A Model of Ethnic Conict Joan Esteban and Debraj Ray Journal of the European Economic Association 2011
Motivation Empirical literature has focused on aggregate inequality and neglected within group inequalities Role of between and within group heterogeneity in explaining conict Is homogeneity in radicalism within groups conducive to conict? How about income heterogeneity, within or between groups? Focus on dierences in income and radicalism of commitment to the group's cause
Main result Within group inequality An increase in income inequality within one ethnic group may make inter-ethnic conict more violent: mobilizing the poor is cheaper, and the rich have more resources to implement that mobilization.
Elements of the model Inputs to conict time devoted to activism nancial contributions
Elements of the model Inputs to conict time devoted to activism nancial contributions "Radicalism" A measure of passion for group success Non-material payo that the group members could derive from a victory over the opponent. Can vary across individuals.
Elements of the model Inputs to conict time devoted to activism nancial contributions "Radicalism" A measure of passion for group success Non-material payo that the group members could derive from a victory over the opponent. Can vary across individuals. Conict over ethnic policies Actions and concessions that are titled towards one group or the other. A shift from the status quo, like razing of a mosque or enforcing aspects of Muslim personal law.
Elements of the model Two ethnic groups: M (Muslim) and H (Hindu)
Elements of the model Two ethnic groups: M (Muslim) and H (Hindu) Radicalism x 0: captures intensity of feelings regarding the proposed shift from the status quo.
Elements of the model Two ethnic groups: M (Muslim) and H (Hindu) Radicalism x 0: captures intensity of feelings regarding the proposed shift from the status quo. A h and A m : total time devoted to activism by individuals of each group
Elements of the model Two ethnic groups: M (Muslim) and H (Hindu) Radicalism x 0: captures intensity of feelings regarding the proposed shift from the status quo. A h and A m : total time devoted to activism by individuals of each group Probability of success of the proposed policy p(a h, A m ) = ψ(a h ) ψ(a m ) + ψ(a h )
Individual decision Individual payo (for an H individual) px + u(w(1 s) + c h s r) x is measure of zeal s is activism time c h is compensation rate for activism r is nancial resources devoted u(.) is strictly concave utility function The higher the wage of individual, the higher the opportunity cost of time devoted to activism
Individual contributions Dene d r/c to be the implicit activist time purchased by a nancial contribution of r. Individual solves max p(a_ + d,a m )x + u((1 s)w + c h s dc) s,d subject to s [0, s] and d 0
Choice of s and d For each type z = (w,x) s(z, c) = s if w < c, 0 if w > c, [0, s] if w = c. Then from Kuhn-Tucker conditions d(z, c) is implicitly given by { cu p 1 (A_+d(z, c), A m )x ((1 s)w + c s d(z, c)c) if w c, cu (w d(z, c)c) if w > c with equality when d(z, c) > 0.
Group equilibrium response Equilibrium response of group H to A m is a pair (A, c) that satises
Group equilibrium response Equilibrium response of group H to A m is a pair (A, c) that satises 1 For each type z = (w,x), s(z, c) and d(z, c) satisfy the rst order conditions above, given (A_, A m ) where A_ A d(z, c).
Group equilibrium response Equilibrium response of group H to A m is a pair (A, c) that satises 1 For each type z = (w,x), s(z, c) and d(z, c) satisfy the rst order conditions above, given (A_, A m ) where A_ A d(z, c). 2 S h z s(z, c)n(z) = D h z d(z, c)n(z) = A
Group equilibrium response Equilibrium response of group H to A m is a pair (A, c) that satises 1 For each type z = (w,x), s(z, c) and d(z, c) satisfy the rst order conditions above, given (A_, A m ) where A_ A d(z, c). 2 S h z s(z, c)n(z) = D h z d(z, c)n(z) = A 1 Individual best response 2 Clearing of the "market" for activism time
Uniqueness of group equilibrium Required assumption: z d(z, c, D) is decreasing in c for every D.
Overall equilibrium An equilibrium is a collection (A h, c h, A m, c m ) such that (A h, c h ) is an equilibrium response to A m and (A m, c m ) is an equilibrium response to A h.
Overall equilibrium An equilibrium is a collection (A h, c h, A m, c m ) such that (A h, c h ) is an equilibrium response to A m and (A m, c m ) is an equilibrium response to A h.
Properties of equilibrium A h rises with A m upto some point (where A h = A m ) and then falls Initial response to opposition with increased activism. Decrease in activism as opposition militancy continues to climb (strain on group resources). Equilibrium is unique
Comparative statics We are going to examine Rise in level of radicalism Change in distribution (inequality) of radicalism Rise in level of income Change in distribution (inequality) of income
One group becoming more aggressive Eects of a change in the parameters of a group that shifts its equilibrium response function outwards (more aggressive)
One group becoming more aggressive Eects of a change in the parameters of a group that shifts its equilibrium response function outwards (more aggressive) If the change took place in the group with smaller number of activists, then equilibrium A h and A m move in the same direction
One group becoming more aggressive Eects of a change in the parameters of a group that shifts its equilibrium response function outwards (more aggressive) If the change took place in the group with smaller number of activists, then equilibrium A h and A m move in the same direction If the change took place in the group with larger number of activists, then equilibrium A h and A m move in the opposite directions
The more radical group becoming more aggressive An increase in radicalism does not necessarily translate into an escalation of conict
Rise in level of radicalism x goes up for every individual in one group
Rise in level of radicalism x goes up for every individual in one group Individuals will increase their contribution to conict { cu p 1 (A_+d(z, c), A m )x ((1 s)w + c s d(z, c)c) if w c, cu (w d(z, c)c) if w > c x = d
Rise in level of radicalism x goes up for every individual in one group Individuals will increase their contribution to conict { cu p 1 (A_+d(z, c), A m )x ((1 s)w + c s d(z, c)c) if w c, cu (w d(z, c)c) if w > c x = d Activism by the group will increase Both as an equilibrium response and in equilibrium.
Change in distribution of radicalism What is the eect on group aggressiveness if radicalism becomes more dissimilar across the population of a given group?
Change in distribution of radicalism What is the eect on group aggressiveness if radicalism becomes more dissimilar across the population of a given group? Radicals become more extremist and moderates become more lukewarm.
Change in distribution of radicalism What is the eect on group aggressiveness if radicalism becomes more dissimilar across the population of a given group? Result Radicals become more extremist and moderates become more lukewarm. 1 If the distribution of radicalism shifts from individuals who are not nancial contributors to those that are, then group activism must increase.
Change in distribution of radicalism What is the eect on group aggressiveness if radicalism becomes more dissimilar across the population of a given group? Result Radicals become more extremist and moderates become more lukewarm. 1 If the distribution of radicalism shifts from individuals who are not nancial contributors to those that are, then group activism must increase. 2 If distribution of radicalism Lorenz-worsens (only) among those who are nancial contributors, then group activism must decline.
Discussion of result 2 The net eect of an increase in the inequality of radicalism on overall aggressiveness is negative. Divide et impera
Discussion of result 2 The net eect of an increase in the inequality of radicalism on overall aggressiveness is negative. Divide et impera Greater homogeneity = stronger sense of identication = heightened social tension
Rise in level of income What is the eect of a uniform rise in wealth levels of all individuals in one group on conict outcome?
Rise in level of income What is the eect of a uniform rise in wealth levels of all individuals in one group on conict outcome? Two eects Increase in nancial contributions to the conict Increase in opportunity cost to mobilization
Rise in level of income What is the eect of a uniform rise in wealth levels of all individuals in one group on conict outcome? Two eects Increase in nancial contributions to the conict Increase in opportunity cost to mobilization The two eects go in opposite direction Overall eect ambiguous
Rise in level of income What is the eect of a uniform rise in wealth levels of all individuals in one group on conict outcome? Two eects Increase in nancial contributions to the conict Increase in opportunity cost to mobilization The two eects go in opposite direction Overall eect ambiguous Rise in wealth level of one group also means higher inequality between groups. The eect of higher across group inequality is ambiguous
Rise in within-group inequality What is the eect of a rise in within-group inequality of income on the level conict?
Rise in within-group inequality What is the eect of a rise in within-group inequality of income on the level conict? Main result A Lorenz-worsening in the income distribution, with the property that some individuals with w below c lose income while those with w above c gain income, leads to higher group activism.
Rise in within-group inequality What is the eect of a rise in within-group inequality of income on the level conict? Main result A Lorenz-worsening in the income distribution, with the property that some individuals with w below c lose income while those with w above c gain income, leads to higher group activism. Higher inequality leads to lower opportunity cost of activism for poorer people, hence increase its supply higher supply of nancial contribution by shifting resources to those who can contribute
Comments Two crucial assumptions 1 The situation is inherently conictual. No attempt to model group decisions into conict in the rst place. 2 Society is already divided along ethnic lines. Extension No reason why. Class conict (outcome of income inequality) may be dominant. Need an explanation for the origin of the conictual alternative (class vs. ethnic objectives)