Decentralization: a double-edged sword?
A comparative case study on decentralization and its influence on ethnic conflict and secessionism
Andrés Durante
Bachelor Thesis
Uppsala University, Fall 2017
Department of Peace and Conflict Research
Supervisor: Espen Geelmuyden Rød
Abstract
This study evaluates the influence of decentralization on ethnic conflict and secessionism.
There is little agreement in the existing literature, as several scholars argue for regional autonomy as an effective method of conflict resolution while other scholars argue that decentralization exacerbates ethnic conflict by reinforcing ethnic identities and promoting ethnic mobilization. By adopting a theoretical framework that distinguishes itself from previous research, an argument is proposed that while regional autonomy concessions provide a direct short-term effect, the indirect consequences of decentralization risk reinforcing ethnic identities and encourage ethnic mobilization long-term. Thus, a combination of territorial and governmental power sharing is necessary for decentralization to be successful in reducing ethnic conflict and secessionism. By adopting a combined structural and group perspective on decentralization, the argument is implemented in a within-case comparison of Spain as well as a between-case comparison with Bosnia and Herzegovina. Main findings suggest that regional autonomy does not provide short-term reduction, rather, a combination of territorial and governmental power sharing with full inclusion into executive power is required for decentralization to successfully reduce ethnic conflict and secessionism. Additional insights on national identity as a compelling variable with potential theoretical inference were also gained.
Keywords: decentralization, ethnic conflict, secessionism, autonomy, power sharing
Table of contents
LIST OF FIGURES AND TABLES ...
1 INTRODUCTION ... 1
2 THEORY ... 2
2.1 Definitions ... 2
2.2 Literature review... 4
2.3 Theoretical framework ... 5
3 METHOD ... 8
3.1 Operationalization of dependent and independent variables ... 9
3.1.1 Dependent variable ... 9
3.1.2 Independent variable ... 9
3.2 Control variables ... 10
3.2.1 Degree of democracy ... 11
3.2.2 Electoral system ... 11
3.2.3 Ethnic and cultural fractionalization ... 12
3.2.4 Ethnic group size ... 13
3.2.5 Socioeconomic status ... 13
3.3 Validity and reliability ... 14
3.4 Case selection ... 15
3.5 Comparative method ... 17
3.6 Sources of empirical material ... 19
4 RESULTS AND ANALYSIS ... 20
4.1 Spain ... 20
4.1.1 Background ... 20
4.1.2 Dependent variable ... 21
4.1.3 Independent variable ... 22
4.1.4 Control variables ... 23
4.1.5 Interpretation ... 24
4.2 Bosnia and Herzegovina ... 26
4.1.1 Background ... 26
4.1.2 Dependent variable ... 27
4.1.3 Independent variable ... 28
4.1.4 Control variables ... 29
4.1.5 Interpretation ... 29
4.3 Consistency with hypothesis... 31
4.4 Limitations ... 31
4.5 Alternative explanations ... 32
5 SUMMARY AND CONCLUSIONS ... 34
REFERENCES ... 36
List of figures and tables
FIGURE 1. Causal path ... 6
FIGURE 2. Theoretical framework... 7
FIGURE 3. Reverse causation ... 32
TABLE 1. Case selection strategy ... 16
TABLE 2. Within-case comparison of Spain ... 20
TABLE 3. Between-case comparison of Spain and Bosnia and Herzegovina ... 26
1 Introduction
The relationship between decentralization and conflict has been subject to intensive study.
Much of the research has focused on its effectiveness at reducing ethnic conflict and secessionism. Yet, there is little agreement in the literature as several scholars argue that regional autonomy indeed eases ethnic tensions and reduces the probability of conflict while other scholars view such concessions as an ineffective method of conflict resolution.
Proponents see decentralization as a successful strategy to target demands for autonomy and secession through an increase of local self-governance, achieved by a division of power between central and regional governments. Opponents argue that decentralization aggravates ethnic conflict by reinforcing ethnic identities and promoting ethnic mobilization. Previous scholarship also suffers from several shortcomings, making it hard to assess the contribution of decentralization to conflict resolution (Cederman et al. 2015: 356). The ability of decentralization to reduce ethnic conflict and secessionism is found to be greater in some states than in others (Brancati 2006). A research gap is identified given the contemporary lack of consensus. To examine this phenomenon, the following research question was developed:
“What is the influence of decentralization on ethnic conflict and secessionism?”
The aim of this study is to analyze the effects of decentralization on ethnic conflict and
secessionism in ethnically divided states. Motivation behind the research question stems from
the current conflicting theoretical viewpoints. Today, decentralization is seen as one of the
state’s most prominent options for confronting demands for self-governance from territorially-
concentrated, ethnically-differentiated groups (Martínez-Herrera 2010). It is generally believed
that multi-ethnic societies are more prone to conflict compared to ethnically homogenous
societies. The question of how states can best accommodate ethnic groups and avoid or reduce
violent conflict is therefore highly relevant (Lijphart 1977). The recent Catalan illegal
independence referendum in 2017 in Spain, and the outbreak of violent conflict between
security forces and Kurds in Iraq following an Iraqi Kurdistan independence referendum in
2017 are both present-day examples of developments that contradict the successful
decentralization reforms implemented to prevent and reduce ethnic conflict and secessionism
in said states. Thus, contributions that expand our understanding of decentralization and its
potential conflict-preventing and conflict-reducing implications are of importance, particularly
in significance for policy – taking into consideration the substantial threat that ethnic conflict poses on peace and stability in the twenty-first century. There is a current need for in-depth research, theory testing and theory building on the role of the state in ethnic conflict, which has largely been absent from conventional theories within the field (Cederman et al. 2010: 87). This study aims at contributing to the debate by adopting a theoretical framework that distinguishes itself from previous research – as the argument asserts that although regional autonomy concessions provide a direct short-term effect, the indirect consequences of decentralization risk reinforcing ethnic identities and encourage ethnic mobilization long-term. Instead, a combination of territorial and governmental power sharing is needed for decentralization to be successful at reducing ethnic conflict and secessionism. With a shared structural and group perspective on decentralization, the argument is applied in a within-case comparison of Spain as well as a between-case comparison with Bosnia and Herzegovina. The findings suggest that regional autonomy does not provide short-term reduction, rather, a combination of territorial and governmental power sharing – with full inclusion into executive power – is required for decentralization to reduce ethnic conflict and secessionism successfully. Additional insights on national identity and its potential implications for theory were also discovered.
The study proceeds as follows: first, relevant theoretical definitions are presented followed by a literature review of previous research. Afterwards, the theoretical framework and its respective hypotheses are introduced, subsequently used to apply the theoretical argument about the relationship between decentralization and ethnic conflict and secessionism.
Methodological section follows, where operationalization of dependent, independent and control variables are outlined, continued with discussion about validity and reliability, case selection, method and sources of empirical material. Then, sections with results, analysis and interpretation are presented followed by a discussion on limitations and alternative explanations. Lastly, a summary of the main findings with remarks on the theoretical implications of the results and suggestions for future research conclude the study.
2 Theory
2.1 Definitions
In the context of this study, ethnic conflict is defined as all forms of small- and large-scale
violent acts between different ethnic groups. Ethnic groups are defined as groups of people that
share and belong to a common category of attribution such as race, ethnicity, tribe, religion, etc. (Horowitz 1985). Secessionism differs from ethnic conflict in that it refers to the desire of groups for an independent state. Secessionism is commonly, albeit not always, associated with violence and ethnic conflict (Hechter 1992).
The definition of decentralization has generated a considerable amount of debate among scholars, primarily due to its ambiguous theoretical nature. Because of this, a variety of definitions can be found in the literature on the subject. There is a general consensus on its most basic characteristics however, which constitutes the guaranteed division of power between central and regional governments (Lijphart 1999: 186; Wolff 2013: 32-33). More specifically, decentralization is a system of government where a vertical division of power is present among multiple levels of government which have independent decision-making power over at least one issue area. As such, decentralized systems deliberately share power and resources between national, regional and local levels of government. These different levels of government can therefore legislate independently on certain matters (Brancati 2006: 654-655).
Given the scholarly practice of viewing decentralization and federalism as theoretical counterparts, it is important to stress their mutual relationship. The terms are closely related to each other and are often seen as synonyms in political science. This is because the purpose of non-centralization of power is decentralization of power, which is regarded as principal characteristics of federalism (Lijphart 1999: 187). However, federalism should be taken as an ideology, a philosophical principle of shared-rule and autonomy rather than an institutional counterpart according to King (1982: 74). Thus, the federalist principle serves as the recommendation, active promotion and support for federation (Burgess 2006: 2). Federation refers to a federal state, as they are unions which combine a central government with strong constituent units – each featuring delegated power through a constitution, each exercising power and each directly elected by its citizens (Watts 1996: 8). Nonetheless, scholars are increasingly replacing the term federalism with the term decentralization due to various reasons – perhaps the most significant being the desire to consider states that do not describe themselves as federal, yet share characteristics such as regional and local governments with independent decision-making powers, as decentralized (Brancati 2006: 654)
Furthermore, the definition of decentralization previously described above – often categorized
as political decentralization – overlaps with Lijphart’s concept of consociationalism. The
fundamental difference being that the definition used in this study is more restrictive. It focuses primarily on ethnic relations, territorial and governmental power sharing, excluding associated dimensions such as economic, legislative, religious and military power sharing commonly found in similar research (Cederman et al. 2015: 355).
2.2 Literature review
Studies done on decentralization have generated two contrasting arguments. According to some scholars, decentralization is likely to reduce ethnic conflict and secessionism by successfully targeting demands for autonomy and secession through an increase of local self-governance.
This is achieved by offering a devolution of power and authority to regional and local level governments. As such, decentralization promotes an accommodation of ethnic diversity while also maintaining the territorial integrity of the state. Advocates of this view assert that decentralization is the most appropriate form of governance for ethnically divided states, as it provides autonomy for ethnic groups in regions predominantly established by themselves as well as cost-effective measures for the central government to maintain state sovereignty (Hechter 2000: 8-9; Erk and Andersson 2009: 193-194). Beyond division of power, decentralization inhibits violence “by bringing the government closer to the people, increasing opportunities to participate in government, and ultimately giving groups control over their political, social, and economic affairs” (Brancati 2006: 655). By enabling communities to exert more influence on policy formulation and implementation, decentralization promotes communal political ownership. As ethnic groups gain greater political autonomy, incentives to seek an independent state diminish, resulting in a reduction of ethnic conflict and secessionism (Hechter 2000; Miodownik and Cartrite 2010: 733-734).
Other scholars adhere to a more pessimistic view of decentralization, regarding it as an ineffective method of conflict resolution, and even as an impediment to peace (Brubaker 1996;
Kymlicka 1998). Snyder (2000: 327) states that autonomy concessions have a “terrible track
record”, referencing in particular to the violent dissolution of former Yugoslavia in the early
1990s. Such observations have given decentralization along ethnic lines a bad reputation. The
critics argue that such arrangements risk reinforcing divisive ethnic identities, as it grants
certain ethnic groups not only a sense of legitimacy through formal recognition, but also
substantial means to pressure the state (Kymlicka 1998; Chapman and Roeder 2007; Elkins and
Sides 2007: 693). Roeder (2009: 214) argues that decentralization supports an inherent
incompatibility though the maintenance of a sovereign central state together with the
autonomous claims of ethnic groups – creating a vulnerable equilibrium which has dangerous implications for potential disintegration of a nation-state. By not addressing the issue of whether ethnic communities belong in a common state or not, decentralization keeps ethnic conflict alive. Following a similar reasoning, Cornell (2002: 276) questions whether decentralization should be used as an ethnic conflict resolution strategy, as autonomous regions are by their very nature conductive to conflict, secessionism and ultimate state dissolution. He argues that decentralization establishes the capacity for ethnic groups to challenge the central state’s authority through subnational political institutions. Through self-governance, ethnic identities are cemented, subsequently promoting ethnic mobilization within communities in order to put pressure on the central state government to maintain or increase regional levels of autonomy.
Thus, decentralization is inherently problematic as it facilitates the capacity for ethnic groups to revolt, ultimately undermining interethnic stability (Cornell 2002: 254-256; 275-276). Critics advocate for partition instead, suggesting ethnic communities should be dissolved into separate sovereignties as a more favorable solution to conflict (Cederman et al. 2015: 355).
Evidently, the relationship between decentralization and ethnic conflict is contentious. Some scholars try to reconcile these two contested viewpoints. Roeder (2009: 203, 208) argues for example that decentralization might have short-term merits in reducing ethnic conflict, but causes political instability long-term as actors become enclosed between two perils:
centralization and dissolution. In his view, experimenting with different institutional power sharing arrangements is unlikely to solve this fundamental issue. Others agree that while decentralization can be an effective solution for ethnically divided societies, it requires restrictive and unlikely conditions such as absence of ethnic dominance and robust democracies (Lake and Rothchild 2005). For decentralization to work, an appropriate distribution of power between the central state government and subnational governments is required. Therefore, demands of autonomy and secession should be addressed by establishing extensive decentralization, as limited reforms will only intensify the initial demands further (Erk and Anderson 2009: 196-198). Lastly, Miodownik and Cartrite (2010: 743) conclude that low levels of decentralization increase the risk of ethnic conflict while high levels decrease it.
2.3 Theoretical framework
Conventional theories on ethnic conflict focus primarily on grievance, where the role of the
state is mostly absent from the causal argument (Cederman et al. 2010: 87, 114). In contrast,
this study’s theoretical claim is based on the central assumption that governments’ maltreatment of minorities incites ethnic insurgencies. The state plays a central role in ethnic conflict, as it often results from competing claims of ethnic groups to state power. Ethnic minorities with no access to state power tend to demand increased independence from the central state (Gurr 2000:
195; Erk and Anderson 2009: 192). As Girardin et al. (2015) argue: “[group] exclusion from government is associated with a higher risk of civil war”. In turn, decentralization has a pacifying influence on ethnic conflict by making governments react and respond to the concerns of discontent minorities and potential secessionist groups (Lustick, Miodownik, and Eidelson 2004). Power sharing mechanisms – either territorial or governmental – grant certain territorially concentrated ethnic groups regional autonomy, or allow group representatives to engage in shared decision making through inclusion within the central executive respectively.
Empirical literature that incorporate this particular framework assert that groups that are excluded from central executive power are more likely to fight the government. This builds on the notion that ethnically divided states that fail to provide self-governance have a greater possibility of ethnic conflict (Hechter 2004; Cederman et al. 2015).
Paradoxically, self-rule has also been shown to exacerbate ethnic conflict. Current literature on ethnic conflict and decentralization does not seem to give a clear answer to this contradiction.
In some cases, decentralization does endorse the claims of its proponents by satisfying ethnic groups with desires of autonomy and secession. In other cases, decentralization does precisely what its opponents affirm, namely that it reinforces ethnic identities and creates incentives and opportunities for ethnic groups to further pursue their secessionist demands (Erk and Anderson 2009: 195-196). Decentralization seems to be a double-edged sword, as illustrated on Figure 1:
-
+ +
Ethnic identities
Decentralization Ethnic conflict
and secessionism
Figure 1
Following the reconciliation attempts on previous research, this study’s theoretical framework adopts a similar yet different standpoint. Therefore, the proposed theory of this study is that decentralization prevents and reduces the risk of ethnic conflict and secessionism directly – by alleviating tensions short-term through regional autonomy concessions, but increases the risk of ethnic conflict and secessionism indirectly – by reinforcing ethnic identities, encouraging ethnic mobilization and creating incentives and opportunities for ethnic groups to further pursue their secessionist demands long-term. However, in contrast with previous research, this study’s theoretical claim asserts that decentralization can also reduce this risk through a combination of territorial and governmental power sharing. As such, long-term risk reduction requires not only regional autonomy concessions but also extensive governmental power sharing as well.
This would explain why decentralization sometimes leads to war and sometimes does not.
This theoretical claim is illustrated below in Figure 2:
Figure 2
+
+ +++
+
- +
The mechanisms behind this theoretical claim are based on the different causal directions associated with decentralization. The short-term direct effects of decentralization are due to the territorial concessions made by the central government, which constitute as a signal of moderate intent – essentially legitimizing the ethnic group and its representative elites, showing an explicit will to cooperate, negotiate and respond to their demands. However, for long-term risk
Decentralization
Regional autonomy
Governmental power sharing
Ethnic identities Ethnic mobilization
Ethnic conflict and secessionism
reduction, this is not enough. By not including the elites of the ethnic group into executive power, it cements ethnic identities and increases the risk of ethnic mobilization as a result. As such, regional autonomy is not enough for long-term risk reduction. It requires a combination of territorial concessions in conjunction with governmental power sharing arrangements to be successful. If the elites of the ethnic group in question are not given appropriate inclusion, their commitment for secessionism increases. It fosters an “us vs them” mentality, primarily due to the lack of influence given to the ethnic group which they believe is warranted.
Paradoxically, because of the regional autonomy granted, ethnic group elites gain enough influence to mobilize and further pursue secessionist demands by pressuring the central government. This can be prevented and avoided all together however, if a combination of territorial and governmental power sharing is implemented – where inclusion into executive power for the ethnic group and its representative elites is granted. This is based on the notion that inclusion of ethnic groups through governmental and territorial power sharing has the highest probability of preventing and reducing the risk of conflict. Regional autonomy also has a pacifying effect on ethnic conflict, by offering more executive influence than complete exclusion, albeit to a lesser degree than inclusion in the central executive (Cederman et al.
2015). In order to test the implications of this theory in an empirical analysis, two testable hypotheses were formulated:
H1. “Included ethnic groups are less likely to rebel against the government than groups which are fully excluded from executive power.”
H2. “Autonomous ethnic groups without central power access are less likely to rebel against the government than groups which are fully excluded from executive power.”
3 Method
The aim of this study is to conduct a theory-driven comparative small-n qualitative empirical
study that seeks to examine the effects of decentralization in conflict-affected states divided
with territorially-concentrated ethnic groups. As decentralization and its aforementioned
theories deal primarily with the distribution of power between central and regional
governments, the unit of analysis is states. Moreover, the combined effect of territorial and
governmental power sharing has rarely been analyzed in previous research (Cederman et al.
2015: 356). By applying the theoretical framework explained earlier, this study seeks to examine this interaction empirically by building on existing decentralization literature. To test and evaluate the established hypotheses, specific independent and dependent variables are defined and used in the analysis.
3.1 Operationalization of dependent and independent variables
3.1.1 Dependent variable
The dependent variable is ethnic conflict and secessionism. It will be measured in conjunction using group-level coding on organized violence and civil war from the Uppsala Conflict Data Program (UCDP). More specifically, the intensity of violence between minority groups and central governments will be operationalized using the concept of armed conflict as defined by the UCDP. Their definition reads:
“An armed conflict is a contested incompatibility that concerns government and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related deaths in one calendar year.” (UCDP 2017)
As such, conflicts will be taken into account if they involve violence between central governments and ethnic groups with explicit claims for either state power, territorial autonomy or independence. Indicators that measure the occurrence of this particular type of conflict are gathered using the information available in said database. The UCDP features yearly datasets covering armed conflicts in the time period between 1946-2016.
3.1.2 Independent variable
The independent variable is decentralization. In order to reliably measure its theoretical power
sharing implications on empirical material, the variable is divided into two subcategories. By
adopting and extending the power proximity logic highlighted by Cederman et al. (2015), this
division allows for territorial and governmental power sharing arrangements to be at the center
of the analysis. Furthermore, it also permits a structural as well as a group level perspective on
decentralization to be studied. The first independent variable – defined as power structure –
will be used to test H1 and H2 by operationalizing the form of government in the unit of
analysis. More specifically, the division of power between central and regional governments
within the state’s structure will be examined. The second independent variable is defined as
power access. H1 and H2 will be tested by using the Geographical Research on War, Unified Platform (GROWup) which provides information about ethnic groups’ access to executive government power as well as their involvement in civil war. This particular power access coding is limited to politically relevant ethnic groups and executive power only (Girardin et al.
2015). The operationalization will be done by identifying and analyzing indicators that measure power sharing arrangements in the unit of analysis. As such, the power access variable would be divided into specific categories provided by GROWup’s dataset, each indicating valid and theoretically relevant types of power sharing configurations.
The first category is labeled “included into the executive” and includes four subcategories, depending on the relative importance of the positions controlled by group elite representatives (Cederman et al. 2010: 99-101; Girardin et al. 2015):
— Monopoly (M). Elite members hold monopoly power in the executive to the exclusion of members other ethnic groups.
— Dominant (D). Elite members of the group hold dominant power in the executive but there is some limited inclusion of “token” members of other groups.
— Senior Partner (S). Representatives participate as senior partners in a formal or informal power
— Junior Partner (J). Representatives participate as junior partners in government.
The second category is labeled “excluded from the executive” and includes three subcategories, depending if elites who claim to represent a particular ethnic group are excluded from participation in central government (Ibid):
— Self-Exclusion (s). Groups that have excluded themselves from central state power, in the sense that they control a particular territory of the state which they have declared independent from the central government
— Powerless (P). Elite representatives hold no political power at either the national or the regional level without being explicitly discriminated against.
— Discriminated (d). Group members are subjected to active, intentional and targeted discrimination, with the intent of excluding them from both regional and national power.
The GROWup project accounts for politically active ethnic groups worldwide from 1946-2013.
3.2 Control variables
For a case comparison to be reliable, control variables need to be identified, specified and applied on the empirical material. This facilitates the identification of their role and the potential causal factors of their relationship with the dependent variable. By applying the logic of elimination, conditions that are held constant in all cases cannot account for the difference in case outcome (George and Bennett 2005: 428). The chosen independent control variables for this study are degree of democracy, electoral system, ethnic and cultural fractionalization, ethnic group size and socioeconomic status. The reasoning behind the selected variables is based on the notion that they all have a potential effect and contribution to ethnic conflict and secessionism, as well as decentralization in selected cases.
3.2.1 Degree of democracy
Brancati (2006: 652) argues that decentralization is only genuine in democracies. Even though a number of non-democratic states have assigned decision-making powers to regional governments in principle, she states that in practice they systematically infringe on these arrangements. Decentralization and democracy is therefore viewed as two concepts of government that are essentially interlaced, mainly due to their fundamental principles of both limiting power and enabling participation, in theory and practice. However, Benz and Sonnicksen (2017: 4-5) argue that this alleged link is profoundly ambivalent. They conclude that decentralization cannot be assumed to be compatible with democracy, as no such connection between the concepts necessarily exists. Furthermore, empirical findings suggest that consolidated democratic states have a lower risk of intrastate conflict compared to semi- and non-democracies (Hegre 2014: 163). Ethnic conflict and secessionism should then be less intense in states where democracy is extensive and political and civil rights are well protected.
In said states, groups should have fewer grievances with the government and they should be better able to resolve such grievances through peaceful means such as protests (Fearon and Laitin 2003). Consequently, the inclusion of degree of democracy as an independent control variable is made using freedom status categorization and scores from Freedom House’s annual reports of political rights and civil liberties. Scores are given from 1-7 on each category, where one is most free and seven is least free.
3.2.4 Electoral system
Likewise, as with degree of democracy, different aspects of a state’s political system may also
affect the probability of ethnic conflict and secessionism. The type of electoral system could
for example help explain why not always an increase of social and economic heterogeneity
incites ethnic conflict (Brancati 2006: 671-672). In this case, proportional representation systems are preferred over plurality or majority systems as they are believed to reduce ethnic conflict and secessionism more. This is because proportional representation systems are generally more inclusive to small parties that represent different ethnic groups (Lijphart 1977).
3.2.2 Ethnic and cultural fractionalization
Ethnic and cultural diversity and their implications on conflict is disputed among scholars.
Some studies suggest that ethnic and cultural diversity in societies can have negative effects on social stability – by impeding cooperation, inciting victimization and hostility towards other groups (Horowitz 1985). Ethnic and cultural divisions are believed to increase the risk of conflict between segments of a society. Thus, inter-group violence is more likely to occur in ethnically divided and multiethnic states compared to mono-ethnic counterparts. The risk is lower on states where the vast majority of the population clearly identifies with the nation-state, as the potential for consensus building is stronger (Karatnycky 2002: 107; Lijphart 1977). Also, Firsh and Brooks (2004: 154-166) find that secessionist movements are less likely, and political decisions easier to accept when they are stipulated by the same ethnic group. This could be explained by the argument that members of an ethnic group tend to favor own group members over non-members due to their closer relationship to them compared to the rest of the population (Vanhanen 1999: 57). However, other studies suggest that there is no significant relationship between heterogeneity and conflict (Fearon and Laitin 2003).
Nevertheless, ethnic and cultural fractionalization is controlled by using common measures of
aggregate ethnic and cultural diversity scores. Ethnic fractionalization is defined as the
probability of two randomly selected individuals from a state will represent different ethnic
groups. Data is indicated by scores between 0 and 1 – where states located near zero are highly
fragmented, states located near the middle are roughly bipolar with two large ethnic groups
dividing most of the population and states located near one are relatively homogeneous,
consisting of a large ethnic majority. Cultural fractionalization is defined as the structural
relationship between languages to estimate cultural proximity between groups in a state. Data
is indicated by scores between 0 and 1 – where states with strong linguistic variance, i.e. several
groups that speak structurally unrelated languages, are located closer to zero. Whereas states
with one language group or a set of ethnic groups that all speak highly similar languages are
located closer to one (Fearon 2003: 196, 206-207, 212). Data is largely from the 1990s, which
could potentially be classified as old and therefore be seen as problematic. Due to the lack of
alternative data and the fact that ethnic and cultural compositions usually stay relatively constant throughout long periods of time, the data is included. Nonetheless, the variable should be interpreted with caution as a result.
3.2.3 Ethnic group size
Scholars on ethnic conflict and secessionism argue that mobilization capacity is a defining factor on the outbreak of ethnic violence. This is because larger ethnic groups are more able to challenge the central government through the displacement of higher numbers of recruitment fighters, while also having a larger potential resource pool to sustain an organizational infrastructure. The larger the ethnic group size, the more legitimacy their political claims receive. This is based on the principles of representativity, as the exclusion of smaller portions of the state populations result in considerably less consequences than the exclusion of larger portions (Cederman et al. 2010: 96). Thus, ethnic group’s demographic size in relationship to their nation-state will be taken into consideration as another independent control variable.
3.2.5 Socioeconomic status
Socioeconomic factors are associated with influencing ethnic conflict and secessionism. For instance, economic development increases the capacity of states to suppress insurgencies, thus reducing the risk of ethnic conflict (Fearon and Laitin 2003). It may also reduce the risk by improving education and social welfare, reducing the population’s vulnerability to extremist ideologies (Lipset 1963). However, if uneven, economic development may increase ethnic conflict and secessionism (Hechter and Levi 1985). As such, socioeconomic status is included as an independent control variable. It is operationalized by using the Human Development Index (HDI) developed by the United Nations as a metric to assess the social and economic development levels of states using indicators such as life expectancy, education and per capita income. It is important to stress its limitations, as the HDI is a mere simplification and limited evaluation of human development. As such, it has received criticism, particularly on its lack of global perspective, its underlying statistics and formula changes (Wolff et al. 2011).
Nevertheless, the index is used as a control variable in this study because of its convenience as
well as its widespread use in academia, media and policy formulation for measuring and
comparing progress in human development between states over time. However, its limited
scope should be taken into account when assessing the variable.
3.3 Validity and reliability
As this study seeks to analyze the influence of decentralization on ethnic conflict and secessionism in ethnically divided states, indicators that not only reflect or capture these concepts of interest but also reliably measure them are of utmost importance. The theoretical definitions of said concepts are well established and clear, being equally applicable across cases, facilitating the matching of adequate indicators. As such, variables are operationalized by using indicators that measure what they are meant to measure. Comparative case studies allow for conceptual refinements that offer a higher level of validity over a small number of cases. While statistical studies usually run the risk of “conceptual stretching” by compiling dissimilar cases to obtain a larger sample, qualitative studies can further engage in conceptual development (George and Bennett 2005: 83-85). In this case, overall concepts including
“conflict”, “groups”, “decentralization” and “government” are theoretically developed using relevant adjectives such as “ethnic”, “territorial”, “governmental” and “federal” to denote important subtypes or subclasses of more general concepts. They are then operationalized by using indicators that explicitly measure these specific concepts. As such, ethnic conflict is measured using UCDP’s definition of armed conflict in combination with data limited to ethnic groups with explicit claims for either state power, territorial autonomy or independence included in UCDP’s database. Decentralization is measured by using structural indicators of the division of power in a state’s government as well as ethnic group’s access to executive power using theoretically relevant categories of power sharing configurations provided by GROWup’s dataset – representing territorial and governmental decentralization respectively.
By using data that corresponds to the relevant theoretical developments studied, validity is increased. Of course, using standard measures such as UCDP’s specific definition of armed conflict affects validity as well. By adopting a definition of armed conflict which includes a specific threshold number of fatalities (25 battle-related deaths per year), data which could potentially be classified as conflict by other types of measurement standards is excluded. While this study uses a simplified definition due to its limited nature, other studies adopt a wider definition that encompasses all forms of violence among ethnic groups – including multiple categories that differentiate between low and high levels of conflict (see Brancati 2006).
Data used in this study is mostly collected directly from major data collection projects such as
the UCDP, GROWup, Freedom House and United Nations. This contributes to a high level of
reliability, as most of this data is well documented. The main advantage of using this type of
data is that it returns the same type of values for a given case to multiple researchers that
evaluate cases according to the established definitions and theoretical concepts. As coding rules are often made public – including references to sources used to create variable values – other scholars can validate and replicate the data, essentially minimizing bias. By using standard measures, findings are not only made comparable, but also sufficiently valid and reliable.
3.4 Case selection
According to George and Bennett (2005: 240-241), the primary objective for a successful case selection strategy is relevance to the research intent of the study in question. To explain phenomena of interest, cases should be selected according to the type of control and variation that is required by the actual research problem. For this to be a viable strategy, a well-defined distribution of population and scope conditions are needed to appropriately select cases. In order to be able to generalize, cases should be representative of the population that is studied.
Cases should also match the unit of analysis in the theory-driven study. Important criticism has been made on case selection in qualitative studies with few cases due to the potential risk for selection bias. Methodologists argue that this approach leave scholars vulnerable to systematic error. Usually, the culprit is the restriction of cases where variation on the outcome is absent or an analysis on a narrow range of variation is present. Qualitative case studies that select only positive or negative occurrences of the phenomenon in question are therefore inherently problematic (Collier and Mahoney 1996: 57; Levy 2008: 8). This type of selection bias is prone to a form of confirmation bias, where the researcher selects only cases where the independent and dependent variables vary according to the suggested hypothesis while ignoring cases that appear to contradict the study’s theory. Therefore, the danger of bias is particularly strong when researchers know the values of the independent and dependent variables of cases beforehand.
However, avoiding knowledge of these values in practice is very difficult, hence the increased risk of bias. A suggested preventive measure to mitigate the problematic nature of selection bias in qualitative studies is to actively search for similar cases where the independent variable of interest is present but the predicted effect is absent, and where the independent variable is absent but the dependent variable is present (Bennett 2004: 40). Thus, “negative cases” are now routinely incorporated into case research designs by qualitative researchers (Levy 2008: 8).
As this is a comparative, small-n empirical study which purpose is to examine the influence of
decentralization on ethnic conflict and secessionism, the selected cases should be relevant and
representative to the applied independent and dependent variables. As the unit of analysis is
states, chosen cases should be based on their applicability with the defined theoretical
framework of the study – being able to provide variation and control as well as adequate representation. For states to be considered relevant, they should also feature a division of territorially-concentrated ethnic groups and currently experience or have previously experienced ethnic conflict and secessionism due to said characteristics. To ensure proper variation, a relatively wide variation on the outcome is needed, including negative or partially negative evidence, as well as variation on the independent variable.
With these criteria in mind, the selected cases for this study are the states of Spain and Bosnia and Herzegovina. Spain was selected because it is formally a unitary state consisting of territorially-concentrated ethnic minorities organized as autonomous communities. Often referred to as “a nation of nations”, Spain adopted a highly decentralized political system following the transition to democracy and subsequent Spanish Constitution of 1978 (Colomer 1998: 40-41). Yet, despite these changes, ethnic conflict and secessionism prevailed (UCDP 2016a). As such, Spain deviates from traditional examples of decentralized democratic states and can be classified as a particularly relevant “negative case” for analysis. Bosnia and Herzegovina was selected because it became formally a federation following a violent conflict outbreak shortly after its independence from former Yugoslavia in 1992. Through the Dayton Peace Agreement in 1995, the ethnically diverse state implemented extensive decentralization as a conflict resolution strategy (Keil 2012: 209-210). Considering the extraordinary case of Spain, finding a comparative case with similar characteristics was particularly difficult.
Nevertheless, Bosnia and Herzegovina corresponds sufficiently to be justifiably relevant. A more in-depth analysis of selected cases, their background and the empirical implications on theory will be presented in the upcoming sections of this study. Case selection strategy – including variation – is illustrated in Table 1:
Table 1
Decentralization Ethnic conflict and secessionism
Yes No
Yes Spain (after) Bosnia and Herzegovina
No Spain (before) -*
* Cell is not needed to test this study’s hypotheses
3.5 Comparative method
The comparative method is a case study method that involves the analysis of a small number of cases in an attempt to approximate the conditions of scientific experimentation. Controlled comparison – the most common variant – involves the study of two or more well-specified phenomena that correspond with each other in every way but one (Lijphart 1971). This is done to provide the function of drawing causal reasoning based on experimental logic. This particular logic is derived from John Stuart Mill’s work, where he discusses what is now known as the
“method of agreement” and “method of difference”. This study will use the method of difference, where cases with different outcomes are compared. By utilizing the logic of elimination, researchers exclude independent variables that are present in all cases as the cause of variance in the outcome (dependent variable). The logic implies that conditions present in all cases cannot account for the differences in case outcomes. At the same time, not present conditions can only be regarded as possible causes of outcome difference. Furthermore, the type of control required is generally difficult to achieve, as finding cases that resemble each other in every way but one is extremely hard. Mill himself emphasized the difficulty in applying the logic of elimination on social science research due to the complex nature of social phenomena (George and Bennett 2005: 419, 425-428). Lijphart (1971: 688) argues however that such objections are founded on the notion of a too exacting scientific standard.
Nevertheless, he stresses the importance in approximating this standard as closely as possible.
Alternative variants of comparative methods have been proposed in an attempt to minimize these limitations. One such alternative is called the “before-after” research design. Instead of finding different cases that are comparable in all ways but one, researchers can achieve control by “dividing a single longitudinal case into two sub-cases” (George and Bennett 2005: 457- 459). Functioning as a quasi-experiment, a phenomenon occurring at some point in time within the sequential development of a longitudinal case allows researchers to identify a before and after configuration. The challenge here lies on controlling for potential causal variables that changed at the same time of occurrence. Furthermore, a combination of longitudinal and cross- sectional designs can serve to gain additional inferential leverage according to Levy (2008: 10).
As such, the methodology used in this study will consist of a combination of these two
comparative methods. First, a within-case comparison of Spain – where the case will be
compared in a before and after manner, measuring the differences over time on the independent
and dependent variables. More specifically, the constitutional reform adopted in the Spanish state in 1978 established decentralization which, for the purpose of this study, makes for an exceptionally relevant case. This historical development provides an opportunity to evaluate the impact of decentralization on ethnic conflict and secessionism as a form of natural experiment (see Putnam 1993; George and Bennett 2005: 459). By holding relevant socioeconomic and cultural variables for the most part constant, the structure of the political institutions that were abruptly altered by the reform can be analyzed. The time period in this case is 1976-1978 as before and 1978-present day as after, marking the phenomenon of interest.
Spain will also be treated as a particular case that is anomalous with respect to existing hypotheses. As such, it will be studied as a single case that is known to deviate and violate theoretical predictions. Spain can be considered what Lijphart (1971: 692) calls a “deviant case”
because despite implementing decentralization, ethnic conflict and secessionism was not reduced. Furthermore, regional autonomy concessions did not provide short-term reduction (e.g. Roeder 2009), nor an effective solution despite the absence of ethnic dominance and presence of a robust democratic system (e.g. Lake and Rothchild 2005), nor a decrease in ethnic conflict and secessionism due to the extensiveness of the reform or the high levels of decentralization established (e.g. Erk and Anderson 2009; Miodownik and Cartrite 2007). Spain is a heavily decentralized state with one of the strongest democracies in the world. By applying this study’s theoretical framework, Spain will be analyzed with the ambition to potentially explain why it deviates from existing theories, why it violates theoretical predictions and to refine, replace or uncover relevant additional variables that were not considered previously.
Thus, such research design can have great theoretical value by not only refining existing theories but also generate additional hypotheses (Lijphart 1971: 692-693; Levy 2008: 5, 13).
Second, an across-case comparison between Spain and Bosnia and Herzegovina – where cases
will be compared using the method of difference in order to test this study’s hypotheses. More
specifically, both cases adopted extensive decentralization reforms, differ on form of
government as well as the presence of ethnic conflict and secessionism and are similar on
selected independent variables relevant to this study. As there is some variation on one
independent variable, while others are held constant, it is likely that this particular independent
variable explains the variation of the outcome. In other words, a qualitative analysis will be
made to seek the potential causal relation between decentralization and ethnic conflict and
secessionism by comparing cases that are classified as “negative” and “positive” respectively.
The time period analyzed in the case of Bosnia and Herzegovina is between 1992, when it first became a sovereign state after its independence from former Yugoslavia, to present-day.
Although this study will not provide a conclusive answer to the question whether decentralization reduces or exacerbates ethnic conflict and secessionism, the analysis will hopefully provide useful insights into different causal directions of decentralization and its influence on ethnic conflict dynamics.
3.6 Sources of empirical material
Material consists primarily of systematic and critically reviewed secondary sources in the form of relevant books and peer-reviewed scholarly articles. Relevant in this case means that intentional decisions were made to include research primarily done in the fields concerning the study in question. In order to produce reliable measurements to adequately test this study’s hypotheses, identifying observable and valid indicators on selected concepts was of the essence.
This includes a relevant body of academic literature done on political science, decentralization, federalism as well as territorial and governmental power-sharing and conflict resolution. Author bias is a notable issue with secondary sources, as an analysis between the original data and reader has been made (George and Bennett 2005: 261).
Unfortunately, due to the nature of this study, primary sources are difficult to obtain. This is mainly because ethnic conflicts are highly intractable at their time of occurrence. Because of this, a selection process involving careful consideration of authors chosen was conducted. As a result, the work of scholars well renowned within their respective disciplines is used.
Furthermore, additional information is used from non-academic sources including reports,
databases and statistics by international organizations and institutions, as well as public data
and official documents by governments. Lastly, news media articles are also used as
complimentary material.
4 Results and analysis
4.1 Spain
Table 2
Spain (before) Spain (after)
Ethnic conflict and
secessionism Yes Yes
Decentralization No Yes
Power structure Unitary state Unitary state
Power access Spanish: Senior Partner (S)
Basques: Powerless (P) Catalan: Junior Partner (J) Galician: Junior Partner (J)
Spanish: Senior Partner (S) Basques: Junior Partner (J) Catalan: Junior Partner (J) Galician: Junior Partner (J)
Degree of democracy Free Free
Electoral system Proportional representation Proportional representation
Ethnic and cultural
fractionalization Medium & High Medium & High
Ethnic group size Basques: 5%
Catalan: 15-16%
Galician: 6-7%
Basques: 5%
Catalan: 15-16%
Galician: 6-7%
Socioeconomic status (HDI) n/a Very High
4.1.1 Background
Spain is an empirical representative of the population in this study and matches the unit of analysis by being a heavily decentralized unitary state. It features an ethnically diverse population represented by territorially-concentrated national minorities which are organized as autonomous communities. Often referred to as the “State of Autonomies”, this unique institutional framework of territorial administration and division of power followed the transition to democracy that occurred between 1976 and 1978. After decades of dictatorship under the rule of Francisco Franco, the first democratic elections were held in 1977. The Spanish Constitution of 1978 was approved in a referendum, establishing political pluralism in Spain. Initially, Catalonia, the Basque Country and Galicia were allowed to establish autonomous governments as these communities had voted for specific institutions in the 1930s.
Soon after, 17 autonomous communities were formed (Colomer 1998: 40-41). With this, Spain
went from being one of the most centralized states in the OECD to being one of the most
decentralized. The current institutional arrangement has a structure unlike any other, being
similar to a federation but not equal (Hernández Lafuente 1999: 50-67). Nevertheless, its fundamental characteristics can be compared to states which are undeniably federal (Marín 2013: 381). Decentralization has shown to be the most innovative institutional attribute in democratic Spain. Yet, it is also a European state with the most visible cleavages on territorial issues. Spain suffers from a dilemma of nationalities, where cultural and political interregional disparities reinforce internal divisions and ethnic and linguistic variances damage its unity.
Identified as “a nation of nations”, its contemporary entity has been subject to ethnic conflict and secessionism due to the historical rivalries among Spanish nationalities and regions (Colomer 1998: 40-41; Moreno 2002: 399).
4.1.2 Dependent variable
Spain experienced ethnic conflict and secessionism soon after the transition to democracy began. The then head of government Adolfo Suárez – appointed by the monarch – began his role as Prime Minister of Spain in 1976 by granting amnesty for several political prisoners, among them Basques. During this period, the Basque Country underwent political turbulence, eventually culminating in violence. Despite the proposed decentralization reform involving autonomy concessions, Basque nationalists withdrew from the congressional meeting. The result was a failed referendum in favor of the Constitution in the Basque Country, being the only region in Spain where the ‘no’ vote triumphed. This rejection contributed to an escalation of conflict, particularly from the Basque separatist group Euskadi ta Askatasuna (ETA), which has fought for secession from Spain, and an independent Basque state since its formation in 1959. The referendum also contributed to the formation of the nationalist Basque political party Herri Batasuna in 1978, which advocated for ‘no’. They were seen as ETA’s political wing, having close links to the armed group while sharing independence ambitions and providing important moral support. An estimated 20% of the Basque population supported Herri Batasuna (UCDP 2016b). The Basque National Party (PNV) eventually approved the Basque statute of autonomy and has governed continuously since 1980 (Colomer 1998: 43-44). However, ETA’s violent campaign against the Spanish government continued throughout most of the 1980s and early 1990s. The violence reached levels of armed conflict during 1978-82, 1985-87 and 1991.
All Spanish governments have condemned the group for attacking Spain’s territorial integrity, rejecting the group as terrorists as a result. Herri Batasuna was eventually banned from Spanish politics but Basque nationalists including leaders of the PNV proclaimed shortly after that they would attempt to “give voice to excluded Batasuna groups without falling into ‘disobedience’”.
The PNV president of the Basque regional government at the time, Juan José Iabrretxe, declared
for the first time openly support for Basque independence from Spain (El País 2003; Birnir 2007: 144). A total number of 412 registered deaths can be attributed to the conflict during this period. The armed conflict officially ended in 2011 when ETA declared the cessation of their armed struggle. Nevertheless, the fundamental dilemma between the warring parties persists, as the group’s desire to achieve secession from Spain and an independent Basque state remains (UCDP 2016a).
4.1.3 Independent variable
Prior to the constitutional reform, the division of power between the central government and regional counterparts was mostly inexistent. The Franco regime had implemented a “sacred unity of the homeland” where any form of federalism, separatism or ethnic identity in the form of language or cultural differentiation was suppressed. As such, the form of government was a highly centralized unitary state. This power structure eventually changed, albeit not formally.
The new constitution in 1978 was drafted in conjunction with all ethnic groups, which led to the compromise of "(1) the idea of an indivisible and solely Spanish nation-state and (2) a concept of Spain as an ensemble of diverse peoples, historic nations, and regions" (Moreno 1997). It was subsequently illustrated as Art. 2 in the constitution, marking the starting point in the territorial organization of Spain (Hernández Lafuente 1999: 50-67):
“The Constitution is based on the indissoluble unity of the Spanish Nation, the common and indivisible homeland of all Spaniards; it recognizes and guarantees the right to self-government of the nationalities and regions of which it is composed and the solidarity among them all.”
