason for this outcome is because the variable that was argued to be an essential variable in indicating a high state capacity, RPE lagged, had a negative correlation to the annual implementation rate of the CPA. While RPR lagged, which was argued to be an important but not as important indicator for state capacity, had a positive correlation to annual implementation rate. What this could indicate is that the variable RPR is actually indicative of another variable within the societies that manage to successfully implement a CPA which will be further discussed in the
alternative explanations.
Continuing, both RPE lagged and RPR lagged indicate much higher statistical significance related to CPA implementation rate than infant mortality rate or GDP per capita does. Based on this finding one can argue that both of the newly presented variables seem to be much more closely related to CPA implementation rate than infant mortality rate or GDP per capita ever was.
Further, if it is true that RPE lagged is more closely related to state capacity than infant mortality rate as has been argued throughout this study. Then, it is instead likely that the statistical significance for infant mortality rate that was found in previous research would indicate rather that infant mortality rate is a proxy for some other variable other than state capacity (Joshi et al., 2017: 1007). Infant mortality rate could be a proxy for gender equality within the implementation process since gender equality could lead to a higher
implementation rate as a bigger and more diverse group of people is able to take part in the implementation process. This could lead to higher support for the operation which would lead to less resistance by the population and thus could lead to a higher implementation rate.
However, the causal mechanism presented or any other that relates to infant mortality rate would have to be researched further.
Other interesting results that were gathered was that the coefficients of both UN
peacekeeping and executive constraints increased in the third model. This observation could indicate that they become more important as the political reach of the state increases.
However, in the fourth model that increase was. It is therefore difficult to know if there is a relationship between these variables or not.
4.4. Alternative Explanations
As the hypothesis presented in this study did not find support in the results it is of interest to discuss alternative explanations to these findings. It was hypothesized in the study that RPE would have a positive correlation with CPA implementation rate. However, the opposite turned out to be true. From a cost-benefit analysis perspective a reason for this negative correlation could be that as the extractive capabilities of the state increases so does their resources. With more resources at its disposal the state might believe that it has a greater chance of successfully rebuilding its military capabilities instead of implementing its
provisions and then being able to renegotiate a better peace agreement for itself after renewed conflict. From this perspective it could also explain why states with lower extractive
capabilities might choose to implement a CPA since they might not have the resources needed for renewed conflict. Therefore, according to their cost-benefit analysis the cost of renewed conflict would be too high, and they would instead choose to implement the provisions in the CPA. This could also explain why dispute resolution might have an increased coefficient, since the need for dispute resolution increases if the government decides to renege on their promises.
The relationship between the political reach of the state and CPA implementation would also need to be further theorized. One possible explanation is that the reasons presented in the operationalization of state capacity for why political reach matters to state capacity are also important for the general implementation of CPA provisions. Namely how integrated the state is in the life of its citizens, how much the population trusts the government and how easily the population is mobilized for the implementation process. These aspects could come to be important in other ways, for example for how much the population believes in the implementation process and how much resistance the population will make to other
provisions such as peacekeeping for example. It could also be that the political reach of the state affects the opportunity structures for spoilers in a society. When more people are part of the formal economy and when the state is able to influence the lives of the population spoilers might have a more difficult time disturbing the CPA implementation process.
All of these theoretical mechanisms that have been presented would need be studied further in future research in order to make a causal claim.
5. Conclusion
The aim of this paper was to study the relationship between state capacity and the
implementation of CPAs. To begin with, the study followed the theory presented by DeRouen et al. (2017: 333–346) that a high state capacity would decrease the cost of implementation for the state which would lead to a higher degree of overall implementation of the CPA. This study argued that previous papers which have tested for this relationship have not done so in a satisfactory manner. These have firstly been proxies of state capacity while the new
measurement is argued to be measuring the states actual capabilities. Second, the previous indicators have also been used as proxies for other variables which make them questionable proxies for state capacity. And lastly, the variable infant mortality rate relies on a false premise that every state will prioritize similar goals. Therefore, a different operationalization of state capacity was introduced which included both the extractive capabilities of the state and its political reach as these factors would give the state as much room as possible to enact policy reform. It was argued that while both indicators would be important, the extractive capabilities of the state would be more important than the political reach of the state.
However, the hypothesis of the paper was not supported in the findings of the study, as the extractive capabilities of the state had a negative correlation with CPA implementation rate while political reach had a positive correlation. Both indicators had a statistically significant relationship to CPA implementation rate.
As an alternative explanation it was theorized that the extractive capabilities of the state affected the cost-benefit analysis of the government in such a way that they perceived that they would benefit more from continued fighting than following through with the agreement.
Additionally, it was theorized that the political reach of the state in itself might not be related to state capacity since political reach and the extractive capabilities had different correlations.
Instead, political reach might be important for CPA implementation as it can increase the trust
and mobilizational capabilities of the population in a country. Or it could be that it affects the opportunity structures of spoilers to disturb the CPA implementation process.
Future research could study both of these alternative explanations further. It could also be a fruitful avenue to study the relationship between infant mortality rate and CPA
implementation rate and how these variables actually relate to each other.
This study is important as it furthers the collective knowledge on which factors increases the implementation rate of CPAs. It also suggests an operationalization of state capacity that is suggested to be more closely related to the conceptualization of state capacity. A policy implication drawn from this study could be to focus on how to improve the political reach of the state in post-war societies so as to increase the chance of implementation success as much as possible.
6. Bibliography:
Arbetman-Rabinowitz, M., Kang, K., Abdollahian, M., Nelson, H., Tammen, R. & Kugler, J.
2012. “Defining and Measuring Extraction, Reach and Allocation”. In Kugler, J. & Tammen, R. (Eds.). 2012. Performance of Nations. Lanham: Rowman and Littlefield. 18–54.
Bekoe, D. 2003. “Toward a Theory of Peace Agreement Implementation: The Case of Liberia”. Journal of Asian and African Studies. 38(2–3). 256–294.
Brinda, E., Rajkumar, A. & Enemark, U. 2015. “Association between gender inequality index and child mortality rates: a cross-national study of 138 countries”. BMC Public
Health. 15(97). 1–6.
Brzoska, M. 2003. “Development Donors and the Concept of Security Sector Reform”.
Geneva: Geneva Centre for the Democratic Control of Armed Forces.
Collier, P. & Hoeffler, A. 2002. “On the incidence of civil war in Africa”. Journal of Conflict Resolution 46(1). 13–28.
Collier, P. & Hoeffler, A. 2004. “Greed and grievance in civil war”. Oxford Economic Papers 56. 563–595.
DeRouen, K. & Bercovitch, J. 2008. “Enduring Internal Rivalries: A New Framework for the Study of Civil War”. Journal of Peace Research. 45(1). 55-74.
