Terrorists and Peace

Terrorists and Peace

A quantitative analysis on terrorist organisations’ effect

on the sustainability of peace agreements.

Martin G. Wallin

Political Science C (Bachelor Thesis) Department of Government

Uppsala University, Fall 2018 Supervisor: Lina M. Eriksson Words: 11,715

Abstract

Do terrorist organisations have an effect on the sustainability of peace agreements? In this paper, I investigate the potential consequences of including armed non-state actors in peace agreements. Specifically, I focus on the most controversial type of actor: terrorist organisations and how they affect the sustainability of peace agreements. I utilise a unique dataset covering all intrastate peace agreements and designated terrorist organisations between 1998 – 2011, and employ a survival analysis through a cox proportional hazard model. I find strong evidence suggesting that the presence of terrorist organisations – both in conflicts and in peace agreements – have a significantly destabilising effect on the sustainability of peace agreements. I also find evidence which suggest that more complex power sharing arrangements could be a viable option to combat the destabilising effect of terrorist actors, although this effect might not be as strong in peace process agreements.

Key words: Peace agreement, terrorist organisation, non-state actors, legitimacy,

Table of Contents

1. Introduction ... 5

2. Theoretical Discussion ... 6

3. Research design ... 15

4. Results and analysis ... 24

4.1 Descriptive statistics ... 24

4.2 Results ... 26

4.3 Discussion ... 32

5. Conclusion ... 35

List of tables

Table 1.

... 25

Table 2.

... 25

Table 3.

... 28

1. Introduction

In a time when presiding institutions and a general conception of statehood is under challenge by a multitude of intrastate conflicts and violence, the role of non-state actors appears to be increasingly important for the stability of a region (Davis, 2009). Armed non-state actors tend to not only shape the condition and outcome of intranon-state conflicts, but also the success or failure of any subsequent peace agreement (Hofmann, 2006). Historically, a lot of regions that have been ravaged by intrastate conflicts witness a resurgence of violence when initial peace settlements collapse (Newman & Richmond, 2006). The role that armed non-state actors play in these peace agreements does, however, seem to vary depending on the condition of the region, and non-state actors tend to assume a larger role in areas under limited statehood (Mehler, Lambach & Smith-Höhn, 2010). The specific type of the non-state actors might also influence their abilities to affect the peace process. As highlighted by Schneckener (2009), it is important to remember that there is an array of various types of non-state actors with different goals, methods, and principles. In this respect, past research has primarily looked at non-state actors as a single form of actor, and consisting of all types bunched together, that is contrasted with the government of a nation-state. However, thus far terrorist organisations, as a specific type of organisation, have not been studied in the context of how they may affect peace processes and peace agreements. It would therefore be interesting to take this research one step further, and see if terrorist organisations, which may be considered the most controversial type of armed non-state actor, actually may affect the sustainability of peace processes and peace agreements.

2. Previous Research and Theory

This section is devoted to discussing the current theoretical ideas and arguments relevant to this paper and to my research. Below, I present some previous research that has been conducted within the field of armed non-state actors and peace agreements. This is important in order to learn from previous academic findings and to find research gaps to which I intend to contribute. I will also discuss the theoretical definitions of ‘terrorist organisation’, ‘legitimacy’, and the ‘removal of spoilers’. Thereafter I build on existing studies and theory to state my research questions, which are followed by a presentation of the hypotheses I intend to test in this paper.

2.1 Previous research

The discussion on armed non-state actors’ roles in peacebuilding has received a growing amount of research and scrutiny. A major part of this research has been focused on non-state actors’ roles as potential future spoilers in the peacebuilding process, and how the international community can deal with this issue (e.g. Hofmann & Scheckener, 2011; Schneckener, 2009; Stedman, 1997). A general conclusion stemming from this research is that a comprehensive peace agreement, where none of the warring parties are excluded, should be more stable, by eliminating the armed non-state actors as future spoilers. However, there has been a clear lack of quantitative research in this area.

agreements through quantitative research, but with a clear distinction between different sorts of armed non-state actors.

Recently, there has been a discussion on the importance of non-state actors as a source of stability in areas of limited statehood. If such areas exist, then the inclusion of non-state actors in peace agreements could be necessary to increase the legitimacy and support for the agreements, given the increased role played by non-state actors in areas of limited statehood (Davis, 2009; Davis, 2010; Mehler, Lambach, & Smith-Höhn, 2010). Furthermore, there has also been a move towards discussing the general legitimacy aspect of including non-state actors in peacebuilding (e.g. Mitchell, 2008; Narten, 2008; Roberts, 2011). According to this literature, the inclusion of non-state actors in peacebuilding processes could potentially increase the legitimacy of the process by making the peacebuilding concept less ‘elitist’ and by connecting it to local actors. Nilsson (2012) applied this idea of legitimacy regarding non-state actors in a quantitative analysis and found statistical support for it. Specifically, Nilsson found that the inclusion of civil society actors in peacebuilding processes leads to more durable peace accords.

Nonetheless, while these latter findings are interesting, I argue that there is a possible overlooked risk in the legitimacy theory. My main concern is that different sorts of non-state actors might affect the legitimacy of a peace agreement differently. The general assumption in past papers has been that the inclusion of terrorist actors in peace processes could legitimise violent means and terrorism, and thus decrease the legitimacy of any subsequent peace agreement (Arnault, 2014; Hofmann, 2006; McCartney, 2006). This follows a logical assumption that the inclusion of terrorists will decrease the legitimacy of an agreement. However, this idea has not, to my knowledge, been substantively tested but instead only assumed. Thus, the question still remains open if terrorist organisations actually affect the legitimacy and the longevity of peace agreements.

research in this field, which makes this paper and the following large-N quantitative analysis focusing on terrorist organisations per se an important contribution.

2.2 Terrorist organisation

There is a certain difficulty in using ‘terrorist organisation’ as a classification variable, given the vast array of definitions and subjective meanings that can be found behind it. Furthermore, there also appears to be a rift between the definitions preferred by state officials compared to those in academia (Freilich et al., 2009). Among the academic definitions relevant to this study is Schneckener (2009), which showcases a possible classification schedule entailing eight different armed non-state actors (including terrorists). However, given the large number of armed non-state actors considered in this paper (147 in total), it is not logistically possible to apply Schneckener’s approach to all of them. Instead, I rely on the US. Department of State’s terrorist-lists, and apply it to the armed non-state actors relevant to my study.1

Some issues regarding the classification of ‘terrorist organisation’ may arise from this approach, and I acknowledge that there is an ongoing discussion about what actually qualifies as ‘terrorism’ (e.g. Freilich et al., 2009; Lentini, 2008). However, this should not be an issue with regards to the findings in this paper, given that I do not aim to provide a

1 _{Given that this research is focusing on the sustainability of peace agreements, and how the inclusion of}

descriptive account for which armed non-state actors that constitute as ‘terrorist organisations’, but rather how the designation of these actors as ‘terrorists’ affects the peace process and the sustainability of peace agreements.

Nevertheless, given that international actors follow different classifications, some concern could be raised with regard to if a terrorist organisation actually partook in a conflict or a subsequent peace agreement. A clear example of this is highlighted in the conflict over Northern Ireland, which is an intrastate conflict that is included in this paper’s dataset. The Real Irish Republican Army (RIRA) was a warring non-state actor in the conflict and had been classified as an illegal terrorist organisation by the United Kingdom through the Northern Ireland (Emergency Provisions) Act (1996; 1998). RIRA was excluded from the peace talks and subsequently rejected the Good Friday Agreement (UCDP, 2018a). However, the organisation was not designated as a ‘Foreign Terrorist Organisation’ by the United States until a few years after the peace agreement was signed (DOS, 2018a). Thus, based on the definitions by the US. Department of State, there were no active terrorist organisations partaking in the Northern Ireland conflict, even if the United Kingdom would beg to differ. Even if it would be interesting to incorporate a more nuanced classification schedule to control for different definitions of a ‘terrorist organisation’, it will not be logistically possible to achieve in this paper, given the large amount of non-state actors analysed. However, it could possibly be an interesting approach for future research to consider.

2.3 Legitimacy

Beyond this, it appears that the legitimate support from not only the local community, but also the international community, has an importance in reaching successful peace agreements (Donais, 2009). There are some non-state actors that the international community would prefer to not legitimise, e.g. terrorist organisations, and the inclusion of such actors in peace accords could thus be seen as a legitimisation of the terrorists’ goals or actions (Arnault, 2014; Hofmann, 2006; McCartney, 2006). This could prove an obstacle for peace, as third-party actors might choose to not engage with designated terrorist organisations in fear of reprimands from the governments that designated those organisations in the first place (Dudouet, 2014). However, if the peace agreement is actually seen as legitimate, then support can be given by e.g. the United Nations (UN) to ease the transition from combat to peace and to ultimately support the peace process. This could be seen in the Guatemalan peace process of the early 1990s (Arnault, 2014). Similar international support might not be present if an agreement includes any designated terrorist organisations, given the increased global commitment to counter-terrorism and state security following 9/11 (Arnault, 2014; Donais, 2009; Tschirgi, 2004).

Research has shown that peace is more likely to be sustained if civil society actors are included in the peace agreement (Nilsson, 2012). According to Nilsson, one of the reasons for this is that it increases the legitimacy for the peace process by, among other things, creating a wider public support for the process domestically. When it comes to armed non-state actors, these usually claim some sort of legitimacy by representing an oppressed group. A clear example of this is how Hezbollah in Lebanon represents, or claims to represent, the Shia minority (Dudouet, 2014). The armed non-state actors can also prove to be a shortcut to promote the peace agreement in areas that are under limited state governance or in more hard-to-reach areas (Mitchell, 2008). However, as highlighted by Hofmann (2006), it can prove to be difficult for state actors to engage with armed non-state actors in peace processes, given that it might legitimise the non-state actors’ actions and/or goals. Even if this is not the case, the inclusion of armed non-state actors might set a dangerous precedent that violent means enables a spot at the negotiating table. This concern becomes an increasingly pressing issue when the state actor is not merely dealing with an armed non-state actor, but with a designated terrorist organisation.

2014). In turn, this could increase the international legitimisation of that actor as a political force, and less as an armed non-state actor. A prime example of this is how the international recognition of Fatah and the PLO changed during the Israeli-Palestinian peace processes in the 1990s. However, this opens the door for other armed non-state actors to take their place and continue an armed struggle. Such events could be seen to have taken place in Palestine with Hamas, and in Northern Ireland with the Real Irish Republican Army (RIRA) and the Continuity Irish Republican Army (CIRA).

In summary, the role of ‘legitimacy’ appears to be highly important in peace building processes. On the one hand, the inclusion of local non-state actors can increase the domestic support for a peace agreement by giving it a wider foundation and thus increasing its legitimacy domestically. Moreover, the legitimisation of a peace agreement internationally can help to bolster its success by increasing support for the peace agreement from third party international actors, as seen in Guatemala. On the other hand, there are concerns that the inclusion of armed non-state actors, and especially designated terrorist organisations, could lead to a decrease in public support for both the peace agreement and the state by legitimising non-state actors’ use of violence and possibly also the act of terrorism. This could also significantly decrease international support if certain armed non-state actors, especially terrorist organisations, are included in the peace process.2

2.4 Removal of spoilers

There are a handful of different understandings of what actually entails ‘spoiling’ of a peace process. However, the main idea of spoiling is that actors, who are either included or excluded from the peace agreement, employ some sort of action to work against the peace process. The reasons for why an actor may wish to spoil the peace process and subsequent peace agreement varies, but it can be seen as a means for the actor to reach its preferred end (Newman & Richmond, 2006). In past research of peace building processes, a lot of thought has been devoted to the removal of spoilers as a way to achieve a more sustainable peace. Several different approaches to dealing with the issue of potential spoilers have been presented. These have covered everything from highlighting the effect of comprehensive peace agreements as a way to combat spoilers (e.g. Hampson, 1996; Nilsson, 2008) to

2 _{For a more in-depth discussion on the roles and implications of legitimacy in peace processes, see: Legitimacy}

discussing different approaches based on the main schools of thought in international relations (e.g. Hofmann & Scheckener, 2011; Schneckener, 2009). In summary, based on past quantitative and qualitative research, the removal of spoilers seems to be an important factor in the creation of a more sustainable peace, even if the preferred approach for removal may differ.

As discussed by Schneckener (2009), it might be necessary to employ different tactics in dealing with different sorts of armed non-state actors, given that they may encompass various goals and actions. It is thus necessary to highlight the particular features of a non-state actor and to not simply view them all as similar actors. The different motivations, whether they are profit-driven or ideological, the different strives, whether they wish to achieve change or to preserve status-quo, and the different tactics, whether they are physical or psychological, all affect an actors’ willingness to spoil an ongoing peace process (Schneckener, 2009). Moreover, another important aspect to consider when it comes to discussing the theoretical aspects of ‘spoilers’ is how the correct removal of spoilers could lead to a transition away from the illegal usage of armed violence and terrorism. The correct removal of the actors’ reasons for spoiling the peace process, potentially through the inclusion of them in the peace agreement, could signal the start of a transitional period from armed non-state actor to legitimate political organisation. A clear example of this is how Fatah and the PLO during the 1990s transformed into an internationally recognised organisation that no longer share the same reasons for spoiling an Israeli-Palestinian peace process.

2.5 Theoretical argument

The first research question is: Does the presence of at least one designated terrorist organisation in the conflict where a peace agreement was signed affect the sustainability of that agreement?

The effect that a presence of terrorist actors in conflicts may have on peace agreements has not been widely discussed in past papers. Instead, the main focus has been on the inclusion of these actors in the actual peace agreements (e.g. Arnault, 2014; McCartney, 2006). It is therefore unclear if there actually is an effect. Accordingly, and as suggested by the above research question, this question remains open. A logical assumption would nevertheless be that the presence of terrorist organisations could constitute a more volatile peace process, given the substantial use of violence against civilians by terrorists, as a way to achieve their goals and to spread fear (Schneckener, 2009). Furthermore, following the increased commitment to counter-terrorism post-9/11, it can be assumed that peace processes supported by the international community are in general not supportive of designated terrorist actors. They might therefore have an increased reason to engage in peace-spoiling behaviour. Finally, terrorist organisations usually represent, or claim to represent, an oppressed part of the country or the population (Dudouet, 2014). This could signal a potential lack of support for the acting government, which in turn could lead a part of the domestic population away from supporting any peace plan constructed by the central-government. The first hypothesis thus follows:

H1: Peace agreements are less sustainable in conflicts where at least one

active warring party is a designated terrorist organisation.

My second research question concerns the actual inclusion of terrorists in peace agreements and follows: Does the inclusion of at least one designated terrorist organisation in a peace agreement affect the agreement’s sustainability?

one hand, the inclusion of armed non-state actors designated as terrorists in the peace process could lead to a decrease of legitimacy for a subsequent peace agreement. Given that the government, and all other included parties, would be negotiating with terrorists, this could be viewed as a legitimisation of the terrorist organisation, and possibly also of their goals. On the other hand, it could also increase the legitimacy of the peace process, at least in the long term, given that the peace agreement will represent a larger proportion of the population, including the oppressed groups the terrorist organisations might (claim to) represent.

All in all, based on previous research into this area and a theoretical discussion, it stems to reason that the inclusion of designated terrorist organisations into peace agreements will lead to a decrease in the sustainability of the peace agreement. The international, and moderate domestic population’s, view of the legitimacy of the peace agreement will be lowered, even if the agreement might be seen as more legitimate by some oppressed minority groups of the domestic population. The removal of spoilers, by incorporating terrorist organisations, will thus likely lead to an increase in the sustainability of the peace agreement in the short term. However, as have been witnessed in both Palestine and Northern Ireland, where one organisation designated as terrorists by some, lay down their violent pursuits, other extreme actors might rise. This will create new potential spoilers of the peace agreement in the long term. The second hypothesis thus follows:

H2: The sustainability of peace agreements will decrease if they include at

3. Research design

In this section, I will discuss and highlight the particular research design I have opted to apply for this paper. I will cover the statistical model applied, discuss the dataset I have created and utilised, as well as the different dependent variables, independent variables, and control variables.

3.1 Choice of method

To analyse the risk of failure for peace agreements, I am conducting a survival analysis by applying a cox proportional hazard model (Cox, 1972). In survival analysis, one is examining how long it takes before a particular event takes place (also called ‘survival time’). In this paper, it will thus be the examination of the survival time of peace agreements from their signings until the agreements broke down or until the observation time ended. If the peace agreement survived until the end of observations, it will be censored at this point in time.3 The reason for conducting a survival analysis over e.g. an ordinary linear regression analysis is, among other things, that the survival analysis can more effectively manage the censoring of observations (Despa, 2018).

The application of a cox proportional hazard model enables me to analyse the effect of several covariates simultaneously. Furthermore, the application of a cox proportional hazard model will be in line with previous quantitative research in this field (e.g. Nilsson, 2008; Nilsson, 2012).4 _{The model looks as follows:}

h(t) = h0(t) × exp(b1x1 + b2x2 + … + bpxp)

Where ‘t’ represents the survival time, (x1, x2,…, xp) are the covariates applied, (b1, b2, …, bp)

are coefficients that measure the size of the effect of the covariates, and ‘h0’ is the baseline

hazard.

3 _{This means that we are only observing the peace agreement until the time of censoring and that we are}

making no comments about the future of the agreement after the observation period ended.

4 _{Some previous research has applied a time-varying model to enable the application of time dependent}

Through this model, it will thus be possible to examine several factors’ simultaneous effect on the rate of a peace agreement breaking down at a particular point in time. Each covariate will receive a ratio indicating that factor’s effect on the survival of peace agreements (also called ‘hazard ratio’ or simply ‘HR’). A hazard ratio of (1) indicates that the particular factor has no effect on the survival time of peace agreements. A hazard ratio less than (1) indicates a decreased risk for the particular event examined to take place (the breakdown of the peace agreements) and thus a positive effect on the survival time of peace agreements. In the same way, a hazard ratio of more than (1) indicates an increased risk for a breakdown of peace agreements. The hazard ratios are reported relative to one (1), e.g. an HR of 1.70 means that the risk of a peace agreement collapsing is increased by 70 %, and an HR of 0.65 means that the risk is decreased by 35 %. Moreover, to test the effect of my selected variables, I will include the Wald statistic, which will be reported in parenthesis under the hazard ratio for each covariate. The Wald test is a way to check if the explanatory variables add anything to the statistical model, i.e. if the covariate is significant.

HR = 1 (no effect)

HR < 1 (decreased risk for breakdown of peace agreements) HR > 1 (increased risk for breakdown of peace agreements)

3.2 Data

To enable an analysis over terrorist organisations’ role in peace processes, I have had to move past the information used in previous research papers. I have thus created a unique dataset, combining all intrastate conflict peace agreements, and the conflicts’ warring parties, with the US. Department of State’s classification of terrorist organisations. Through this approach, I will be able to examine if the incorporation of terrorist organisations in peace agreements affect their sustainability, as well as controlling for the durability of peace agreements in conflicts where terrorist organisations are active warring parties of the conflict. I have utilised the Uppsala Conflict Data Program (UCDP) and their unique collection of data and information on conflicts and peace agreements.5 _{The UCDP Peace Agreement}

5 _{For the purpose of this paper, I will follow the UCDP’s definition of a conflict: “a contested incompatibility that}

Dataset has functioned as a core foundation for this research, providing a detailed collection on every peace agreement between 1989 – 2011 that has been signed by at least two warring parties that are on opposite sides of the conflict (UCDP, 2011a). I have excluded all peace agreements related to interstate conflicts and have only included those covering intrastate conflicts. These include disputes over territory and/or government, and could either be partial or full agreements. Some previous quantitative research in the field of peace agreements (e.g. Nilsson, 2008) have opted to exclude ‘peace process agreements’, which are agreements that “initiate[s] a process that aims to settle the incompatibility” (UCDP, 2011b, p. 10) instead of actually settling the incompatibility. Through the creation of two separate models where one includes and one excludes ‘peace process agreements’, I will be able to build on previous research while also utilising a larger number of cases for my study. Worth noting is that there is a clear discrepancy in the number of peace agreements signed per conflict. Some conflicts registered several peace agreements during the time period monitored, e.g. Uganda with 8 agreements. At the same time, some conflicts only registered one (1) peace agreement during the time studied, e.g. Senegal.

Building on the existing information provided by the UCDP, I collected information on every non-state warring party in the conflicts where a peace agreement subsequently had been signed. These non-state actors were then compared to the US. Department of State’s (DOS) lists covering designated terrorist organisations or entities. To ensure that the classification of ‘terrorism’ could have had an impact on the peace process, the non-state actors have been compared to the designated terrorist-lists prior to the peace process. E.g. if a peace agreement was signed in the year 2000, the relevant actors were compared to the DOS’ lists published before this date, most likely in the year 1999 or previously. Given that the DOS only keeps online recorded information on ‘terrorist designations’ from 1997 and forward, I only include peace agreements signed in 1998 and forward (i.e. peace agreements signed 1998 – 2011).

2018b; DOS, 2018c). Thus, it becomes clear that terrorist organisations have become an important factor in post-Cold War intrastate conflicts.

3.3 Dependent variable

For the purpose of this paper, I have created a dependent variable called: ‘Peace agreement sustainability’, to examine how designated terrorist organisations affect the sustainability of peace agreements. This dependent variable measures two different variables, namely the survival time of peace agreements and whether or not a peace agreement broke down during the time of observation. The survival time variable measures the amount of time in months (rounded down)6 _{that a peace agreement was in effect, from its signing until the day} it broke down or until the observation time period ended. To incorporate whether or not a peace agreement broke down during the time of observation, a dummy variable was created. This was coded (0) if the peace agreement remained intact throughout the time period observed, and (1) if the peace agreement broke down during the observation period. In this paper, since I use the UCDP data I have also opted to follow UCDP’s definition of when a peace agreement broke down. This differs from some previous research which, instead of applying the definition of UCDP, opted for a definition based on when a conflict post-peace agreement reached a level of 25 battle-related deaths (e.g. Nilsson 2008; Nilsson, 2012). According to the UCDP Peace Agreement Dataset Codebook, a peace agreement has ended:

If the validity of the agreement is contested by one or more of the warring parties that signed. A peace agreement cannot, from the UCDP perspective, survive if the primary parties are no longer party to it. If a party officially withdraws from a peace agreement, it is considered to have ended. (UCDP, 2011b, p.5).

While the option to incorporate the UCDP definition, and not the one used in previous research, will limit my ability to compare my results to past research, it nonetheless holds some advantages. This is based on the fact that previous research is applying a different

6 _{The only exception were the peace agreements that broke down in less than a month. They were coded as}

dependent variable. I am measuring the life span of peace agreements, whereas past research has looked at the longevity of peace. That said, the benefit of my approach is that it allows for a more detailed definition of the survival time of peace agreements. Moreover, through this approach, I will be able to measure the survival time of peace agreements on a monthly basis, instead of on a yearly basis. It is thus important to remember that the two options imply two different dependent variables, measuring two different aspects of the end of peace and peace agreements. In this respect, it could also be argued the measurement used here is more precise.

3.4 Independent variables

For the purpose of this paper, I have created two key independent variables that aim to capture the effect that terrorist organisations have on the sustainability of peace agreements. The first one is ‘Terrorist actors in conflict’, which is a dummy variable that captures whether or not any designated terrorist organisations were active warring parties of the conflict where a peace agreement was signed, at the time of signing. The variable is coded (0) if no designated terrorist organisations were present at the time of signing and (1) if at least one designated terrorist organisation were active in the conflict, at the time of signing. Similarly, ‘Terrorist actors in peace agreements’ is a dummy variable that captures if at least one designated terrorist organisation was a signatory party to the peace agreement. The variable is coded (1) if at least one signatory party was a designated terrorist organisation and (0) if none of the signatories were terrorist organisations. Through both independent variables, I have made sure that the official designation of ‘terrorist organisation’ has occurred prior in time to the signing of the peace agreement, to remove the risk of reversed causality.

application of a single definition of the term. Secondly, given the rather ambiguous definition of ‘terrorism’, any study related to this term faces the risk of several reliability issues. However, this should not be of concern for this paper, since I have opted to apply the DOS’ lists of terrorist actors. This enables a fairly strict and straightforward process to divide the armed non-state actors into two groups, those that the DOS considers as terrorist actors and those that they do not consider to be terrorists. Furthermore, through the application of the timestamps from each terrorist designation, it is also possible to control for whether or not the terrorist designation occurred before the signatory of the peace agreement. All in all, there are several concerns that could be raised regarding my classification of terrorist organisations in the independent variables. However, given the scope of this paper, and with regards to the present lack of logistically applicable solutions to this issue, this should still be seen as an appropriate first step in the analysis of terrorist organisations’ effect on peace agreements.

3.5 Control variables

In order to make sure that the effect of terrorist organisations on the sustainability of peace agreements is isolated, a few control variables are incorporated. These will be applied to the cox proportional hazard model to avoid a spurious result and to allow me to further discuss and analyse the effect of terrorist organisations on peace agreements.

Conflict issue

Conflict intensity

The intensity of the conflict, where the peace agreement has been signed, has previously shown to be of importance for the sustainability of the peace process (e.g. Hartzell & Hoddie, 2003; Nilsson, 2012). To construct a manageable covariate for this factor, I have divided the conflicts into two levels of intensity: those that have at least once reached the cut-off point of 1,000 battle-related deaths in a single calendar year, and those that have not. The reasoning behind choosing “1,000 battle-related deaths” as the cut-off point is to relate to previous research (e.g. Nilsson, 2012) which have used this level, and because the UCDP applies this level of battle-related deaths as the cut-off point between a minor conflict and war (UCDP, 2018b). If the conflict reached the stated level of 1,000 battle related deaths in a single calendar year, at least once, it will be designated as (1) and if not it will receive (0). The data covering the amount of battle-related deaths, and thus the intensity level of the conflict, has been retrieved from the UCDP (2018c). For the purpose of this paper, I will rely on a dichotomous control variable for this factor, however, it would be interesting for future research to expand on this and apply a continuous and time dependent covariate to control for a conflict’s intensity.

Inclusive peace agreement

There has been a lot of previous research discussing the importance of inclusive peace agreements, as a way to remove armed non-state actors as potential future spoilers (e.g. Hampson, 1996). Nilsson (2008) also found statistical evidence for how the exclusion of at least one warring party from a peace agreement leads to a heightened risk of violence resurrecting. The inclusion of all warring parties in peace agreements will thus be an important factor to control for in my analysis. The covariate ‘inclusive’ will be coded as a dummy variable with (0) if at least one warring party is excluded from the peace agreement, and (1) if all warring parties are included in the peace agreement. If the agreement is not comprehensive, it means that at least one non-state actors was excluded from it, or chose not to engage in the peace process7_{. The data used to construct and code this variable has} been gathered from UCDP (2011a).

7 _{Every peace agreement included in the UCDP Peace Agreement Dataset had a government or government}

Power sharing

Previous research has highlighted how power sharing provisions could have a significant impact on the durability of peace agreements (e.g. Hartzell & Hoddie, 2003; Nilsson, 2012). It will thus be important to control for this factor. To be able to do this, I created a dummy variable ‘Power sharing’ reflecting whether or not a peace agreement contained any power sharing arrangements. This data was retrieved from the UPCD Peace Agreement Dataset and reflects if a peace agreement included any political, military, or territorial power sharing arrangements. The agreements that contained at least one power sharing arrangement were coded (1) and those that did not were coded (0).

However, following the rather poor variation in the previously mentioned dummy variable8_{, and to more in-depth look at the effect of power sharing arrangements, I} have created a second control variable: ‘Power sharing complexity’. This variable considers the varying levels of complexity that the power sharing provisions might have in peace agreements by looking at the number of different power sharing arrangements that are present. It is coded (0) if there are no power sharing arrangements included, (1) if one division of arrangements is included, (2) if two different divisions of arrangements are included, and (3) if all three different divisions of arrangements are included. These could either be military provisions, political provisions, or territorial provisions. The military provisions include, among other variables, whether or not a DDR (disarmament, demobilisation, reintegration) arrangement was included and if the agreement provided for a ceasefire between the warring parties. The political provisions include, among other variables, the provisions for an upcoming election, the right for non-state groups to become political parties, and the arrangements for national talks. The territorial provisions include, among other variables, the creation of local power sharing arrangements, border demarcation, and the granting of independence for specific regions.9 _{The peace agreement}

8 _{5 out of 65 peace agreements in the dataset without ‘peace process agreements’ and 9 out of 83 peace}

agreements including ‘peace process agreements’ did not contain any power sharing provisions and were thus coded (0) in the dummy variable.

9 _{For a complete list of the included variables, see the codebook for the UCDP Peace Agreements Dataset.}

will qualify for the inclusion of a power sharing arrangement if at least one of the sub-variables in one of the divisions (military, political, territorial) is included.

Discussion: Control variables

First of all, it is worth noting that I am occasionally utilising different data and definitions compared to previous quantitative research on the durability of peace agreements, which limits the comparability of my results with theirs. However, despite this, the results and discussions in this paper should still showcase a developed discussion about armed non-state actors’, and in particular terrorist organisations’, roles and effects in peace agreements. Given the relative simplicity and how straightforward the selected control variables are, there should not be any major validity concerns with these measurements. It is possible to highlight a potential reliability concern given that all but one of the control variables are coded as dummies. However, based on the dichotomous nature of the factors controlled for, and how the variables are defined, no severe reliability issue should emerge and should not be of concern for the results and upcoming analysis.

4. Results and analysis

This section will be dedicated to the presentation and discussion of the results that followed the survival analysis. I will briefly cover the descriptive statistics of my chosen variables, the results that were produced in the cox proportional hazard model, and finally discuss how these results either support or contradict the two hypotheses of this paper.

4.1 Descriptive statistics

While reviewing the two datasets used in this paper, where the first excludes and the second includes peace process agreements, it becomes clear that the presence of terrorist actors is not limited to a small proportion of intrastate conflicts. Designated terrorist organisations by the US. Department of State were active warring parties in 4 out of 22 conflicts (18 %) and in 6 out of 27 conflicts (22 %) when including the conflicts where a peace process agreement was signed. Thus, 14 out of 65 peace agreements (21,5 %) were signed in conflicts where at least one designated terrorist actor was present at the time of signing the peace agreement. This figure increases to 21 out of 83 peace agreements (25,3 %) when peace process agreements are included. Furthermore, 8 out of 65 peace agreements (12,3 %) had at least one designated terrorist organisation as a signatory of the agreement. When including signed peace process agreements, this figure increases to 15 out of 83 agreements (18 %).

Table 1. Descriptive statistics (excluding ‘peace process agreements’)

4.2 Results

Table 3

Table 3 summarises the results of the statistical analyses made on the first dataset, which excludes peace process agreements. The hazard ratios (HR) are reported relative to one (1), e.g. an HR of 2.10 means that the risk of a peace agreement collapsing is increased by 110 %, and an HR of 0.70 means that the risk is decreased by 30 %. Model 1 through 3 focuses on the effect of a presence of terrorist organisations in the conflict where the peace agreement was signed, and model 4 to 6 at the effect of including terrorist organisations in the actual peace agreement. Model 1 has an HR of 3.575, which indicates that the presence of at least one terrorist organisation in a conflict has a significantly destabilising effect on subsequent peace agreements in that conflict. The presence of terrorist organisations increases the risk of a peace agreement failing by 257,5 %, which is statistically significant at the 0.01 level. This destabilising effect is decreased narrowly to 232,7 % (an HR of 3.327) when controlling for the conflict issue, the intensity of the conflict, if the peace agreement was inclusive, and if any power sharing provisions were included in the agreement (showcased in model 2). The effect remains statistically significant, but now at a 0.05 level. None of the control variables indicates a statistically significant effect on the sustainability of peace agreements (model 2). However, given that the variation on the dummy variable ‘power sharing’ is so low10_{, I exchange this covariate for one that measures the complexity of different power} sharing arrangements (shown in model 3). First of all, ‘power sharing complexity’ is now shown to have a statistically significant effect on the sustainability of peace agreements at the 0.10 level and decreases the risk of a peace agreement failing with 41 % (an HR of 0.59). Secondly, it causes no significant effect on the other control variables, however, it does produce some very interesting results on our key independent variable, the presence of terrorist actors. When controlling for the complexity of power sharing provisions, the presence of terrorist actors in the conflict is no longer statistically significant. It has also neutralised some destabilising effect caused by terrorist actors, thus lowering the destabilising effect by the key independent variable to 129,4 %. Thus, it seems that the more power sharing dimensions included in the peace agreement, the more stability is to be find

10 _{92.3 % of the peace agreements, excluding ’peace process agreements’ (table 1), and 89.2 % of the peace}

in the longevity of the actual agreements, and the destabilising effect caused by the presence of terrorist actors is neutralised.

Moving forward to the effect on the sustainability of peace agreements by including designated terrorist actors in the actual agreements, we find similar results (model 4 – 6). Peace agreements that include designated terrorist actors, compared to those that do not – regardless if terrorist actors are actually present in the conflict – will increase the risk of a peace agreement collapsing by 271,8 % (model 4). This is statistically significant at the 0.01 level. When controlling for the conflict issue, the intensity of the conflict, if the peace agreement included all warring parties, and if any power sharing provisions were included, the results remain nearly the same (model 5). The inclusion of terrorist actors will, according to model 5, increase the risk of the peace agreement failing by 278,1 % (an HR of 3.781), and is now statistically significant at the 0.05 level.

Table 3. Survival

analysis (excluding

Table 4

Table 4 summarises the results of the statistical analyses made on the second dataset, which includes peace process agreements. Following the structure laid out in table 3, model 1–3 looks at the presence of terrorist actors in the conflicts where the peace agreements were signed and model 4–6 at the inclusion of terrorist actors in the actual agreements. Model 1 indicates that the presence of terrorist actors in the conflict where a peace agreement was signed, will increase the risk of that peace agreement ending with 157 % (an HR of 2.570). This effect is statistically significant at the 0.05 level. When controlling for the conflict issue, the intensity of the conflict, if the peace agreement was inclusive, and if there were any power sharing provisions included, this effect remains largely unchanged (model 2). The presence of terrorist actors in the conflict will now increase the risk of a peace agreement collapsing with 177,1 % (an HR of 2.771), which is statistically significant at the 0.10 level. None of the control variables are statistically significant. When including the covariate for ‘power sharing complexity’ (model 3) we, once again, get some interesting results. Although this time, none of the covariates, neither ‘power sharing complexity’ nor ‘terrorist actors in conflict’, are statistically significant. The effect of the presence of terrorist actors has been significantly neutralised and is now at 99,9 % (an HR of 1.999).

When looking at the effect of including terrorist actors in peace agreements (model 4), the effect is statistically significant at the 0.01 level. The inclusion of terrorist actors in peace agreements increases the risk of failure with 227,6 % (an HR of 3.276). This effect is strengthened (280,3 %) when the control variables are included, although the effect is now statistically significant at the 0.05 level (model 5). As witnessed in model 2, none of the control variables are statistically significant. In model 6, where the complexity of the power sharing provisions is included, the effect of including terrorist actors in peace agreements is lowered, although it remains statistically significant at the 0.10 level. The inclusion of terrorist actors in peace agreements will now result in an increased risk of the agreement collapsing with 174,9 % (an HR of 2.749). No other covariate has a statistically significant effect on the sustainability of peace agreements in this model.

Table 4. Survival

analysis (including

4.3 Discussion

The results presented in table 3 and table 4 above highlight the significant effect designated terrorist actors have on the sustainability of peace agreements in intrastate conflicts. The strong destabilising effect is present both when including and excluding peace process agreements, although there is a slight difference in how powerful the effect is. Furthermore, the destabilising effect appears to be even more powerful when at least one terrorist actor is included in the peace agreement. This difference is even larger when peace process agreements are included. Moreover, it is very fascinating to find such a strong neutralising effect on terrorist actors by the covariate measuring the complexity of power sharing provisions.

The second hypothesis of this paper states: “The sustainability of peace agreements will decrease if they include at least one actor designated as a terrorist organisation”. This hypothesis is supported by model 4 and 5 in table 3, which indicates that the inclusion of terrorist actors in peace agreements increases the risk of peace agreements failing with 371,8 % and 378,1 % respectively. The second hypothesis is also supported by model 4, 5, and 6 in table 4, which indicates an increased risk of peace agreements collapsing by 327,6 %, 380,3 %, and 274,9 % respectively.

However, in model 6 of table 3, the destabilising effect of including terrorist actors in peace agreements is neutralised, and no longer statistically significant, when controlling for the complexity of power sharing provisions. This goes against the hypothesis, but is in line with the results related to the presence of terrorist actors in conflicts which was discussed above. It thus indicates that the inclusion of terrorist actors in peace agreements creates a destabilising effect, however, the incorporation of power sharing provisions decreases this destabilising effect. This highlights from a public policy perspective, the importance of including several power sharing arrangements in peace agreements related to terrorist actors, to work against the destabilising effect they might otherwise have.

One important aspect of the results, concerning the inclusion of terrorist organisations in peace agreements, is that no support can be found for the idea that it would remove potential spoilers and thus decrease the risk of a breakdown of peace. Instead, the destabilising effect of terrorist actors was larger than the destabilising effect of the presence of terrorist actors in the conflict, in all models in both datasets. This directly contradicts the theoretical discussions about spoiler removal and suggests that it might be some additional factors interfering with the results. Unfortunately, it is not possible to distinguish these potential factors in this paper and the discovery of them requires additional research with a more advanced model. A possible solution is to apply a divided survival model that is looking at the signatories of a peace agreement separately from the excluded parties, similarly to the model used by Nilsson (2008; 2012). Due to the many limitations of a bachelor thesis, I was not able to implement such a model. This would nevertheless be a natural continuation for future research and enable a more detailed discussion on the spoiler effect of terrorist organisations.

5. Conclusion

In this paper, I have attempted to research the effect that terrorist organisations have on the sustainability of peace agreements in intrastate conflicts. To expand on previous research, while also maximising potential observations, I have worked with two separate datasets throughout this paper; one that excludes peace process agreements and one that includes them. My two starting hypotheses were that “Peace agreements are less sustainable in conflicts where at least one active warring party is a designated terrorist organisation” and “The sustainability of peace agreements will decrease if they include at least one actor designated as a terrorist organisation”. To distinguish ‘terrorist organisations’ among the different types of armed non-state actors present in intrastate conflicts, I applied the ‘terrorist designation lists’ by the US. Department of State. To enable a theoretical discussion about the results and possible underlying reasons, I have considered the legitimacy aspect of peace agreements and the potential spoiler effect that terrorist organisations might have.

have not increased enough to provide for a more durable peace, given that peace process agreements are only in the beginning of the peace process.

6. Works cited

Arnault, J. (2014). Legitimacy and peace processes: International norms and local realities. In A. Ramsbotham & A. Wennmann (Eds.), Legitimacy and peace processes: From coercion to consent, (pp. 21-25). London, England: Conciliation Resources. https://www.c-r.org/downloads/Accord25WEB.pdf

Clements, K. P. (2014). What is legitimacy and why does it matter for peace? In A.

Ramsbotham & A. Wennmann (Eds.), Legitimacy and peace processes: From coercion to consent, (pp. 13-16). London, England: Conciliation Resources. https://www.c-r.org/downloads/Accord25WEB.pdf

Cox, D. (1972). Regression Models and Life-Tables. Journal of the Royal Statistical Society. Series B (Methodological), 34(2), 187-220. Retrieved from

http://www.jstor.org/stable/2985181

Davis, D. E. (2009). Non-State Armed Actors, New Imagined Communities, and Shifting Patterns of Sovereignty and Insecurity in the Modern World. Contemporary Security Policy, 30:2, 221-245, DOI: 10.1080/13523260903059757

Davis, D. E. (2010). Irregular Armed Forces, Shifting Patterns of Commitment, and

Fragmented Sovereignty in the Developing World. Theory and Society, 39:397-413,

https://doi.org/10.1007/s11186-010-9112-6

Despa, S. (2018). What is survival analysis? StatNews #78. Cornell Statistical Consulting Unit. Cornell University. Accessed December 2018.

https://www.cscu.cornell.edu/news/statnews/stnews78.pdf

Donais, T. (2009). Empowerment or Imposition? Dilemmas of Local Ownership in Post- Conflict Peacebuilding Processes. Peace & Change, 34:1, 3-26, DOI:

https://doi.org/10.1111/j.1468-0130.2009.00531.x

DOS, US. Department of State. (2001). Executive Order 13224. Accessed November 2018. https://www.state.gov/j/ct/rls/other/des/122570.htm

DOS, US. Department of State. (2004). Terrorist Exclusion List. Accessed November 2018. https://www.state.gov/j/ct/rls/other/des/123086.htm

DOS, US. Department of State. (2018b). Individuals and Entities Designated by the State Department Under E.O. 13224. Accessed November 2018.

https://www.state.gov/j/ct/rls/other/des/143210.htm

DOS, US. Department of State (2018c). Terrorism Designations Press Releases. Accessed November 2018. https://www.state.gov/j/ct/rls/other/des/index.htm

Dudouet, V. (2014). Transformation of coercive actors. In A. Ramsbotham & A. Wennmann (Eds.), Legitimacy and peace processes: From coercion to consent, (pp. 91-94). London, England: Conciliation Resources.

https://www.c-r.org/downloads/Accord25WEB.pdf

EU. (2009). COUNCIL COMMON POSITION 2009/468/CFSP. Official Journal of the European Union, 16/6-2009, p. 45-50.

https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32009E0468&qid=1412596355797&from=EN EU. (2018). COUNCIL DECISION (CFSP) 2018/1084. Official Journal of the European Union,

31/7-2018, p. 144-146. https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32018D1084&from=EN

Freilich, J. D., et al. (2009). Surveying American State Police Agencies About Terrorism

Threats, Terrorism Sources, and Terrorism Definitions. Terrorism and Political Violence, 21:3, 450-475, DOI: 10.1080/09546550902950324

Hampson, F. O. (1996). Nurturing Peace: Why Peace Settlements Succeed or Fail. Washington, DC: United States Institute of Peace Press.

Hansen, A. S. (1997). Political legitimacy, confidence-building and the Dayton peace agreement. International Peacekeeping, 4:2, 74-90, DOI:

10.1080/13533319708413667

Hartzell, C., & Hoddie, M. (2003). Institutionalizing Peace: Power Sharing and Post-Civil War Conflict Management. American Journal of Political Science, 47(2), 318-332. Doi: 10.2307/3186141 https://www.jstor.org/stable/3186141

Hofmann, C. (2006). Engaging Non-State Armed Groups in Humanitarian Action. International Peacekeeping, 13:3, 369-409, DOI: 10.1080=13533310600824090 Hofmann, C., & Schneckener, U. (2011). Engaging non-state armed actors in state- and

Koehler, J., & Zürcher, C. (2007). Assessing the Contribution of International Actors in

Afghanistan: Results from a Representative Survey. SFB-Governance Working Paper Series, No. 7, https://www.files.ethz.ch/isn/95432/WP7e.pdf

Lentini, P. (2008). Understanding and combating terrorism: Definitions, origins and strategies. Australian Journal of Political Science, 43:1, 133-140, DOI:

10.1080/10361140701842615

McCartney, C. (2006). Engaging armed groups in peace processes: reflections for practice and policy from Colombia and the Philippines. Conciliation Resources.

https://c-r.org/downloads/CR_Engaging_armed_groups2006.pdf

Mehler, A., Lambach, D., & Smith-Höhn, J. (2010). Legitimate oligopolies of violence in post- conflict societies with particular focus on Liberia and Sierra Leone. Forschung DSF, 23, https://nbn-resolving.org/urn:nbn:de:0168-ssoar-260456

Mitchell, C. (2008). The Limits of Legitimacy: Former Loyalist Combatants and Peace-Building in Northern Ireland. Irish Political Studies, 23:1, 1-19,

DOI: 10.1080/07907180701767922

Narten, J. (2008). Post-Conflict Peacebuilding and Local Ownership: Dynamics of External- Local Interaction in Kosovo under United Nations Administration. Journal of Intervention and Statebuilding, 2:3, 369-390, DOI: 10.1080/17502970802436361 Newman, E., & Richmond, O. (2006). Peace building and spoilers. Conflict, Security &

Development, 6:1, 101-110, DOI: 10.1080/14678800600590728

Nilsson, D. (2012). Anchoring the Peace: Civil Society Actors in Peace Accords and Durable Peace. International Interactions, 38:2, 243-266, DOI:

10.1080/03050629.2012.659139

Nilsson, D. (2008). Partial Peace: Rebel Groups Inside and Outside of Civil War Settlements. Journal of Peace Research, 45:4, 479-495, http://www.jstor.org/stable/27640711 Northern Ireland (Emergency Provisions) Act 1996.

http://www.legislation.gov.uk/ukpga/1996/22/pdfs/ukpga_19960022_en.pdf Northern Ireland (Emergency Provisions) Act 1998.

http://www.legislation.gov.uk/ukpga/1998/9/pdfs/ukpga_19980009_en.pdf

Roberts, D. (2011). Post-Conflict Peacebuilding, Liberal Irrelevance and the Locus of Legitimacy. International Peacekeeping, 18:4, 410-424,

DOI: 10.1080/13533312.2011.588388

Schneckener, U. (2009). Spoilers or Governance Actors? Engaging Armed Non-State Groups in Areas of Limited Statehood. SFB-Governance Working Paper Series, No. 21,

http://www.sfb-governance.de/en/publikationen/sfb-700-working_papers/wp21/SFB-Governance-Working-Paper-21.pdf

Stedman, S. (1997). Spoiler Problems in Peace Processes. International Security, 22(2), 5-53. DOI: 10.2307/2539366

Tschirgi, N. (2004). Post-Conflict Peacebuilding Revisited: Achievements, Limitations,

Challenges. Prepared for the WSP International/IPA Peacebuilding Forum Conference, New York.

UCDP. (2011a). UCDP Peace Agreement Dataset. The Uppsala Conflict Data Program. http://ucdp.uu.se/downloads/

UCDP. (2011b). UCDP Peace Agreement Dataset Codebook. The Uppsala Conflict Data Program. http://ucdp.uu.se/downloads/

UCDP. (2018a). RIRA. The Uppsala Conflict Data Program. Accessed November 2018. http://ucdp.uu.se/#/actor/283

UCDP. (2018b). UCDP/PRIO Armed Conflict Dataset Codebook version 18.1. The Uppsala Conflict Data Program. Accessed December 2018. http://ucdp.uu.se/downloads/ UCDP. (2018c). UCDP/PRIO Armed Conflict Dataset version 18.1. The Uppsala Conflict Data