Abstract
While the study of the causes of civil war is a well-established subdiscipline in international relations, the effects of civil war on society remain less understood. Yet, such effects could have crucial implications for long-term stability and democracy in a country after the reaching of a peace agreement. This article contributes to the understanding of the effects of warfare on interethnic relations, notably attitudes of ethno-nationalism. Two hypotheses are tested: first, that the prevalence of ethno-nationalism is higher after than before the war, and second, that individuals who have been directly affected by the war are more nationalist than others. The variation in ethno-nationalism is examined over time, between countries, and between ethnic groups. Three countries that did not experience conflict on their own territory serve as a control group. The effect of individual war exposure is also tested in the analysis. Sources include survey data from the former Yugoslavia in 1989, shortly before the outbreak of war in Croatia and Bosnia and Herzegovina, and in 2003, some years after the violence in the region ended. Contrary to common beliefs, the study shows that ethno-nationalism does not necessarily increase with ethnic civil war. The individual war experiences are less important than expected.
Introduction
The end of war does not necessarily mean the beginning of a stable, peaceful society. This article seeks to expand our understanding of how civil war affects the general population and people’s disposition to live together despite diversity. To increase our understanding of post-conflict society, we also need to incorporate the individuals who live in the society and see how they are affected by armed conflict. While some of the consequences of civil war are relatively well described (Collier et al., 2003), the effects of war on individual attitudes and behavior have received comparably less attention, despite several apparently good reasons to improve our knowledge on how civil war affects the civil population. At the social level, we know that violent conflict leads to psychological traumas, higher infant mortality rates, and other demographic changes (Brunborg & Tabeau, 2005; Brunborg & Urdal, 2005). The effect of violent conflict on individuals’ disposition to live together with people who belong to a different ethnic or religious group is less well understood. The obvious, almost commonsensical assumption would be that warfare cements individual identities, increases negative sentiments towards the ‘others’ or outgroups, and strengthens nationalism, but this assumption is rarely tested. Wood (2008) argues that civil conflict has the potential to polarize social identities and increase local cleavages, but we do not know how strong or durable these effects are, or whether the whole population is equally affected.
In this article, I examine ethno-nationalism in the Balkans before and after the breakup of Yugoslavia. Previous studies have treated ethnic animosity mainly as an explanatory factor leading to war. While I do not exclude the possibility that tense interethnic relations can increase the risk of civil war, I reverse the question and focus instead on how ethnic civil war can affect ethno-nationalism. To answer this question, I analyze the change in aggregate ethnic attitudes in six countries in a period where four of them experienced various degrees of violent conflict. I also test if individuals who have experienced warfare are more nationalist than individuals who were not directly exposed to violence. The rationale is that an increased knowledge of the effect of war on interethnic attitudes may also increase our understanding of the problems postwar societies are facing.
Through a short period, the former federation of Yugoslavia was the battleground of several devastating wars, which resulted in the formation of seven new states. 1 I use survey data from 1989, that is, shortly before the outbreak of war in Croatia and Bosnia and Herzegovina (Bosnia), and from 2003, about eight years after the fighting had ended in Croatia and Bosnia, and only a couple of years after a ceasefire was reached in Kosovo and Macedonia. Interestingly, no clear effect of war exposure on nationalism can be found. In the cases of Croatia and especially in Bosnia and Herzegovina, prewar ethno-nationalism was very low, while Macedonia and especially Kosovo also saw high levels of nationalism before the war. In Bosnia and Croatia, ethno-nationalism increased significantly from 1989 to 2003, while Macedonia and Kosovo saw no change. Notably, it is only in Croatia that individuals who report having been directly exposed to war are significantly more nationalist. The findings show that the relationship between ethnic conflict and nationalism is more complicated than commonly assumed. The residual of ethno-nationalism in society in the aftermath of ethnic civil war represents a potential that nationalist politicians can tap into, which could complicate the further process of stabilization and peacebuilding.
The remainder of the article is organized in five sections. In the next section, I define central concepts and discuss how conflict can cement ethnic identities as well as increase nationalism. Section 3 introduces the ethnic composition of Yugoslavia as of 1989 and offers a short review of the wars in the 1990s. In section 4, I explain the research design and the data sources. Section 5 presents the analyses and results, while section 6 offers some concluding remarks.
Ethnic civil war and nationalism
Before I proceed, some definitions are required. First, I follow Chandra (2006), who defines ethnicity as an identity based on traits associated with or believed to be associated with descent. Moreover, as discussed by Fearon & Laitin (2000) and Brubaker (2009), as well as illustrated by my data (elaborated in the analysis section), ethnicity is not a constant, but a complex identity whose salience and defining features depend on the context. Neither are identities always mutually exclusive, as they may be nested within one another (Pickering, 2007).
In the literature, nationalism is a fuzzy concept used in a number of different ways. Much of the literature fails to distinguish between nationalism as loyalty to a (nation-)state and loyalty to a nation or a group (for a discussion, see Connor, 1972). I therefore prefer the term ethno-nationalism to refer to the latter, which I take to mean a desire to keep the (ethnic) nation homogeneous and separated from other groups, where loyalty to the group is stronger than loyalty to the state. This definition is similar to that of Dekker, Malová & Hoogendoorn (2003: 347). They define nationalism as feeling a sense of belonging to a particular ‘nation’ with a common origin, wanting to keep that origin, wanting to keep that ‘nation’ as pure as possible, and desiring to establish and/or maintain a separate and independent state for that particular nation.
Ethno-nationalism, as I have defined it, falls under the umbrella of outgroup evaluations. 3 Such evaluations are often viewed as a basic predisposition in the individual, and are generally quite stable over time (Kinder & Kam, 2009). Explanations of outgroup evaluations usually operate at the group level or the individual level. For the purpose of this article, intergroup explanations are more relevant, particularly realistic conflict theory (Sherif, 1966) and social identity theory (Tajfel & Turner, 1979), which will serve as the point of departure for this section. I propose two main mechanisms through which the experience of armed ethnic conflict should increase ethno-nationalism and negative outgroup evaluations.
According to social identity theory, strong ingroup identification leads to negative evaluations of other groups in order to maintain a positive self-esteem. This implies that increased salience of a certain group identity enhances a negative outgroup bias. Ethnic identities are not always salient or well defined, and often, people may simply not define themselves in terms of ethnic or national groups (see Haug, 2001). During the course of mobilization, targeted people may be forced to choose a side, and so individual identities that may not have been clearcut at the outbreak of the conflict become increasingly cemented or ‘hardened’ during the course of civil war (Fearon & Laitin, 2000; Gurr, 2000; Wood, 2008). In some cases, the very process of civil conflict contributes to a growing polarization of attitudes, which in turn can increase the risk of continued instability and conflict, as ethnicity becomes the main organizing principle of politics. Even if much of the violence at the local level has its origin in local disputes, which may have little or nothing to do with the main cleavage at the national level, the violence may be interpreted in terms of the national cleavage (Kalyvas, 2003, 2008).
Wood (2008) describes three distinct but mutually enforcing mechanisms through which mobilization and violence can increase local cleavages and polarize local identities in a civil war, as it becomes increasingly difficult to remain neutral in the conflict. First, local actors may choose a side in order to gain local advantage; second, civilians may be forced to join one side in order to gain protection, especially if violence is indiscriminate; and finally, the violence itself may cause moral outrage, which makes previously neutral civilians take a side. Whether because of opportunism, need for protection, or outrage, the result of the side-taking could be an increased salience and polarization of local identities, followed by a deteriorated interethnic relationship.
Realistic conflict theory proposes that competition between groups leads to negative outgroup bias (Quillian, 1995). Thus, threat to the ingroup, whether perceived or real, should increase the negative evaluation of the outgroup. The terrifying experience of war should most likely contribute to such perceptions. In a cross-national study, Hutchison & Gibler (2007) also show that territorial threat increases conformist attitudes and reduces political tolerance. Political propaganda used by nationalist leaders could further add to this effect (see Somer, 2001).
All taken together, there is reason to expect that ethno-nationalism increases during a civil war, even if interethnic or intergroup relations were relatively good before the war broke out. All other things equal, we should find that ethno-nationalism increases with the experience of armed ethnic conflict.
The study of the micro effects of war is an emerging field, and the number of empirical studies is limited. Whitt & Wilson (2007) use experimental data to examine the effect of ethnic violence on norms of fairness. They find that even though preferential treatment of own ethnic group exists, the level of favoritism is less than expected. Recently, a handful of studies also link the experience of political violence and post-traumatic stress disorder in Israel to prejudices towards minorities (Besser & Neria, 2009; Canetti-Nisim et al., 2009). Few studies exist on the effect of warfare on the general population. Using survey data from Macedonia, Dyrstad et al. (2011) find limited evidence that geographical proximity to fighting reduces support for some aspects of the Macedonian peace agreement.
In a study from Croatia, Strabac & Ringdal (2008) found little effect of the individual experience of war, but a stronger regional effect, that is, the influence of the local intensity of fighting on ethnic prejudice. O’Loughlin (2010) found little effect of war experiences on willingness to engage in interethnic friendships in postwar Bosnia and Herzegovina, but in a study of reconciliation in North Caucasus, Bakke, O’Loughlin & Ward (2009) found that those relatively close to fighting were more willing to forgive people of other nationalities for the violence.
Based on the arguments and findings above, I propose the following hypotheses:
H1: Ethno-nationalism is higher shortly after than before an ethnic civil war.
H2: Individuals who were exposed to violence display a stronger sense of ethno-nationalism than others.
The former Yugoslavia: A historical backdrop
The wars that accompanied the dissolution of Yugoslavia were the deadliest in Europe since World War II. A full account of causes and main actors is outside the scope of this article; 4 however, a short historical backdrop is necessary in order to assess the consequences of the conflicts. In 1990, Yugoslavia was a federation consisting of the republics of Slovenia, Croatia, Bosnia and Herzegovina, Macedonia, Montenegro, and Serbia. From the 1974 constitution, Vojvodina and Kosovo, which were part of Serbia, enjoyed an extensive political autonomy, including representation in the Yugoslav parliament. Yugoslavia had six official nationalities: Slovenes, Croats, Serbs, ethnic Muslims, Montenegrins, and Macedonians, which were all the largest group in one of the republics (Urdal, 2001). 5 Minority groups included Albanians, Bulgarians, Hungarians, Italians, Roma, Turks, and Vlah, among others (see online Appendix A for a map of the current ethnic composition in the countries). Language, religion, and to some extent history were the main defining criteria of the nationalities (Urdal, 2001). The official language in Yugoslavia was Serbo-Croatian, spoken by Croats (mainly Catholics), Serbs (predominantly Orthodox), ethnic Muslims, and Montenegrins. 6 The main difference between Croats, Serbs, and ethnic Muslims was therefore religious (Urdal, 2001). Actually, in the early period after World War II, the Muslims were believed to be either Serbs or Croats of a different faith (Ramet, 2006: 285). Although similar to Serbo-Croatian, both Macedonian and Slovenian are separate languages. Montenegrins and Serbs share both language and religion, and differences between the two groups are therefore more a question of history (Urdal, 2001). 7
Albanians, also predominantly Muslims and with a distinct, non-Slavic language, formed the majority in the province of Kosovo, but were not considered a constitutive nationality. Montenegro and Macedonia also had large Albanian minorities. In the second Autonomous Province, Vojvodina, Serbs were the largest groups, but there were also large groups of Hungarians, Croats, Slovenes, and others.
Before war broke out, interethnic relations were described as generally good (Massey, Hodson & Sekulić, 2006), although the extent of contact between groups varied across regions. The principal exception to this was Kosovo, where both Serbs and Albanians reported harassment and abuse. Kosovo became part of the Kingdom of Serbia as late as 1913, contrary to the will of the Albanian majority, and Albanians were underrepresented in political positions (Ramet, 2008). The region was the poorest in Yugoslavia, with high illiteracy rates and a lack of basic infrastructure (Ramet, 2006: 296). In 1968, Kosovo achieved a more autonomous position within Yugoslavia, which was reinforced by the 1974 constitution (Trix, 2010). This autonomy was lost in 1989. In short, the situation in Kosovo was very tense before the dissolution of Yugoslavia, and the tension in Kosovo contributed to the collapse of Yugoslavia and the Wars of Yugoslav succession as it increased Serb nationalism and helped consolidate the power of the Serb president, Slobodan Milošević.
Formally, the War of Yugoslav succession refers to the 1991–95 wars in Slovenia, Croatia, and Bosnia and Herzegovina, which ended with the Dayton Peace Accords in November 1995 (Hoare, 2010). After the skirmishes that followed Slovenia’s declaration of independence, the first serious conflict erupted in August 1991 when Croatia declared independence. The fighting broke out between the Croatian army and the Serb minority, who wished to recess from Croatia and to remain part of Yugoslavia. The Serb rebels, who received support from the Serb-dominated Yugoslav national army (JNA), were defeated in 1995 after the massive ‘Operation Storm’, and most of the Serb minority fled the country. A general agreement on the number of killed on each side does not exist, but it is clear that several thousands were killed on each side and hundreds of thousands displaced, both Serbs and Croats. By 1995, most of the Serb population in Croatia had left the country (Paris, 2004: 109). The bloodiest conflict took place in Bosnia and Herzegovina from 1992–95. Serb forces and the JNA fought the Croat and Bosnian forces. During the course of war, the fragile alliance between Croats and Bosniaks failed, which led to fighting between Croat and Bosnian forces (Hoare, 2010). About one million people fled from Bosnia and Herzegovina during the war. After the war, the International Criminal Tribunal for the former Yugoslavia (ICTY) was established to investigate the war crimes, notably the Srebrenica massacre, where at least 7,500 male Bosniaks were executed in 1995. No agreement exists on the total number of casualties. Early estimates reported about 200,000 victims (Brunborg, Lyngstad & Urdal, 2003), while a more recent report estimated 100,000 victims (Research and Documentation Center Sarajevo, 2007).
The Kosovo war refers to the war in 1998–99, which was the culmination of more than a decade of riots and clashes between Serb security forces and Albanian secessionists. In early 1998, the Kosovo Liberation Army started to attack Serb security forces, which led to strong reprisals. Thousands of Albanians were arrested, and the Serbs accused the Albanians of ‘pursuing a policy of “genocide” against non-Albanians’ (Ramet, 1998). In 1999, NATO initiated a bombing campaign against Serbia, and by June the same year, Kosovo was technically a UN protectorate (Trix, 2010). Kosovo declared independence from Serbia in 2008.
Macedonia declared independence from Yugoslavia peacefully in late 1991 following a referendum and was not directly involved in the War of Yugoslav succession. The country was severely affected by the Kosovo war in 1999, when about 300,000 Kosovo Albanians fled to Macedonia through the northern border of Macedonia, where there was already a large Albanian minority (Trix, 2010: 364). Tensions between ethnic Macedonians and Albanians in Macedonia already existed before the Kosovo war (Brunnbauer, 2004), and in 2001, clashes broke out between Albanian insurgents and Macedonian security forces. The conflict ended with the Ohrid peace agreement, signed in August 2001. Compared with the other armed conflicts, however, the violence in Macedonia was limited in time and space, as well as in number of casualties. The insurgency lasted less than a year, and the fighting was concentrated in a handful of municipalities along the border of Kosovo. While the number of losses is disputed, no claim exceeds 100 casualties.
Only Serbia and Montenegro did not experience conflict on their own territory, although both were involved the wars in Bosnia and Croatia, and Serbia in Kosovo. As the last republic, Montenegro declared independence from the union with Serbia in June 2006 after a referendum. Based on this information and the theoretical argument outlined in the previous section, I expect that ethno-nationalism in Kosovo was already high in 1989 and stayed high in 2003. Bosnia and Herzegovina and Croatia should both see a clear increase in ethno-nationalism. Due to the more limited fighting, the increase in Macedonia should be smaller. In the countries that did not experience civil war, I expect only smaller changes. Finally, individuals directly affected by violence should display more nationalist attitudes than others.
Data and methods
The analysis in this article is two-fold. To test H1, I compare ethno-nationalism over time and between a group of countries that did experience civil war and a group of countries that did not. To perform these tests, I use ordinary least square (OLS) regression, with observations clustered in country strata. 8 I also examine the change in ethno-nationalism in each country as well as within the different ethnic groups to control for a possible composition effect, since large demographic changes took place during the wars. Here, I compare prewar and postwar group means and use a t-test of equality of means to see whether the differences are statistically significant. To test H2, I use OLS regression to test the effect of individual exposure to warfare in the countries that did experience conflict.
The data sources used in the analyses consist of two different surveys conducted in 1989 and 2003 in the former Yugoslav area. The prewar data is a survey collected in late 1989 and early 1990 by the Consortium of Social Research Institutes of Yugoslavia (Kunovich & Hodson, 1999). As postwar data I use the South-East European Social Survey Project (SEESSP), a survey from late 2003 to early 2004 that covered seven Balkan countries (Simkus, 2007). 9 This survey will also be used to test H2. Both surveys used a random sample of households and door-to-door interviews. Fortunately, the SEESSP survey was especially designed to be comparable with previous Yugoslav surveys, so several questions from the 1989 survey are also included in the SEESSP survey (Simkus, 2007). Variable-specific details on the questionnaire are provided below. For the first part of the analysis, the aggregate units are Kosovo, Bosnia and Herzegovina, Croatia, Macedonia, Montenegro, Serbia, and Vojvodina, reflecting both the federal division of Yugoslavia and the structure of the 1989 survey. Slovenia was included in the 1989 survey but not in the 2003 survey, and is therefore excluded. As mentioned, both Kosovo and Vojvodina enjoyed a large degree of autonomy until 1989. The 2003 survey treated Kosovo independently and Vojvodina together with Serbia. However, Vojvodina’s history and ethnic composition are quite different from the rest of Serbia, which makes it interesting to analyze apart. The first four countries represent the test group and the latter three the control group. The second part of the analysis focuses on the most war-affected countries only, that is, Bosnia, Croatia, and Kosovo. 10
Some methodological considerations
Conceivably, some of the changes that are visible in the immediate postwar period are brief and will not persist. The Dayton peace agreement for Bosnia and Croatia was signed in 1995, which means that there is a time lag of about eight years from the end of the conflict to the measurement of its possible effects. The advantage of such a time lag is that the data do not capture the most immediate and possibly short-lived effects, but rather long-term changes. The disadvantage is, of course, that other factors have also changed between the two surveys, interfering in the causal relationship. For example, the Dayton agreement in Bosnia has been criticized for organizing politics along ethnic lines (McMahon & Western, 2009), giving room for nationalist propaganda as seen in recent elections. 11 In the cases of Kosovo and Macedonia, the conflicts are more recent. Moreover, postwar recovery, which should make it easier to leave the war behind and look forward, has been more rapid in Croatia than in Bosnia and Herzegovina. These differences represent an omitted variable bias, which cannot be controlled for, given the small number of cases. 12
Comparing surveys that were conducted by different organizations is difficult and must be done with caution. In this case, the second survey was designed in order to ensure comparability with the first. Still, some variables are coded differently in the two surveys, partly due to differences in the political situation between 1989 and 2003. This is especially important for variables like respondent ethnicity, where the relevant answer categories changed with the dissolution of Yugoslavia. Changes in the demographic composition in the countries complicate the comparison over time, but the disaggregation into ethnic groups provides an extra ‘layer’ of comparison, so that some general trends can be captured. In Croatia, for example, the Serb minority is drastically reduced from 1989 to 2003. To the extent that Serbs and Croats in Croatia have different attitudes, changes in the relative size of the two groups will affect the overall country response profile. Still, the disaggregation of the Croatian population into Serbs and Croats allows for estimations of changes within each group, and it is therefore possible to assess whether the 1989–2003 change in the aggregate country profile is a composition effect or a real change. Importantly, minorities were oversampled in the 2003 survey, so that it is possible to estimate and compare the mean response even for small minorities like Serbs in Kosovo and Croatia. When comparing ethnic groups, unweighted data are used. Estimations for country populations are made using a population weight that improves the overall sample representativity (Simkus, 2007).
The use of ethnicity in survey and census data presupposes the meaningful existence of such categories or groups (Haug, 2001). Critics of ‘groupism’, notably Brubaker (2002, 2004; see also Fearon, 2003), argue that this might not be a valid presupposition. This might be true; however, previous research from the former Yugoslavia has showed that ethnic identity is indeed an important determinant of a number of different attitudes, and should be included in the analyses (Dyrstad et al., 2011; O’Loughlin, 2010). While we do not yet know what it is about the ethnic identification that produces such attitudinal differences, the differences are real, and therefore, self-reported ethnic identity should be included in the models to account for these differences.
Because of the large demographic changes in the countries and the lack of updated census data for several of the countries, it is hard to be sure that the two survey samples are equally representative. The 2003 sample is generally younger than the 1989 sample, and men are overrepresented in the 1989 sample, whereas the education distribution is quite similar in the two samples. Not surprisingly, the 2003 sample is more urban than the prewar sample, despite what appears to be a small overrepresentation of the republic capitals in 1989.
Dependent variable
Factor analysis of ethno-nationalism, factor loadings, 1989 and 2003 surveys
a Kaiser-Meyer-Olkin measure of sampling adequacy.
Four statements that tap into nationalist attitudes are included in both the 1989 and the 2003 surveys. These are (1) ‘Nationally mixed marriages must be more unstable than others’, (2) ‘Men can feel completely safe only when the majority belong to his nation(ality)’, (3) ‘Among nations it is possible to create cooperation, but not full trust’, and (4) ‘Every nation(ality) should have its own state’. Response categories ranged from ‘strong disagreement’ to ‘strong agreement’. Crucially, statements and answer categories are very similar across the two surveys. 13 In previous studies, some of these items have also been taken to measure ethnic intolerance (Hodson, Sekulić & Massey, 1994; Kunovich & Hodson, 1999; Massey, Hodson & Sekulić, 1999, 2006), ethnic prejudice (Hodson & Kunovich, 2002; Strabac & Ringdal, 2008), and ethnic exclusionism (Simkus, 2007). I argue that the term ethno-nationalism captures more accurately the content of these items.
Principal component analysis indicates that the items described above scale well. Cronbach’s alpha is 0.67 (1989 survey) and 0.77 (2003 survey). 14 Factor loadings range from 0.47 to 0.77. The good scalability in both surveys increases the overall reliability of the scale. Based on this, I computed two scales: a simple additive scale and a scale based on factor scores. Since the two scales had a 0.99 correlation, I chose to use the simpler scale, which has an easier interpretation as it can be rescaled to an interval of 1 (strongly disagree) to 5 (strongly agree).
Independent variables
To measure individual exposure to warfare I rely on a dummy variable that captures whether the respondent has had a close friend or family member killed or disappeared. There are several other variables available in the dataset that also measure war exposure, for example, being captured by enemy force, being forced to emigrate, and having a close family member or friend wounded. I use the arguably strongest measure of war exposure and use the others only as a robustness check. Unfortunately, in the Kosovo sample, these questions were only included in the Albanian questionnaire, so the Serb respondents have been coded to zero. Several indicators that tap into post-traumatic stress disorder also exist. Since these indicators also contain an element of how well individuals have coped with the traumatic events after the war, I prefer relying on an actual war-related event that the respondent has experienced.
Descriptive statistics by country, 2003 samplea
a Data weighted for increased representativity.
In addition to the variables described here, the regression model of ethno-nationalism includes a set of common demographic variables that the literature on outgroup evaluations has shown as relevant. 16 Male takes the value of 1 for men and 0 for women. Education measures the respondents’ education in years of completed education, while age measures the respondents’ age in years at the time the surveys were conducted. The scales of Current rural residence and rural residence at age 14 measure the size of the respondent’s place of residence where a value of 1 represents a capital city and 8 a small place consisting of a few farms. Income and religiosity are measured as standardized variables with a mean of 0 and a standard deviation of 1. Table II shows descriptive statistics of the variables in the last part of the analysis (Bosnia, Croatia, and Kosovo). 17
Analysis
The purpose of the present analysis is to test two propositions; first, that ethno-nationalism increases with the experience of ethnic civil war, and second, that individuals who were exposed to violence display a stronger sense of ethno-nationalism than others.
A box plot, with median and first and third quartiles, provides a first overview of ethno-nationalism in 1989 and 2003 in the former Yugoslavia (Figure 1). Starting with Kosovo, the level of ethno-nationalism was already high in 1989 and hardly changed by 2003. Bosnia, Croatia, and, to a lesser extent, Macedonia saw an increase in ethno-nationalism from 1989 to 2003. Countries that did not experience civil war on their own territory also saw an increase in ethno-nationalism, but not as strongly as Bosnia and Croatia. In most countries, there appears to be a convergence of attitudes, that is, the variation is smaller in 2003 than in 1989.
Ethno-nationalism, by country and year.
Predicted ethno-nationalism, by year and civil war experiencea
a Predicted from OLS regression. N=21,423; R2 = 0.096.
** Pr(|T| > |t|) < 0.01; * Pr(|T| > |t|) < 0.05. Observations are clustered on country to account for within-country dependence between observations.
Pre- and postwar ethno-nationalism
Table III shows the estimated values of ethno-nationalism for both groups of countries at both points of time. The values are predicted from the beta coefficients of an OLS regression containing two dummy variables, year and civil war experience, and their interaction term. Before the violent dissolution of Yugoslavia, the level of ethno-nationalism was slightly higher in the countries that were to experience ethnic civil war, but the difference is small. Ethno-nationalism increased during the war in both groups of countries, but the increase was significantly larger in the countries that did experience ethnic civil war, that is, in Kosovo, Bosnia and Herzegovina, Croatia, and Macedonia.
However, in the group that experienced ethnic civil war, the increase in ethno-nationalism is relatively small. In the postwar period, the mean difference between the two groups is about 0.3 points on a 5-point scale. Thus, while H1 finds some support, the evidence is not very convincing. To get a better understanding of the changes that took place, I disaggregate the two groups of countries into individual countries and ethnic groups in each country. As discussed previously, the wars that took place were quite different in duration as well as intensity. The large migrations that took place during and after the wars also make it useful to compare subnational groups.
Table IV shows pre- and postwar means as well as the difference between them. To check whether the changes are statistically significant, I used a series of t-tests of equality of means (indicated by asterisks).
Starting with the group of countries that did experience ethnic civil war, only two of them actually saw an increase in ethno-nationalism from the prewar to the postwar period. The increase is largest in Bosnia and Herzegovina, but notable also in Croatia. In Kosovo and Macedonia, on the other hand, no change took place. In comparison, all the countries in the control group saw a small increase in ethno-nationalism.
Looking more closely at the specific groups in each country, we see that in Kosovo, both Albanians and Serbs have very high levels of ethno-nationalism in 1989. This is as expected, due to the longstanding, low-intensity conflict in Kosovo throughout the 1980s. While Albanian ethno-nationalism stayed equally high in 2003, the Serb minority actually saw a small reduction. As already noted, the 1980s saw violent clashes between Albanian insurgents and Serbian security forces in Kosovo. Although they were not actually at war, the situation in Kosovo was also turbulent when the 1989 survey was conducted, so the classification of 1989 Kosovo as a ‘prewar country’ might be misleading. Moreover, the level of ethno-nationalism was already so high in 1989 that it could hardly increase any further. Therefore, the lack of change could also be interpreted as a ceiling effect.
In Bosnia and Croatia, all ethnic groups saw an increase in ethno-nationalism, notably self-declared Serbs in Bosnia and Herzegovina. Both countries remained less polarized than Kosovo, however. Macedonia saw relatively high levels of ethno-nationalism in 1989, and overall, this remained unchanged in 2003. However, while self-declared Albanians actually had a reduction in ethno-nationalism from 1989 to 2003, the self-declared Macedonians saw an increase during the same period, so the net effect is no change. The share of Albanians in the sample is also larger in 2003 than in 1989, giving the Albanian reduction a relatively greater weight than the Macedonian increase in ethno-nationalism. Recall that the violence in Macedonia was limited both in time and in space compared to the successive attempts of ethnic cleansing that took place in Kosovo, Bosnia, and Croatia.
Change in ethno-nationalism (mean), by ethnic group and country
Group means are excluded if the number of observations is less than 20. Two-tailed t-test of equality of means, equal variance assumed. ** Pr(|T| > |t|) < 0.01; * Pr(|T| > |t|) < 0.05.
For the countries in the control group, there was a small increase among all groups except the Albanians in Montenegro. 18 Thus, the results for my first hypothesis are mixed. Four countries experienced armed conflict on their own territory in the period between the two surveys. Of these, Bosnia and Herzegovina and Croatia started out with relatively low levels of ethno-nationalism and saw a large increase in a time span of 14 years, while Macedonia and Kosovo saw no change. Kosovo was already highly polarized along ethnic lines in 1989, probably due to the low-intensity conflict between Serbs and Albanians, and this did not change much. Macedonia had moderately high levels of ethno-nationalism in 1989 and a very nationalist minority, and saw no net change from 1989 to 2003.
Yugoslavia was still considered one country by the time conflict started. The Yugoslav national army was involved in the wars in Bosnia and Croatia, and Serbia took part in the Kosovo war. As parts of the federation, all the republics were to some extent affected by the ongoing violence, even if the violence did not spread to their own territory. Therefore, it is not surprising that the other three cases also saw an increase in ethno-nationalism during the period. Previous research has also found a spatial contagion effect of conflict risk in neighboring countries, in particular when ethnic boundaries do not follow state boundaries (Gleditsch, 2007).
It is difficult to estimate a possible effect of the different time lags from the conflict to the post-conflict survey (eight years in the case of Croatia and Bosnia, four years in the case of Kosovo, and two years in the case of Macedonia). The only increase in ethno-nationalism is experienced by the two countries with longest time lag, but as already discussed, the conflict trajectories vary from case to case, and it is hard to make general observations beyond the cases examined.
Effect of war exposure on ethno-nationalism
Having analyzed the changes in ethno-nationalism over time, I now turn to the effect of the individual experience with warfare in order to test the second hypothesis. Table V shows a set of OLS regression models on a pooled sample of Kosovo, Bosnia and Herzegovina, and Croatia. 19
Model (i) tests the effect of having a close relative killed when controlling for other relevant factors. Model (ii) is a variation of Model (i) where the measure of war exposure is allowed to vary between countries. The control variables confirm the impression from Table IV; ethnic identity is an important determinant of ethno-nationalism, and the differences are substantial. The other control variables seem to be less important. Only religiosity turns out to be statistically significant on a 0.05 level. People who report being more religious also share more ethno-nationalist attitudes. However, as pointed out by Kunovich & Hodson (1999), religiosity is also related to strength of ethnic identity, so ethno-nationalism and religiosity could be mutually reinforcing. Overall, the socio-demographic variables have surprisingly little effect (see Quillian, 1995).
Determinants of ethno-nationalism (OLS regression)
a Croatia is reference category. b Self-reported Serb ethnicity is the reference category.
Robust standard errors in parentheses. ** Pr(|T| > |t|) < 0.01; * Pr(|T| > |t|) < 0.05.
According to the first model, the respondents’ sense of ethno-nationalism is not affected by exposure to war-related violence. In the second model, we see that in Croatia, the effect is actually statistically significant and quite strong, whereas in Bosnia and Herzegovina and Kosovo, no effect can be found.
Turning to the second hypothesis, again, the evidence is mixed. According to Model (i), having a close friend or family member killed does not affect the respondents’ reported level of ethno-nationalism. However, when allowed to vary between countries, we see that in Croatia, the effect is statistically significant and relatively strong, whereas in Bosnia and Kosovo, no effect can be identified. The results remain stable through a series of model specifications with different proxies for war exposure (being forced to emigrate, strong symptoms of post-traumatic stress disorder), with a set of disaggregate dependent variables, with country-specific samples, and with an ML hierarchical model with ethnic subgroups as level two. 20 While exposure to war in Croatia significantly increases ethno-nationalism, it does not appear to have any effect in the two countries where the fighting was most widespread.
Despite the lack of effect on the individual level, the large increase in ethno-nationalism in Bosnia and Herzegovina from 1989 to 2003 is hard to explain without reference to the background of the 1992–95 war. How can these seemingly contradictory findings be explained? Several possible explanations exist. Previous studies have emphasized the role of fear in explaining negative outgroup evaluations (Feldman & Stenner, 1997; Quillian, 1995). In a conflict setting, it is likely that the perceived threat or the fear of exposure to violence work in the same direction as the actual exposure to war-related activities, especially if the fighting is widespread and indiscriminate and civilians are targeted. In Croatia, the conflict affected only some regions directly; the others remained relatively peaceful and their inhabitants protected. In comparison, in Bosnia and Kosovo, where the warfare was much more massive, the whole population was affected more indiscriminately, and almost nobody was spared from fear and insecurity. This could explain why in Croatia, there is a marked difference between the attitudes of those who were exposed to war and those who were not, while in Bosnia and Herzegovina and Kosovo, no such difference can be found.
Conclusion
The current situation in Bosnia and Herzegovina illustrates the need to understand the legacy of war on individual attitudes. Fifteen years after the peace agreement was signed in Dayton in 1995, the country remains highly segregated along ethnic lines. The complex political system that stems from the Dayton agreement was designed to secure the protection and inclusion of the three main ethnic groups, but in practice, it also bars political reform, and ethnicity remains the most dominant cleavage in the political landscape (McMahon & Western, 2009).
In multi-ethnic states, where the principle of congruence between the political and the national units (Gellner, 2006) is bound to be violated, attitudes of nationalism can represent a serious problem. My analysis indicates that ethnic civil war has the potential to increase ethno-nationalism and polarize the society, but it is also clear that this effect is context-dependent and sometimes behaves in unexpected ways. Before the war in Yugoslavia, ethno-nationalism was weak in most of the republics, notably in Bosnia and Herzegovina, but also Croatia, Serbia, Montenegro, and the province of Vojvodina. Multi-ethnic Bosnia and Herzegovina underwent a radical change from very liberal to relatively nationalist in just a few years. The same change, but to a lesser degree, occurred in Croatia, where the warfare was less widespread. The changes might not be explained by the wars alone, but it is hard to understand the complicated interethnic relations in the Western Balkans today without reference to the recent conflict.
The case of Kosovo is quite different from that of Bosnia and Herzegovina. While prewar ethno-nationalism was strong, it did not change much from 1989 to 2003. Most likely, people were mobilized along ethnic lines long before the actual outbreak of war. One interpretation of the finding is that there is a limit to how much ethno-nationalism can increase, that is, there is a ceiling effect at work. By 1989, the level of mutual distrust and segregation between Serbs and Albanians in the Serb province was already so high that the events in the 1990s hardly changed anything. It is also likely that there always will be small groups of liberal ‘hardliners’ who refuse to take on a nationalist attitude, no matter what.
Hegre & Sambanis (2006) find a strong positive ‘neighborhood effect’ on the onset of civil wars. They argue that even though conflict in a neighboring country makes civil war more likely, the exact mechanism of this spillover effect is not well understood. As the case of Montenegro makes clear, ethno-nationalism can also increase in countries that do not experience armed conflict on their own territory. Especially in regions where ethnic and state boundaries do not concur, ethnic conflict in one state could cause ethnic relations to deteriorate in neighboring countries, making it easier for rebels to gain support (see Gleditsch, 2007).
Perhaps the most puzzling finding is the lack of difference between the most and the least affected individuals and their level of ethno-nationalism in Bosnia and Herzegovina. In Kosovo, this non-finding is not so surprising, simply because ethno-nationalism is so dominant that there is hardly any variance to explain. In Bosnia and Herzegovina, on the other hand, one would expect that as in Croatia, the most affected individuals are also the most nationalist. One possible explanation for the difference between the two can be found in the intensity and geographical extension of the violence. The widespread fighting in Bosnia and Herzegovina made the whole population a possible target of violence, which could have reduced the difference between those who actually experienced war-related violence and those who only feared it. Future studies of other post-conflict settings could help answer questions like these.
To conclude, it is hard to explain this set of cases without reference to context-specific explanations. At least, it seems clear that war has no deterministic effect on ethno-nationalism; context matters. It is also very likely that local policy responses as well as development in the postwar recovery process influence how individuals cope with and react to the recent violence. Further complicating the picture, the actual meaning or implication of ethno-nationalism could be context-dependent. In Bosnia, a demand for separatism based on ethnicity could possibly lead to more violence for all involved parties, whereas for the Albanian majority in Kosovo, such a demand could simply lead to increased independence for the de facto state of Kosovo. Future research should concentrate on expanding the number of cases with comparable data, so that these questions can be answered with a greater level of certainty.
Footnotes
Replication data
Acknowledgements
A previous version of this manuscript was presented in Stockholm at the SGIR 7th Pan-European International Relations Conference, 9–11 September 2010. I would like to thank the editor of JPR, three anonymous reviewers, Ola Listhaug, and Halvard Buhaug for helpful comments.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
