Social cleavages,wartime experience,and ethnic cleansing in Europe

Abstract

What explains ethnic cleansing? Recent research has used systematic evidence to explore the causes of civilian victimization and mass killings. Yet, comparable studies that focus on ethnic cleansing are still rare. This article conceptualizes ethnic cleansing as a group-level phenomenon that is distinct from civilian victimization or mass killings and studies its causes by using systematic evidence from Europe 1900–2000. The article makes two theoretical moves. First, it highlights the salience of non-ethnic cleavages such as social class as a background condition that has the ability to hinder ethnic cleansing. Second, it distinguishes between two causal mechanisms, one that considers wars as ‘strategic environments’ and the other as ‘transformative experiences’, that relate to the proximate causes of ethnic cleansing. Using original data from 20th-century Europe, the empirical analysis offers two main findings. First, it shows that salient social cleavages, measured through levels of land inequality, political competition, and support for left-wing parties, substantially decrease the risk of ethnic cleansing. Second, the analysis suggests that the arguments that underscore psychological mechanisms related to wartime experiences provide a better explanation for ethnic cleansing than the arguments that emphasize the role of strategic wartime aims. This finding is further supported by a brief discussion of key cases in which both causal mechanisms predict ethnic cleansing. The results highlight the importance of treating ethnic cleansing as a conceptually separate phenomenon and offer implications for the debate on democracy and mass ethnic violence.

Keywords

civilian victimization democracy and violence ethnic cleansing genocide mass killings social cleavages

Introduction

What explains the wholesale deportation or murder of ethnic groups? Recent years have witnessed the proliferation of systematic studies that explore the causes of civilian victimization in civil or international wars (Downes, 2006; Kocher, Pepinsky & Kalyvas, 2011; Valentino, Huth & Balch-Lindsay, 2004; Schneider & Bussman, 2013; Eck & Hultman, 2007; Wayman & Tago, 2010). Despite the existence of studies that treat ethnic cleansing as conceptually separate, comparable empirical analyses of this phenomenon are hard to come by.¹ In the absence of such studies, it is difficult to ascertain whether and to what extent ethnic cleansing and related phenomena such as mass killings or civilian victimization share the same causes.²

This article contributes to the literature on mass violence in general and ethnic cleansing in particular in several ways. Conceptually, it utilizes a definition of ethnic cleansing that is analytically and empirically distinct from related phenomena. In particular, the article argues that, since in cases of ethnic cleansing the target is by definition an ethnic group, this concept only includes events in which a large proportion of an ethnic group becomes the target of violence. This conceptualization distinguishes ethnic cleansing from instances of mass killings and/or civilian victimization, which are typically defined by the absolute number of victims.

Theoretically, the article makes two moves that improve upon the existing literature. First, it highlights the salience of non-ethnic cleavages such as social class as a background condition that has the ability to hinder ethnic cleansing. This approach constitutes a fresh understanding as the existing literature on mass ethnic violence tends to focus on the nature of interethnic relations as the only relevant background factor (Kaufman, 2001; Fein, 1979; Mann, 2005). Second, the article unpacks the relationship between wars and ethnic cleansing by distinguishing between two causal mechanisms. The first, inspired by the literature on civilian victimization, treats ‘wars as strategic environments’, and argues that ethnic cleansing is a means to winning or avoiding losing wars (Downes, 2008; Valentino, 2004: 69). The other mechanism treats wars as ‘transformative experiences’, and argues that events that transpire during wars lead to ethnic cleansing by altering the dominant ethnic groups’ perceptions and expectations that relate to other ethnic groups (Petersen, 2002; Midlarsky, 2005, 2011; Bulutgil, forthcoming).³

Empirically, the article uses systematic data from Europe, where the unit of analysis is the country-group-year during the period 1900–2000, to analyze the causes of ethnic cleansing. The study is among the first to undertake this type of systematic cross-national analysis of ethnic cleansing and one of its goals is to stimulate similar studies that would assess its findings by using data from other regions.⁴ The theories under evaluation, however, are meant to apply across geographical contexts.

Two main findings follow. First, the background conditions that have the most significant impact on ethnic cleansing are those that relate to competition along socio-economic cleavages. The results indicate that countries with higher levels of socio-economic inequality, higher levels of political competition, and higher levels of support for left-wing parties are less likely to use ethnic cleansing against the groups within their territory. Thus, briefly put, the structural conditions that impact ethnic cleansing are far from being exclusively about ethnicity. Second, on the relationship between war and ethnic cleansing, the findings suggest that, while both causal mechanisms play some role, the one that considers wars ‘transformative experiences’ constitutes the more typical path that results in ethnic cleansing.

The following section provides the conceptual and operational definition of ethnic cleansing and distinguishes the concept from related ones. Next I outline the theoretical landscape and present the testable implications of the approaches. The ensuing section describes the data and presents the results. Finally, the article concludes by summarizing the broader implications of the findings.

Defining ethnic cleansing

Three criteria are particularly important in developing a conceptual definition for ethnic cleansing: the target, the perpetrator, and the methods that the perpetrator uses to carry out the violence. The first essential characteristic of ethnic cleansing is that the intended target of violence is an ethnic group as a whole rather than members selected based on some other criteria. The ideal typical case of ethnic cleansing would be the one in which all members of an ethnic group are victimized; and those cases in which a significant portion of a group is victimized would be closer to the ideal type than those in which a few members of the group are targeted. The challenge for the operationalization of this dimension is setting a threshold that specifies what percentage of a group needs to be targeted before we can reasonably assume that the goal was to victimize the entirety or at least a substantial part of it. There is no theoretically infallible way of setting such a threshold. Thus, the empirical section picks a plausible percentage and then varies it to ensure that the results are robust to different threshold specifications.

The second issue is the perpetrator that implements ethnic cleansing. One approach is to suggest that the perpetrator could be any organized group that controls and governs a given territory. Therefore, policies carried out by states as well as non-state actors that have the ability to govern territory could count as cases of ethnic cleansing. There are, however, serious practical problems with this approach. The theoretical explanations for ethnic cleansing need to be tested using indicators on the size and characteristics of ethnic groups as well as the levels of political and socio-economic competition in the territory in question. It is close to impossible to identify systematic information on these indicators across territories controlled by non-state actors. Thus, the units of analysis in the empirical section are the ethnic groups in internationally recognized states and the cases of ethnic cleansing included are those in which an internationally recognized state is at least one of the perpetrators.

The third definitional issue is the method used to remove ethnic groups from a specific territory. States use a variety of methods including massacres, expulsion of populations to other countries, compulsory population exchange agreements, and internal deportations. This article categorizes all the events that fulfill the criteria discussed above as instances of ethnic cleansing regardless of the methods used.

Figure 1 displays the relationship between ethnic cleansing and related concepts.⁵ The broadest category corresponds to mass victimization, which includes deportations and killings that target a large number of people. The main criterion for the operationalization of mass victimization would be the absolute number of victims rather than their proportion within a group. Therefore, ethnic cleansing can be thought of as a subcategory of mass victimization where the target is an ethnic group. The next relevant concept is mass killing, which is identical to mass victimization except that it only includes cases in which the method of victimization is killing. Ethnic cleansing cases in which a substantial part of an ethnic group is murdered, or genocides, also count as instances of mass killing.⁶ However, since the concept of mass killing includes cases in which the target is not an ethnic group and ethnic cleansing includes cases in which deportations are the main method of targeting, this overlap between the two concepts remains only partial. Finally, civilian killings during military conflict, which are a subcategory of mass killings, coincide with the cases of ethnic cleansing that both occur during military conflicts and take the form of killings.

Theoretical landscape: Background conditions and proximate causes

The article categorizes potential explanations for ethnic cleansing along two dimensions: those that focus on background conditions and those that focus on proximate causes. The theories that emphasize the background factors provide predictions on the underlying conditions that make some contexts more prone to ethnic cleansing than others, but they have little to say about when exactly ethnic cleansing occurs. The theories that focus on proximate causes provide specific predictions about timing, but they treat the underlying structural conditions as a black box. Given their different foci, the explanations that fall under the two dimensions are best treated as complementing rather than competing with each other. As discussed below, however, there are potentially competing arguments within each dimension.

Background conditions: Ethnic relations and social cleavages

These arguments depart from the idea that there are exogenous structural conditions that determine the extent to which dominant ethnic groups in different contexts are inclined to use ethnic cleansing. Such conditions are theoretically relevant as in cases where the underlying proclivity is already high, potential triggers might easily result in ethnic cleansing, whereas in contexts where this proclivity is low to begin with, ethnic cleansing might not follow even when the potential triggers are in place. This section focuses on two approaches that relate to these baseline conditions.

Ethnic relations

The study of mass ethnic violence has long highlighted the potential importance of dislike or prejudice as a potential cause. Some studies argue that long-standing myths generate suspicions against certain groups and, under conditions of instability, provide advantages to extreme nationalist leaders (Kaufman, 2001; Fein, 1979; Staub, 1990). Others contend that, in ethnically heterogeneous contexts, the emergence of the modern nation state coupled with democratic rule or limited resources result in ethnic exclusion (Mann, 2005; Wimmer, 2002). In these approaches, the ethnic groups that evoke the highest levels of suspicion on the part of the dominant group are those that possess the most plausible chance of reversing the ethnic hierarchy.

Several factors can contribute to the ability of a non-dominant group to revise the existing power relations. Demographic comparison between the dominant and non-dominant groups might play a role given that those groups that have the numerical strength to challenge the dominant group might be more likely to change the ethnic hierarchy. Organizational and cultural factors might also enhance the ability of the non-dominant group to withstand assimilation. For example, groups with a separate religious organization would not only have a pre-existing cultural basis on which to build a national identity but also have access to an organization through which the leaders can disseminate the nationalist message. The existence of significant linguistic differences

Figure 1.

Ethnic cleansing and related concepts

might also hinder assimilation and provide a unifying symbol for the non-dominant groups. Finally, non-dominant groups might be more likely to retain their identity and challenge the dominant group if they share ethnic affinity with a neighboring state.⁷

Social cleavages and political competition

The arguments discussed above use a one-dimensional picture of politics in multi-ethnic contexts that solely concentrates on ethnicity. This approach is both theoretically and empirically questionable. There is a long tradition in comparative politics that draws a link between the existence of competing social cleavages and the mitigation of political conflict (Dahl, 1982; Dahrendorf, 1959; Lipset, 1959; Taylor & Rae, 1969; Lipset & Rokkan, 1969; Lijphart, 1977). Recently, the idea that competing social cleavages might reduce the intensity of specifically ethnic conflict has been used in works that analyze democratic stability, political party behavior, voting, and civil wars (Chandra, 2005; Dunning & Harrison, 2010; Cederman, Weidmann & Gleditsch, 2011; Gubler & Selway, 2012). Historical works on multi-ethnic contexts in Europe also show that politics in these settings were informed not only by ethnicity but also by divisions that related to other cleavages (Rothschild, 1974; Kuru, 2009). Despite this literature, studies that focus on ethnic cleansing either ignore competing social cleavages or assume them to be invariably tangential.⁸

This section argues that the existence of salient cleavages within dominant groups might affect the likelihood of ethnic cleansing by generating political and ideological incentives on the part of the dominant group members to cooperate with the non-dominant groups. In contexts with cross-cutting cleavages, this cooperation would take the form of political alliances between ideological counterparts within the dominant and non-dominant groups (for example, cooperation between socialists within two ethnic groups). But political cooperation could be possible even in the absence of ideological counterparts within the non-dominant group. First, the members of the dominant group who are repressed by their co-ethnics on another political dimension might be ready to give ethno-linguistic concessions to the non-dominant group in return for their support.⁹ Second, these repressed members might also be ideologically more inclined to harbor suspicions towards the national project and its exclusionary goals.

When confronted with events that might trigger ethnic cleansing, the reactions of the political factions that cooperate with the non-dominant groups might differ from those of the others. Due to the possibility of future cooperation, even when faced with potential triggers, these individuals might prefer to select out specific members of the non-dominant group rather than victimizing the group as a whole. Furthermore, based on their past experience, the individuals in the cooperative faction might hold more positive attitudes towards the non-dominant group and hence be more reluctant to approve ethnic cleansing. Therefore, salient divisions within the dominant groups could decrease the probability of ethnic cleansing.

The main distinction between this argument and the one that emphasizes the nature of interethnic cleavages is that this argument puts ethnicity in the context of other cleavages. Thus, the argument is not inconsistent with the expectation that, all else being equal, in contexts where ethnic cleavages are more salient, ethnic cleansing might be more likely. However, what distinguishes this argument from the former is that it suggests an additional empirical implication that the previous argument does not:

H1: The more salient is the competition along socio-economic (or other non-ethnic) cleavages within the dominant ethnic group, the less likely is ethnic cleansing against the non-dominant ethnic groups.

Proximate causes: War as strategic environment and war as transformative experience

As discussed above, it is plausible to assume that the members of the dominant groups vary in the extent to which they are inclined to use ethnic cleansing against other groups. The theories that focus on proximate causes offer hypothetical stories about how specific events tilt the balance towards those members that favor ethnic cleansing. These causal stories can be grouped into two types that ascribe two theoretically distinct roles to wars.

War as strategic environment

According to this causal logic, wars serve as potential triggers for ethnic cleansing by shortening the time-horizon of the leaders vis-à-vis certain ethnic groups. Specifically, the leaders of the dominant group become more inclined to use this policy as the immediate goal of winning (or avoiding losing) a war outweighs the other, more long-term, goals that they might have (Downes, 2008, 2006: 170; Valentino, 2004). These long-term goals might include the possibility of future political cooperation with the non-dominant group or any other opportunity cost that might follow from using ethnic cleansing. Two main predictions follow from this argument. First, the targets of ethnic cleansing are groups that are either anticipated to or actually hinder the war effort. Second, since the timing of ethnic cleansing is linked to the short-term horizon generated by the goal of winning a war, once the war is over, these arguments no longer predict ethnic cleansing.

The related literature offers several potential explanations on how leaders identify groups that pose strategic risks during wars. One is an extension of the idea that minorities that have access to a neighboring state in which their co-ethnics constitute the dominant group are potentially dangerous for their host state. This argument does not predict the timing of ethnic cleansing given that ethnic affinity with neighboring states is a condition that remains relatively constant overtime. However, in the case of a war between the neighboring and the host states, this type of link might increase the likelihood of ethnic cleansing by leading the host state to anticipate that the ‘co-ethnic’ minorities would join the enemy forces or aid them behind the front lines (Downes, 2008: 35–36).

Leaders might also consider specific ethnic groups as strategically risky if these groups actually militarily collaborate with an enemy state or organize a rebellion during a war (Valentino, 2004: 69; Downes, 2008). Once again the presumed motivation in these cases would be to prevent the ethnic group from undermining the war effort by indirectly diverting resources from the front line or directly increasing the strength of the enemy. This argument is consistent with the studies that indicate that balance of power considerations exert a significant impact on the levels of civilian victimization during civil wars or military occupations (Kalyvas, 2006; Wood, 2010; Valentino, Huth & Balch-Lindsay, 2004; Eck & Hultman, 2007). Yet we do not know whether the logic of these arguments also accounts for ethnic cleansing.

Put together, these arguments lead to the following expectation:

H2: Ethnic groups that (a) are the ethnic kin of an enemy state in war, (b) engage in military collaboration with an enemy state in war, or (c) rise in rebellion during a war should face an elevated risk of ethnic cleansing during the war in question.

War as transformative experience

The arguments that treat wars as transformative experiences suggest that events that the members of dominant ethnic groups encounter during wars alter the extent to which they desire to cooperate and live with the other ethnic groups in the long run. Several causal stories fall under this argument. One is the idea that status reversals that rearrange the ethnic hierarchy result in feelings of resentment on the part of the group that loses status and increases the likelihood of mass ethnic violence if and when this group regains power (Petersen, 2002; Midlarsky, 2005, 2011). A related mechanism is revenge, which refers to cycles of violence between ethnic groups that eventually reach levels of wholesale removal for one or both of the groups (Balcells, 2010; Kennan, 1993). War and postwar contexts are especially conducive to these types of outcomes as they allow repeated exchanges of territory, thereby exacerbating feelings of resentment and vengefulness.

Experiences during wars might also change the dominant ethnic group’s assessments on the possibility of peaceful coexistence after the war. Evidence from behavioral economics indicates that individuals are likely to take into account their most recent experiences when making decisions (Tversky & Kahneman, 1973). In immediate postwar environments, ethnic groups might be inclined to discount earlier experiences of cooperation and exaggerate their recent experiences that relate to wartime. To the extent that these experiences entail status-reversal and/or collaboration with enemy forces, wars might lead to the pessimistic assessment that peaceful coexistence is no longer possible and that the other group(s) constitute a substantial threat in the long run. Moreover, the recent experience of war itself might lead the leaders to exaggerate the future likelihood of wars, further exacerbating the perceived threat.¹⁰ Based on these causal mechanisms, the arguments that treat wars as transformative experiences generate the following hypothesis:

H3: Those groups that experience status promotion or engage in wartime collaboration should face an elevated risk of ethnic cleansing either in the course of these events or soon after these events come to an end.

As the discussion shows, there are empirical differences as well as commonalities between the ‘war as transformative experience’ and ‘war as strategic environment’ arguments. Given its retrospective rather than anticipatory logic, the ‘war as transformative experience’ approach differs from the other one in two main respects. First, it expects ethnic cleansing to target groups that retain collaborative ties with enemy states during or in the aftermath of wars. Second, it does not predict ethnic cleansing against those groups that share the same ethnic identity as the dominant groups in enemy states unless these groups actually collaborate with the enemy state in question.

The two approaches also share one empirical implication: they both expect leaders to use ethnic cleansing against groups that militarily collaborate with a wartime enemy state. The underlying logic for this expectation is different in the two arguments. In the case of the ‘war as strategic environment’ argument, the underlying reason is the immediate strategic considerations during wars; whereas in the case of the ‘war as transformative experience’ argument, the underlying reasons are the motivations for revenge as well as the changing understanding of long-term security threats. Given this commonality, the empirical section follows a two pronged strategy to evaluate these arguments. First, it compares the empirical performance of the two arguments in the context of statistical analysis. Second, it provides a discussion of the specific cases in which the two types of approaches have identical predictions.

Data and results

The empirical analysis uses data from 20th-century Europe to test the arguments presented.¹¹ The dataset is based on dyads of states and non-dominant groups in each year, where a non-dominant ethnic group is defined as any group which is not the most populous in a given state. This definition makes conceptual sense given that the non-dominant groups in Europe were almost always also numerically smaller.¹²

Systematic data on the ethnic groups and their population shares do not exist for the first half of the 20th century. The datasets that cover later periods cannot be presumed to apply to this era precisely because of the ethnic cleansing episodes and border changes that substantially changed the demographic landscape of Europe. To generate a new list, I first included the ethnic groups mentioned under each country heading in the relevant issues of Encyclopedia Britannica. These data were then refined by consulting books and articles on specific countries.¹³ Ethnic groups are counted as residing within a state if the land they lived on belonged to the state since independence or if it was incorporated in the state’s territory through annexation.¹⁴

Ethnic cleansing is defined as an event in which a state exterminates or forcefully and permanently deports at least 20% of an ethnic group on its territory from their current location to another within three years. The starting year of each event in which at least 20% of a group was removed within three years was coded as 1. In cases where the ethnic cleansing campaign took longer than one year, the subsequent minority-state-years were dropped from the dataset. (The online appendix provides the list of the cases.)

The independent variables correspond to the background conditions and proximate causes outlined above. To test H1, which posits that higher levels of political competition along socio-economic (or other non-ethnic) cleavages should result in lower likelihood of ethnic cleansing, the analysis uses three indicators. First, Family farms measures the area of family farms as a percentage of the total area of agricultural holdings (Vanhanen, 2003). The higher the portion of land held by family farms rather than sizable estates, the more equal the distribution of land, and the lower the level of inequality. Since the expectation here is that higher levels of socio-economic equality should increase the likelihood of ethnic cleansing, the expected sign on this variable is positive. The extent of class conflict in a given country is likely to be the outcome of pre-existing rather than current levels of inequality; therefore the variable is lagged for ten years.

A counter-intuitive aspect of the coding criteria for this variable is that the lowest values belong to the communist countries, suggesting that these were the countries with the deepest socio-economic cleavages. This is because the coding rule counts the communist states themselves as large land owners. However, since in these countries the land distribution was not unequal in the conventional sense, the coding does not accurately capture the depth of land inequality. Therefore, the models below are also run while excluding most of the communist countries.¹⁵

The other two variables capture the organizational aspects of the socio-economic divisions. Political competition measures the extent of competition in the political system of a given country. This variable is calculated by subtracting the percentage of votes gained by the largest political party in parliamentary and/or presidential elections from 100% (Vanhanen, 2003). In cases where there are no elections or only one party competes in the elections, it takes the value of 0. The variable is available in ten-year intervals starting from 1898. The article uses the level of competitiveness that corresponds to the beginning of each ten-year interval.

Political competition and Family farms are not perfect proxies for the extent of socio-economic competition within the dominant groups because they do not distinguish between competition within the dominant group and competition between groups. Specifically, both variables are likely to overstate the extent of competition within the dominant group in ethnically more heterogeneous parts of Europe such as central and eastern regions. Since these are also the regions in which ethnic cleansing episodes tended to occur, the overestimation actually makes the analysis less likely to support H1. Thus, in the absence of systematic group-level socio-economic data for the period under study, the two variables are good indicators to test H1.

The last indicator for levels of socio-economic competition is Left-wing vote, measured as the percentage of votes captured by political parties with a socialist, communist, or social democratic platform in the closest competitive parliamentary elections to date within the past ten years.¹⁶ If no competitive elections occurred in the last ten years then the indicator takes the value of 0. This variable specifically measures the extent to which political parties that prioritized socio-economic divisions and downplayed ethnic ones enjoyed political clout. One complication with this variable is that in the context of the post-communist countries, some of the so-called socialist parties in practice adopted a primarily nationalistic rather than economic platform. The most obvious example is the Socialist Party of Serbia, which remained socialist in name but not in ideology. Given this ideological ambiguity, the models that include Left-wing vote are also run while excluding the post-communist countries.

The article also employs four indicators to control for the impact of interethnic relations on ethnic cleansing. Religious difference ranges from 0 to 2 and is coded 0 if most members of the non-dominant group belong to the same religious organization as most members of the dominant group; 1 if most members of the non-dominant belong to a religious denomination (e.g. Protestantism, Catholicism, and Orthodox Christianity) different from most members of the dominant group; and 2 if most members of the non-dominant group belong to a religion (e.g. Christianity, Judaism, Islam) different from most members of the dominant group. The second, Linguistic difference, measures the extent to which the non-dominant group language falls in the same category as the dominant group language. If the languages belong to different families (e.g. Indo-European, Uralic, Altaic), the variable is coded 2. If they belong to different branches of the same family (e.g. Germanic and Latin), the variable is coded 1. If they fall within the same branch of a language family, the variable is coded 0. The third, Demographic balance, corresponds to the ratio of the dominant group to that of the non-dominant group. Since ethnic groups that are larger might also be costlier to remove, I use the log of this variable. Finally, Ethnic affinity with neighbor is coded 1 for groups which share the language of the dominant group in a neighboring state and 0 otherwise.¹⁷

The analysis also uses a number of binary variables to test the theories on proximate causes.¹⁸ To evaluate the ‘war as strategic environment’ argument, the article uses a variable labelled Fifth column. This variable is coded 1 under three conditions: (a) if a group shares the language of the dominant group in a neighboring country with which their host state is at war;¹⁹ (b) if group members rebel during a war;²⁰ (c) if group members serve in units linked to the army of an enemy state during a war.²¹ This variable is coded 1 until the war ends.

To test the ‘war as transformative experience’ argument, the article uses two variables. The first, Promoted group, measures whether a given group experiences political or military promotion compared to the dominant ethnicity in the context of a war. The variable is coded 1 if (a) a non-dominant group’s language replaces that of the previously dominant group’s language in parts of the country that are occupied by foreign armies; (b) an enemy state enlists the members of the group to fight against the army of their host state and/or to police and govern the dominant group in occupied territories.²² The variable is coded 1 for two years beginning with the year of enlistment or reversal of linguistic hierarchy. If a state did not have access to the promoted group until after it recovered its territory, the variable is coded 1 for the year of recovery and the following year.

The second variable, Defending group, captures the possibility of ethnic cleansing in newly acquired territories. When a group plays a prominent role in the defense of a country, the adversary might engage in revenge seeking if it succeeds in capturing the territory. In this case the primary targets would be the groups that formed the backbone of the defending state’s army and other groups that formed volunteer units to serve in the defending army. Therefore, Defending group is coded 1 for (a) groups that had been numerically and administratively dominant in the army of the defending state before an annexation; (b) groups that fought against the annexing state in separate volunteer units. The groups that fulfill either of these criteria are coded 1 for the year that the annexation takes place and for the year after.

Finally, the analyses also control for several other factors that might impact the probability of ethnic cleansing. Spillover controls for the possibility that the promotion of one non-dominant group also increases the likelihood of ethnic cleansing for the other non-dominant groups. Thus, if one group in a given state was coded 1 for Promoted group, all the other groups in this state were coded 1 for Spillover. Literacy captures the extent to which states possess the capacity to utilize policies such as assimilation that could constitute alternatives to ethnic cleansing (Vanhanen, 2003). Finally, the models also control for Population, Per capita income, and a Year counter variable that starts with 0 for year 1900 goes up to 100 for year 2000.²³

Results

To estimate the effect of the proximate and structural factors on ethnic cleansing, the article uses logit analysis. Table I displays the models in which the dependent variable is Ethnic cleansing measured at the 20% threshold. The models were also run using 10% and 40% thresholds for measuring ethnic cleansing as well as using year or country fixed effects. These models are mentioned in the article if they alter the results in a notable way and are available through the online appendix.

Background conditions: Social cleavages and political competition

The results support the argument that the non-ethnic social cleavages, especially once they are organized and competitive, considerably reduce the risk of ethnic cleansing in a given country.²⁴ Political competition is statistically significant throughout all the specifications including those that control for year or country fixed effects and use different levels of threshold for measuring ethnic cleansing. Left-wing vote is statistically significant in the model that excludes the former communist countries due to the ideological ambiguity of the socialist parties in these contexts. This finding is robust to using the 40% threshold or fixed effects models. Family farms, which captures equality of land distribution, also has the expected impact on the likelihood of ethnic cleansing. Compared to relatively direct indicators such as Political competition or Left-wing vote, equality of land distribution is a more distant proxy for the extent to which the actors that focused on socio-economic cleavages actually wielded political influence in a given context. Despite this, Family farms is statistically significant in the model that includes the entire dataset. Family Farms continues to be statistically significant in the model that uses the 10% level threshold but not when using the 40% threshold or the fixed effects models.

To appreciate the substantive impact of competition along socio-economic cleavages, consider how the predicted probability of ethnic cleansing changes for groups that are coded 1 on Promote group depending on the levels of Political competition, Left-wing vote, or Family farms.²⁵ Under this assumption, the predicted probability of ethnic cleansing for groups that are located in countries with the highest level of political competition is 0.01, as opposed to 0.10 for the groups located in countries with the lowest score of political competition. The impact of Family farms on the probability of ethnic cleansing is likewise substantial. The predicted probability of ethnic cleansing is 0.07 for groups that had the highest percentage of family farms and hence the least unequal distribution of land, whereas it is 0.018 for groups that had the smallest percentage of family farms and hence the most unequal distributions of land.²⁶ When excluding the communist countries (Model 5), the predicted probabilities are even more remarkably different: 0.10 for the least unequal societies and 0.02 for the most unequal ones. The impact of Left-wing vote on the probability of ethnic cleansing is also noteworthy especially when excluding post-communist countries. For groups located in a country with the highest level of support for left-wing parties, the predicted probability of ethnic cleansing is 0.009, as opposed to 0.04 for countries with the lowest level of support for left-wing parties.²⁷

These findings also shed light on historical contrasts between otherwise similar contexts. Consider, for example, the policies of the Austro-Hungarian and Ottoman Empires toward ethnic minorities during wars. Both of these multinational empires included groups that collaborated during wars as well as groups that were ethnically and politically linked to neighboring enemy countries with irredentist goals (Bullock & Bujeiro, 2009; Tucker, Wood & Murphy, 1999; Kennan, 1993; Mann, 2005). Yet, while the Ottoman Empire used ethnic cleansing against various groups, the Austro-Hungarian Empire by and large avoided this type of collective targeting. In light of the findings, this variation can be attributed

Table I.

Logit analysis of ethnic cleansing

DV: Ethnic cleansing	Model 1	Model 2	Model 3 (post-communist countries excluded)	Model 4	Model 5 (communist countries excluded)
Demographic Balance (log)	−0.168 (0.139)	−0.175 (0.136)	−0.219 (0.145)	−0.211 (0.144)	−0.154 (0.179)
Linguistic difference	−0.133 (0.321)	−0.101 (0.305)	−0.213 (0.339)	0.203 (0.291)	0.081 (0.301)
Religious difference	0.847** (0.299)	0.641** (0.243)	0.763** (0.280)	0.707** (0.261)	0.421 (0.284)
Ethnic affinity w/Neighbor	1.448** (0.439)	1.349** (0.455)	1.647** (0.499)	1.349** (0.434)	0.933* (0.465)
Literacy	0.014 (0.012)	−0.001 (0.011)	0.010 (0.012)	0.002 (0.011)	−0.003 (0.012)
Political competition	−0.043* (0.019)
Left-wing vote		−0.010 (0.011)	−0.028* (0.014)
Family farms				0.016^† (0.01)	0.020^† (0.011)
Fifth column	1.270* (0.629)	1.310* (0.643)	1.335* (0.670)	1.502* (0.648)	0.980 (0.763)
Promoted group	4.084** (0.594)	3.773** (0.629)	4.019** (0.634)	3.702** (0.607)	3.800** (0.615)
Defending group	2.397** (0.761)	2.813** (0.825)	2.506** (0.973)	2.334** (0.817)	2.658** (0.916)
Spillover	1.428* (0.618)	1.534** (0.594)	1.445* (0.632)	1.343* (0.622)	1.288 (0.831)
Per capita income	−0.052 (0.11)	−0.144 (0.113)	−0.312* (0.126)	−0.194 (0.136)	−0.144 (0.120)
Population	−0.004 (0.004)	0.001 (0.003)	0.001 (0.003)	0.001 (0.004)	−0.010 (0.008)
Year counter	−0.015 (0.013)	0.002 (0.010)	0.001 (0.010)	0.004 (0.010)	0.001 (0.01)
Constant	−7.998 (0.926)	−7.942 (0.903)	−8.286 (1.069)	−9.476 (1.049)	−8.134 (1.298)
N	11714	11204	10601	11684	8268
Pseudo R-squared	0.378	0.366	0.401	0.362	0.335

^† p < 0.1, *p < 0.05, **p < 0.01. Robust standard errors in parentheses.

to the differences in the political structure of the two countries. The Habsburg Empire hosted a number of German and minority group parties on the left and right of the political spectrum. In this context, even ‘the radical nationalists typically focused more on competing with rival parties within their own national camps than on combating their so-called national enemies’ (Cohen, 2007: 267). The situation in the Ottoman Empire was markedly different. The political competition within the Turkish community had never been as well organized as the intra-German competition in Austro-Hungary. Furthermore, by the beginning of World War I, a small clique of leaders with a nationalist agenda had already taken over the government and forced the bulk of within-Turkish political opposition to exile.

Background conditions: Ethnic relations

The results on the structural variables that relate to ethnic competition are more mixed. Religious difference and Ethnic affinity with neighbor both elevate the probability of ethnic cleansing and these findings are in general robust to different specifications. However, the substantive impact of the two variables on the probability of ethnic cleansing is smaller than that of Political competition. When keeping Promoted group at 1 and other variables at their mean, the probability of ethnic cleansing was 0.02 for those coded 0 on Religious difference and 0.08 for those coded 2. Likewise, keeping Promoted group at 1 and other variables at their mean, the likelihood of ethnic cleansing was 0.09 for groups that had ethnic affinity with a neighboring country and 0.02 for groups that did not.

Neither Linguistic difference nor Demographic balance has the expected impact on the likelihood of ethnic cleansing. The former variable is statistically insignificant in all the models and the latter, though significant in some, has a negative rather the expected positive sign. More refined measures of ethnic balance of political power, such as the share of minority groups in the state bureaucracy, might support the hypothesized relationship.²⁸ In the absence of such data, this finding raises doubts about the explanations that highlight the importance of exogenous ethnic competition without completely dismissing them.

Proximate factors: War as strategic environment and war as transformative experience

The results provide some support for both the ‘war as strategic environment’ and ‘war as transformative experience’ arguments. However, the analyses also suggest that the latter causal mechanism constitutes the more typical and prevalent path that links wars to ethnic cleansing. The three variables that capture the proximate causes of ethnic cleansing all come out as statistically significant and are generally robust to alternative specifications. However, the size of the impact is different in each case.

The two variables that test the ‘war as transformative experience’ argument, Promoted group and Defending group, have a larger impact on the probability of ethnic cleansing compared to Fifth column. To demonstrate, compare two hypothetical groups. Both are religiously different from the dominant group in their country of residence, located in a context without effective political competition, and ethnically linked to the dominant group in a neighboring country. One of the groups enjoys political or military promotion (coded 1 on Promoted group), whereas the other does not. Keeping the other factors at their mean, the first group would face a 41% probability of ethnic cleansing whereas the comparable probability for the group that does not experience promotion would be 1%.²⁹ If we were to replace Promoted group in the foregoing hypothetical with Defending group, then the corresponding probabilities would be 13% for the group coded 1 on Defending group and 1% for the other. Lastly, the corresponding probabilities would be 6% when the group is coded 1 on Fifth column and 1% otherwise.

As discussed in the theory section, both the ‘war as transformative experience’ and the ‘war as strategic environment’ arguments expect leaders to use ethnic cleansing during wars against groups that militarily collaborate with an enemy state. To further compare the empirical performance of the two causal paths, it is useful to discuss the ethnic cleansing cases that fit this description.³⁰ Four out of eight of these groups were located in the Northern Caucasus region of the USSR, which came under German occupation between September 1942 and January 1943.³¹ In all these cases, the Soviet authorities carried out the ethnic cleansing campaigns after the German armies pulled out of the region. However, since the war continued in the western regions, their goal might have been to prevent the recurrence of collaboration in case the Germans returned or to prevent rebellions that could have distracted resources from the front line. If this strategic logic holds, we would expect the ethnic cleansing episodes to occur during periods of higher threat from the Germans and uncertainly about Soviet victory.

To evaluate this expectation, it is useful to divide the period after the Soviet army’s recapture of the North Caucasus into two. First, between January and September 1943, there was still a minor possibility that the Germans would come back to the region. The German army retained a small foothold in Northern Caucasus and some German leaders continued to boast about having potential allies in the area (Dallin, 1981). During this period, the Soviet forces engaged in a mixture of policies such as violent campaigns against selected villages or working with former insurgent groups that switched to the Soviet side after the German retreat (Comins-Richmond, 2002; Burds, 2007). They did not, however, yet engage in wholesale deportations that targeted the entirety of these groups. During the second period after September 1943, the Germans abandoned their last foothold in North Caucasus and the possibility of a German comeback became all but non-existent (Dallin, 1981; Burds, 2007). It was during this period of relative security that all the deportations in Northern Caucasus were conducted. In short, though the deportations in Northern Caucasus took place during World War II, within the context of the war, they coincided with periods of relative Soviet ascendancy rather than decline and insecurity.

Conclusion

The main conclusions of the article are twofold. First, the analysis shows that, in the context of 20th-century Europe, socio-economic cleavages measured through land inequality, political competition, and strength of left-wing parties substantially decreased the risk of ethnic cleansing. Second, as far as the proximate causes of ethnic cleansing go, the findings suggest that the causal logic that emphasizes the role of wars as transformative experiences provides a fuller explanation compared to the causal logic that emphasizes the role of wars as strategic environments.

Both these findings have implications for the broader literature. The first implication concerns the debate on the relationship between democracy and mass ethnic violence. One side of this debate maintains that democracy, when applied to ethnically heterogeneous contexts, increases the likelihood of ethnic cleansing because in these contexts ethnic groups that are difficult to assimilate come to be excluded from the demos (Mann, 2005). The proponents of the argument draw attention to ethnic cleansing episodes that occur after free and fair elections such as the one in Bosnia or to colonial ethnic cleansing campaigns that were carried out by relatively democratic states.³² This argument is also compatible with studies that show varying degrees of relationship between elections and political violence or demonstrate that democracies are no less likely to target civilians in wars than dictatorships (Snyder, 2000; Collier, 2009; Downes, 2007; Cederman, Gleditsch & Hug, 2013; Steele, 2011). The other side of the debate offers cross-national evidence showing that democracies are less likely to engage in mass killings (Rummel, 1995). Their argument is that democracies hinder these types of events by generating norms that conflict with mass violence or by imposing institutional constraints on leaders.

The findings here support the argument that democracy, to the extent that it enhances competition along non-ethnic cleavages, acts as an obstacle against ethnic cleansing. In addition, the logic of the argument allows us to account for the cases of ethnic cleansing that were conducted by relatively democratic countries or not long after competitive elections. In Yugoslavia and Bosnia, 45 years of communist rule had largely eliminated significant class divisions that could have served as an alternative rallying point against ethnicity.³³ Moreover, the politicians who emphasized ethnicity had an additional advantage in that they could rely on existing religious organizations. Given the absence of salient alternative cleavages and the presence of organizational advantages for ethnic parties, it is not surprising that the transition to a competitive political system did not prevent ethnic cleansing in this context. In the case of the colonial ethnic cleansing campaigns, the targeted populations lacked citizenship rights and hence were excluded from the electoral process. This meant that the political incentives that ordinarily motivate competing actors within the dominant group to cooperate with the non-dominant groups did not extend to those groups in the colonies.³⁴ Hence, political competition in the colonizer could not have prevented ethnic cleansing against the colonized populations.

The findings also offer implications for the question of whether ethnic cleansing and other categories of mass violence, such as civilian victimization, share the same causes. As discussed above, the cross-national empirical literature on civilian victimization generally supports the idea that strategic considerations during wars are among the main predictors of civilian victimization during military conflict. In contrast, the findings here suggest that, at least in Europe, the causal mechanisms that emphasize the psychological reactions that are generated by wartime experiences capture the path that leads to ethnic cleansing better. Based on this evidence, the distinctions drawn between ethnic cleansing and other phenomena at the beginning of the article appear to be not only conceptually but also empirically justified.

Footnotes

Replication data

The dataset, codebook, and do-files for the empirical analysis in this article, as well as the online appendix, can be found at .

Acknowledgements

The author would like to thank the Editor and three anonymous reviewers for their insightful comments and suggestions. She is also grateful to Carles Boix, Stathis Kalyvas, Jacob Shapiro, and Nils B Weidmann for their comments on early versions of the article.

Notes

References

Ahmad

Feroz

(1982) Unionist relations with the Greek, Armenian, and Jewish communities of the Ottoman Empire, 1908–1914. In: Braude

Benjamin

Lewis

Bernard

(eds) Christians and Jews in the Ottoman Empire: The Functioning of a Plural Society. New York: Holmes & Meier, 401–434.

Balcells

Laia

(2010) Rivalry and revenge: Violence against civilians in conventional civil wars. International Studies Quarterly 54(2): 291–313.

Banks

Arthur S

(1997) Cross national time series: A database of social, economic, and political data (http://www.databanksinternational.com/32.html).

Bell-Fialkoff

Andrew

(1996) Ethnic Cleansing. New York: St Martins.

Bisogno

Marcelo

Chong

Albert

(2002) On the determinants of inequality in Bosnia and Herzogovina. Economics of Transition 10(2): 311–337.

Boix

Carles

(2008) Economic roots of civil wars and revolutions in the contemporary world. World Politics 60(3): 390–437.

Brubaker

Rogers

(1996) Nationalism Reframed. New York: Cambridge University Press.

Bullock

David

Bujeiro

Ramiro

(2009) The Czech Legion 1914–20. Oxford: Osprey.

Bulutgil

H Zeynep

(forthcoming) The Roots of Ethnic Cleansing in Europe. New York: Cambridge University Press.

10.

Burds

Jeffrey

(2007) The Soviet war against fifth columnists: The case of Chechnya, 1942–4. Journal of Contemporary History 42(2): 267–314.

11.

Cederman

Lars-Erik

Gleditsch

Kristian Skrede

Girardin

Luc

(2009) Ethnonationalist triads: Assessing the influence of kin groups on civil wars. World Politics 61(3): 403–437.

12.

Cederman

Lars-Erik

Gleditsch

Kristian Skrede

Hug

Simon

(2013) Elections and ethnic civil war. Comparative Political Studies 46(3): 387–417.

13.

Cederman

Lars-Erik

Weidmann

Nils B

Gleditsch

Kristian Skrede

(2011) Horizontal inequalities and ethno-nationalist civil war: A global comparison. American Political Science Review 105(3): 478–495.

14.

Chandra

Kanchan

(2005) Ethnic parties and democratic stability. Perspectives on Politics 3(2): 235–252.

15.

Cohen

Gary

(2007) Nationalist politics and the dynamics of state and civil society in the Habsburg Monarchy, 1867–1914. Central European History 40(2): 241–278.

16.

Collier

Paul

(2009) Wars, Guns, and Votes: Democracy in Dangerous Places. New York: Harper.

17.

Comins-Richmond

Walter

(2002) The deportation of the Karachays. Journal of Genocide Research 4(3): 431–439.

18.

Dahl

Robert Alan

(1982) Dilemmas of Pluralist Democracy: Autonomy vs. Control. New Haven, CT: Yale University Press.

19.

Dahrendorf

Ralf

(1959) Class and Class Conflict in Industrial Society. Stanford, CA: Stanford University Press.

20.

Dallin

Alexander

(1981) German Rule in Russia, 1941–1945: A Study of Occupation Policies. Boulder, CO: Westview.

21.

Downes

Alexander B

(2006) Desperate times, desperate measures: The causes of civilian victimization in war. International Security 30(4): 152–195.

22.

Downes

Alexander B

(2007) Restraint or propellant? Democracy and civilian fatalities in interstate wars. Journal of Conflict Resolution 51(6): 872–904.

23.

Downes

Alexander B

(2008) Targeting Civilians in War. Ithaca, NY: Cornell University Press.

24.

Dunning

Thad

Lauren

Harrison

(2010) Cross-cutting cleavages and ethnic voting: An experimental study of cousinage in Mali. American Political Science Review 104(1): 21–39.

25.

Eck

Kristine

Hultman

Lisa

(2007) One-sided violence against civilians in war: Insights from new fatality data. Journal of Peace Research 44(2): 233–246.

26.

Fearon

James D

(2003) Ethnic and cultural diversity by country. Journal of Economic Growth 8(2): 195–222.

27.

Fein

Helen

(1979) Accounting for Genocide: National Responses and Jewish Victimization during the Holocaust. Chicago, IL: University of Chicago Press.

28.

Gordon

Raymond G

ed. (2005) Ethnologue: Languages of the world. Fifteenth Edition. Dallas, Texas: SIL International. Online version (http://www.ethnologoue.com).

29.

Groth

Alexander J

(1968) The legacy of three crises: Parliament and ethnic issues in prewar Poland. Slavic Review 27(4): 564–580.

30.

Gubler

Joshua R

Selway

Joel Sawat

(2012) Horizontal inequality, crosscutting cleavages, and civil war. Journal of Conflict Resolution 56(2): 206–232.

31.

Harff

Barbara

(2003) No lessons learned from the Holocaust: Assessing risks of genocide and mass murder since 1955. American Political Science Review 97(1): 57–73.

32.

Kalyvas

Stathis

(2006) The Logic of Violence in Civil War. Cambridge: Cambridge University Press.

33.

Kaufman

Stuart J

(2001) Modern Hatreds: The Symbolic Politics of Ethnic War. Ithaca, NY: Cornell University Press.

34.

Kennan

George F

(1993) Report of the international commission of inquiry into the causes and conduct of the Balkan Wars. In: Kennan

George F

(ed.) The Other Balkan Wars: 1913 Carnegie Report. Washington, DC: Carnegie Endowment.

35.

Kocher

Adam

Pepinsky

Thomas

Kalyvas

Stathis

(2011) Aerial bombing and counterinsurgency in the Vietnam War. American Journal of Political Science 55(2): 201–218.

36.

Kuru

Ahmet

(2009) Secularism and State Policies toward Religion: The United States, France, and Turkey. New York: Cambridge University Press.

37.

Lieberman

Benjamin

(2006) Terrible Fate: Ethnic Cleansing in the Making of Modern Europe. Chicago, IL: Ivan R Dee.

38.

Lijphart

Arend

(1977) Democracy in Plural Societies: A Comparative Exploration. New Haven, CT: Yale University Press.

39.

Lipset

Martin

(1959) Political Man: The Social Bases of Politics. Garden City, NY: Doubleday.

40.

Lipset

Seymour M

Rokkan

Stein

(1969) Cleavage structures, party systems, and voter alignments. In: Lipset

Seymour M

Rokkan

Stein

(eds) Party Systems and Voter Alignments: Cross-National Perspectives. New York: Free Press, 1–63.

41.

Mann

Michael

(2005) The Dark Side of Democracy: Explaining Ethnic Cleansing. New York: Cambridge University Press.

42.

Midlarsky

Manus

(2005) The Killing Trap: Genocide in the Twentieth Century. New York: Cambridge University Press.

43.

Midlarsky

Manus

(2011) Origins of Political Extremism: Mass Violence in the Twentieth Century and Beyond. New York: Cambridge University Press.

44.

Mylonas

Harris

(2013) The Politics of Nation-Building: The Making of Co-Nationals, Refugees, and Minorities. New York: Cambridge University Press.

45.

Naimark

Norman

(2001) Fires of Hatred: Ethnic Cleansing in Twentieth-Century Europe. Cambridge, MA: Harvard University Press.

46.

Petersen

Roger

(2002) Understanding Ethnic Violence: Fear, Hatred and Resentment in Twentieth-Century Eastern Europe. Cambridge: Cambridge University Press.

47.

Rothschild

Joseph

(1974) East Central Europe between Two World Wars. Seattle, WA: University of Washington Press.

48.

Rummel

Rudolph J

(1995) Democracy, power, genocide, and mass murder. Journal of Conflict Resolution 39(1): 3–26.

49.

Saideman

Stephen

Ayres

William

(2008) For Kin and Country: Xenophobia, Nationalism, and War. New York: Columbia University Press.

50.

Sarkees

Meredith Reid

(2000) The Correlates of War data on war: An update to 1997. Conflict Management and Peace Science 18(1): 123–144.

51.

Schneider

Gerald

Bussmann

Margit

(2013) Accounting for the dynamics of one-sided violence: Introducing KOSVED. Journal of Peace Research 50(5): 635–644.

52.

Snyder

Jack L

(2000) From Voting to Violence: Democratization and Nationalist Conflict. New York: Norton.

53.

Staub

Ervin

(1990) The Roots of Evil. Cambridge: Cambridge University Press.

54.

Steele

Abbey

(2011) Electing displacement: Political cleansing in Apartadó, Colombia. Journal of Conflict Resolution 55(3): 423–445.

55.

Straus

Scott

(2001) Contested meanings and conflicting imperatives: A conceptual analysis of genocide. Journal of Genocide Research 3(3): 349–375.

56.

Taylor

Michael

Rae

Douglas

(1969) An analysis of crosscutting between political cleavages. Comparative Politics 1(4): 534–547.

57.

Ther

Philipp

(2014) The Dark Side of Nation States: Ethnic Cleansing in Modern Europe. New York: Berghahn.

58.

Tir

Jaroslav

Schafer

Philip

Diehl

Paul F

Goertz

Gary

(1998) Territorial changes, 1816–1996. Conflict Management and Peace Science 16(1): 89–97.

59.

Toft

Monica Duffy

(2007) Population shifts and civil war: A test of power transition theory. International Interactions 33(3): 243–269.

60.

Tucker

Spencer

Wood

Laura Matysek

Murphy

Justin D

(1999) The European Powers in the First World War: An Encyclopedia. New York: Taylor & Francis.

61.

Tversky

Amos

Kahneman

Daniel

(1973) Availability: A heuristic for judging frequency and probability. Cognitive Psychology 5(2): 207–232.

62.

Valentino

Benjamin

(2004) Final Solutions: Mass Killing and Genocide in the 20th Century. Ithaca, NY: Cornell University Press.

63.

Valentino

Benjamin

Huth

Paul

Balch-Lindsay

Dylan

(2004) ‘Draining the sea’: Mass killing and guerilla warfare. International Organization 58(2): 375–407.

64.

Vanhanen

Tatu

(2003) Democratization and Power Resources 1850–2000. FSD1216, version 1.0 (2003-03-10). Tampere: Finnish Social Science Data Archive [distributor] (https://services.fsd.uta.fi/catalogue/FSD1216?lang=en&study_language=en).

65.

Wayman Frank

Tago

Atsushi

(2010) Explaining the onset of mass killing, 1949–87. Journal of Peace Research 47(3): 3–13.

66.

Weiner

Myron

(1971) The Macedonian Syndrome: An historical model of international relations and political development. World Politics 23(4): 665–683.

67.

Wimmer

Andreas

(2002) Nationalist Exclusion and Ethnic Conflicts: Shadows of Modernity. Cambridge: Cambridge University Press.

68.

Wood

Reed

(2010) Rebel capability and strategic violence against civilians. Journal of Peace Research 47(5): 601–614.

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.

0.00 MB