VoxEU Column Politics and economics

Breakdowns of social consensus: The political economy of culture and ethnicity

The current refugee crisis has highlighted the importance of understanding how ethnic and cultural differences affect social cohesion. This column investigates the links between ethnicity and culture, and the relationship between diversity and civil conflict. It finds that globally, there appears to be little overlap between ethnic identity and cultural identity. Also, ethnic diversity per se has no effect on civil conflict. It is when differences in culture coincide with differences in ethnicity that conflict becomes more likely.

Are ethnic cleavages associated with deep differences in preferences, norms, values, and attitudes? Many people think so. In developing countries, ethnic divisions are often blamed for lack of agreement on the broad goals of society – leading to dysfunctional governance and conflict. Countries with fractionalised ethnic compositions, especially in sub-Saharan Africa, South Asia and Southeast Asia, are prone to social tensions and civil conflicts. In developed countries, the recent rise of populist movements has brought to the fore issues of cultural identity, and there is a growing perception that immigration and multiculturalism may lead to the breakdown of social consensus. The debate over Syrian refugees in Europe is at least as much about cultural values as it is about economic interests.

Many of these concerns over ethnic divisions are based on the underlying assumption that people agree about values and norms within ethnic groups and disagree across ethnic groups, but this assumption has not been subjected to empirical scrutiny. The relationship between ethnic diversity and political economy outcomes has been extensively analysed in widely cited contributions by Easterly and Levine (1997), Alesina et al. (1999), Fearon and Laitin (2003) and Alesina et al. (2003) (these are but a few examples in a vast and growing literature).

Yet much less is known about the underlying mechanism. Does ethnic diversity per se impact growth, public goods, redistribution, and conflict? Does cultural diversity act as the mediating channel? Or is it only when ethnic groups are culturally distinct that political economy outcomes are negatively impacted? In recent research, we conducted a systematic investigation of the links between ethnicity and culture (Desmet et al. 2016).

A definition of culture

We think of an individual’s culture as his set of norms, values, attitudes, and preferences. To measure culture empirically, we use the World Values Survey. This survey asks individuals across a wide variety of countries a set of questions about their values, norms, and attitudes. The survey also provides information about an individual’s ethnicity and language. This allows us to analyse the extent to which culture and ‘ethnolinguistic’ identity coincide.

Social antagonism, culture and ethnicity

Ethnic differences may exacerbate social antagonism, but what is not well-understood is why this might be the case. We argue that there could be three different channels at work, depending on the factors underlying social tensions. When discussing these three channels, consider a hypothetical society with two ethnic groups – the Amhara and the Oromo – and one cultural trait – how obedient children should be.

  • Cultural antagonism

Consider a society where antagonism arises from cultural differences, and where ethnicity plays no role. No one cares about the other’s language, colour, or ethnicity. People only feel tension when they are with others who have different preferences. In our simple example, no one cares about whether you are Amhara or Oromo, but people do care about how you educate your children. In this case, we can measure tensions by the probability that two randomly chosen individuals disagree about how obedient a child should be. We refer to this measure of social antagonism as cultural fractionalisation.

  • Ethnic antagonism

Now consider a society where antagonism only stems from ethnic differences. No one cares about how the others think or what their attitudes are, they only care about whether they have the same skin colour or speak the same language. This could be due to pure racial animosity, or it could come from a lack of trust or communication barriers between groups. In our simple example, no one cares about whether you think children should be obedient, but people do care about whether you are Amhara or Oromo. In this case, we can measure tension by the probability that two randomly drawn individuals are of a different ethnic group. We refer to this measure of social antagonism as ethnic fractionalisation.

  • The overlap between ethnicity and culture

A third possible society is one where individuals feel antagonism towards other ethnic groups, but only to the extent that those other groups are culturally different from their own. That is, people do not feel tension towards culturally different individuals from their own ethnic group or towards other ethnic groups that are culturally similar. They only experience antagonism towards other ethnicities if they are culturally different. In our simple example, you only dislike people of a different ethnic group if they have different views on how to educate children. 

A very intuitive way of measuring this overlap between culture and ethnicity is by using the fixation index, FST, a measure first developed by population geneticists. In our context it is the ratio of between-group cultural heterogeneity to total cultural heterogeneity. It measures the share of the overall variation in culture that is due to differences between ethnic groups. We refer to this measure of social antagonism as FST, or the overlap measure.

If FST or overlap is high, then knowing someone’s ethnicity provides a lot of information about his culture. If, on the other hand, FST is low, then knowing someone’s ethnicity tells me little about his culture. To illustrate this idea, go back to our child obedience example. If two-thirds of the Amhara think children should be obedient and one-third don’t, and the same is true for the Oromo, then knowing someone’s ethnicity does not reveal anything about his culture. The variation in culture in the society at large is identical to the variation in culture within the Amhara and within the Oromo. In that case, the FST value would be zero – in population genetics this would be the result of perfect interbreeding, or ‘panmixia’. If, instead, all the Amhara think children should be obedient and none of the Oromo do, then knowing someone’s ethnicity gives us perfect information about her culture. In that case, the FST value would be one – we have perfect fixation of the cultural trait to ethnic groups.

When economists and political scientists use ethnic diversity as a proxy for cultural diversity, they typically have this latter situation in mind. More to the point, if there is perfect fixation, then ethnic diversity and cultural diversity would be indistinguishable. If we were to find that ethnic diversity causes conflict, we would not be able to know whether this is because of ethnic antagonism or because of cultural antagonism. But as argued before, so far this has been an unproven assumption.

What do the data show?

Figure 1 shows a map of cultural fractionalisation, measured by the probability that two randomly drawn individuals of a country answer a randomly drawn question of the WVS differently. The most culturally diverse country is Zambia and the least culturally diverse country is Jordan. Examples of countries that are culturally highly diverse include France and India, and examples of countries that are culturally much more homogeneous include Egypt, Indonesia, and China.

Figure 1 Cultural fractionalisation

Similarly, Figure 2 displays ethnic fractionalisation. A common assumption in the literature is that cultural diversity is highly correlated with ethnolinguistic heterogeneity. Comparing Figure 2 and Figure 1, it becomes immediately obvious that no such correlation is present in the data. Countries such as Pakistan and Egypt have high levels of ethnolinguistic heterogeneity but low levels of cultural heterogeneity. At the other extreme are countries such as Germany and South Korea, which are ethnolinguistically fairly homogeneous but culturally diverse. The lack of a relationship between both types of heterogeneity is not limited to these few examples – the correlation between both measures is zero.

Figure 2 Ethnolinguistic fractionalisation

The absence of a correlation between cultural and ethnic fractionalisation suggests that there is little overlap between ethnic identity and cultural identity. Figure 3 shows FST, the share of the variation in cultural fractionalisation that occurs between groups. Perhaps surprisingly, the average FST is around 0.012. That is, only 1.2% of the variation in culture is between groups, so 98.8% occurs within groups. Yet there is interesting variation in FST across countries: Western Europe and Latin America display particularly low values of FST whereas South Asia and Sub-Saharan Africa have high values.

Figure 3 Variation in cultural fractionalisation

Culture, ethnicity and civil conflict

We now return to one of our initial questions. What drives the relationship between diversity and civil conflict? Is it cultural diversity and the differences in opinions about life? Is it ethnic diversity per se that worsens outcomes? Or is it when ethnic groups have divergent views that outcomes are affected? We find it is the latter.

In contrast to a commonly accepted view, we find that ethnic diversity per se has no effect on civil conflict. Instead, it is when differences in culture coincide with differences in ethnicity that conflict becomes more likely. Cultural diversity, on the other hand, has, if anything, a pacifying effect. The magnitude of the effect of overlap on conflict is large – a one standard deviation increase in our overlap measure leads to a 17% increase in the probability of conflict. We have applied this same methodology to analyse the provision of public goods and the level of development. As with civil conflict, we find that ethnic diversity per se does not matter. Instead, it is when the overlap between culture and ethnicity is high that public goods provision suffers and economic development lags behind.


Our analysis leads to four important conclusions. First, ethnic diversity cannot be viewed as a proxy for cultural diversity. The correlation between both is essentially zero. Second, differences between ethnic groups only explain 1.2% of the overall cultural heterogeneity we observe in societies, making it hard to predict someone’s culture based on knowing her ethnic group. Third, there is interesting variation in the degree to which ethnicity predicts culture. In rich, developed democracies, as well as in Latin America, the association between the two has essentially been erased, whereas in poor societies of sub-Saharan Africa and South Asia, there is a stronger link between cultural values and ethnic identity. Fourth, although the overlap between culture and ethnicity is small, variation in the degree of overlap is relevant as a predictor of civil conflict, public goods and development.

Our research suggests that if modernity allows cultural values to become further unmoored from ethnic identity, there is hope that the deleterious effects of ethnic antagonism will fade away as countries become richer and more democratic.


Alesina, A, A Devleeschauwer, W Easterly, S Kurlat, and R Wacziarg (2003), "Fractionalization", Journal of Economic Growth, 8 (2), 155-194

Alesina, A, R Baqir, and W Easterly (1999), "Public Goods and Ethnic Divisions", Quarterly Journal of Economics, 114 (4), 1243-1284

Desmet, K, I Ortuño-Ortín, and R Wacziarg (2016), "Culture, Ethnicity and Diversity", CEPR Discussion Paper 10451

Easterly, W, and R Levine (1997), "Africa's Growth Tragedy: Policies and Ethnic Divisions", Quarterly Journal of Economics, 112 (4) 1203-1250

Fearon, J, and D Laitin (2003), "Ethnicity, Insurgency, and Civil War", American Political Science Review, 97 (1), 75-90

6,824 Reads