Ethnic Polarization, Potential Conflict, and Civil Wars

The correlation between the index of fractionalization obtained using alternative data sources is very high (over 0.8).

“In general the correlation between the index of fractionalization obtained using these alternative data sources is very high (over 0.8).”

• cites Montalvo and Reynal-Querol (2000)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Alesina et al (2003) sourcePartly supported

  Section 2.2 reports a correlation of 0.88 between the paper’s linguistic index and the Soviet index, supporting one instance of the claim. But the corresponding ethnic-index correlation is 0.76, and the paper says the religious index has much lower correlations, so it does not support the generalization that correlations exceed 0.8. (LLM, conf=0.88)

  Experimental — accuracy not yet measured
• cites Fearon (2003)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The correlation with the index of Alesina et al. (2003) is 0.83.

“The correlation with the index of Alesina et al. (2003) is 0.83.”

• cites Alesina et al. (2003) sourceSupported

The index of ethnic polarization is a significant variable in the explanation of the incidence of civil wars.

“The empirical section shows that this index of ethnic polarization is a signi cant variable in the explanation of the incidence of civil wars.”

• Also stated in: § 1 Introduction · p. 3 · plain

  “In the empirical section we show that the index of ethnic polarization is a signi cant explanatory variable for the incidence of civil wars.”

The new measure of ethnic heterogeneity satisfies the basic properties associated with the concept of polarization.

“We develop a new measure of ethnic heterogeneity that satis es the basic properties associated with the concept of polarization.”

The index of ethnic fractionalization calculated with our data has a correlation of 0.86 with the index obtained using the Atlas Nadorov Mira (ELF).

“The index of ethnic fractionalization calculated with our data has a correlation of 0.86 with the index obtained using the Atlas Nadorov Mira (ELF).”

For low levels of ethnic fractionalization, the correlation between ethnic fractionalization and polarization is positive and high.

“It shows that for low levels of fractionalization the correlation between ethnic fractionalization 20 and polarization is positive and high.”

When moving from two groups to three groups, the relationship between FRAC and Q breaks down.

“When we move from two groups to three groups the relationship between FRAC and Q breaks down.”

Religious fractionalization (RELFRAC) is not statistically significant.

“Column 4 shows that religious fractionalization (RELFRAC) is not statistically signi cant.”

In a specification with both ethnic and religious polarization, only the coefficient of ethnic polarization is statistically significant.

“In column 7 we include together the index of ethnic polarization and religious polarization. Only the estimated coe¢ cient of the rst one is statistically signi cant.”

Only three out of the ten countries with the highest level of religious fractionalization suffered a civil war.

“However only three out of the ten countries with the highest level of religious fractionalization su/ered a civil war.”

An increase in one standard deviation (0.24) of the average polarization increases the probability of conflict by 67%.

“An increase in one standard deviation (0.24) of the average polarization increases the probability of con ict by 67%.”

Ethnic polarization (ETHPOL) has a positive and statistically significant effect on the incidence of civil wars.

“However if we substitute the index of ethnic fractionalization by the Q index of ethnic polarization, ETHPOL, we nd (column 2) a positive and statistically signi cant e/ect on the incidence of civil wars.”

If the level of polarization increases from the average (0.51) to the level of Guinea (0.84), the probability of conflict almost doubles.

“Using the results in column 3, if the level of polarization increases from the average (0.51) to the level of Guinea (0.84) then the probability of con ict almost doubles.”

Nigeria has a high level of religious polarization between Christians (49%) and Muslims (45%).

“For instance in Nigeria there is a high level of religious polarization between Christians (49%) and Muslims (45%)”

The Q index satisfies the condition that the highest risk is associated with the middle range of ethnic diversity.

“The Q index satis es this condition.”

The index of fractionalization and polarization are very different independent of the source of data used in their calculation.

“In this section we describe the alternative data sources for ethnic and religious heterogeneity and we show that the index of fractionalization and polarization are very di/erent independent of the source of data used in their calculation.”

The sensitivity of the Q index is highest when groups are close to 50%.

“By contrast the sensitivity of the Q is the highest when groups are close to 50%.”

Bosnia has a high level of religious polarization between Christians (50%) and Muslims (40%).

“similar to the case of Bosnia (50% Christians and 40% Muslims).”

There is only one level of polarization sensitivity (α = 1) for which the discrete polarization measure satisfies the properties of polarization.

“In particular, and in contrast with the polarization index of Esteban and Ray (1994), there is only one level of polarization sensitivity ( α = 1 ) for which the discrete polarization measure satis es the properties of polarization.”

Ethnolinguistic fractionalization (ETHFRAC) has no statistically significant effect on the incidence of civil wars.

“The rst column shows that the index of ethnolinguistic fractionalization (ETHFRAC) has no statistically signi cant e/ect on the incidence of civil wars.”

Seven out of ten countries with the highest level of religious polarization suffered a civil war during the sample period.

“During the same sample period civil wars occurred in 7 out of 10 countries with the highest level of religious polarization.”

Religious polarization (RELPOL) is not statistically significant in column 5.

“Neither is the coe¢ cient of religious polarization (RELPOL) in column 5.”

There is only one value of k (k = 4) such that the index DP ranges between 0 and 1.

“In addition there is only one value of k ( k = 4 ) such that the index DP ranges between 0 and 1.”

If all groups have equal size, the proportional contribution of each group is equal to its relative size in both fractionalization and polarization.

“If all the groups have equal size the proportional contribution of each of the groups is equal to its relative size in both, fractionalization and polarization, that is c i = ˜ c i = π i .”

Ethnic polarization is robust to eliminating Latin American countries.

“Columns 5 and 6 show the robustness of ethnic polarization to eliminating from the sample the Latin American countries.”

The Q index of polarization calculated using Atlas Nadorov Mira data has a positive and statistically significant effect on the probability of civil wars.

“When we run the regression with the Q index of polarization calculated using the row data of the Atlas Nadorov Mira, we nd out that it has a positive e/ect (estimated coe¢ cient=2.35 and z=3.33) on the probability of civil wars”

Using the data of Doyle and Sambanis (2000) and Fearon and Laitin (2003), the importance of initial level of democracy is much larger than using the dataset of PRIO.

“Finally we should notice that using the data of Doyle and Sambanis (2000) and Fearon and Laitin (2003) the importance of initial level of democracy is much larger than using the dataset of PRIO.”

The product of the largest ethnic group by primary exports has a coefficient not significantly different from 0.

“In column 4 we can see that this variable has a coe¢ cient that is not signi cantly di/erent from 0.”

For intense civil wars (PRIO1000), the coefficient on ethnic polarization calculated using Alesina et al. (2003) data is 1.95 (z=2.22).

“For instance for intense civil wars (PRIO1000 de nition) the coe¢ cient on ethnic polarization calculated using the data of Alesina et al. (2003) is 1.95 (z=2.22).”

Ethnic polarization continues to be significant even in the presence of the size of the largest minority.

“while ethnic polarization continues to be signi cant even in the presence of this new variable (column 7).”

Ethnic dominance (ETHDOM) does not have any significant effect in the core specification.

“In column 2 of table 2 we show that ethnic dominance (ETHDOM) does not have any signi cant”

The statistical significance of the coefficient on ethnic polarization is robust to the use of the other two definitions of PRIO.

“Columns 3 and 4 show that the statistical signi cance of the coe¢ cient on ethnic polarization is robust to the use of the other two de nitions of PRIO.”

The coefficient measuring the effect of ethnic polarization on the probability of civil wars increases monotonically with the intensity of the conflict (2.05, 2.28, 2.33).

“In fact it is interesting to notice that the coe¢ cient that measures the e/ect of ethnic polarization on the probability of civil wars increases monotonically with the intensity of the con ict (2.05 including minor con icts; 2.28 for intermediate and high intensity con icts, and 2.33 for the most violent con icts).”

For intense civil wars, the coefficient on ethnic polarization calculated using ANM data is 1.98 (z=2.63).

“If ethnic polarization is calculated using the ANM then its estimated coe¢ cient on the incidence of intense civil wars is 1.98 (z=2.63).”

The index of ethnolinguistic polarization is significantly different from zero in cross-section regressions with or without regional dummy variables.

“Table 6 shows that the index of ethnolinguistic polarization is signi cantly di/erent from zero with (column 1) or without including the regional dummy variables (column 2) 33 .”

Ethnic polarization is statistically significant when using the Doyle and Sambanis (2000) definition of civil wars.

“Columns 1 shows that ethnic polarization is statistically signi cant when we use as dependent variable the de nition of civil wars of Doyle and Sambanis (2000).”

The effect of ethnic polarization on civil wars is robust to eliminating Asian countries.

“Finally, columns 7 and 8 con rm that the e/ect of ethnic polarization on civil wars is robust to the elimination from the sample of the Asian countries.”

The result is robust to the use of different datasets for civil wars like Doyle and Sambanis (2000) or Fearon and Laitin (2003).

“The result is robust to the use of di/erent dataset for civil wars like Doyle and Sambanis (2000), columns 3 and 4, or Fearon and Laitin (2003), columns 5 and 6.”

Ethnic fractionalization is not significant in the presence of regional dummies.

“which is not signi cant.”

In both cases, ethnic fractionalization is not statistically significant.

“In both cases ethnic fractionalization is not statistically signi cant.”

In a specification with ethnic and religious polarization plus ethnic and religious fractionalization, only the coefficient of ethnic polarization is significantly different from 0.

“If we add also as explanatory variables the degree of ethnic fractionalization and religious fractionalization (column 8) only the coe¢ cient of ethnic polarization is signi cantly di/erent from 0.”

The statistical significance of ethnic polarization in the explanation of civil wars is robust to the use of alternative datasets for the endogenous variable even without DEMOCRACY.

“Columns 6 to 9 show that the statistical signi cance of ethnic polarization in the explanation of civil wars is robust to the use of alternative datasets for the endogenous variable even if we do not consider the DEMOCRACY variable in the speci cation.”

When eliminating African countries and including ethnic fractionalization, the coefficient on ethnic polarization is not significant.

“If we eliminate those African countries and include in the regression the index of ethnic fractionalization, column 4, then the coe¢ cient on ethnic polarization is not signi cant.”

The results of table 1 are robust to the exclusion of the DEMOCRACY variable.

“Column 5 shows that the results of table 1 are robust to the exclusion of the DEMOCRACY variable”

The index of polarization is significant even when the product of the largest ethnic group by primary exports is included.

“In column 5 we show that the index of polarization is signi cant even when the product of the largest ethnic group by primary exports is included as an explanatory variable.”

The elimination of Sub-Saharan African countries does not affect the statistical significance of ethnic polarization.

“The elimination from the sample of the countries in Sub-Saharan Africa, column 3, does not a/ect the statistical signi cance of ethnic polarization.”

The size of the largest minority (LARGMINOR) has a coefficient that is not statistically significant.

“Column 6 shows that the coe¢ cient on this new variable is not statistically signi cant”

The coefficient of initial GDP per capita is robust in columns 1 to 4 of table 4.

“Another interesting fact in columns 1 to 4 of table 4 is the robustness of the coe¢ cient of initial GDP per capita.”

The index of fractionalization calculated using Atlas Nadorov Mira data is not statistically significant.

“while the index of fractionalization calculated with the same dataset is not statistically signi cant (estimated coe¢ cient=1.20 and z=1.41).”

Ethnic polarization is marginally statistically significant when using the Fearon and Laitin (2003) definition of civil war.

“Column 2 shows that ethnic polarization is marginally statistically signi cant if we use the de nition of civil war of Fearon and Laitin (2003).”

Many empirical studies find no relationship between ethnic fractionalization, ethnic conflicts, and civil wars.

“However many empirical studies nd no relationship between ethnic fractionalization 1 , ethnic con icts, and civil wars.”

• cites Fearon and Laitin (2003) sourceNot checked — no full text of the source
• cites Collier and Hoeffler (2002)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The channel from ethnic fractionalization to growth is not through an increase in potential ethnic conflict.

“Although ethnic fractionalization may be important for growth, we argue that the channel is not through an increase in potential ethnic con ict.”

The result that ethnic polarization is significant for civil wars is robust to other indicators of ethnic heterogeneity, alternative data sources, and alternative civil war data.

“This result is robust to the presence of other indicators of ethnic heterogeneity, alternative sources of data for the construction of the indicators, and alternative data for civil wars.”

The empirical evidence supporting the effect of ethnic fractionalization on the incidence of civil conflicts is very weak.

“However the empirical evidence supporting the e/ect of ethnic fractionalization on the incidence of civil con icts is very weak.”

• cites Fearon and Laitin (2003) sourceNot checked — no full text of the source
• cites Collier and Hoeffler (2002)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Only four out of the ten most ethnically fractionalized countries suffered a civil war during the sample period (1960-95).

“In the case of ethnic fractionalization only four out of the ten most fractionalized countries have su/ered a civil war.”

• cites [22] sourceNot checked — no full text of the source

The index of fractionalization has at least two theoretical justifications based on completely different contexts.

“The index of fractionalization has, at least, two theoretical justi cations based on completely di/erent contexts.”

Income per capita is a proxy for state's overall financial, administrative, police and military capabilities.

“Fearon and Laitin (2003) argue that income per capita is a proxy for state s overall nancial, administrative, police and military capabilities.”

• cites Fearon and Laitin (2003) sourceNot checked — no full text of the source

Using periods of five years for civil wars and GDP per capita at the beginning of each period reduces the potential endogeneity problem.

“We use periods of ve years for civil wars and the GDP per capita at the beginning of each period. This set up reduces also the potential endogeneity problem.”

The previous results do not depend on the source of data used in the construction of the indices.

“Figures 3 and 4 con rm that the previous results do not depend on the source of data used in the construction of the indices.”

Esteban and Ray (1999) show that a two-point symmetric distribution of population maximizes conflict.

“Esteban and Ray (1999) show, using a behavioral model and a quite general metric of preferences, that a two-point symmetric distribution of population maximizes con ict.”

• cites Esteban and Ray (1999) sourceNot checked — no full text of the source

Conditional on a given degree of polarization, more religious diversity decreases the probability of a civil war.

“This means that, conditional on a given degree of polarization, more religious diversity decreases the probability of a civil war.”

A low level of income per capita reduces the opportunity cost of engaging in a civil war.

“In addition a low level of income per capita reduces the opportunity cost of engaging in a civil war.”

• cites Fearon and Laitin (2003) sourceNot checked — no full text of the source

The correlation between fractionalization and polarization is low when there is a high degree of heterogeneity.

“Therefore the correlation is low when there is a high degree of heterogeneity, which is the interesting case.”

An index of polarization should capture better the likelihood or intensity of potential conflict than an index of fractionalization.

“If this is so then an index of polarization should capture better the likelihood of con icts, or the intensity of potential con ict, than an index of fractionalization.”

One possible reason for the lack of explanatory power of ethnic heterogeneity on the probability of armed conflicts and civil wars is the measure for heterogeneity.

“We argue that one possible reason for the lack of explanatory power of ethnic heterogeneity on the probability of armed con icts and civil wars is the measure for heterogeneity.”

Nine out of the ten most ethnically polarized countries suffered a civil war during the sample period (1960-95).

“A previous version of this paper 22 shows that nine out of the ten most ethnically polarized countries have su/ered a civil war during the sample period (1960-95).”

• cites [22] sourceNot checked — no full text of the source

The correlation of the index of ethnic fractionalization using Fearon's ethnic distances with the original ELF is very high, 0.82.

“As in the case of alternative data sources, the correlation of the index of ethnic fractionalization, using these distances , with the original ELF is very high, 0.82.”

• cites Fearon (2003)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The weak explanatory power of ethnic heterogeneity on the incidence of civil wars found by several recent studies is due to the use of an index of fractionalization instead of an index of polarization.

“Therefore it seems that the weak explanatory power of ethnic heterogeneity on the incidence of civil wars found by several recent studies is due to the use of an index of fractionalization instead of an index of polarization.”

• Also stated in: § 7 Conclusions · p. 24 · hedged

  “Therefore it seems that the weak explanatory power of ethnic heterogeneity on the incidence of civil wars found by several recent studies is due to the use of an index of fractionalization instead of an index of polarization.”

Collier and Hoeffler (1998) note coordination cost would be at their lowest when the population is polarized between an ethnic group identified with the government and a second, similarly sized ethnic group, identified with the rebels.

“Collier and Hoe› er (1998) note coordination cost would be at their lowest when the population is polarized between an ethnic group identi ed with the government and a second, similarly sized ethnic group, identi ed with the rebels.”

• cites Collier and Hoeffler (1998)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

An increase in ethnic polarization has an indirect negative effect on growth because it increases the incidence of civil wars and public consumption and reduces the rate of investment.

“However they also nd that an increase in ethnic polarization has an indirect negative e/ect on growth because it increases the incidence of civil wars and public consumption and reduces the rate of investment.”

• cites Montalvo and Reynal-Querol (2005) sourceNot checked — no full text of the source

The measure of ethnic heterogeneity appropriate to capture potential conflict should be a polarization measure.

“We argue that the measure of ethnic heterogeneity appropriate to capture potential con ict should be a polarization measure.”

It is unlikely that ethnic fractionalization affects economic development through an increase in the probability of conflicts.

“Therefore it is unlikely that ethnic fractionalization a/ects economic development through an increase in the probability of con icts.”

The relevance of ethnic polarization in the explanation of civil wars is robust to the presence of other indicators of ethnic heterogeneity like fractionalization, dominance or the product of the size of the largest group by the proportion of primary exports.

“The previous section has shown that the relevance of ethnic polarization in the explanation of civil wars is robust to the presence of other indicators of ethnic heterogeneity like fractionalization, dominance or the product of the size of the largest group by the proportion of primary exports.”

The presence of ethnic fractionalization increases the standard error of the ethnic polarization estimated coefficient.

“However, as we argued before, since ethnic fractionalization is not statistically signi cant it seems clear that its presence increases the standard error of the ethnic polarization estimated coe¢ cient.”

The index of ethnic fractionalization is not able to capture the idea that severe conflicts arise where a large ethnic minority faces an ethnic majority.

“The index of ethnic fractionalization is not able to capture this idea appropriately.”

The relative weakness of the coefficient of initial GDP per capita in tables 1 and 2 is due to the definition of civil war used.

“It seems that the relative weakness of the coe¢ cient of this variable in tables 1 and 2 is due to the de nition of civil war used (intermediate and high intensity types following PRIO).”

Ethnic polarization has a robust and powerful explanatory power on civil wars in the presence of other indices of fractionalization and polarization.

“It seems clear that ethnic polarization has a robust and powerful explanatory power on civil wars in the presence of other indices of fractionalization and polarization”

The statistical relevance of religious polarization depends on the particular specification.

“while the statistical relevance of religious polarization depends on the particular speci cation”