Measuring social diversity in economic literature: an overview for cross-country studies

Received: 26 October 2020 Revised: 6 October 2021 Accepted: 25 October 2021 DOI: 10.1111/joes.12484 ARTICLE Measuring social diversity in economic literature: An overview for cross-country studies Eva M. Buitrago1M. Ángeles Caraballo2 1Economía Aplicada III, University of Seville Faculty of Economics and Business Sciences, Sevilla, Spain 2Economía e Historia Económica, University of Seville Faculty of Economics and Business Sciences, Sevilla, Spain Correspondence M.ÁngelesCaraballoPou,Economíae HistoriaEconómica,UniversityofSeville Faculty of Economics and Business Sciences,RamónyCajal,1,Sevilla41018, Spain. Email:[email protected] Fundinginformation PAIDI.JuntadeAndalucía,Grant/Award Number:ProjectP20-00628; Junta de Andalucía,Grant/AwardNumber: Research Group SEJ258 Abstract Over recent decades, the growing research on social diversity at country level has striven to explain several outcomes such as the differences in income across countries and the origins of civil conflicts. The literature employs a wide range of indices to measure social diversity that hinders the comparison of the results with respect to their effects on socio-economic performance. This paper intends to disentangle such a variety of indices and their applications. To achieve this goal, we have collected the social diversity indices used in cross-country studies, and have ascertained not only their similarities and differences, and the relationships between them, but also their main applications. Studies at country level have been selected that construct their own indices and that make their databases available. We show that the dimension and the index chosen to measure social diversity, the level of disaggregation of the social groups, and the geographical unit of analysis explain the tangle of indices and the mixed results achieved by the literature focused on socio-economic outcomes of social diversity. This paper enables the suitable evaluation and comparison of the effects of social diversity and the selection of the appropriate index depending on the analysis to be carried out. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2021 The Authors. Journal of Economic Surveys published by John Wiley & Sons Ltd. JEconSurv.2021;1–55. wileyonlinelibrary.com/journal/joes 1 2BUITRAGO and CARABALLO POU KEYWORDS conflicts, economic growth, fractionalization, polarization, social diversity indices 1 INTRODUCTION A quick look at income data worldwide shows that the gap in mean income across countries has continued to widen in recent decades. The GDP per capita of 10 of the richest countries in 1970 was 34.3 times the GDP per capita of 10 of the poorest at that time, but, by 2010, this difference had risen to a multiple of 104.9 (see Table A1 in Appendix I). Explaining why some countries are trapped in poverty while others are steadily growing constitutes one of the most highly studied topics in the growth and development area of economics, and remains the most crucial and challenging issue for economists in the field. A relevant strand of the literature aims to explain this issue based on socialidiversity as one of the main causes that may be underlying these differences in economic performance. From a theoretical point of view, the effects of social diversity on economic growth can be both positive and negative. Alesina and La Ferrara (2005) summarize the pros and consii involved. On the one hand, they point out that heterogeneity gives rise to a diversification in abilities and skills due to the variety in experiences and cultures. This may encourage innovation and creativity, which foster productivity, and, in turn, may boost economic growth. The benefits of diversity enter their model through the production function of private goods. On the other hand, diversity could give rise to conflicts of preferences regarding the provision of public goods that cause social unrest and economic damage, thereby harming economic growth. In the same vein, by focusing on ethnolinguistic diversity as a dimension of social diversity, Mauro (1995), Alesina et al. (2003), Alesina and Zhuravskaya (2011), and Papyrakis and Mo (2014), among others, assert that institutions in heterogeneous societies tend to be of lower quality by arguing that ethnic favoritism encourages corruption. If politicians and bureaucrats intend to favor certain groups, or to receive favors from such groups, then an inefficient allocation of resources is induced, which, in turn, affects economic growth. Alesina and La Ferrara (2005) introduce these costs in the model through the utility function. They consider that utility depends on the consumption of private and public goods. As mentioned above, private goods include the positive effects of heterogeneity. However, public goods may bring about the negative effects due to the conflicts that may arise from the different preferences shown by the various groups on the type of public goods that should be provided. As the number of groups increase, the utility from the consumption of public goods decreases. The fact that one effect prevails over another depends on several elements, which explain why the literature achieves conflicting results with respect to the impact of social diversity on social and economic outcomes, such as civil conflicts and economic growth. These issues are addressed in detail in Section 5. The first task to be tackled involves the definition of the category “social”, because it can be considered from different perspectives, such as language, race, religion, culture, and birthplace. In this respect, since the influential contribution of Easterly and Levine (1997), the dimension related to ethnicity has attracted special attention in the empirical literature at country level. Easterly and Levine (1997) strive to show that ethnolinguistic diversity may help explain why BUITRAGO and CARABALLO POU 3 certain countries choose the correct policies to promote growth, while others fail. This result can be illustrated with the so-called African growth tragedy. To measure ethnolinguistic diversity, Easterly and Levine (1997) employed a fractionalization index, which is defined as the probability that two individuals randomly drawn from a population belong to a different ethnolinguistic group, that is, the higher the index, the more diverse the society. At first glance, according to this index, low-income countries are more than twice as diverse as are high-income countries. In fact, from the 29 low-income countries for which the fractionalization index is available, 23 are African countries with a fractionalization index over 0.5 (see Table A2, Appendix I). This evidence has triggered growing interest in the effects of ethnolinguistic heterogeneity on socio-economic performance. The question that immediately arises involves the criteria that have to be employed to delimit the ethnic groups. For instance, the main differentiating criterion between ethnicities in the index used by Easterly and Levine (1997) is that of language. However, other criteria, such as race, religion, and place of birth (Horowitz, 1985), can also be employed. Moreover, social diversity can be regarded from dimensions other than ethnicity. In this vein, we refer to genetics, and also to culture due to its significance achieved in the literature. As discussed in Section 5, several alternative perspectives towards defining social diversity give rise to different social and economic outcomes. Second, there are social outcomes that can be explained from a number of the aforementioned dimensions of social diversity. In fact, the negative impact of ethnolinguistic fractionalization on economic growth found in the literature (Alesina et al., 2003; Easterly & Levine, 1997;Montalvo& Reynal-Querol, 2005b, among others) has been interpreted as a result of the fact that conflicts are more likely in highly fractionalized societies which, in turn, harm economic growth. However, while the negative relationship between ethnolinguistic fractionalization and growth is widely supported by various studies, the ethnolinguistic fractionalization does not appear as a significant element in the explanation of civil conflicts (Collier & Hoeffler, 1998, 2004; Fearon & Laitin, 2003; Montalvo & Reynal-Querol, 2005a, 2005b). Nevertheless, this cannot lead to the conclusion that ethnicity is not one of the possible sources of conflict, but rather indicates that fractionalization is not the appropriate index to measure the ethnic divisions that can generate conflict (Esteban & Ray, 1999; Esteban et al., 2012; Montalvo & Reynal-Querol, 2005a). Instead, the polarization index, initially developed by Esteban and Ray (1994) and Wolfson (1994), appears capable of collecting those aspects of ethnolinguistic diversity that can explain civil war and civil conflict: A polarized society with two ethnic groups of the same size shows a low index of fractionalization but is more prone to conflicts than a society with many small ethnic groups, that is, with a high fractionalization index. In short, the choice of the appropriate indicator to measure social diversity, depending on the socio-economic result to be analyzed, appears as a second key task. Furthermore, there are two additional issues that have to be taken into account in the empirical analysis. On the one hand, the level of aggregation selected to delimit the groups determines the value of the diversity index and hence its potential ability to explain socio-economic outcomes (Desmet et al., 2012). On the other hand, following Alesina and La Ferrara (2005) and Montalvo and Reynal-Querol (2020), the level of spatial disaggregation at which the analysis is performed can change the effects of diversity. For instance, the negative effects of ethnolinguistic fractionalization on economic growth found by the literature at country level are less intense in studies across localities, to such an extent that a positive association can be found between ethnolinguistic heterogeneity and economic performance (see, for example, Alesina and La Ferrara (2005)and Sparber (2010) for US cities). 4BUITRAGO and CARABALLO POU Therefore, the way in which social diversity is defined, the appropriate index for its measurement, the level of disaggregation of the groups, and the geographical unit of analysis, all contribute towards the tangle of indices and the various results achieved in the literature on the socio-economic effects of social diversity. This paper is intended to untangle the variety of indices and their applications. We aim to collect these social diversity indices employed in cross-country studies, and to show their similarities and differences, and the main applications of each index. The present paper enables a suitable evaluation and comparison of the effects of social diversity and the selection of the appropriate index depending on the analysis to be carried out. Since this paper is focused on the literature on the socio-economic impacts of social diversity at country level, and in order to achieve our goals, we have chosen those aspects of social diversity that have attracted the most attention in this literature: diversity due to the various dimensions of ethnicity (ethnolinguistic, religious, and birthplace); and genetic and cultural diversity. We have also chosen two main indices to measure diversity: fractionalization and polarization. With respect to socio-economic results, our focus is on economic growth and conflicts. Studies at country level have been selected that construct their own indices and that make their database available. In an increasingly heterogeneous and complex world, the proper assessment of the impacts of diversity and their management have become key issues in guiding and designing policies that make it possible to take advantage of their beneficial effects while mitigating the possibly adverse effects. The paper is organized as follows. Section 2is devoted to the simple fractionalization index, widely employed in the literature. This index considers a discrete distance between groups, that is, whether the individual either belongs to a group or not. Moreover, this section addresses the criteria to characterize a group, and considers classifications based on one dimension (ethnolinguistic, religion, and birthplace) and those based on a set of characteristics (genetic and cultural groupings). However, the simple fractionalization index fails to take into account that there are certain groups that can be closer to each other than can others. In other words, it could be considered that two ethnicities that share a religion and employ a language with the same roots are more similar to each other than two groups with different religions and a totally different alphabet (Fearon, 2003). The measure of a continuous distance as the degree of similarity between groups is analyzed in Section 3. The polarization indices with discrete and continuous distances are described in Section 4. Section 5aims to clarify the main applications at a country level of the indices described in the economic literature, and pays special attention to endogeneity issues. Finally, conclusions are drawn and further lines of research are outlined in Section 6. 2 FRACTIONALIZATION INDICES WITH DISCRETE DISTANCE The fractionalization indices describe the variety in a society. As mentioned in the Introduction, when referring, for example, to ethnicity, this variety is measured as the probability that two individuals, randomly drawn from a population, belong to different ethnicities. In order to obtain the fractionalization indices, let us consider that a population Tcan be divided into several Igroups. Denote tias the population of group i,andπi=ti/T as the share of population of group i. Therefore 1- πiis the probability that an individual of population Tdoes not belong to group i.Tocalculate the probability that two randomly selected individuals do not belong to the same group, we obtain BUITRAGO and CARABALLO POU 5 the weighted average: 𝐹𝑅𝐴𝐶 = 𝐼 ∑ 𝑖=1 (1−𝜋 𝑖)𝜋𝑖=1− 𝐼 ∑ 𝑖=1 𝜋2 𝑖(1) From (1), the usual expression of the fractionalization indices can be obtained, which can also be interpreted as a Herfindahl index. This index is a measure of market concentration. It calibrates the degree of competition between the firms in an industry as the sum of the squares of the market shares of each firm with respect to the industry. The index ranges from 0 (maximum degree of competition) to 1 (monopoly). For the fractionalization index, the market share would be the share of population of a group within the total population, and is defined as one minus the Herfindahl index. This measure therefore varies from 0, which corresponds to the monopoly situation (i.e., there is only one ethnicity in the whole population), to 1, which is the maximum degree of fractionalization. Expression (1) is a measure of diversity that can be applied to any field of study depending on the type of Igroups considered. Broadly speaking, an exogenous and an endogenous way can be regarded to distinguish between the groups. As far as social diversity is concerned, the former could be characterized through ethnicity and then, once the classification of groups is made, individuals are assigned to each of them. In this case, what matters are the variables selected to capture diversity across ethnicities. According to Horowitz (1985), an inclusive interpretation of ethnicity requires linking ethnicity to race, language, religion, or some other dimension of common origin such as place of birth. As Reynal-Querol (2002) points out, it is necessary to analyze which of these attributes are more appropriate in explaining the specific social and/or economic outcome under study, since each dimension can have different effects. In this respect, she considers religion as the main feature causing social conflicts, while Bove and Elia (2017) focus on birthplace as the relevant characteristic to explain differences in economic growth. On the other hand, that which Arbatli et al. (2020) denote as interpersonal population diversity can be used, where groupings are based on a set of individual characteristics. In this vein, Bossert et al. (2011) provide a theoretical foundation for a generalized fractionalization index, of which the ethnolinguistic fractionalization index is a particular case. In order to obtain the index, information regarding several characteristics of the individuals is employed. Among these characteristics, such as income, employment, and education, ethnicity can also be included. Therefore, individuals are not pre-assigned to a precise group, since groups are formed endogenously depending on the characteristic selected. Table 1summarizes the most relevant aspects of the country-level fractionalization indices with discrete distance analyzed: type of diversity, level of disaggregation (number of groups), databases used, year, and number of countries for which they are available. Each aspect is described below, both for the indices that use exogenous groupings based on one dimension, and for those that are based on a set of individual characteristics. 2.1 Grouping based on one dimension The exogenous grouping previously requires the definition of the groups in order to later assign the individuals. This task has to be carried out carefully since the delimitation made in each database 6BUITRAGO and CARABALLO POU TABLE 1 Fractionalization indices with discrete distance FRAC =𝟏− ∑𝑰 𝒊=𝟏𝝅𝟐 𝒊(𝝅𝒊=share of population of group i) Paper Diversity (Type) Groups (Number) Countries Data source Years Taylor and Hudson (1972) Ethnic-linguistic i=1....1600 93 AtlasNarodovMira(1964) Early60s Alesina et al. (2003)Linguistic i=1....1055 201 Encyclopedia Britannica (2001) Early to mid-1990s Ethnic i=1...650 190 EncyclopediaBritannica(2001);CIA(2000); Levinson (1998); Minority Rights Group International (1997); National Census Early to mid-1990s Religious i=1...294 215 (including dependencies) Encyclopedia Britannica (2001) Early to mid-1990s Fearon (2003) Ethnic i=1...822 160 CIA’sWorldFactbook(mainsource).Encyclopedia Britannica; Library of Congress Country Study; Country-specific sources; Minority at Risk; Scarrit and Mozafar (1999) Early 1990s Montalvo and Reynal-Querol (2005a, 2005b) Ethnic-linguistic Vanhaven’s ethnolinguistic families 138 World Christian Encyclopedia (main source). Vanhaven (1999) and World Factbook 1980-1990 Religious i=1...12 138 L’EtatdesReligionsDansleMonde(mainsource). The Statesman’s Yearbook 1970-75-80 Desmet et al. (2012)Linguistic i=1...691215 disaggregation levels 103 Ethnologue Database (2005; 15th edition) 2005 Esteban et al. (2012) Linguistic i=1...6912 137 EthnologueDatabase(2005;15thedition) 2005 Alesina and Zhuravskaya (2011)1 Linguistic i=1....1055 92 Country/regionally-specific sources Close to 2000 Ethnic i=1...650 97 Country/regionally-specificsources Closeto2000 Religious i=1...294 78 Country/regionally-specific sources Close to 2000 (Continues) BUITRAGO and CARABALLO POU 7 TABLE 1 (Continued) FRAC =𝟏− ∑𝑰 𝒊=𝟏𝝅𝟐 𝒊(𝝅𝒊=share of population of group i) Paper Diversity (Type) Groups (Number) Countries Data source Years Desmet Ortuño-Ortín, and Wacziarg (2017) Ethnic-linguistic identity i=(max 20-min 2) 76 Integrated World Values Survey - European Values Survey 1981 to 2008 Alesina, Harnoss, and Rapoport (2016) Birthplace i=1...195 195 Artuc et al. (2015)1990 and 2000 Bove and Elia (2017) Birthplace i=1...230 230 WorldBank(dataonmigrantstocks); Ratha et al. (2016) 1960-70-80-90- 2000-10 Ashraf and Galor (2013a)Genetic i=Variants of a particular gene or DNA locus(averaged over 783 loci) 21 (Observed genetic diversity) 145 (Predicted genetic diversity) Human Genome Diversity Cell Line Panel (HGDPCEPH) (53 ethnic groups/21 countries) Ramachandran et al. (2005); World Migration Matrix (Putterman & Weil, 2010) 1500 and 2000 Desmet, Ortuño-Ortín, and Wacziarg (2017) Cultural attitudes i=Different answers to questions on values 76 Integrated World Values Survey - European Values Survey (WVS-EVS): 175 questions (answers: 49 binary, 26 unordered, 100 scale) 1981 to 2008 Notes:1National fractionalization is calculated by aggregating the regional fractionalization indices. 8BUITRAGO and CARABALLO POU and the level of aggregation chosen may have significant implications. Let us consider the example of religion as a dimension of social diversity. If we deem the group “Christians” as unique, we will be including Catholics, Anglicans, and the various branches of Protestantism, which would hide the relevant religious diversity in regions such as Northern Ireland. The same argument can be applied to languages and ethnic groups. Likewise, another problem to consider is that the definition of groups and the self-identification of individuals with the groups change over time as a function, among others, of economic, social, or political factors (this issue is discussed in Section 5). Moreover, the dimension chosen to identify social groups also has important implications, as discussed below. We have selected three dimensions of ethnicity that are widely studied in the literature (ethnolinguistic, religious, and birthplace) and focus especially on the first dimension since it has attracted special attention among researchers. 2.1.1 Ethnolinguistic groups In this case, we consider those indices that use language and/or racial dimensions to delimit ethnicity. As shown in Table 1, we can distinguish between ethnic-linguistic groups (Desmet, Ortuño- Ortín, & Wacziarg., 2017; Montalvo & Reynal-Querol, 2005a, 2005b; Taylor & Hudson, 1972), ethnic groups (Alesina & Zhuravskaya, 2011; Alesina et al., 2003;Fearon,2003), and linguistic groups (Alesina & Zhuravskaya, 2011; Alesina et al., 2003;Desmetetal.,2012; Esteban et al., 2012).iii One of the first and most widely used measurements of ethnolinguistic diversity was provided by Taylor and Hudson (1972). They computed an ethnic-linguistic fractionalization index (ELF), which employed the Narodov Mira Atlas (1964), in which the delimitation of ethnicity is based on language as the main differentiating criterion. This criterion has been widely criticized since language is only one of the many features that anthropologists and ethnologists employ to define ethnicity. Alesina et al. (2003) exemplify certain problems of ELF by showing the existence of groups that speak the same language and have different racial origins or different skin color (e.g., in the US, if only language is taken into account, then black people and white people could be in the same group, that is, English speakers). Alesina and his co-authors constructed two indices to overcome these shortcomings. On the one hand, they computed one based exclusively on language using the Encyclopedia Britannica. On the other hand, they calculated a broad ethnicity index, combining linguistic aspects with racial aspects of physical characteristics. To this end, they employed several primary databases and carried out in-depth fieldwork in order to determine, in each case, which delimiter element was the most relevant. Fearon (2003) developed similar work but based it on different databases, which allow a higher level of disaggregation of ethnic groups (822 vs. 650). In contrast, Montalvo and Reynal-Querol (2005a, 2005b) computed a new index by employing the World Christian Encyclopedia (WCE) and by disaggregating only the ethnolinguistic families of Vanhanen (1999) (see Table 1). As Montalvo and Reynal-Querol (2005b) point out, they follow Vanhaven (1999) in order to identify the relevant level of disaggregation, while Alesina et al. (2003) strive to capture the most disaggregated level (for a thorough review of the advantages and disadvantages of each database, see Montalvo and Reynal-Querol (2005a, 2005b). Regardless of the database used, the correlation between the indices computed by Alesina et al. (2003) and Montalvo and Reynal-Querol (2005a, 2005b) is high (over 0.7 for our sample, see Table A4, Appendix I). A debate on the implications of the level of aggregation or degree of coarseness of ethnolinguistic classifications was opened from the aforementioned research. In this respect, the work carried out by Desmet et al. (2012) was conclusive. They computed linguistic diversity measures at vari- BUITRAGO and CARABALLO POU 9 ous levels of aggregation, by exploiting the information of language trees (15 levels of aggregation are available in Ethnologue, with level 15 being integrated by 6912 languages). Their results confirm that the level of aggregation is an important factor that also enables the historical dimension to be introduced. High levels of aggregation describe cleavages that go back thousands of years, while finer divisions are the result of cleavages of a more recent nature. In fact, the changes in the level of aggregation markedly affect the value of the diversity indices. For our sample, the correlation between the index with the highest and lowest level of aggregation is only 0.43, and even lower (below 0.33) on examining the correlation between the former and the indices proposed by Alesina et al. (2003) and Montalvo and Reynal-Querol (2005a, 2005b). In a later work, Desmet, Ortuño-Ortín and Wacziarg (2017) calculate an ethnolinguistic fractionalization index supported by self-reported ethnic and/or linguistic identity by individuals, based on data from the integrated World Values Survey—European Values Survey (1981–2008). The groups were defined according to race and/or self-reported language based on the previous literature and on the territory considered. Thus, most ethnographers agree that language is decisive in determining groups in many African countries, while race is a more decisive factor in Latin America.iv More recently, with the aim of introducing possible long-term changes in the ethnic composition of countries, Dražanová (2020) builds the Historical Index of Ethnic Fractionalization dataset. This dataset provides annual ethic fractionalization estimates for 162 countries for the years 1945−2013, based on the Cline Center for Democracy Composition of Religious and Ethnic Groups. 2.1.2 Religious groups Religious diversity has also been measured as an aspect that, in certain cases, constitutes an element of differentiation between ethnicities (Alesina & Zhuravskaya, 2011; Alesina et al., 2003; Montalvo & Reynal-Querol, 2005a, 2005b). According to Montalvo and Reynal-Querol (2005b), the main sources for the classification of religious groups are Barret’s (1982) World Christian Encyclopedia, Encyclopedia Britannica, and, in particular, the Britannica World Data, L’Etat des Religions Dans le Monde, The Statesman’s Yearbook, and The World Factbook. These differ mainly in their inclusion of Animist and Syncretic cults. Two of the most widely used religious fractionalization indices are those of Alesina et al. (2003) and of Montalvo and Reynal-Querol (2005a, 2005b). The former construct their index using the information from the Encyclopedia Britannica, and the latter construct their index using L’Etat des Religions Dans le Monde as a primary source and The Statesman’s Yearbook as a secondary source (Table 1). Moreover, Alesina and Zhuravskaya (2011) compute the fractionalization index for ethnic, linguistic, and religious groups for both national and subnational levels. From the comparison of the two indices, the relationship between ethnicity and geographical distribution can be addressed, in an effort to analyze whether the different groups are located in separate territories. This issue can be measured by the segregation indices (see Reardon and Firebaugh (2002) for a review). As Alesina and Zhuravskaya (2011) note, the cross-country literature barely touches on this issue due to the scarcity of available data. These authors calculate a segregation index at a national level, by employing one of the six indices reported by Reardon and Firebaugh (2002). The index is based on the difference between fractionalization indices at regional and national levels. By denoting m=1...Mas the number of regions, tmas the population of region min the country, and πmi as the share of group iin total population of region m, the segregation index employed by 16 BUITRAGO and CARABALLO POU the other hand, this antagonism can be enhanced by a feeling of identification of the individual within the group of individuals of similar characteristics. These authors introduce the identification function I=I(πi), where the higher the number of members there are in the group, the greater the identification that exists between them. The interpersonal antagonism EAij is therefore the joint result of the sense of identification of individual iwith their own group, which depends on the size of group i, and the feeling of alienation towards members of group jas captured by the distance dij: 𝐸𝐴ij =𝑇 (𝐼(𝜋𝑖),𝑎(𝑑ij)) (7) Therefore, the societal antagonism is the sum of all effective antagonism: 𝑆𝐴 = 𝑛 ∑ 𝑖=1 𝑛 ∑ 𝑗=1 𝜋𝑖𝜋𝑗𝑇(𝐼(𝜋𝑖),𝑎(𝑑𝑖𝑗)) (8) A particular measure of polarization is yielded once a specific functional form for T(⋅), I(⋅), and a(⋅) is selected. Esteban and Ray (1994) derive the concept of polarization of a distribution from three axioms and obtain the following measure of polarization: 𝑃𝑂𝐿 = 𝛽 𝐼 ∑ 𝑖=1 ∑ 𝑗≠𝑖 𝜋1+𝛼 𝑖𝜋𝑗𝑑𝑖,𝑗 (9) Expression (9) is a specific measure of societal antagonism. The identification function is I=I(πi)=πiα, where parameter αis defined as the degree of “polarization sensitivity”. This is a positive parameter that should lie between zero and an optimum value, which Esteban and Ray (1994) show to belong to the interval (0,1.6]. The more αincreases, the greater the importance given to identification between individuals. This parameter αis what makes the polarization measure significantly different from fractionalization. βis a positive parameter used for population normalization. The application of this concept of polarization to dimensions such as religion or ethnicity is not straightforward. As pointed out by Montalvo and Reynal-Querol (2008), the polarization index developed by Esteban and Ray (1994) measures distance between groups using a continuous dimension such as income. However, when considering religious, linguistic, or ethnical aspects, the way in which the distance across groups can be measured remains unclear. On the one hand, this distance can be discretized, as shown in Section 4.1. On the other hand, the concept of distance between groups described in Section 3has also been used for polarization indices, as discussed in Section 4.2. Furthermore, Table 3summarizes the main features of the country-level polarization indices with discrete and continuous distance. 4.1 Polarization indices with discrete distance Following Montalvo and Reynal-Querol (2008), several reasons justify the use of discrete distance to construct ethnolinguistic and religious polarization indices. On the one hand, no measures are yet available and generally accepted regarding distance across ethnolinguistic and religious groups. On the other hand, the attempt to measure distances may give rise to larger measurement BUITRAGO and CARABALLO POU 17 TABLE 3 Polarization index Polarization index with discrete distance POLRQ =𝟒∑𝑰 𝒊=𝟏∑𝒋≠𝒊𝝅𝟐 𝒊𝝅𝒋(𝝅𝒊and 𝝅𝒋=population share of groups 𝒊and 𝒋) Paper Diversity (Type) Groups (No.) Countries Data source Years Montalvo and Reynal-Querol (2005a, 2005b) Ethnic-linguistic Vanhaven’s ehnolinguistic families 138 World Christian Encyclopedia (main source). Vanhaven (1999)and World Factbook 1970-75-80 Religious i=1...12 138 L’Etat des Religions Dans le Monde (main source). The Statesman’s Yearbook 1970-75-80 Desmet et al. (2012) Linguistic i=1...6912 15 disaggregation levels (L1...L15) 103 Ethnologue Database (2005; 15th edition) 2005 Esteban et al. (2012)Ethnic i=1...822 138 Fearon (2003)Early 1990s Esteban et al. (2012) Linguistic i=1...6912 137 EthnologueDatabase(2005;15th edition) 2005 Desmet, Ortuño-Ortín, and Wacziarg (2017) Ethnic-linguistic identity i=(max 20-min 2) 76 Integrated World Values Survey–European Values Survey (WVS-EVS) 1981 to 2008 Bove and Elia (2017) Birthplace i=1...230 230 WorldBank(dataonmigrant stocks) Ratha et al. (2016) 1960-70-80- 90-2000-10 Desmet, Ortuño-Ortín, and Wacziarg (2017) Cultural attitudes i=Answers to questions on values 76 Integrated World Values Survey–European Values Survey (WVS-EVS): 175 questions (answers: 49 binary, 26 unordered, 100 scale) 1981 to 2008 (Continues) 18 BUITRAGO and CARABALLO POU TABLE 3 (Continued) Polarization index with discrete distance POLRQ =𝟒∑𝑰 𝒊=𝟏∑𝒋≠𝒊𝝅𝟐 𝒊𝝅𝒋(𝝅𝒊and 𝝅𝒋=population share of groups 𝒊and 𝒋) Paper Diversity (Type) Groups (No.) Countries Data source Years Polarization index with continuous distance 𝐶𝐷𝑃𝑂𝐿 = 𝐼 ∑ 𝑖=1 ∑ 𝑗≠𝑖 𝜋2 𝑖𝜋𝑗𝑑𝑖,𝑗 𝑑𝑖𝑗 =1−( 𝑐𝑏 𝑚)𝛿(dij =distance i-j; cb =common branch; δ=parameter relevance similarities) Esteban et al. (2012)Ethnic (dij =Distance linguistic; m=15; δ=0.5 and 0.05) iand j=1...822 138 Diversity: Fearon (2003) Correction: Ethnologue Database (2005; 15th edition) Early 1990s Linguistic (dij =Distance linguistic; m=15; δ=0.5 and 0.05) iand j=1...6912 137 DiversityandCorrection: Ethnologue Database (2005; 15th edition) 2005 BUITRAGO and CARABALLO POU 19 errors than those of the discrete measures. In fact, as pointed out by Duclos et al. (2004), for certain dimensions, such as religion (Reynal-Querol, 2002), the simple criterion belong/not belong can accurately reflect the interest of the individuals (the same reasons could be applied analogously for fractionalization indices considered in Section 2). Moreover, Montalvo and Reynal-Querol (2008) stress the endogeneity problems that can arise when, for example, the strength of the sentiment of identity is employed to measure the distance between groups, since it may be the conflict itself that is causing that sentiment of identity. Therefore, there are arguments in favor of using the index of polarization with discrete distance. In this vein, Reynal-Querol (2002) proposed the following indexxii: 𝑃𝑂𝐿𝑅𝑄 = 1 − 𝑁 ∑ 𝑖=1 (0.5 − 𝜋𝑖)2𝜋𝑖 0.25 (10) which achieves its maximum value when there are two groups of the same size, as in Esteban and Ray (1994) and Wolfson (1994), and fulfills the usual properties of these kinds of indices (see Reynal-Querol (2002, p. 53), for details). Moreover, we can relate the polarization index with discrete distance with the polarization measure in (9). If we consider a discrete distance in the polarization measure of Esteban and Ray (1994), then dij =0if i=j,anddij =1if i≠j, and therefore, from expression (9), we obtain: 𝐷𝑃𝑂𝐿 = 𝛽 𝐼 ∑ 𝑖=1 ∑ 𝑗≠𝑖 𝜋1+𝛼 𝑖𝜋𝑗(11) Montalvo and Reynal-Querol (2008, p. 1841) show that the only value of αfor which DPOL fulfils the basic properties of a discrete polarization index is 1, and βshould be equal to 4 for DPOL to lie between 0 and 1, which gives the POLRQ index in expression (10): 𝑃𝑂𝐿𝑅𝑄 = 1 − 𝑁 ∑ 𝑖=1 (0.5 − 𝜋𝑖)2𝜋𝑖 0.25 =1− 𝑁 ∑ 𝑖=1 (0.25 + 𝜋2 𝑖−𝜋 𝑖)(𝜋𝑖 0.25 ) =1− 𝑁 ∑ 𝑖=1 𝜋𝑖+4 𝑁 ∑ 𝑖=1 𝜋2 𝑖(1−𝜋 𝑖)=4 𝐼 ∑ 𝑖=1 ∑ 𝑗≠𝑖 𝜋2 𝑖𝜋𝑗 (12) It is worth remarking that this does not signify that this is the only discrete polarization measure. In other words, what Montalvo and Querol (2008) are showing is that this precise index is one of those that fulfills the properties of a polarization index. Furthermore, if β=1andα=0inDPOL is considered, then the discrete fractionalization index (FRAC in expression 1)xiii is obtained. As can be observed, FRAC increases with the number of groups, while the maximum for POLRQ is attained when there are two groups. In fact, in the case of two groups, both indices FRAC and POLRQ attain the same value, but this is no longer true when the number of groups increases. This is due to the fact that, for POLRQ, the probabilities of two individuals belonging to different groups are weighted by the relative size of each group, while for FRAC they are not. Furthermore, the FRAC index does not account for the effect of group size on the sense of identity (α=0), while the POLRQ does, by assuming that identification increases with the size of the group. Several authors have calculated various polarization indices at country level by employing different databases and different delimitations of ethnicity (see Table 3). Thus, Montalvo and 20 BUITRAGO and CARABALLO POU Reynal-Querol (2005a, 2005b) provide an ethnolinguistic polarization index by using the ethnolinguistic families described in Vanhanen (1999), and a religious polarization index by using the L’Etat des Religions Dans le Monde as their main source. Desmet, Ortuño-Ortín, and Wacziarg (2017) calculate a polarization index using the self-reported ethnic and/or linguistic identity by individuals based on surveys on values. As for the fractionalization index, Desmet et al. (2012) compute linguistic polarization measures at different levels of aggregation. Furthermore, Bove and Elia (2017) also obtain a birthplace polarization index. Empirically, Montalvo and Reynal-Querol (2005a, 2005b) show the relationship between POLRQ and FRAC for their own data. They obtain a positive correlation for low levels of fractionalization, zero for the medium range, and negative correlation for high levels of fractionalization. Moreover, Desmet et al. (2012) show that the correlation between FRAC and POLRQ, at the same levels of aggregation, decreases as the level of disaggregation increases. Thus, for our sample, the correlation between the two indices rises from 0.43, at the highest level of disaggregation, to 0.78 at a medium level, and reaches a value of 0.98 at the lowest level. The intuition behind these results is that, for lower disaggregation, fewer groups remain, and, therefore, the distinction between the two indices vanishes. When the levels of disaggregation differ, the correlation is even lower. For instance, between FRAC and POLRQ for the lowest (highest) and highest (lowest) level of disaggregation respectively, the correlation is 0.3 (0.43) (see Table A8, Appendix I). 4.2 Polarization indices with continuous distance BasedonEstebanandRay(1994) and Duclos et al. (2004), Esteban and Ray (2011) derive a polarization measure with continuous distance for discrete distributions: 𝐶𝐷𝑃𝑂𝐿 = 𝐼 ∑ 𝑖=1 ∑ 𝑗≠𝑖 𝜋2 𝑖𝜋𝑗𝑑𝑖,𝑗 (13) Using this index, Esteban et al. (2012) obtain an ethnic polarization index that utilizes the grouping of Fearon (2003) and a linguistic polarization index from the Ethnologue Database (see Table 3). From the above expression, they calculate both ethnic and linguistic polarization indices for δ=0.5 and δ=0.05 in di,j (see expression (5)). Moreover, Esteban and Mayoral (2011) construct both ethnic and religious polarization indices withcontinuousdistancesmeasured fromtheintensityinthe ethnic and religiousattitudes,which is derived from the answers to a set of questions that refer to religious or ethnic feelings collected by the World Values Survey.xiv This way of measuring distance enables the capture of the fact that a significant number of individuals in a group may feel no antagonism towards individuals belonging to another group, and hence belonging to different groups generates no social polarization. Therefore, in this case, a polarization index with discrete distance or with a continuous distance in the spirit of Fearon (2003), instead of that proposed by Esteban and Mayoral (2011), can reflect antagonism that is not actually taking place in society. Finally, it is also worth noting that, using the polarization index in expression (9), Desmet, Ortuño-Ortín, and Weber (2017) propose the peripheral heterogeneity index to analyze the ethnolinguistic conflicts arising from the existence of a central dominant group and several minority groups. Considering that there are i=1...Iminority groups and a dominant group denoted by 0, BUITRAGO and CARABALLO POU 21 the peripheral heterogeneity index can be defined as: 𝑃𝐻 = 𝐼 ∑ 𝑖=1 𝜋1+𝛼 𝑖𝑑0𝑖 + 𝐼 ∑ 𝑖=1 𝜋𝑖𝜋1+𝛼 0𝑑0𝑖 (14) where d0i is the distance between the dominant group and each minority group, which Desmet, Ortuño-Ortín, and Weber (2017) compute using the Dyen et al. (1992) distance matrix. In expression (14), the first addition is the effective antagonism from the minorities to the dominant group and the second addition is the effective antagonism from the dominant group to the minorities. From the set of axioms established by Desmet, Ortuño-Ortín, and Weber (2017), they derive that αcan be either positive or negative. When αis positive, the larger the group, the greater the sense of identification. However, as Desmet, Ortuño-Ortín, and Weber (2017) point out, the reverse could also be true, that is, the larger the group, the smaller the sense of identification and hence αbecomes negative. They show that for negative values of α, the PH indexisameasureof variety (fractionalization) and for positive values, it can be interpreted as a polarization measure. 5SOCIAL DIVERSITY AND SOCIO-ECONOMIC OUTCOMES In addition to constructing the social diversity indices described in previous sections, most of the papers cited employ them to carry out empirical analyses with the aim of assessing the effects of diversity on social and economic outcomes, especially economic performance and conflicts. Tables 4and 5summarize the main results obtained in the literature on the relationship between socialdiversity and these outcomes (Tables A9 andA10 in Appendix Iinclude detailed information on the variables used in the empirical analysis). As can be observed from these tables, the social and economic implications differ in certain cases depending on the indices. These can be due to any of the several reasons discussed in greater detail below: 1. The different ways of grouping. On the one hand, as addressed in Section 2, distinction is made between exogenous and endogenous groups, which itself has consequences since the former only capture inter-group differences while the latter also capture intra-group diversity (Arbatli et al., 2020). On the other hand, each dimension chosen to delimit the groups may have different consequences. For instance, religious and ethnolinguistic fractionalization exert diverse effects on growth. 2. The level of disaggregation when considering the definition of the groups. According to Desmet et al. (2012), the level of aggregation of the groups is a significant factor which allows the historical dimension to be introduced. 3. The level of spatial disaggregation for which the analysis is made. Although this paper is focused on contributions at country level, differences in the results are also considered when a more disaggregated geographical level is observed. 4. The selection of the accurate index, fractionalization versus polarization, to explain certain issues, such as conflicts. 22 BUITRAGO and CARABALLO POU TABLE 4 Applications of fractionalization indices Indices Diversity Growth and Income Conflicts Fractionalization with discrete distance Taylor and Hudson (1972)1Ethnic-linguistic (- Sig.): GDPpc Growth Alesina et al. (2003)Linguistic (- Sig.): GDPpc Growth Ethnic (- Sig.): GDPpc Growth Religious (No correlation) Fearon (2003)2Ethnic (+Sig.): Conflict Variables Montalvo and Reynal-Querol (2005a, 2005b) EthniclinguisticReligious (- Sig): GDPpc Growth(No correlation) No correlation: Civil Wars(- NS): Civil Wars Desmet et al. (2012) Linguistic (- Sig.): Growth (L6 to L15) (+Sig.): Onset Civil Wars (L1) (NS): Onset Civil Wars(L6 to L15) Esteban et al. (2012)Linguistic (+NS): Conflict Variables Alesina and Zhuravskaya (2011) Linguistic Ethnic Desmet, Ortuño-Ortín, and Wacziarg (2017) Ethnic-linguistic identity (+NS): GDPpc (- NS): Civil Conflict Alesina, Harnoss, and Rapoport (2016) Birthplace (+Sig.): GDPpc(inverse U-Shaped) Bove and Elia (2017) Birthplace (+Sig.): GDPpc Ashraf and Galor (2013a)3Genetic (+Sig.): GDPpc(inverse U-Shaped) (+Sig.): Civil Conflict Desmet, Ortuño-Ortín, and Wacziarg (2017) Cultural attitudes (+Sig.): GDPpc (- Sig.): Civil Conflict Fractionalization with continuous distance Fearon (2003)2 (Linguistic distance δ =0.5) Ethnic (- Sig.): Conflict Variables Esteban et al. (2012)(Linguistic distance δ =0.05) Ethnic (- Sig.): Conflict Variables Esteban et al. (2012)(Linguistic distance δ =0.5 and δ=0.05) Linguistic (- NS): Conflict Variables Alesina, Harnoss, and Rapoport (2016)(Genetic and GDP distances) Birthplace (+Sig.): GDPpc Desmet, Ortuño-Ortín, and Wacziarg (2017)(Distance values) Cultural attitudes (- Sig.): Civil Conflict Notes:(1) The results for Growth were obtained by Easterly and Levine (1997) and those for Corruption by Mauro (1995). Both studies used the Taylor and Hudson (1972)index.(2) The results were obtained by Esteban et al. (2012)usingtheFearon(2003) index. (3) The results for conflicts were obtained by Arbatli et al. (2020) using the Ashraf and Galor (2013a)index. (+Sig.): positive and significant results, (-Sig.): negative and significant results, (NS): not significant, (Mix): mixed results, (No correlation): coefficients are positive and negative in different specifications and non-significant in all cases. BUITRAGO and CARABALLO POU 23 TABLE 5 Applications of polarization indices Indices Diversity Growth and Income Conflicts Polarization with discrete distance Montalvo and Reynal-Querol (2005a, 2005b) Ethnic-linguistic Religious (+NS): GDPpc Growth (+NS): GDPpc Growth (+Sig.): Civil Wars (+Sig.): Civil Wars Desmet et al. (2012)Linguistic (- NS): Growth (L1 to L15) (+Sig.): Onset Civil Wars (L1) (- NS): Onset Civil Wars (L6 to L15) Esteban et al. (2012)Ethnic (+Sig.): Conflict Variables Linguistic (NS): Conflict Variables Desmet, Ortuño-Ortín, and Wacziarg (2017) Ethnic-linguistic identity (- NS): Civil Conflict Bove and Elia (2017) Birthplace (+Sig.): GDPpc Desmet, Ortuño-Ortín, and Wacziarg (2017) Cultural attitudes (- Sig.): Civil Conflict Polarization with continuous distance Esteban et al. (2012)(Linguistic distance.δ=0.5 and δ=0.5) Ethnic Linguistic Notes:(+Sig.): positive and significant results, (-Sig.): negative and significant results, (NS): not significant, (Mix): mixed results. Moreover, the empirical literature on social diversity faces several problems of endogeneity. With regard to grouping based on one dimension as described in Section 2.1, two main problems arise with respect to ethnolinguistic groups. On the one hand, ethnolinguistic diversity evolves endogenously in the very long-term for a variety of reasons. First, high ethnic diversity in developing countries could be the result of a legacy of colonialism and the artificial mapping of borders. In this vein, Leeson (2005) concludes that colonial policy in Africa was often purposely designed to increase ethnic heterogeneity while Ahlerup and Olson (2012) find that the length of the colonial period is positively associated with ethnic fractionalization. They argue that this result could be indicative of the consequences of the colonial policies of “divide-and-rule”, which was deliberately implemented to maintain control of colonies. Second, these authors find that the timing of the initial settlement by modern humans constitutes a key factor in explaining much of the current ethnic diversity. Third, Ahlerup and Olson (2012) also point out the relevance of the history of the modern state. They argue that the length of statehood is negatively associated to ethnic diversity since states might have incentives to actively design policies aimed at achieving populations of a more homogenous nature. If this argument is followed while using the present national borders as the units within which ethnic diversity is measured, then certain endogeneity problems may arise since we would be mixing, for instance, African countries that have gained their independence since 1950, with Western European and North American countries with a much longer history. If this is the case, then the former could be more ethnically fragmented because these nations have had very little time to impose homogenizing policies as did Western European and North American countries back in the 1800s. Furthermore, the wars and genocides that have been perpetrated against certain ethnic groups throughout the process of consolidation of nation states should also be borne in mind. In fact, there could be countries that might seem ethnically more homogeneous 24 BUITRAGO and CARABALLO POU today (consider, for instance, Australia with a fractionalization index of 0.09 (Alesina et al., 2003) due to the genocide against the natives at the time of conquest). Finally, Michalopoulos (2012) stresses the role of geographic variability, captured by the variation in the terrain and differences in land endowments in explaining the changes of ethnolinguisticdiversityovertime On the other hand, changes in social, economic, and political variables can, in turn, generate changes in the definitions of ethnolinguistic groups and the self-identification of individuals with these groups. Alesina et al. (2003) cite Somalia as an example of this latter concern. Before the 1991 civil war, this country could be considered as ethnically homogeneous, with 85% of the population identified as Somalis. Nevertheless, during and after the civil war, Somalia has become in a highly fragmented society since the population no longer self-identifies with the Somali ethnicity but instead with the multiple clans that existed prior to the war. However, despite these two problems, the literature focused on the social and economic outcomes of ethnolinguistic fractionalization takes fractionalization indices as exogenous in crosscountry regressions (Alesina et al., 2003; Bluhm & Thomsson, 2020;Desmetetal.,2012; Montalvo & Reynal-Querol, 2005a, 2005b). As Alesina et al. (2003, p. 160) argue, this statement is based on the assumption that ethnolinguistic diversity is historically determined before any of the outcomes intended to be explained arise. Thus, they deem it reasonable to assume that ethnolinguistic diversity is persistent in a 30-year horizon, which is the usual horizon in cross-country analysisxv. Nonetheless, this assumption might not hold for religious diversity indices since this type of diversity might apparently change in a shorter horizon. For instance, the transition from a democracy to an authoritarian regime could lead religious diversity to a spurious decrease since it would be impossible to count individuals belonging to non-authorized religions. Likewise, it has to be taken into account that religion can be considered as a dynamic process. In fact, the ways of interpreting and practicing by the religious groups may change over time, as well as the individual’s sense of belonging to a precise group. Therefore, whenever data became available, the accurate measurement of religious diversity would require dynamic indices. Furthermore, when focusing on conflicts, another problem of endogeneity arises since the antagonism captured by the polarization index can be due to previous conflicts. This issue is addressed in Section 5.2. As far as birthplace diversity is concerned, this is a more frequently changing index due to the increasing migratory flows. In this case, the endogeneity problems cannot be left aside. The positive relationship with economic performance found by Alesina, Harnoss, and Rapoport, (2016) and Bove and Elia (2017) can be due to the good performance of the high-income destination that encourages immigration and not the other way around. To tackle these issues, both papers use an instrumental variable approach. These studies specify a gravity model to predict the size and diversity of a country’s immigration by employing bilateral geographic/cultural variables. They construct instruments for birthplace diversity using the predicted bilateral migration shares estimated from gravity models. The robustness of their initial findings is subsequently confirmed in IV-regressions. With respect to groupings based on a set of individual characteristics, endogeneity problems have also been taken into account. In fact, potential endogeneity is a problem that has repeatedly preoccupied researchers dealing with the effects of genetic diversity on both economic development (Ashraf & Galor, 2013a; Depetris-Chauvin & Özak, 2020) and conflict (Arbatli et al., 2020). Genetic diversity could reflect historical processes such as interregional migrations that were, in turn, determined by historical patterns of comparative development. Thus, in the course of human history, both economic growth differences and conflicts have plausibly altered the observed levels BUITRAGO and CARABALLO POU 25 of genetic diversity. In surmounting these potential issues of endogeneity, most of the research on this topic has implemented one of several empirical strategies based mainly on the Out-of-Africa hypothesis and on the serial founder effect, as analyzed in Section 2.2. On the one hand, the predicted genetic diversity (precolonial and contemporary) is estimated by exploiting the explanatory power of migratory distance from Africa for observed genetic diversity. Since the migratory distance of a country’s prehistorically indigenous population from East Africa has no direct effect on current development or contemporary conflicts, potential endogeneity problems are mitigated. On the other hand, once the exogeneity of the migratory distance has been tested, this variable is used as an instrument in the estimates of the relationships between genetic diversity and economic development or conflicts. Moreover, research into the effects of genetic diversity on the contemporary era addresses the possible endogeneity bias that can generate the impact of great cross-country migrations in the post-1500 era. In order to tackle this problem, empirical investigations carry out regressions for restricted samples, from which those countries that can cause disturbances are eliminated. The models for panels centered on a subsample of countries in the Old World are thereby estimated, since while post-1500 population flows exerted a marked effect on the genetic diversity of national populations in the Americas and Oceania, the diversity of populations in Africa, Europe, and Asia remained largely unaltered. These strategies towards addressing potential endogeneity have also been employed in the study of the relationships between genetic diversity and other measures of fragmentation (Ashraf & Galor, 2013b). On the other hand, Desmet, Ortuño-Ortín, and Wacziarg (2017) are also aware of a potential problem of endogeneity, whereby the cultural diversity index may change when the population suffers civil conflicts since these can lead people to change their values, beliefs, and preferences, and, therefore, to modify their responses to the questions from the World Values Survey. In order to address this endogeneity concern, they rerun the regressions while performing two strategies. First, they include cultural indices that select only the questions with a higher degree of persistence, as these tend to be less influenced by external events such as conflicts. Second, they limit the sample, on the one hand, to the post-1970 period and, on the other hand, to respondents born before 1950 and to the post-1970 sample. In the two cases, if endogeneity were an issue, then there would be major changes in the results of the estimations with the restricted samples in comparison with those of the full sample. The results obtained after applying these approaches remain similar to those of the previous approaches. The most relevant economic and social results that can be explained through the social diversity indicators included in this work can now be discussed. Our focus is mainly on two outcomes that have attracted the most attention in the empirical literature on social diversity: economic growth and conflicts. 5.1 Diversity, economic growth, and income A first look at the descriptive statistics and simple correlations for our sample can help illustrate the relationship between diversity and income. For ethnolinguistic fractionalization and segregation, the mean of the indices decreases from low-income countries to high-income countries, while the correlation with the income per capita remains negative. More precisely, it is above 0.33 in absolute values for the indices computed by Alesina et al. (2003), Montalvo and Reynal-Querol (2005a, 2005b), and Desmet et al. (2012) for the highest level of disaggregation and also for ethnic and linguistic segregation indices constructed by Alesina and Zhuravskaya (2011)(seeTables A2 32 BUITRAGO and CARABALLO POU 2005b), and they become positive and significant when considering birthplace diversity (Alesina, Harnoss, & Rapoport, 2016; Bove & Elia, 2017). In the case of groupings based on individual characteristics (genetic and cultural diversity), the results are different from those previous since these indices capture, in addition to the diversity inter-groups, the diversity intra-groups. In this respect, research on genetic diversity documents a persistent hump-shaped effect on comparative economic development (Ashraf & Galor, 2013a; Ashraf et al., 2014), while cultural diversity is positively associated to economic performance (Desmet, Ortuño-Ortín, & Wacziarg, 2017). The way of grouping is also important in explaining the different results concerning the effects of diversity on conflicts. In this case, while the studies that use fractionalization indices based on one dimension (ethnolinguistic, religion, and birthplace) find no significant results (Desmet et al., 2012; Desmet, Ortuño-Ortín, & Weber, 2017; Esteban et al., 2012; Montalvo & Reynal-Querol, 2005b), those that are based on genetic diversity report positive and significant causal results on conflicts (Arbatli et al., 2020). (2) The results of the reviewed literature differ when fractionalization or polarization indices are considered, for both economic performance and conflicts (Esteban et al., 2012;Montalvo& Reynal-Querol, 2005b). Although most of the studies find that the relationship between ethnolinguistic diversity and economic growth is negative and significant when diversity is measured with fractionalization indices, this relationship is no longer significant when polarization indices are used. In contrast, the relationship between ethnolinguistic diversity and conflicts is significant only for polarization indices. As has been discussed throughout the paper, these results are due to the fact that each type of index enables the capture of different dimensions of diversity. (3) Regarding disaggregation, on the one hand, we can consider the level of disaggregation when defining the groups. In this respect, the findings of Desmet et al. (2012) are conclusive: the relationship between diversity and growth is only significant at the most disaggregated levels of the index, while the relationship between diversity and conflicts is only significant for the most aggregated levels. The most aggregated levels capture deep cleavages, originating thousands of years ago, which lead to better predictors of conflict. In contrast, growth is more influenced by finer divisions that condition the ability to coordinate. On the other hand, the level of spatial disaggregation at which the analysis is performed also appears as a major issue. As shown by Montalvo and Reynal-Querol (2020), the positive effects of social diversity on economic growth through productivity dominate at higher levels of spatial disaggregation, where the issues related to conflicts on the provision of public goods and the low quality of institutions are less significant. This explains why studies across localities find a positive association between ethnolinguistic heterogeneity and economic performance, while, in most of the studies at country level reviewed in the present paper, the negative effects of ethnolinguistic diversity on economic growth prevail. The work carried out in this paper contributes to the research on the effects of social diversity. In an increasingly heterogeneous world, where xenophobic attitudes are spreading, this research has become highly relevant. It is necessary to precisely ascertain the consequences of diversity and the mechanisms that can take full advantage of their effects. This task will provide a suitable guide to social and economic policies. In further research, it is worth continuing research into ascertaining which dimensions of social diversity do indeed generate antagonism between individuals. Diversity is necessary and will always exist, and hence specifying which of its dimensions lead to that feeling of antagonism that may trigger negative socio-economic effects constitutes a relevant and challenging question, since those dimensions that arouse rejection among individuals change both over time and across geographical areas. In this respect, the overlapping indices employed by Desmet, Ortuño-Ortín, BUITRAGO and CARABALLO POU 33 and Wacziarg (2017) to analyze the relation between ethnicities and culture might well be inspiring because they can be applied to other crucial dimensions in today’s societies, such as the level of education achieved, sexual orientation, and political ideology. Moreover, the ability of each society to manage diversity is also crucial for the positive effects to outperform the negative effects. Van Staveren and Pervaiz (2017) stress social cohesion as a tool to handle diversity successfully. In this vein, Buitrago et al. (2019) highlight tolerance as a key issue in achieving social cohesion in an ethnically diverse society, since it contributes to the integration of all members of a community, which helps overcome frictions caused by diversity. Likewise, Caraballo and Buitrago (2019) point out the decisive role of education in promoting social cohesion since it facilitates social networks and fosters tolerance and cooperativeness. In fact, they show that education enhances the positive effects of diversity and calculate the level of education required to reverse the possible adverse effects of ethnolinguistic diversity on income. Future research should therefore be focused on a deeper exploration of these and other possible mechanisms, with the aim of guiding the implementation of educational and socio-political institutions towards successfully addressing diversity. ORCID EvaM.Buitrago https://orcid.org/0000-0003-4113-5836 M.ÁngelesCaraballo https://orcid.org/0000-0002-8753-1562 ENDNOTES iThe term “social” diversity is used instead of “population” diversity in order to leave aside the diversity between individuals due to purely economic reasons. ii Alesina and La Ferrara (2005) focus on ethnolinguistic diversity as a dimension of social diversity. However, their arguments can be applied in general to the broader concept of social diversity. iii In order to avoid confusion, we use the terms ethnic, ethnic-linguistic, and linguistic to classify diversity depending on whether the criteria to delimit ethnicity are based, respectively, mainly on racial aspects of physical characteristics, mainly on language, or only on language. When we employ ethnolinguistic diversity, we refer to any (or all) of these categories. iv Many Latin American countries are relatively homogeneous in terms of the language spoken (often that of a former colonizer), but heterogeneity increases when factors related to ethnicity such as skin colour and race are considered. For instance, in Bolivia, Alesina et al. (2003, p. 159) identified the following groups based on race: Whites (10.13%), Aymara (30.38%), Quechua (30.38%), mestizos (25.32%), and other groups (3.80%), and that 87.65% of the population speaks Spanish. Thus, in Latin America, the ethnic fractionalisation index computed by Alesina et al. (2003) is much higher (0.42) than that computed for linguistic fractionalization (0.16). In contrast, they also find the highest linguistic heterogeneity is in Africa (e.g., Liberia (0.903) and Cameroon (0.889)). vIn this regard, Alesina et al. (2003, pp. 161) argue that the stability of ethnic fractionalization data at country level poses no problem for a 20- to 30-year horizon. Nevertheless, as we have already pointed out, Dražanová (2020) makes a first attempt in offering a historical index of ethnic fractionalization. vi It is worth noting that this argument holds when it is assumed that country borders are fixed. However, depending on when the birthplace diversity is measured, there could be cases where people born in different countries (e.g., Slovenia and Croatia) could have received their education in the same school system (Yugoslavia). vii Depetris-Chauvin and Özak (2020) find that the migratory distance to East Africa alone explains 72% of the variation in intra-ethnic diversity. Moreover, they test the robustness to origin of serial founder effect by considering alternative origins located in South Africa (as suggested by Chan et al., 2019). The research concludes that the effect of distance from the origin of the SFE on intra-ethnic diversity is practically identical for all origins. viii Ashraf and Galor (2013b) estimate the relationship between genetic diversity and the following measures of ethnoliguistic fragmentation: the log number of ethnic groups, ethnic fractionalization indices of Fearon (2003) 34 BUITRAGO and CARABALLO POU and Alesina et al. (2003); linguistic fractionalization and polarization indices of Desmet et al. (2012) to level 1; and ethnolinguistic polarization indices of Esteban et al. (2012) and Montalvo and Reynal-Querol (2005a). The authors test the robustness of the results by considering a large number of control variables and they address potential endogeneity problems through two strategies: using a panel from Old World countries, and using the instrumental variable migratory distance from East Africa. ix Depetris-Chauvin and Özak (2020) construct intra-ethnic diversity indices by combining geocoded linguistic and genetic data. Specifically, they use the Ethnographic Atlas and the World Atlas of Language Structures to achieve an index of observed intra-ethnic diversity for 116 ethnic groups and a predicted intra-ethnic diversity index for 1265 ethnicities. Arbatli et al. (2020), using a geo-referenced genetic dataset, construct an observed genetic diversity index for 207 ethnic groups and a predicted index for 901 ethnicities. xHowever, as discussed in Section 4.1., there are arguments in favor of discrete distance instead of continuous distance to measure ethnolinguistic and religious diversity (Montalvo and Reynal-Querol, 2008). xi The first attempts of providing a precise definition of polarization are due to Esteban and Ray (1994) and Wolfson (1994). The latter focuses on the difference between inequality and polarization and constructs a theoretical measure of polarization that can be interpreted using the Lorenz curve. We focus on the contribution of Esteban and Ray (1994) due to its relevance in research on social polarization. xii Montalvo and Reynal-Querol (2005b, pp. 302-304) argue that the rent-seeking models also justify the use of this index to explain conflicts. They obtain the POLRQ index from a simple model of rent-seeking. xiii As noted by Esteban and Mayoral (2011, p. 6), both indices are conceptually closely related. FRAC is the probability that two people drawn randomly belong to different groups, while POLRQ is the probability that, out of three people, two belong to the same group and the third to any other group. xiv Esteban and Mayoral (2011) are aware of the endogeneity problems that this way of measuring distance may involve. This is addressed in Section 5. xv Regarding the case of Somalia, they argue that “While the example of Somalia is interesting, in our sample period, such examples are rare and ethnic fractionalization displays tremendous time persistence” (Alesina et al., 2003, p. 161). xvi However, the comparison between the results obtained by studies at local level with those at country level should be carried out cautiously, since studies at local level often use data on nationalities or birthplace, and, if this is the case, the literature at the country level also finds a positive relationship (See Alesina, Harnoss, & Rapoport, 2016 and Bove and Elia, 2017). xvii The birthplace fractionalization index is also uncorrelated with segregation and polarization indices (see Tables A6 and A8 in Appendix I). xviii See Ashraf and Galor (2018) for a full discussion of the relationship between genetic diversity and development. xix A brief description of the indices used by Desmet, Ortuño-Ortín, & Wacziarg (2017)toanalyzetheoverlap between ethnicity and culture is included in Appendix II. xx Conflicts are usually measured as a binary variable, although there are continuous indices to measure the level of social unrest. See Sambanis (2004), Montalvo and Reynal-Querol (2005a), and Esteban and Ray (2012)for details. xxi More recently, Huber and Mayoral (2019) have found a robust relationship between income inequality within ethnic groups and the severity of civil war. xxii As Esteban and Mayoral (2011, p.9) argue “Distance measures based on linguistic differences appear to be reasonably free from endogeneity as the possible tensions that lead to the split took place hundreds, if not thousands, of years ago. This helps to substantiate a causality relationship between the independent and dependent variables. Yet, this potential causality is, by the nature of the measure quite remote and cannot be directly taken as an immediate cause of the onset of a conflict. Linguistic distances have always been there, but only in some historical instances and in some countries these differences have become activated and developed into a relevant social cleavage”. ACKNOWLEDGMENTS The authors thank an Associate Editor and three anonymous referees for excellent comments which greatly improved the paper. The usual disclaimer applies. The authors also acknowledge the financial support from the Universidad de Sevilla (for covering the open access APCs) and BUITRAGO and CARABALLO POU 35 the Junta de Andalucía (Project P20-00628 (PAIDI 2020) and Research Group SEJ258-Research in Applied Economics). REFERENCES Alesina, A., & La Ferrara, E. (2005) Ethnic diversity and economic performance. Journal of Economic Literature, 43(3), 762–800. https://doi.org/10.1257/002205105774431243 Alesina, A., & Zhuravskaya, E. (2011) Segregation and the quality of government in a cross section of countries. American Economic Review,101(5), 1872–1911. https://doi.org/10.1257/aer.101.5.1872 Alesina, A., Devleeschauwer, A., Easterly, W., Kurlat, S., & Wacziarg, R. (2003) Fractionalization. Journal of Economic growth,8(2), 155–194. https://doi.org/10.1023/A:1024471506938 Alesina, A., Harnoss, J., & Rapoport, H. (2016) Birthplace diversity and economic prosperity. Journal of Economic Growth,21(2), 101–138. https://doi.org/10.1007/s10887-016-9127-6 Alesina, A., Michalopoulos, S., & Papaioannou, E. (2016) Ethnic inequality. Journal of Political Economy,124(2), 428–488. https://doi.org/10.1086/685300 Ager, P., & Brüeckner, M. (2013) Cultural diversity and economic growth: Evidence from the US during the age of mass migration. European Economic Review,64,76–97.https://doi.org/10.1016/j.euroecorev.2013.07.011 Ager, P., & Brüeckner, M. (2018) Immigrants’ genes: Genetic diversity and economic development in the United States. Economic Inquiry,56(2), 1149–1164. https://doi.org/10.1111/ecin.12540 Ahlerup, P., & Olson, O. (2012) The roots of ethnic diversity. Journal of Economic Growth,17(2), 71–102. https: //doi.org/10.1007/s10887-011-9075-0 Arbatli, C. E., Ashraf, Q. H., Galor, O., & Klemp, M. (2020) Diversity and conflict. Econometrica,88(2), 727–797. https://doi.org/10.3982/ECTA13734 Artuc, E., Docquier, F., Özden, C., & Parsons, C. (2015) A global assessment of human capital mobility: Therole of non-OECD destinations. World Development,65,6–26.https://doi.org/10.1016/j.worlddev.2014.04.004 Ashraf, Q., & Galor, O. (2013a) The ‘Out of Africa’ hypothesis, human genetic diversity, and comparative economic development. American Economic Review,103(1), 1–46. https://doi.org/10.1257/aer.103.1.1 Ashraf, Q., & Galor, O. (2013b) Genetic diversity and the origins of cultural fragmentation. American Economic Review, Papers and Proceedings,103(3), 528–533. https://doi.org/10.1257/aer.103.3.528 Ashraf, Q., & Galor, O. (2018) The macrogenoeconomics of comparative development. Journal of Economic Literature,56(3), 1119–1155. https://www.jstor.org/stable/10.2307/26494242 Ashraf, Q., Galor, O., & Klemp, M. (2014) The out of Africa hypothesis of comparative development reflected by light intensity. Working Paper, MPRA paper_55634. https://mpra.ub.uni-muenchen.de/55634/ Accessed April 21, 2021. Atlas Narodov M. (1964) Moscow: Miklukho-Maklai Ethnological Institute at the Department of Geodesy and Cartography of the State Geological Committee of the Soviet Union. Bakens, J., Mulder, P., & Nijkamp, P. (2013). Economic impacts of cultural diversity in the Netherlands: Productivity, utility, and sorting. Journal of Regional Science,53(1), 8–36. https://doi.org/10.1111/jors.12012 Barret, D. (ed.) (1982). World Christian encyclopedia. Oxford University Press. Bluhm, R., & Thomsson, K. (2020) Holding on? Ethnic divisions, political institutions and the duration of economic declines. Journal of Development Economics,144, 102457. https://doi.org/10.1016/j.jdeveco.2020.102457 Bossert, W., D’Ambrosio, C., & La Ferrara, E. (2011) A generalized index of fractionalization. Economica,78(312), 723–750. https://doi.org/10.1111/j.1468-0335.2010.00844.x Bove, V., & Elia, L. (2017) Migration, diversity, and economic growth. World Development,89, 227–239. https://doi. org/10.1016/j.worlddev.2016.08.012 Buitrago, E. M., Caraballo, M. A., & Roldán, J. L. (2019) Do tolerant societies demand better institutions?. Social Indicators Research,143(3), 1161–1184. https://doi.org/10.1007/s11205-018-2002-4 Cavalli-Sforza, L. L., Menozzi, P., & Piazza, A. (1994) The history and geography of human genes. Princeton University Press. Caraballo, M. A., & Buitrago, E. M. (2019) Ethnolinguistic diversity and education. A successful pairing. Sustainability,11(23), 6625, 1–18. https://doi.org/10.3390/su11236625 36 BUITRAGO and CARABALLO POU Chan, E. K., Timmermann, A., Baldi, B. F., Moore, A. E., Lyons, R. J., Lee, S.-S., Kalsbeek, A. M. F., Petersen, D. C., Rautenbach, H., Förtsch, H. E. A., Bornman, M. S. R., & Hayes, V. M. (2019). Human origins in a southern African palaeo-wetland and first migrations. Nature,575(7781), 185–189. https://doi.org/10.1038/s41586-019-1714-1 Collier, P., & Hoeffler, A. (1998) On economic causes of civil wars. Oxford Economic Papers,50,563–573. https://doi.org/10.1093/oep/50.4.563 Collier, P., & Hoeffler, A., (2004) Greed and grievance in civil war. Oxford Economic Papers,56(4), 563–595. https://doi.org/10.1093/oep/gpf064 Cook, C. J., & Fletcher, J. M. (2018) High-school genetic diversity and later-life student outcomes: Microlevel evidence from the wisconsin longitudinal study. Journal of Economic Growth,23(3), 307–339. https://doi.org/10. 1007/s10887-018-9157-3 Desmet, K., Gomes, J. F., & Ortuño-Ortín, I. (2020) The geography of linguistic diversity and the provision of public goods. Journal of Development Economics,143,https://doi.org/10.1016/j.jdeveco.2019.102384 Desmet, K., Ortuño-Ortín, I., & Wacziarg, R. (2012) The political economy of linguistic cleavages. Journal of Development Economics,97(2), 322–338. https://doi.org/10.1016/j.jdeveco.2011.02.003 Desmet, K., Ortuño-Ortín, I., & Wacziarg, R. (2017) Culture, ethnicity, and diversity. American Economic Review, 107(9), 2479–2513. https://doi.org/10.1257/aer.20150243 Desmet, K., Ortuño-Ortín, I., & Weber, S. (2017) Peripheral diversity: Transfers versus public goods. Social Choice and Welfare,49,787–823.https://doi.org/10.1007/s00355-016-1017-5 Desmet, K., Weber, S., & Ortuño-Ortín, I. (2009) Linguistic diversity and redistribution. Journal of the European Economic Association,7(6), 1291–1318. https://doi.org/10.1162/JEEA.2009.7.6.1291 Depetris-Chauvin, E., & Özak, Ö. (2020) The origins of the division of labor in pre-industrial times. Journal of Economic Growth,25(3), 297–340. https://doi.org/10.1007/s10887-020-09179-2 Dražanová, L. (2020) Introducing the Historical Index of Ethnic Fractionalization (HIEF) dataset: Accounting for longitudinal changes in ethnic diversity. Journalofopenhumanitiesdata,6(1), 6. https://hdl.handle.net/1814/ 68317 Duclos, J. Y., Esteban, J., & Ray, D. (2004) Polarization: Concepts, measurement, estimation. Econometrica,72(6), 1737–1772. https://doi.org/10.1111/j.1468-0262.2004.00552.x Dyen, I., Kruskal, J. B., & Black, P. (1992) An Indoeuropean classification: A lexicostatistical experiment. Transactions of the American Philosophical Society,82(5), iii-iv+1-132. https://doi.org/10.2307/1006517 Easterly, W., & Levine, R. (1997) Africa’s growth tragedy: Policies and ethnic divisions. The Quarterly Journal of Economics,112(4), 1203–1250. https://doi.org/10.1162/003355300555466 Esteban, J., & Mayoral, L. (2011) Ethnic and religious polarization and social conflict. Mimeo. 1–47. http://hdl.handle. net/10261/35507 Esteban, J. M., Mayoral, L., & Ray, D. (2012) Ethnicity and conflict: An empirical study. American Economic Review, 102(4), 1310–42. https://doi.org/10.1257/aer.102.4.1310 Esteban, J. M., & Ray, D. (1994) On the measurement of polarization. Econometrica: Journal of the Econometric Society,62,819–851.https://doi.org/10.2307/2951734 Esteban, J. M., & Ray, D. (1999) Conflict and distribution. Journal of Economic Theory,87,379–415.https://doi.org/ 10.1006/jeth.1999.2549 Esteban, J., & Ray, D. (2011) Linking conflict to inequality and polarization. American Economic Review,101(4), 1345–74. https://doi.org/10.1257/aer.101.4.1345 Fearon, J. D. (2003) Ethnic and cultural diversity by country. Journal of Economic Growth,8(2), 195–222. https://doi.org/10.1023/A:1024419522867 Fearon, J. D., & Laitin, D. D. (2003). Ethnicity, insurgency, and civil war. American Political Science Review,97(1), 75–90. https://www.jstor.org/stable/3118222 Greenberg, J. H. (1956) The measurement of linguistic diversity. Language,32(1), 109–115. https://doi.org/10.2307/ 410659 Horowitz, D. (1985). Ethnic groups in conflict. University of California Press. Huber, J. D., & Mayoral, L. (2019) Group inequality and the severity of civil conflict. Journal of Economic Growth, 24(1), 1–41. https://doi.org/10.1007/s10887-019-09162-6 Lazear, E. P. (1999). Globalisation and the market for team-mates. The Economic Journal,109(454), 15–40. https://doi.org/10.1111/1468-0297.00414 BUITRAGO and CARABALLO POU 37 Leeson, P. T. (2005). Endogenizing fractionalization. Journal of Institutional Economics,1(1), 75–98. https://doi.org/ 10.1017/S1744137405000056 Levinson, D. (1998) Ethnic groups worldwide: A ready reference handbook. Oryx Press. Mauro, P. (1995) Corruption and growth. Quarterly Journal of Economics,110(3), 681–712. https://doi.org/10.2307/ 2946696 Michalopoulos, S. (2012) The origins of ethnolinguistic diversity. American Economic Review,102(4), 1508–1539. https://doi.org/10.1257/aer.102.4.1508 Montalvo, J. G., & Reynal-Querol, M. (2005a) Ethnic polarization, potential conflict, and civil wars. American Economic Review,95(3), 796–816. https://doi.org/10.1257/0002828054201468 Montalvo, J. G., & Reynal-Querol, M. (2005b) Ethnic diversity and economic development. Journal of Development Economics,76(2), 293–323. https://doi.org/10.1016/j.jdeveco.2004.01.002 Montalvo, J. G., & Reynal-Querol, M. (2008) Discrete polarisation with an application to the determinants of genocides. The Economic Journal,118(533), 1835–1865. https://doi.org/10.1111/j.1468-0297.2008.02193.x Montalvo, J. G., & Reynal-Querol, M. (2020) Ethnic diversity and growth: Revisiting the evidence. Review of Economics and Statistics,1–43.https://doi.org/10.1162/rest_a_00901 Nathan, M. (2011). The long term impacts of migration in British cities: Diversity, wages, employment and prices.SERC Discussion Paper, 67. Spatial Economics Research Centre (SERC), London School of Economics and Political Sciences. http://eprints.lse.ac.uk/id/eprint/33577 Papyrakis, E., & Mo, P. H. (2014) Fractionalization, polarization, and economic growth: Identifying the transmission channels. Economic Inquiry,52(3), 1204–1218. https://doi.org/10.1111/ecin.12070 Putterman, L., & Weil, D. N. (2010) Post-1500 population flows and the long-run determinants of economic growth and inequality. The Quarterly Journal of Economics,125(4), 1627–1682. https://doi.org/10.1162/qjec.2010.125.4. 1627 Ramachandran, S., Deshpande, O., Roseman, C. C., Rosenberg, N. A., Feldman, M. W., & Cavalli-Sforza, L. L. (2005) Supportfrom therelationshipof genetic andgeographic distancein humanpopulations fora serial foundereffect originating in Africa. Proceedings of the National Academy of Sciences,102(44), 15942–15947. https://doi.org/10. 1073/pnas.0507611102 Ratha, D., Eigen-Zucchi, C., & Plaza, S. (2016) Migration and remittances Factbook 2016. World Bank Publications. Reardon, S. F., & Firebaugh, G. (2002) Measures of multigroup segregation. Sociological Methodology,32(1), 33–67. https://doi.org/10.1111/1467-9531.00110 Reich, D. (2018). Who we are and how we got here? Ancient DNA and the new science of the human past.Oxford University Press. Reynal-Querol, M. (2002) Ethnicity, political systems, and civil wars. Journal of Conflict Resolution,46(1), 29–54. https://doi.org/10.1177/0022002702046001003 Sambanis, N. (2004). What is civil war? Conceptual and empirical complexities of an operational definition. Journal of Conflict Resolution,48(6), 814–858. https://doi.org/10.1177/0022002704269355 Spolaore, E., & Wacziarg, R. (2009) The diffusion of development. Quarterly Journal of Economics,124(2), 469–529. https://doi.org/10.1162/qjec.2009.124.2.469 Scarritt, J. R., & Mozaffar, S. (1999). The specification of ethnic cleavages and ethnopolitical groups for the analysis of democratic competition in contemporary Africa. Nationalism and Ethnic Politics,5(1), 82–117. https://doi.org/ 10.1080/13537119908428555 Sparber, C. (2010) Racial diversity and macroeconomic productivity across US states and cities. Regional Studies, 44(1), 71–85. https://doi.org/10.1080/00343400802360436 Suedekum, J., Wolf, K., & Blien, U. (2014). Cultural diversity and local labour markets. Regional Studies,48(1), 173–191. https://doi.org/10.1080/00343404.2012.697142 Taylor, C. L., & Hudson, M. C. (1972) World handbook of political and social indicators. Yale University Press. Van Staveren, I., & Pervaiz, Z. (2017) Is it ethnic fractionalization or social exclusion, which affects social cohesion?. Social Indicators Research,130(2), 711–731. https://doi.org/10.1007/s11205-015-1205-1 Vanhanen, T. (1999) Domestic ethnic conflict and ethnic nepotism: A comparative analysis. Journal of Peace Research,36(1), 55–73. https://doi.org/10.1177/0022343399036001004 Wolfson, M. C. (1994). When inequalities diverge. The American Economic Review,84(2), 353–358. https://www. jstor.org/stable/2117858 38 BUITRAGO and CARABALLO POU How to cite this article: Buitrago, E. M., & Caraballo, M. Á. (2021) Measuring social diversity in economic literature. An overview for cross-country studies. Journal of Economic Surveys,1–55.https://doi.org/10.1111/joes.12484 APPENDIX I TABLE A1 Descriptive statistics. GDP per capita GDPpc1970 GDPpc1980 GDPpc1990 GDPpc2000 GDPpc2010 LOW-INCOME COUNTRIES Mean 692.48 629.10 472.26 410.32 492.63 Median 602.46 531.43 474.97 421.02 429.59 Maximum 1624.33 1461.13 668.33 553.58 896.57 Minimum 265.22 279.95 329.15 228.24 231.19 Observations 10 10 10 10 10 HIGH-INCOME COUNTRIES Mean 23760.06 30538.86 38342.90 47136.70 51697.20 Median 23969.42 29264.47 36400.84 43930.94 47910.68 Maximum 32267.08 48538.24 60268.66 81709.66 87770.27 Minimum 17934.19 21865.13 28691.29 35576.77 38893.02 Observations 10 10 10 10 10 Notes: GDPpc1970, GDPpc1980, GDPpc1990, GDPpc2000, and GDPpc2010: gross domestic product per capita, in PPP (purchasing power parity) in constant 2011 international dollars, taken from the World Bank for the years 1970, 1980, 1990, 2000, and 2010, respectively. The classification of the countries by income level has been made in accordance with the criteria of the World Bank. We have selected ten countries that have always featured in the categories of low income and high income during the period for which the classification of the World Bank has been available: 1987–2020. Low-income countries: Burkina-Faso, Central African Republic, Chad, the Democratic Republic of the Congo, Liberia, Madagascar, Mali, Niger, Togo, Sierra Leone. High-Income countries: Australia, Canada, Denmark, Finland, Germany, Japan, Sweden, Norway, the United Kingdom, the United States of America. APPENDIX II The overlap between ethnicity and other features, such as income and culture, enables an analysis to be carried out into whether belonging to an ethnic group yields additional information regarding the characteristics of an individual. The overlap is also a way of measuring antagonism, since individuals feel antagonism towards other individuals belonging to different ethnic groups not due to the ethnicity itself, but because different ethnic groups involve different incomes and/or different cultures. In this vein, Desmet, Ortuño-Ortín, and Wacziarg (2017) study whether different ethnicities do in fact imply different cultures, and therefore different preferences, which is the assumption underlying the negative economic effects of ethnolinguistic diversity. In order to address this issue, they focus on the cultural features extracted from the answers to selected questions of the World Values Survey related to norms, values, attitudes, and preferences (see Table 2). They use two indices. The first index measures the social antagonism (SA) with a χ2index. When the BUITRAGO and CARABALLO POU 39 TABLE A2 Descriptive statistics and correlations: Ethnolinguistic fractionalization indices ALE ALL ETFRAC ELF1 ELF6 ELF15 SRELF CDFRAC TOTAL SAMPLE Mean 0.48 0.41 0.48 0.16 0.39 0.50 0.38 0.32 Median 0.52 0.40 0.50 0.06 0.40 0.54 0.37 0.32 Obs. 108 108 108 108 108 108 74 108 LOW-INCOME COUNTRIES Mean 0.73 0.69 0.70 0.16 0.54 0.75 0.57 0.47 Median 0.75 0.79 0.75 0.05 0.62 0.83 0.66 0.54 Obs. 24 24 24 24 24 24 524 LOWER-MIDDLE-INCOME COUNTRIES Mean 0.52 0.48 0.56 0.15 0.45 0.59 0.49 0.33 Median 0.53 0.46 0.60 0.034 0.50 0.69 0.54 0.28 Obs. 25 25 25 25 25 25 18 25 UPPER-MIDDLE-INCOME COUNTRIES Mean 0.48 0.33 0.48 0.20 0.28 0.37 0.40 0.27 Median 0.54 0.33 0.48 0.12 0.26 0.32 0.38 0.29 Obs. 25 25 25 25 25 25 21 25 HIGH-INCOME COUNTRIES Mean 0.26 0.23 0.26 0.12 0.32 0.34 0.27 0.23 Median 0.16 0.16 0.21 0.04 0.20 0.31 0.16 0.18 Obs. 34 34 34 34 34 34 30 34 LOW POLITICAL STABILITY Mean 0.57 0.48 0.55 0.18 0.46 0.57 0.47 0.37 Median 0.65 0.55 0.63 0.12 0.50 0.66 0.49 0.43 Obs. 65 65 65 65 65 65 38 65 HIGH POLITICAL STABILITY Mean 0.34 0.30 0.36 0.12 0.29 0.40 0.28 0.24 Median 0.25 0.25 0.35 0.04 0.15 0.36 0.16 0.18 Obs. 43 43 43 43 43 43 36 43 CORRELATIONS GDPpc00 −0.50 −0.36 −0.46 −0.21 −0.17 −0.28 −0.25 −0.27 GDPpc10 −0.51 −0.37 −0.47 −0.19 −0.18 −0.32 −0.27 −0.28 PS −0.44 −0.31 −0.36 −0.21 −0.33 −0.33 −0.43 −0.31 Notes: ALE and ALL: ethnic and linguistic fractionalization indices, respectively, obtained by Alesina et al. (2003); ETFRAC: ethnolinguistic fractionalization index provided by 2005aMontalvo and Reynal-Querol (2005a, 2005b); ELF1, ELF6 and ELF 15: linguistic fractionalization indices computed by Desmet et al. (2012) disaggregated by levels 1, 6, and 15 respectively, available in Ethnologue; SRELF: ethnolinguistic fractionalization index based on self-reported ethnic and/or linguistic identity by individuals obtained by Desmet, Ortuño-Ortín, and Wacziarg (2017); CDFRAC: ethnolinguistic fractionalization index with continuous distance computed by Fearon (2003); GDPpc00 and GDPpc10: gross domestic product per capita, in PPP (purchasing power parity) in constant 2011 international dollars, taken from the World Bank for years 2000 and 2010 respectively. The classification (Continues) 40 BUITRAGO and CARABALLO POU TABLE A2 (Continued) of the countries by income level has been made in accordance with the criteria of the World Bank (2016). PS: political stability index taken from the World Governance Indicators. This index varies from −2.5 (minimum political stability) to 2.5 (maximum political stability. We have considered countries below (over) 0 as countries of low (high) stability. ethnic groups reflect the population distribution of answers, then the index takes low values, while higher values indicate an increasing group-specificity. In order to obtain the SA index, let us consider i=1...I ethnic groups, q=1...Q questions selected from the WVS, and r(q) =1...R(q)possible answers to each question q. Desmet, Ortuño- Ortín, and Wacziarg (2017) define the antagonism felt by an individual belonging to group iwho gives an answer rj(q) to a question q. The individual feels antagonism when the members of their group igive an answer that differs from rj(q) to a question q. Social antagonism for a given question qand a reply rj(q) is therefore defined as the antagonism felt by an individual when the members of any ethnic group give an answer different from rj(q) for a question q(see Table A11 for the definitions of these two measures). Finally, the SA index captures the antagonism arising from the different answers given to the same question, and it is defined as a weighted average of the social antagonism for a given question q: 𝑆𝐴 = 1 𝑄 𝑄 ∑ 𝑞=1 𝐼 ∑ 𝑖=1 𝑅(𝑞) ∑ 𝑟(𝑞)=1 𝜋𝑖(𝜋𝑞,𝑟(𝑞) 𝑖−𝜋 𝑞,𝑟(𝑞))2 𝜋𝑞,𝑟(𝑞) where πq,r(q) is the percentage of the overall population that gives answer r(q) to a question q,and πiq,r(q) is the percentage of individuals of ethnic group ithat give answer r(q) to a question q. The second index is a fixation index that captures the share of the total variation in culture due to the differences between ethnic groups. It ranges from zero, which indicates that the ethnic group of an individual gives no information regarding the answers, to one, which indicates that the ethnic identity can accurately predict the answers. Desmet, Ortuño-Ortín, and Wacziarg (2017) consider the within-cultural diversity (CUFW) for all ethnic groups as the weighted average of the culturalfractionalizationwithin a groupI(see Table A11).Finally, the Fixation Indexis the cultural fractionalization that cannot be explained by the within-cultural diversity across ethnic groups: 𝐹𝑆𝑇 =𝐶𝑈𝐹 − 𝐶𝑈𝐹𝑊 𝐶𝑈𝐹 Desmet, Ortuño-Ortín, and Wacziarg (2017) conclude that ethnic diversity is not a good proxy of cultural diversity, given the low values of SA and the Fixation Index obtained in their study. BUITRAGO and CARABALLO POU 41 TABLE A3 Descriptive statistics and correlations: Religious, birthplace, genetic and cultural fractionalization indices ALR RFRAC MIG00 MIG90 PDIV CUF TOTAL SAMPLE Mean 0.44 0.31 0.12 0.13 0.72 0.52 Median 0.43 0.30 0.04 0.05 0.73 0.53 Observations 108 108 108 108 108 74 LOW-INCOME COUNTRIES Mean 0.47 0.46 0.05 0.08 0.74 0.52 Median 0.54 0.52 0.03 0.05 0.74 0.54 Observations 24 24 24 24 24 5 LOWER-MIDDLE-INCOME COUNTRIES Mean 0.43 0.35 0.06 0.08 0.72 0.50 Median 0.42 0.36 0.02 0.02 0.74 0.50 Observations 25 25 25 25 25 18 UPPER-MIDDLE-INCOME COUNTRIES Mean 0.38 0.31 0.06 0.07 0.70 0.53 Median 0.31 0.29 0.03 0.02 0.71 0.53 Observations 25 25 25 25 25 21 HIGH-INCOME COUNTRIES Mean 0.46 0.17 0.26 0.24 0.72 0.54 Median 0.42 0.05 0.19 0.12 0.73 0.54 Observations 34 34 34 34 34 30 LOW POLITICAL STABILITY Mean 0.39 0.35 0.08 0.10 0.72 0.51 Median 0.37 0.36 0.02 0.04 0.73 0.52 Observations 65 65 65 65 65 38 HIGH POLITICAL STABILITY Mean 0.51 0.24 0.19 0.17 0.72 0.54 Median 0.55 0.10 0.11 0.10 0.73 0.54 Observations 43 43 43 43 43 36 CORRELATIONS GDPpc00 0.04 −0.43 0.48 0.38 0.00 0.34 GDPpc10 0.05 −0.45 0.46 0.37 −0.01 0.34 PS 0.18 −0.26 0.29 0.23 −0.13 0.45 Notes: ALR religious fractionalization index obtained by Alesina et al. (2003); RFRAC: religious fractionalization index provided by Montalvo and Reynal-Querol (2005a, 2005b); MIG00 and MIG90: birthplace diversity indices for years 2000 and 1990 respectively, computed by Alesina , Harnoss, and Rapoport (2016); PDIV: predicted genetic diversity provided by Ashraf and Galor (2013a); CUF: cultural fractionalization index. GDPpc00 and GDPpc10: gross domestic product per capita, in PPP (purchasing power parity) in constant 2011 international dollars, taken from the World Bank for years 2000 and 2010 respectively. The classification of the countries by income level has been made in accordance with the criteria of the World Bank (2016). PS: political stability index taken from the World Governance Indicators. This index varies from −2.5 (minimum political stability) to 2.5 (maximum political stability. We have considered countries below (over) 0 as countries of low (high) stability. 48 BUITRAGO and CARABALLO POU TABLE A9 (Continued) Indices Diversity Dependent variables Control variables (-NS): Public Consumption Initial income; Revolutions & coups; Assassinations; Civil wars; Democracy index; Fractionalization (religious) and polarization (ethnic and religious) indices. (+NS): Civil Wars Initial income; Population; Index of democracy; Regional dummies; Fractionalization (religious) and polarization (ethnic and religious) indices. Religious Not correlation: GDPpc Growth (1960 to 1989 organized in 5-year intervals) Initial income; Real government consumption/real GDP; Revolutions & coups; Assassinations; Absolute deviation PPP value investment deflator from sample mean; Real domestic investment/GDP, Primary and secondary-school enrolment rates; Civil wars; Regional dummies; Fractionalization (ethnic) and polarization (ethnic and religious) indices. (-NS): Investment (Real domestic investment/GDP) (-NS): Public Consumption Initial income; Revolutions & coups; Assassinations; Civil wars; Democracy index; Fractionalization (ethnic) and polarization (ethnic and religious) indices. (+NS): Civil Wars Initial income; Population; Index of democracy; Regional dummies; Fractionalization (ethnic) and polarization (ethnic and religious) indices. Desmet et al. (2012) Linguistic (-Sig.): Growth (L6 to L15 disaggregation) Initial income; Investment/GDP, Schooling years; Population, Openness; Legal origins dummies; Regional dummies (-Sig.): Redistribution (Transfers and subsidies/GDP) (-Sig. L1; - NS L6 to L15 disaggregation) GDPpc 1985−95; Population 1985−95; Population above 65; Small Island dummy; Latitude; Legal origins dummies; Regional dummies (Continues) BUITRAGO and CARABALLO POU 49 TABLE A9 (Continued) Indices Diversity Dependent variables Control variables (Mix): Output of Public Goods (Infant mortality (+Sig. L1 to L15); School attainment (-NS L1 to L15); Illiteracy rate (+Sig. L6 to L15); Infrastructure quality index (-NS L1 to L15); health care (NS); measles immunization rates for children (-Sig L1 to L15); access to public services (-Sig. L6 to L15); clean water (-Sig L15); infrastructure (NS)). Legal origins dummies; Regional dummies; Latitude; GNPpc (+Sig.): Onset Civil Wars (+Sig. L1; NS L6 to L15 desegregation) Lagged civil war; Lagged GDP; Lagged population; %Mountainous; Non-contiguous state; Oil exporter; New state; Instability; Democracy lagged; Legal origins dummies; Regional dummies Esteban et al. (2012) Linguistic (+NS): Conflict Variables (Armed conflict, Intermediate armed conflict, War, Conflict intensity, Index of social conflict–NS-) GDPpc; Initial population; %Mountainous, Non-contiguous state; Oil/Diamond; Oil reserves pc; Political rights; Institutionalized democracy; Executive constraints; Group cohesion; Institutionalized autocracy; Civil liberties; Publicness index Alesina and Zhuravskaya (2011) LinguisticEthnic (Mix): Quality of Government (Voice, Political stability, Government effectiveness, Regulatory quality, Rule of law, control of corruption). (-Sig.) No controls: (+NS) All controls and both full simple and excluded dictatorships. CANALS LINGUISTIC: (NS) Generalized trust; (+Sig.) Separatist movements; (NS) Subnational budget revenue %. CANALS ETHNIC: (-NS) Ethnic party dummy GDPpc; Population; Average size of region; Religion dummies; Latitude; Legal origins dummies; Regional dummies; Democratic tradition; %Mountainous (Continues) 50 BUITRAGO and CARABALLO POU TABLE A9 (Continued) Indices Diversity Dependent variables Control variables Desmet Ortuño-Ortín, and Wacziarg (2017) Ethnic-linguistic identity (+NS): GDPpc Lagged war; Lagged GDPpc; Lagged population; Mountain %; Non-contiguous territory; Oil; New state; Instability; Democracy lagged; Regional dummies; GDP growth and lagged (+NS): Public Goods (Variables idem Desmet et al., 2012) (-NS): Civil Conflict (Incidence of conflict, Onset of conflict) Alesina, Harnoss, and Rapoport (2016) Birthplace (+Sig.): GDPpc (inverse U-Shaped) (+Sig.) All simple and rich countries; (NS) poor countries Exogenous (Land area, Landlocked country, Latitude, Population within 100 km from ice free coast; Mean temperature; Mean precipitation); Endogeneous (Genetic diversity, Malaria, Yellow fever and Tuberculosis incidence, Polity2 institutional quality index, Trade openness, Trade diversity, Average GDP/capita at immigrants’ origin, Years of schooling, Population size, Regional & Colonial FE) Bove and Elia (2017) Birthplace (+Sig.): GDPpc GDPpc t0; Investment/GDP; Population growth; Average years schooling; Government consumption/GDP; Trade openness/GDP; Ethnic inequality index; Regional dummies Ashraf and Galor (2013a)3 Genetic (observed and predicted) (+Sig./ hump-shaped relationship.): Development Outcomes historical and contemporary (population density in 1500; GDPpc 2000) CANALS: (+Sig.) Scientific productivity; (-Sig.) Trust (+Sig.): Civil Conflict (Frequency, Incidence, and Onset of civil conflict) Neolithic transition timing; Land productivity controls; Institutional and cultural controls; Geographical controls; Country-specific disturbance; Legal origins dummies; Regional dummies; Ethnic fractionalization. Geographical Characteristics; Institutional Factors; Ethnolinguistic Fragmentation; Natural Resources and Development Outcomes. Desmet, Ortuño-Ortín, and Wacziarg (2017) Cultural attitudes (+Sig.): GDPpc Lagged war; Lagged GDPpc; Lagged population; Mountain %; Non-contiguous state; Oil; New state; Instability; Democracy lagged; Regional dummies; GDP growth and lagged; Genetic diversity (Continues) BUITRAGO and CARABALLO POU 51 TABLE A9 (Continued) Indices Diversity Dependent variables Control variables (+Sig.): Public Goods (Variables idem Desmet et al, 2012) (-Sig.): Civil Conflict (Incidence of conflict, Onset of conflict) Fractionalization with continuous distance Fearon (2003)2Ethnic (Linguistic distance. δ=0.5) (-Sig.): Conflict Variables (Armed conflict, Intermediate armed conflict, War, Conflict intensity, Index of social conflict) GDPpc; Initial population; %Mountainous, Non-contiguous state; Oil/Diamond; Oil reserves pc; Political rights; Institutionalized democracy; Executive constraints; Group cohesion; Institutionalized autocracy; Civil liberties; Publicness index Esteban et al. (2012) Ethnic (Linguistic distance. δ=0.05) (-Sig.): Conflict Variables (Armed conflict, Intermediate armed conflict, War, Conflict intensity, Index of social conflict) Linguistic (Linguistic distance. δ=0.5 and δ=0.05) (-NS): Conflict Variables (Armed conflict, Intermediate armed conflict, War, Conflict intensity, Index of social conflict) Alesina, Harnoss, and Rapoport (2016) Birthplace (Genetic and GDP distances) (+Sig.): GDPpc Effects largest for immigrants from richer countries and from intermediate levels of cultural proximity Exogenous (Land area, Landlocked country, Latitude, Population within 100 km from ice free coast; Mean temperature, Mean precipitation); Endogeneous (Genetic diversity, Malaria, Yellow fever & Tuberculosis incidence, Polity2 institutional quality index, Trade openness, Trade diversity, Average GDP/capita at immigrants’ origin, Years of schooling, Population size, Regional & Colonial FE) Desmet, Ortuño-Ortín, and Wacziarg (2017) Cultural attitudes(Distance values) (+Sig.): Public Goods (Variables idem Desmet et al., 2012) (-Sig.): Civil Conflict (Incidence of conflict, Onset of conflict) Lagged war; Lagged GDPpc; Lagged population; Mountain %; Non-contiguous state; Oil; New state; Instability; Democracy lagged; Regional dummies; GDP growth and lagged; Genetic diversity Notes:1The results for Growth were obtained by Easterly and Levine (1997) and those for Corruption by Mauro (1995). Both studies used the Taylor and Hudson (1972)index.2The results were obtained by Esteban et al. (2012)usingtheFearon(2003)index.3The results for Conflicts were obtained by Arbatli et al. (2020) using the Ashraf and Galor (2013a)index. (+Sig.): positive and significant results, (-Sig.): negative and significant results, (NS): not significant, (Mix.): mixed results. 52 BUITRAGO and CARABALLO POU TABLE A10 Applications of polarization indices (extended) Indices Diversity Dependent variable Control variables Polarization with discrete distance Montalvo and Reynal-Querol (2005a, 2005b) Ethnic-linguistic (NS): GDPpc Growth (1960 to 1989 organized in 5-year intervals). Indirect effect (CANALS Sig.: Investment, Public consumption, Civil wars) Initial income; Real government consumption/real GDP; Revolutions & coups; Assassinations; Absolute deviation PPP value investment deflator from sample mean; Real domestic investment/GDP, Primary and secondary-school enrolment rates; Civil wars; Regional dummies; Fractionalization (ethnic and religious) and Polarization (religious) indices. (-Sig.): Investment (Real domestic investment/GDP) (+Sig.): Public Consumption Initial income; Revolutions & coups; Assassinations; Civil wars; Democracy index; Fractionalization (ethnic and religious) and Polarization (religious) indices. (+Sig.): Civil Wars Initial income; Population; Index of democracy; Regional dummies; Fractionalization (religious) and Polarization (ethnic and religious) indices. Religious (+NS): GDPpc Growth (1960 to 1989 organized in 5-year intervals). Indirect effect (CANALS Sig.: Investment, Public consumption, Civil wars) Initial income; Real government consumption/real GDP; Revolutions & coups; Assassinations; Absolute deviation PPP value investment deflator from sample mean; Real domestic investment/GDP, Primary and secondary-school enrolment rates; Civil wars; Regional dummies; Fractionalization (ethnic and religious) and Polarization (ethnic) indices. (-Sig.): Investment (Real domestic investment/GDP) (+Sig.): Public Consumption Initial income; Revolutions & coups; Assassinations; Civil wars; Democracy index; Fractionalization (ethnic and religious) and Polarization (ethnic) indices. (+Sig.): Civil Wars Initial income; Population; Index of democracy; Regional dummies; Fractionalization (ethnic and religious) and Polarization (ethnic) indices. (Continues) BUITRAGO and CARABALLO POU 53 TABLE A10 (Continued) Indices Diversity Dependent variable Control variables Desmet et al. (2012) Linguistic (-NS): Growth(L1 to L15 desegregation) Initial income; Investment/GDP; Schooling years; Population; Openness; Legal origins dummies; Regional dummies (-Sig.): Redistribution (Transfers and subsidies/GDP) (-Sig. L1; -NS L6 to L15 desegregation) GDPpc 1985−95; Population 1985−95; Population above 65; Small island dummy; Latitude; Legal origins dummies; Regional dummies (Mix): Output of Public Goods (Infant mortality (+Sig. L1 to L15); School attainment (-NS L1 to L15); Illiteracy rate (+Sig. L6 to L15); Infrastructure quality (-NS L1 to L15); Health care (NS); Measles immunization rates children (-Sig. L1 to L15); Access public services (-Sig. L6 to L15); clean water (-Sig. L1 to L6); infrastructure (NS)) Legal origins dummies; Regional dummies; Latitude; GNPpc (+Sig.): Onset Civil Wars (+Sig. L1; - NS L6 to L15 desegregation) Lagged civil war; Lagged GDP; Lagged population; %Mountainous; Non-contiguous state; Oil exporter; New state; Instability; Democracy lagged; Legal origins dummies; Regional dummies Esteban et al. (2012)Ethnic (+Sig.): Conflict Variables (Armed conflict, Intermediate armed conflict, War, Conflict intensity, Index of social conflict–NS-) GDPpc; Initial population; %Mountainous, Non-contiguous state; Oil/Diamond; Oil reserves pc; Political rights; Institutionalized democracy; Executive constraints; Group cohesion; Institutionalized autocracy; Civil liberties; Publicness index Linguistic (NS): Conflict Variables (Variables idem Ethnic) (Continues) 54 BUITRAGO and CARABALLO POU TABLE A10 (Continued) Indices Diversity Dependent variable Control variables Desmet Ortuño-Ortín, and Wacziarg (2017) Ethnic-linguistic identity (+NS): Public Goods (Variables idem Desmet et al., 2012) Lagged war; Lagged GDPpc; Lagged population; Mountain %; Non-contiguous state; Oil; New state; Instability; Democracy lagged; Regional dummies; GDP growth and lagged; Genetic diversity (-NS): Civil Conflict (Incidence of conflict, Onset of conflict) Bove and Elia (2017) Birthplace (+Sig.): GDPpc (All time 1960-2010 and sub-periods; -Sig. 2000-10) GDPpc t0; Investment/GDP; Population growth; Average years schooling; Government consumption/GDP; Trade openness/GDP; Ethnic inequality index; Regional dummies Desmet, Ortuño-Ortín, and Wacziarg (2017) Cultural attitudes (-Sig.): Civil Conflict (Incidence of conflict, Onset of conflict) Lagged war; Lagged GDPpc; Lagged population; Mountain %; Non-contiguous state; Oil; New state; Instability; Democracy lagged; Regional dummies; GDP growth and lagged; Genetic diversity (+NS): Public Goods (Variables idem Desmet et al., 2012) Polarization with continuous distance Esteban et al. (2012) Ethnic and Linguistic (Linguistic distance. δ=0.5 and δ=0.05) (+Sig.): Conflict Variables (Armed conflict, Intermediate armed conflict, War, Conflict intensity, Index of social conflict) GDPpc; Initial population; %Mountainous, Non-contiguous state; Oil/Diamond; Oil reserves pc; Political rights; Institutionalized democracy; Executive constraints; Group cohesion; Institutionalized autocracy; Civil liberties; Publicness index Notes:(+Sig.): positive and significant results, (-Sig.): negative and significant results, (NS): not significant, (Mix.): mixed results. BUITRAGO and CARABALLO POU 55 TABLE A11 Overlapping indices (Desmet, Ortuño-Ortín, and Wacziarg, 2017) Index Expression Definition Individual antagonism (vir(q))𝜋𝑟(𝑞) 𝑖−𝜋𝑟(𝑞) 𝜋𝑟(𝑞) Antagonism felt individual belonging to group ifor a question q with answer r(q) Social Antagonism for a question q(SAq) 𝐼 ∑ 𝑖=1 𝑅(𝑞) ∑ 𝑟(𝑞)=1 𝜋𝑖(𝜋𝑟(𝑞) 𝑖−𝜋𝑟(𝑞))2 𝜋𝑟(𝑞) The social antagonism for a given question qis defined as a χ2 index. Social Antagonism (SA)1 𝑄 𝑄 ∑ 𝑞=1 𝐼 ∑ 𝑖=1 𝑅(𝑞) ∑ 𝑟(𝑞)=1 𝜋𝑖(𝜋𝑞,𝑟(𝑞) 𝑖−𝜋𝑞,𝑟(𝑞))2 𝜋𝑞,𝑟(𝑞) The social antagonism for all selected questions q=1...Qis aχ2index Cultural fractionalization (CUF) 1 𝑄 𝑄 ∑ 𝑞=1 (1 − 𝑅(𝑞) ∑ 𝑟(𝑞)=1 𝜋2 𝑞,𝑟(𝑞))Probability that two randomly drawn individuals from a population give different answers to a question Cultural fractionalization within a group i(CUFi) 1− 𝑅(𝑞) ∑ 𝑟(𝑞)=1 (𝜋𝑟(𝑞) 𝑖)2Probability that two members of the same ethnic groups give a different answer to a question Within cultural fractionalization (CUFW) 𝐼 ∑ 𝑖=1 𝜋𝑖𝐶𝑈𝐹𝑖Weighted average for the Igroups of CUFi Fixation Index (FST)𝐶𝑈𝐹−𝐶𝑈𝐹𝑊 𝐶𝑈𝐹 Cultural fractionalization that is not explained by the within cultural diversity across ethnic groups Notes: i =1...I;ethnicgroups;q=1...Q : questions selected from the WVS; r(q) =1...R(q) :possible answers to each question q; πr(q): percentage of the population that gives answer r(q); πir(q): percentage of individuals of ethnic group ithat gives answer r(q); πq,r(q): percentage of the overall population that gives answer r(q) to a question q; πiq,r(q): percentage of individuals of ethnic group ithat gives answer r(q) to a question q.