Cite as:

Yaneer Bar-Yam, How can we stop ethnic violence?, New England Complex Systems Institute (November 25, 2016).

Ethnic violence is a global problem that has claimed many millions of lives. Countless acts of violence have been fueled by tensions between groups of disparate cultures, religions, and languages. Finding a way to alleviate ethnic tension would go a long way towards a more peaceful world. Traditional studies of violence focus on the details of historical animosities, economic disparities, leadership, or social structures of different regions. Our research has identified a method to accurately predict propensity to violence based only on the large scale structure of a society. This insight provides guidance for how the worst of ethnic violence can be prevented.

The former Yugoslavia is an important case study for our understanding of ethnic violence. Yugoslavia was formed at the end of World War I as a union of smaller states with diverse cultures. Throughout its history, Yugoslavia experienced frequent periods of economic and political turmoil. The state broke apart in the 1980s and 90s, including a period of severe ethnic violence. Many modern day countries are similar amalgams of groups that self-identify as distinct but were combined by the forces of colonialism and conquest. Understanding the fate of Yugoslavia has broad relevance.

In our research we found that ethnic geography is the key property for predicting and explaining ethnic violence [1]. Ethnic geography refers to the places members of distinct groups of people live in a particular region. Regardless of the history of a society or whether the groups are different ethnicities, religions or languages, there are fundamental principles that describe how they settle in geographic patches.

Individuals have a natural tendency to choose to live near people similar to themselves. Even if they start from a random spatial distribution (see Figure 1), the desire to live near others of the same type causes them to form patches of a particular group that grow in area over time. This behavior is universal, it does not depend on the specific details of what is happening or where. As long as people move around and prefer to be near others of the same type, any system will display the same universal dynamic — a general process that applies to diverse social systems and even materials formed of atoms or biological organisms formed of cells. Migrations of a group of people with a shared history and population displacements can increase this tendency to form patches when a group settles together in a specific area.

Because of this universal process, geographic aggregates of people are seen everywhere in the world. It also means we should start to describe how groups of people interact with each other based upon the geographic patches that are formed. The way groups behave toward each other is determined by how individual behaviors combine together, a property that is itself related to the way groups are distributed in space. This is the reason we investigated the connection between patch size and ethnic violence.

We should realize, however, that the idea of universal dynamics that applies to many different types of system does not mean that it happens in every instance. Every group of people has a mixture of individual differences. For some types of differences individuals do not have a preference and so become naturally mixed. In other cases they actually choose to be near specific other types (certain types of professionals that need to work together, or men and women, for example). We can tell which is happening by looking at their geographical distribution and these cases should be treated differently.

If the patch size is important in predicting violence, then it seems reasonable that there is a specific size of patches where violence will take place. A well mixed neighborhood would correspond to the case when people are fine with living together. Said in the other direction, when people are mixed they know each other personally so they won’t tend to fight. On the other hand, when the patch size is very large, most people don’t encounter members of the other groups, so they wouldn’t be involved in fights. Intermediate patch sizes, patches of a particular size, are likely to be the culprit.

We performed case studies in Yugoslavia, India, and Switzerland and found indeed that when patches of ethnic groups are above a certain size, ethnic violence does not occur. When patches are below a certain size, violence does not occur either. These thresholds are approximately above 60 kilometers and below 20 kilometers. This is a natural size, as 20 kilometers is the longest distance a person might walk in a day — a natural measure of the size of a community. When islands or peninsulas of one ethnic group fall within the critical range, surrounded by another group or groups, ethnic violence is likely to occur.

Figure 2 shows our analysis of ethnic geography and violence in Yugoslavia. The areas predicted as vulnerable to violence and actual incidents of violence have a 90 percent geographical correlation.

Large ethnic patches occur in a well separated / segregated society. Each ethnic group has its own spatial region in which its own values, customs, and traditions are expected to apply in public spaces by default. In a well mixed society, with small patches, no one group has or expects to have a monopoly over behavior in public spaces. In the critical range for violence, expectations about public spaces are in conflict. An ethnic group’s sense of autonomy is challenged by members of groups in bordering neighborhoods who may regularly pass through public spaces. Consequently, the border regions of these patches show the highest propensity towards violence.

This suggests that as groups choose to separate the only solution to violence is to separate them even more. However, there is another choice — boundaries that allow for more local autonomy so that groups can regulate their public spaces.

Switzerland is an example of a state with well placed boundaries [2]. While Switzerland has a reputation as a peaceful country, it has a diverse population of Catholics and Protestants and French, German and Italian speakers. In other places these can be in conflict, yet as Figure 3 shows, little violence is predicted by our analysis, and this analysis is consistent with the reality. Switzerland’s multidimensional sub-populations, combining both languages and religions, are separated by both political borders in the form of cantons that are part of a federal system of governance, and physical boundaries formed by the country’s natural abundance of mountains and lakes. Still, there is one region where violence was predicted and in fact has been historically observed. This region occurs near a too easily navigable mountain range between French and German speakers. This violence led to the creation of a new canton in 1979, a political border created to promote separation and regional autonomy.

Yugoslavian borders are quite different from those in Switzerland. Figure 4 shows the predictions for ethnic violence both with and without borders. There are two areas where the borders do work as in Switzerland. When the model doesn’t include the borders around Slovenia and Macedonia we predict violence there, but when the model includes them that violence is gone. There isn’t violence in the real world, which does have those boundaries. However, the presence or absence of borders in the rest of the country has little effect on the geography of violence. This means that these borders are not in the right places. Accordingly, moving the boundaries would have been a peaceful alternative to the disastrous violence that took place.

During Yugoslavia’s existence, it was variously called the Kingdom of Yugoslavia, Democratic Federal Yugoslavia, Federal People’s Republic of Yugoslavia, and Socialist Federal Republic of Yugoslavia. All of these governments attempted to keep the peace between the country’s diverse ethnic groups. While they may have succeeded for a time, violence eventually broke out. The idea that local autonomy is a good way to organize a government is present in other places, including not just Switzerland but also the United States, which has municipal, district, state and federal government systems. This model of governance is not being used in many places of the world where national governments have most of the governing authority.

A well integrated society is often held up as the ideal. Efforts to form boundaries and separate conflicting groups are viewed negatively. Our results suggest that in many cases boundaries and local autonomy are a constructive alternative to severe violence. Historically, attempts at separation didn’t recognize that there is a specific patch size for which the introduction of boundaries is the most critical. They also generally didn’t recognize the use of subnational boundaries for this purpose. For example, the separation of Pakistan and India left the patchy area of Kashmir that is the site of violence. The division of Northern Ireland from the rest is similar.

Sub-national boundaries aligned with natural communities can reduce ethnic tensions and prevent violence. Self-governance at the neighborhood, municipal or district-level can offer communities a sense of autonomy and safety. This may not be the embodiment of a cultural melting pot, but well thought out boundaries and the way local decision making serves each community can allow for peaceful coexistence. Such boundaries are not needed if people choose to integrate, which is also present in many places.

If Yugoslavia had respected local political autonomy of groups, it might have looked a lot more like Switzerland. As we consider the many places in the world where patchy ethnic areas are the locations of friction, conflict, suffering and refugees the recognition of a way to achieve peaceful coexistence can provide better alternatives.

We can also consider the idea of patches as its own kind of respect for individual differences. We respect individual differences in a society because of what each individual can contribute. Complexity science adopts a multiscale perspective on society. In this view, groups of various sizes are like individuals contributing to the global society that we are all part of. Respecting group decisions about public spaces, giving them the local autonomy to do so, is part of this respect. The idea of self-determination is already one of the concepts that are widely discussed for inter-group relationships.

Given the death and suffering caused by ethnic violence, we can identify places where people are in geographic patches of different sizes, and rather than forcing them to become integrated, we can provide for peaceful coexistence. The natural tendency to self-aggregate can be leveraged for peace if communities are given proper borders and self-governance over their own neighborhoods. Additional thought and research are needed into where and what form the boundaries and the autonomy should take. Switzerland and other existing federal systems of governance can serve as starting points for this inquiry.


  1. M. Lim, R. Metzler, Y. Bar-Yam, Global pattern formation and ethnic/cultural violence. Science 317, 5844 (2007).

  2. A. Rutherford, D. Harmon, J. Werfel, S. Bar-Yam, A.S. Gard-Murray, A. Gros, Y. Bar-Yam, Good fences: The importance of setting boundaries for peaceful coexistence. PLoS ONE 9(5), e95660 (May 21, 2014). doi:10.1371/journal.pone.0095660

Figure 1: Simulation of type separation with two types of agents [(A) to (E) show the system at 8, 64, 512, 4096, and 32768 attempted moves per particle, respectively]. Patches of a certain size that are surrounded by the other type are highlighted. We identify such regions with a high likelihood of conflict.

Figure 2: (A) Census data from 1991 shown here in map form were converted into a spatial representation shown in (B). Our prediction of populations likely to be in conflict with neighboring groups [red overlay, (C) and (D)] agrees well with the location of cities reported as sites of major fights and massacres [yellow dots, (D)]. The actual violence can be in the immediate vicinity of the locations of those populations that we predict are likely to be involved in violence.

Figure 3: Linguistic groups and topographical boundaries in Switzerland. Maps of Switzerland showing (A) proportion of linguistic groups according to the 2000 census, (B) elevation within Switzerland, (C) overlay of linguistic groups onto a digital elevation model, and (D) topographical features including lakes (blue) and ridges extracted using edge detection (cyan). Comparison of calculated propensity (color bar) to violence between linguistic groups without (E) and with (F) the inclusion of topographical features as boundaries using a characteristic length scale of 24 km. Mercator projection, except C which is the Europe Albers projection. The distance scale is approximate.

Figure 4: Ethnic groups, political borders, and topographical boundaries, in the former Yugoslavia. (A) Map of the area of the former Yugoslavia showing administrative provinces. Propensity to violence calculated without (B) and with (C) administrative boundaries, using a characteristic length of 21 km. Locations of boundaries are shown on both plots as solid and dashed yellow lines respectively. Sites of reported violence are shown as red dots [18]. Spurious violence is predicted along the borders of Slovenia and Macedonia when boundaries are not included. Province labels: SL: Slovenia, CR: Croatia, VO: Vojvodina, B&H: Bosnia & Herzegovina, SR: Serbia, MN: Montenegro, KO: Kosovo, MA: Macedonia.