Terrorism and the Clash of Civilizations�

Esther Haukyand Hannes Muellerz

Preliminary Draft, June 2010

Abstract

This article builds a micro founded model of the clash of cultures.The clash is de�ned as the parent�s fear of a trait change by his child inan overlapping generations model with intergenerational transmissionof cultural traits. The extent of the clash is manipulated by culturalleaders who bene�t from the cultural education e¤ort by parents. Weidentify three channels through which the leaders can increase theclash of cultures: (i) by providing bene�cial cultural values, (ii) byclaims of cultural superiority and (iii) by cultural alienation, i.e. byinducing cultural dislike towards their own group which can be e¤ec-tively induced by terrorism. We show that all three channels can bein the leader�s interest but channels (ii) and (iii) reduce the utility ofthe leader�s goup members. Additionally, the model provides a newexplanation why recent religiously motivated terror attacks were allperformed by relatively well-o¤Muslims. Finally, we o¤er an interpre-tation of the clash. Violence and terror on both sides are supported /used by opposing leaders to the bene�t of leaders in both camps andto the loss of the population at large.

�Both authors acknowledge the support of Barcelona GSE Research and the govern-ment of Catalonia. Hauk thanks the economics department of the London School of Eco-nomics for its hospitality and acknowledges �nancial support from CICYT project numberECO2009-12695 and of the programa de movilidad de profesorado project PR2009-0375.

yInstitute of Economic Analysis (IAE-CSIC), Campus UAB, 09103 Bellaterra. E-mail:esther.hauk@iae.csic.es

zInstitute of Economic Analysis (IAE-CSIC), Campus UAB, 09103 Bellaterra. E-mail:hannes.mueller@iae.csic.es

1

1 Introduction

The expression "clash of civilizations" or "clash of cultures" has become awinged word after the attacks on the twin towers on 11 September 2001.Originally brought up as term by the historian Bernard Lewis1 the expres-sion was made famous by Hungtington�s book "The Clash of Civilizationsand the Remaking of a World Order" (1996). His main hypothesis is thatthe fundamental source of con�ict in the post cold war period will be alongcultural and religious lines. In his words: "The fault lines between civiliza-tions will be the battle lines of the future" (Huntington 1993). The debatesurrounding the work has received a sense of urgency after the September11 attacks. Its merit has been debated - not the least for the di¢ culty ofde�ning a "civilization".2 Nonetheless it had an important role in framingthe debate on a whole range of issues.This article builds a micro motivated model of the clash of civilizations

in order to explain its importance for public debate and policy actions. Tobuild a model of the clash we adopt the framework of cultural transmission ofpreferences by Bisin and Verdier (2001). In their model parents maximize theutility of their children by choosing an educational e¤ort that determines theprobability that the children receive the parent�s cultural trait. If educationfails the child adopts the cultural trait of a randomly selected member of theparent�s generation (oblique transmission). The "clash" in this framework isthe parent�s fear that children will adopt a di¤erent trait.However, a realistic micro model of the con�ict needs to account for the

fact that the propagation of culture and education employs a host of profes-sional agents (Madras school teachers, the pope, politicians, book authors...)who we call "cultural leaders". They receive money from the population atlarge in exchange for the provision of cultural characteristics. In our modelthe extend of the clash is therefore determined by these cultural leaders. Ifcultural leaders cannot be controlled perfectly this delegation of trait provi-sion implies a principal agent problem. This raises the question: what arethe incentives of these "cultural leaders"?We assume that cultural leaders bene�t directly from the cultural e¤ort

of parents - through wage payments, book sales or simply the ego rents of

1Lewis used the term �rst in his article in the September 1990 issue of The AtlanticMonthly titled "The Roots of Muslim Rage".

2See also the criticism by Sen (2006) who stresses the complexity of human identities.According to Sen a "clash" on one of these dimension is a choice not a destiny.

2

power and in�uence. This is our simple conceptual contribution to Bisin andVerdier (2001). In their model, the extent of the clash is determined by somerandom source and parents react by adapting the educational e¤ort. Nobody,except for the child, bene�ts from this e¤ort. In our model cultural leadersreceive the bene�ts of the educational e¤ort and have some control over theclash. The model therefore allows for a detailed analysis of the channels thatleaders have to in�uence the clash.In the model the clash, namely the parent�s fear of a trait change, is quan-

ti�ed by the parent�s utility loss should the child change the cultural trait.The total value of a certain cultural trait does not only depend on people�sfeelings towards this trait but also on the economic opportunities peoplewith this trait are facing. We capture this by allowing for trait dependentwages. The fear of a trait change is therefore characterized by two compo-nents: a purely cultural component which we will refer to as the "culturaldislike" parameter and by an economic component capturing the economic(dis)advantage of one cultural trait over the other.Cultural leaders can in�uence the clash of cultures via the cultural compo-

nent in three ways (i), by providing cultural goods that can only be consumedwithin the culture, (ii) by claiming cultural superiority, iii) by alienating theother cultural group. While ingroup manipulation is direct, outgroup manip-ulation is typically indirect. An e¤ective way to induce cultural alienationis violence (terror attacks, formation of mobs...) aimed at the outgroup, en-couraged or at least tolerated (not condemned) by the cultural leader. Ouranalysis of cultural alienation can therefore be interpreted as an analysis ofthe incentives to engage in terrorism.Increasing the cultural dislike towards the other group (providing cultural

goods and claiming superiority) is always in the leader�s interest because itincreases both the spending per parent and the equilibrium group size. Ahigher cultural dislike of the other group towards the cultural leader�s group(alienation) induces a higher education e¤ort in the outgroup. In the resultingsteady state the leader�s group will have shrunk but both groups will exerthigher educational e¤ort. Hence, alienation means that the leader cashesin more but from fewer people. We show that the resulting radicalizationof the environment can increase the cultural leader�s income if his group issu¢ ciently large.3

3A corollary of this �nding is that modernization (members leaving the group) is con-sistent with a growing group of radicals.

3

A corollary of these results is that a clash of civilizations caused by nega-tive views on the respective outgroup is truly a con�ict between the benefac-tors of the con�ict and the population at large. The fact that cultural leadersbene�t from radicalization per se, even if it harms the size of their group,leads to perverse incentives for leaders in both camps. Leaders have an in-centive to support violence against the outgroup and can use the atrocitiesconducted against their own group to claim superiority over the outgroup.The fact that leaders might play into each others hands can explain whycultural con�ict is so hard to tackle and calm.Our theory also allows us to analyze the interaction between economic

factors and the use of violence to achieve alienation. We show that alien-ation is not a feasible strategy for the leader if his group is economically verydisadvantaged or if alienation leads to even higher economic disadvantages.This has two important implications: (i) institutional/structural limits todiscrimination might encourage alienation and hence the use of terror; (ii)a minority that is less disadvantaged economically will be culturally moreaggressive. This latter implication might explain why recent religiously mo-tivated terror attacks were all performed by relatively well-o¤ Muslims.The remainder of the paper is organized as follows. In the next section

we will defend the main building block of our model that rents generated inthe production of culture are a powerful force in the creation of sentimentsof cultural superiority and dislike towards other cultures. In Section 3 wedefend our thesis that terror attacks aimed at the outgroup are a meansto induce cultural dislike towards the perpetrator�s group and might resultin economic disadvantages. Section 4 develops the basic components of ourmodel. Section 5 contains our main results on cultural leaders and the clash ofcivilizations. Section 6 discusses the existence of destructive leaders. Section7 concludes.

2 Cultural rents and cultural leaders

The main argument in this article is that the existence of "cultural" rentscan be a strong incentive in the creation of a sentiment of fear and hatred.But what are these cultural rents and who bene�ts from them? To illustratewhat we have in mind we provide two examples of important cultural rentsenjoyed by religious leaders.

Example 1 Pakistan is a good example of a country where the state leaders

4

have lost the cultural leadership and educational role to other parties. It is es-timated that Madrasas and Mosques in Pakistan gather over 70 billion rupees(around U.S. $ 1.1 billion) from resident Pakistanis each year. The aim ofMadrasas is explicitly the promotion of an Islamic ethos in students. Whilehuman capital accumulation (learning how to read and write for example) ispart of the education a major criticism of these schools is that they do notprepare students for a life outside the Madras and Mosque world.Attempts of changing the curriculum of Madrases has been met with �erce

resistance by the ulema (teachers running the seminaries). The InternationalCrisis Group summarizes in its report:"Any suggestion of change in the traditional sector of Islamic instruction

makes the clergy suspicious of government intentions. They are willing toteach non-religious subjects but "secularisation" is their worst fear, and theyvow �ercely to resist it.", page 3In fact, the rents (in terms of in�uence and �nance) from running the

cultural formation in Pakistan are so large that the Pakistani government hasbecome dependent on the groups running it.

Example 2 The Vatican is another example of cultural leadership with sub-stantial rents. A rough idea of the amount of revenue channeled through theCatholic church is the budget of the Holy See - the episcopal jurisdiction ofthe Catholic Church in Rome. It alone commands revenue above 230 millionEuros per year. Altogether, the Vatican estimated employment of 408,024priests of the Catholic Church in 2007.4 In addition, the pope enjoys consid-erable media attention and in�uence. His word is taken as representative forover a billion members of the church.

But religious leaders are not the only ones who bene�t from culturale¤orts of the population. Journalists, politicians and even academics bene�tdirectly from the desire of the population to be culturally educated. Thehistory is crowded with cases where politicians / journalists or academicswho act as cultural leaders directly bene�t from a cultural con�ict in thein�uence or even resources they receive. A simple example is the book "Clashof Civilizations" itself which has received considerably more attention after11 September 2001.

4Annuario Ponti�cio 2007, Vatican City: Vatican Publishing House.

5

3 Terrorism leads to cultural alienation anddiscrimination

In this section we try to motivate our assumption that terrorist attacks andthe use of violence can be interpreted as indirect manipulation of the out-group by the cultural leader. The leader�s aim is to induce fear against hisown group leading to the cultural alienation of his group and might resultin economic discrimination. Hence, the issue arises whether terrorist attacksdo indeed have these e¤ects and why.A terrorist attack clearly creates a threat situation. Most animals nat-

urally respond to threat with either "�ight or �ght". Evolutionarily we areprogrammed to feel the urge to �ght when �ight is not an option. Anger andhatred of the enemy is a mental preparation for a �ghting survival situation.In the complex modern world this evolutionary response fuels a "us/them"mentality because in a �ght situation we need an enemy to hate (Freyd(2002)). Hence, terror attacks are likely to emphasize cultural identities andcultural di¤erences. Moreover, terror attacks remind us of the possibility ofour own death creating anxiety.5 Since cultural world views and self-esteemact as anxiety bu¤ers reminders of one�s own mortality intensify e¤orts touphold cultural world-views and self-esteem,6 in particular they increase na-tionalism and prejudice7 and the support of charismatic leaders.8

Apart from psychological experiments there is a substantial body of em-pirical evidence that seems to con�rm that terror increases prejudice. Ac-cording to the FBI hate crimes targeting Muslims increased by 1600% from2000 to 2001. Also hate crimes against people of Middle Eastern origin in-creased from 354 attacks in 2000 to 1501 attacks in 2002 (Oswald, 2005).Sheridan (2006) investigates levels of self-reported racial and religious dis-

5For a nice description of the psychology of terror see McCauley (2007).6This theory called "terror management theory" (TMT) is due to Greenberg et al.

(1986) and Solomon et al. (1991). TMT has been empirically con�rmed in over 175published experiments (Cohen et al. (2004)).

7Das et al. (2009) show that terrorist news increase prejudice about outgroups. Bar-Taland Labin (2001) study the e¤ects of terrorist attacks carried out by Palestinian extremistson Israel�s stereotypes, social distance etc towards Palestinians, Jordanians and Arabs.While in general adolescents� perceptions were di¤erentiated towards the three targetgroups, most negative towards Palestians, most positive towards Jordans (no con�ict withIsrael) and intermediate towards Arabs, the terrorist attacks caused all stereotypes of allthree groups to change in a negative direction.

8see Cohen et al (2004) and Landau et al. (2004))

6

crimination in a sample of 222 British Muslims. Respondents indicate thatfollowing September 11th, 2001, levels of implicit or indirect discriminationrose by 82.6% and experiences of overt discrimination by 76.3%. Kam andKinder (2007) provide evidence that American ethnocentrism de�ned as thecommonplace human tendency to partition the social world into virtuous in-groups and nefarious outgroups powerfully underwrites support for the waron terrorism, and the strength of the relationship between ethnocentrismand opinion is in�uenced in part by the extraordinary events of 9/11. Hitlanet al. (2007) examine prejudice and perceived threat toward Mexican andArab immigrants/immigration along the U.S./Mexico border before and af-ter 9/11. While American identity is shown to be positively related to higherprejudice and threat perceptions, prejudice and symbolic threat perceptionof Arab immigrants increased sharply after the September 11 attacks. ForIsrael Canetti-Nisim et al. (2009) provide evidence that terrorism inducesnondemocratic attitudes that threatens minorities.Prejudice might not only a¤ect cultural dislike parameters but widen the

economic gap between di¤erent cultural traits by leading to increased labormarket discrimination. A terror attack leads to a higher visibility of theperpetrator�s group emphasizing negative cultural aspects - the group mightbe perceived as dangerous - and might increase the wage gap by increasingan existing taste for discrimination.9 But even in the absence of a taste fordiscrimination, a terror attack might lead to increased wage discriminationif employers have to use interviews to predict an employee�s productivity.Typically interviews with somebody belonging to a di¤erent culture producenoisier signals than interviews with members from the same culture.10 Terrorattacks increase the noise of the signal of the a member of the other cultureeven further because there is a small probability that this person might bedangerous which could lead either to direct damage or to a productivity lossfollowing an arrest. Since the employer would like to be compensated for thisrisk, the wage o¤er to this worker drops.While the prediction on the theory side is clear, what does the data tell us?

Kaushal et al. (2007) �nd a real wage decline of 9-11% and reduced interstatemobility for �rst- and second generation Arab and Muslim men in the US

9This will happen in Becker�s (1971) seminal work on wage discrimination. Similare¤ects will be found in equilibrium search models with �rm discrimination (Black (1995),Rosen (1997)), co-worker discrimination (Sasaki (1999)).10Examples of the statistical discrimination literature are: Phelps (1972), Aigner and

Cain (1977), Morgan and Vardy (2009) and Cornell and Welch (1996).

7

following 9/11. Similarly, Dávila and Mora (2005) �nd evidence of wage dropof Arab men in the US after 9/11. Rabby (2007) uses household survey datasets to estimate the e¤ects of 9/11 and the July 2005 London bombing onthe labor market outcomes of Arabs, South-Asians and Muslims living in theUS and in the UK in a di¤erence-in-di¤erences framework. Both terroristattacks did not a¤ect employment of the target groups - male immigrantsfrom Muslim-majority countries and for the UK also men who are Muslimsby religious a¢ liation - across all ages. However, in the US labor market, therelative earnings of the target group decreased; in particular for those withnativity pro�les closer to the terrorists�. Employment of very young (age 16to 25) target-group men decreased after 9/11 both in the USA and in Britainand after the July 2005 London bombings in Britain. Hence these studiescon�rm that terrorist attacks associated with the minority group can lead toworse economic opportunities for this group.

4 The model

Our basic model is based on Bisin and Verdier�s (2001) model of culturaltransmission of preferences. In this model a society of size 1 has two possiblecultural traits, trait 1 and trait 2. The fraction of individuals having trait 1is q and the fraction of individuals having trait 2 is q2 = 1 � q. We followHauk and Saez-Marti (2002) in the way overlapping generations are modeled:a Poisson birth and death process keeps the population size of active agentsconstant.11

We adopt Assumption 1 from Bisin and Verdier (2001), that parents haveimperfect empathy: they can correctly evaluate the economic perspectives oftheir children but evaluate their child�s future utility through their own eyesand therefore would like their children to adopt their own cultural trait. Inother words, let V ij be the utility of a type i parent of having a trait j child.Imperfect empathy implies that V ii > V ij so that �1 = V

11 � V 12 > 0 and�2 = V

22�V 21 > 0. The parameters �1 and �2 can therefore be interpretedas the fear of a trait change within the family.

11The survival probability of an active agent is � each period. With probability 1�� anactive agent has a child without any predetermined preferences who will becomes activethe next period. For simplicity we look at life-time values. This saves on notation. If weonly looked at per period values everything would have to be multiplied by 1

1�� to getlife-time utilities.

8

Parents choose their education e¤ort di which determines the probabilitythat their child will be of their own type. We assume that the costs of edu-cation are C(di) = 1

2d2i . If education fails, the child bumps into a randomly

chosen member of the parent�s society and copies his preferences. Thereforethe probability P ij of a type i parent to have a type j child are given by:

P 11 = d1 + (1� d1)qP 22 = d2 + (1� d2)(1� q)P 12 = (1� d1)(1� q)P 21 = (1� d2)q

and di is chosen to maximize

maxdiUi = P

iiV ii + P ijV ij � 12d2i (1)

which for the di¤erent groups can be written as

maxd1V 11 � (1� q)(1� d1)�1 �

1

2d21 (2)

maxd2V 22 � q(1� d2)�2 �

1

2d22 (3)

leading to the following optimal education e¤orts

d1 = �1(1� q) (4)

d2 = �2q (5)

which means that educational e¤ort is higher the greater the fear of a traitchange and the smaller the respective trait group. Direct and oblique social-ization are cultural substitutes.The population dynamics can be written as follows:

qt+1 = �qt + (1� �)(qtP 11t + (1� qt)P 21t )= �qt + (1� �)qt((d1 + (1� d1)qt) + (1� qt)(1� d2))

qt+1 � qt = (1� �)qt(1� qt)(d1 � d2)Hence in steady state

0 = (1� �)q(1� q)(d1 � d2)

9

which has three rest points, q = 0 and q = 1 and the interior rest pointd1 = d2 which by Proposition 1 in Bisin and Verdier (2001) is the only stablerest point. Hence the stable steady state is determined by

�1 (1� q�) = �2q�

so that the equilibrium group size of trait 1 is given by

q� =�1

�1 +�2

: (6)

We will concentrate our analysis on this equilibrium q� and consider whichsteady state outcome would be the most preferred outcome of the culturalleader of group 2. We will refer to trait 1 as the "majority" trait and trait 2as the "minority" trait.Note that the equilibrium group size and the educational e¤ort can be

expressed as a function of �1 and �2, the fear of a trait change within thefamily. We therefore �rst examine what determines this parameter.

4.1 Economic opportunities and cultural dislike

We assume that each trait has an intrinsic value uij which is the pleasurederived by a type i parent if his child holds cultural value j. Since by as-sumption parent�s care about the survival of their cultural trait uii � uij. Wewill refer to fi � uii � uij as cultural dislike. In order to model the clash ofcivilizations we assume that cultural leaders can a¤ect these values. Detailsare discussed in the following section.Apart from cultural values the utility of parents is also determined by

the economic opportunities of their o¤springs. We assume that these areobjecti�ed, i.e. the perspective on economic opportunities is independent ofthe trait. In general, we will refer to the life-time economic opportunities aswages wi where i refers to the cultural trait of the child. We assume thatw1 � w2. This assumption can re�ect two di¤erent economic environments:(i) everybody works in the same sector but there is wage discriminationagainst cultural trait 2 (or equivalently this trait has a lower productivity),(ii) the di¤erent traits work in separated sectors and the minority sector isless productive.As a �rst step it is useful to combine the cultural and economic factors

in the expressions of V ij (the utility of a type i parent of having a trait j

10

child). They are

V 11 = u11 + w1

V 12 = u12 + w2

= u11 � f1 + w2V 22 = u22 + w2

V 21 = u22 � f2 + w1:

where we have rewritten the utility after a trait change with the help of ourde�nition of cultural dislike. The fear of a trait-change for the majority, �1,can be written

�1 = f1 + (w1 � w2)where the �rst term re�ects the cultural dislike while the second term re�ectsthe loss of economic opportunity of the child. Keep in mind that the fear ofa trait-change is what gives parents an incentive to educate their children.Hence, majority parents are motivated both by their cultural preferences andthe anticipation of their children�s economic well-being. In the minority, feartakes the form

�2 = f2 � (w1 � w2)which follows the logics of the majority except for now, unequal economicopportunities demotivate educational e¤ort. For cultural survival to be pos-sible �2 > 0. Hence (u22 � u21) > (w1 � w2) is needed for the survival oftrait 2. In other words, the minorities dislike for the majority trait has tocompensate the wage di¤erential if the minority trait is to survive in the longrun.The overall level of fear in society is

�1 +�2 = f1 + f2

and as shown above the equilibrium size of the majority is given by

q� =f1 + (w1 � w2)

f1 + f2

which implies that the equilibrium size of the majority is increasing in theextent of wage discrimination against the minority and the cultural dislikeof the majority towards the minority.In what follows we use these results to explain how cultural leaders and

the clash of civilizations are linked.

11

5 Cultural leaders and the clash of civiliza-tions

Our explanation of the clash of civilizations hinges on two assumptions. First,we assume that cultural leaders bene�t directly from the educational e¤ortsof their group, i.e. by increasing the attention, power or income they receive.In order to model this we assume that the leader maximizes the sum ofcultural expenses in the minority given by

U2leader =1

2d22(1� q) =

1

2�22q2(1� q): (7)

Intuitively cultural leaders bene�t from the extent of fear in their group �2

and the size of the group (1 � q). The term q2 captures the existence of apublic good problem in cultural transmission. Parents have an incentive tofree-ride on the education of society at large. The larger the group of theleader, the more free-riding opportunities society o¤ers. Hence, per capitale¤ort is decreasing in group size.The second assumption is that cultural leaders can a¤ect the extent of

the clash (directly or indirectly). Our discussion of the fear parameters aboveallows us to identify three channels for this in�uence.Cultural Goods: leaders can provide cultural goods that can only be

consumed within the setting of the trait. In the model this implies a riseof uii. Examples are the provision/teaching of rules or knowledge that helpindividual or community life, creation of art that are complementary to otherworks in the culture or simply entertainment that is accessible only for thosehaving the trait.Cultural Superiority: if the possibility of providing utility-enhancing

cultural goods is very costly, leaders might try to lower, uij, which is the per-ception that parents have of the other culture. Examples are the descriptionof the other cultural trait as barbaric, stressing the lack of freedom in thatgroup or the use of religious arguments (the prospect of hell for example)to instill fear of a trait change. An important driver of this sentiment arestories of injustices or atrocities towards the ingroup as in Glaeser (2002).Cultural Alienation : another possibility of cultural leaders is to in-

crease the fear within the other group, uji: Most prominently this is achievedthrough violent acts/terror against members of the other trait which makeexplicit reference to the fact that the Ingroup trait is responsible for the vi-olence. The Ingroup trait will be regarded as radical, harmful or barbaric

12

by the Outgroup. To some degree this is what happened after the Septem-ber 11 attacks. But, of course, there are other possibilities for alienation.Provocations, threats or public displays of hatred in the name of the trait byreligious and political leaders can be e¤ective ways to alienate other traits.Indeed, the clash of civilizations features an incredible amount of culturalleaders harming their own group with messages of intolerance and hatred.We now analyze the incentives of the cultural leaders to follow di¤erent

clash strategies. Both production of cultural goods and claims of culturalsuperiority over the outgroup are indeed common by both political or reli-gious leaders. Our model con�rms that they are in the interest of culturalleaders.12

Proposition 1 If cultural leaders can a¤ect f2 � u22 � u21 they will try tomaximize it. The minority population always supports an increase of u22 andnever supports a decrease of u21:

Proof. See appendix.In words, the cultural leader has a large interest in raising cultural dislike

of the minority group.13 This claim can be either supported through theprovision of cultural values (raising u22) or through making the other cultureappear inferior (lowering u21). Since parents have to live with a chance thattheir children change culture, the latter strategy is less desirable from theirperspective as it increases anxiety.More surprisingly, perhaps cultural leaders can have an interest to alienate

the Outgroup.Violence by the minority can either trigger increased cultural dislike to-

wards the minority or economic discrimination. Formally, the religious leadercan a¤ect either f1 or (w1 � w2) through his/her support for violence. Notethat per se both these changes do not play in the minority�s favor.

Proposition 2 Alienation, i.e. an increase of cultural fear (lower u21) to-wards the minority is supported by the cultural leader as long as the minorityis su¢ ciently large (q� < 2

3): Alienation it is not in the minorities interest.

12In our main analysis we abstract from the cost side of changing f1 or f2. The analysisreveals when the marginal bene�t of increasing f1 or f2 is positive, and hence might bein the leader�s interest. If increasing f1 and f2 is costly, leaders would choose the pointwhere this marginal bene�t equals the marginal cost. We will discuss costly increases off1 and f2 later in this section.13For costly changes in f2, the result can be interpreted as saying that cultural leaders

will maximize f2 subject to the cost of producing it.

13

Proof. See appendix.Proposition 2 is our main �nding. The intuition for this result is that

the leader prefers a small radical group to a large unmotivated group. Anincrease in dislike towards the minority leads to a radicalization and increasedindividual e¤ort in education. The optimum from the point of view of theleader is reached when the minority is su¢ ciently small.The appendix shows that a rise in cultural dislike towards the minority

a¤ects the minority indirectly. It increases the e¤ort of majority parents,shrinks the minority and therefore forces minority parents to become moreradical to make up for the decrease in free-riding possibilities.One interesting question is whether increasing inequality will increase or

decrease cultural alienation. To answer that question note that by

q� =f1 + (w1 � w2)

f1 + f2:

the wage gap (w1�w2) increases the size of the majority q�: Remember thatfor optimal alienation the leader strives for q� = 2

3. If q� increases due to an

increase in (w1 � w2) then f1 needs to fall to decrease q� again.

Corollary 1 If cultural leaders maximize their utility using alienation (f1),alienation decreases with the wage gap (w1 � w2). In other words, a minoritythat is less disadvantaged economically will be culturally more aggressive.

Corollary 1 might appear puzzling. It stems from the fact that, in theleader�s optimum, wage discrimination and alienation are substitutes. If theleader�s group is economically very disadvantaged the pressures on the mi-nority to change trait are large. This implies that alienation as a strategyof increasing revenue is not feasible. The fact that recent religiously mo-tivated terror acts were all performed by relatively well-o¤ Muslims couldbe seen as evidence that this concept of a "capacity for alienation" is nottoo far fetched.14 Our model therefore provides a new explanation for thisphenomenon.15

14Empirical studies demostrate that terrorist operatives tend not to be from societies�worst-o¤ socioeconomic group. On the contrary, they have above average educationalachievements and are less likely to live in poverty than the average person (Krueger andMalechova (2003), Berribi (2007)).15Bueno de Mesquita (2005) suggests that this high quality of terrorists is due to the

terror organization screening the candidates and only picking the best of them. In his

14

If terror is a form to implement the strategy of alienation, Corollary 1also suggests that the attractiveness of using violence drops if the majoritydoes not only react with alienation but also increased wage discriminationtowards the minority. This has the following implication.

Corollary 2 Restrictions on wage discrimination make alienation more tempt-ing for the cultural leader.

Proof. See appendix.We can think of two di¤erent reasons why wage discrimination does not

react to alienation. (i) Institutional constraints prevent majority employersfrom discriminating against the minority. (ii) The minority has developed itsown sub-economy that is independent of cultural sentiments in the major-ity. It is not uncommon that the minority develops an informal sector thatrelies on minority social networks for enforcement etc.16 Corollary 2 tells usthat both types of restrictions on wage discrimination - instead of helpingthe minority - might back�re because they increase the leader�s capacity foralienation.We now turn to the issue when the leader prefers ingroup manipulation

(changing f2) to outgroup manipulation (changing f1).

Proposition 3 If an increase of f2 and an increase of f1 have the same costto the leader, the leader prefers to increase f2 for all q� > 2

5:

Proof. See appendix.It will us useful to illustrate Proposition 3 for the case in which both

increasing f1 and increasing f2 is costless for the leader. Proposition 2 statedthat outgroup violence was desirable for the leader for all q < 2

3. Hence the

leader will switch from supporting violence to increasing f2 at q = 25. It

is important to note that the e¤ect of these two strategies on q� is exactlyopposite. While increasing f2 raises q�, increasing f1 lowers it. The result isa mixture between provision of cultural goods and alienation at a minoritysize 1� q� = 3

5:

model economic opportunities and the ability to carry out terrorist tasks are correlated,but the highest types will not engage in terrorism. Our model suggests that it is thehighest types for which the strategy of cultural alienation is most successful.16In an international context the lack of an economic backlash is most obvious in oil

producing countries. Economic opportunity in these resource rich countries remains inde-pendant from the cultural feeling of the West towards these countries.

15

If changing f1 and f2 is not costless, Proposition 3 still tells us that themarginal bene�t from increasing f2 is bigger than the marginal bene�t fromincreasing f1 if and only if q� > 2

5. But now the leader will also have to

evaluate the marginal cost of each strategy. Recall that raising f2 can bedone in two ways, by the provision of utility enhancing cultural goods andby successful claims of cultural superiority leading to utility losses of theleader�s group. While the provision of positive cultural goods is likely to becostly, claims of cultural superiority might be cheap, especially if they can berooted in actions from the other group. The present "clash" has generated ahost of atrocities on both sides that can be used to claim cultural superiority.Hence, it is reasonable to assume that at least at some point increasing f2 ischeaper through claims of cultural superiority leading to a con�ict of interestsbetween the leader and his group.Increasing f1 is a costly strategy if the leader has to sponsor violence

and terror. However, if all that is required is approval of an already exist-ing (maybe even independent), terror organization and words of encourage-ment, increasing f1 can be surprisingly cheap. This might explain why terrororganizations with a strong religious and cultural emphasis are not easilycondemned by leaders from their own culture.Claims of cultural superiority and attempts to alienation are not bene�cial

for the population at large. A comparison of propositions 1 and 2 allows usto make quite bold statements about the con�ict of interest between culturalleaders and the population at large. Clearly, due to symmetry cultural leadersof the majority will have the same interest as their peers in the minority.Similarly the majority population shares the interest in uii with the minority.

Corollary 3 Cultural leaders of both majority and minority share an interestin escalating the clash as long as both groups are not too unequal in size.Members of the groups prefer a setting with high uii and low uij:

Corollary 3 delivers a striking re-interpretation of the "clash of civiliza-tions". The clash is to a large degree due to the existence of cultural leadersin both groups who have an incentive to provoke intolerance and fear on bothsides of the clash. While the population bene�ts from the production of pos-itive value it is a¤ected negatively by the fear created by an atmosphere ofintolerance. Increasing radicalization on both sides requires and increasingamount of investment in education - this furthers the interest of leaders.An even more sinister view can be reached if we combine our results with

Glaeser (2005) who shows that claims of superiority are facilitated by violent

16

attacks on the group. Leaders on the two sides could then play in each othershands by supporting violence/terror and using the other group�s violence todraw the picture of a barbaric culture on the other side of the clash. Giventhat real atrocities occur, it is very hard for the population to ignore thesemessages.

6 Destructive leaders

The above model takes the existence of leaders as given. In this context itshows that leaders can bene�t from cultural alienation and cultural supe-riority at the expense of the utility of the population. Hence, in as far asalienation and claims of cultural superiority can be e¤ective, cultural leaderswill use them. But this harms the population at large. So why does thepopulation fail to replace these destructive leaders?While we will hint at a few reasons below, a detailed answer to this

question is left for future research. It would require to model competitionbetween leaders. This competition can be internal or external. If there isinternal competition within the trait to become a cultural leader potentialcandidates would o¤er "trait platforms" and are then selected by the popu-lation. If parents foresee the consequences of their leader choices correctly,internal competition is likely to be an e¤ective tool to lower abuses. However,once a leader is in o¢ ce, he will have access to the cultural rents and mightuse these rents to buy o¤ potential challengers and to establish a monopolyposition and then renege on his platform. Moreover, even if leaders do notactively engage in cultural alienation and cultural superiority claims and onlyprovide positive cultures, they have no incentives to silence any attempts ofalienation and cultural superiority claims that arise among their followers.They will free-ride on any radical that exists. Additionally, leader selection isoften institutionalized and leaders might manipulate these institutional rulesin their favor.17

External competition to a cultural leader arises from the possibility tolaunch a new cultural trait. This way of creating competition is less likely

17Two striking features of the selection of especially religious leaders are that oftenthese leaders are elected for life and that there exist culturally entrenched institutionsthat prohibit material wealth to these leaders. However, this restriction is often onlynominal: the leader does not own personal wealth but manages at least to some extentthe institutional wealth.

17

to align incentives between population and leaders as the incentive to claimcultural superiority might even be ampli�ed with rising competition. Asshown by Bisin et al. (2008) and Montgomery (2008) if there are morethan two cultural traits �i > 0 is not su¢ cient for cultural survival. Usingthe same assumption on educational costs as in the present article, thesepapers show that the two types with the highest �i - the two most "radical"types - survive but types with lower �i might become extinct.18 Hence,successful entry of a new type requires either strong economic advantagesassociated with the type or a high cultural dislike towards other types orboth. Cultural traits with less economic possibilities need to emphasize theircultural superiority: otherwise they might simply be eliminated in the �ghtfor cultural survival.Whatever the reasons why the population tolerates destructive leaders,

destructive cultural leadership is an empirical fact. Alienation and claims tosuperiority implies a rise in con�ict and numerous historical examples haveshown that rising con�ict has the perverse e¤ect of rallying people behindits leaders.19 As explained above, these e¤ects are especially important ifcultural leaders from both sides of the "clash" collude by simultaneouslyincreasing alienation.20

7 Conclusion

Our �ndings indicate that if the clash of cultures can be manipulated byleaders who gain rents from educational e¤orts then there is a danger ofconsiderable radicalization. However, our model allows for a more in-depthanalysis of the di¤erent characteristics of this radicalization. We identify

18The proofs require a constant �i for all types. This is the case in the present paper.19The rally around the �ag e¤ect is mainly studied for leaders involved in interstate

con�ict. See e.g. Baker and Oneal (2001) for rally around the �ag e¤ects in the US andLai and Reiter (2005) for rally around the �ag e¤ects in the UK.Colaresi (2004) uses event history techniques to analyze the likelihood of all state leaders

between 1950-1990 to be replaced from o¢ ce depending on whether they were more orless aggressive than the leader of the rival state in a con�ict situation. He �nds strongevidence that leaders who are less agressive than their rivals pay an electorial price (aremore likely to lose o¢ ce).20The 1994 Rwanda Genocide is an important example how e¤ective hate messages can

be. Yanagizawa (2009) estimates that 9% of the killings can be explained by the infamoushate radio station Radio RTLM which increased violence by 65 to77% in villages withcomplete radio coverage.

18

mostly three channels through which the leader can increase the clash ofcultures, namely

1. provision of bene�cial cultural values that increase utility in-group;

2. claims of cultural superiority: the spread of negative attitudes towardsother cultures that leads parents to be afraid of a trait change of theirchildren;

3. cultural alienation: the spread of negative attitudes in other groupsagainst the leader�s own culture which leads to a shrinking in size ofthe leader�s group but higher per capita e¤ort.

By revealing these channels our paper uncovers a serious commitmentproblem between the population at large and its cultural leaders. On theone hand, the population can bene�t massively from the provision of culturalgoods (art, moral guidance, rituals, entertainment); this channel is whatmakes the world go round. It cannot be stressed enough that the deeplyrooted desire of humans to consume culture and to grow in that sense hascarried wonderful blossoms throughout the years. On the other hand, thepopulation is harmed by channels 2 and 3: cultural leaders have the incentiveof spreading fear within the group and outside of the group to increase percapita expenditure. Precisely because we need to rely on cultural leaders forthe provision of positive cultural goods (channel 1) it so hard to abolish thosecultural leaders who abuse their position to amass riches through channels 2and 3.By claiming cultural superiority the second channel serves as a barrier

towards adopting another culture. Our model shows that it is optimal forcultural leaders to maximize this barrier. This provides an explanation whycultural leaders put a lot of emphasis on their trait appearing to be verydi¤erent and incompatible with other traits. There are some indicators thatthey can be successful. Many cultural/religious con�icts are characterized byrules of rejection or even harm of children who attempt to cross-breed withother cultures or religions. This goes as far as that parents prefer the deathof their children to a change of trait.While it is easy to understand that cultural leaders want barriers to avoid

children of their �ock changing trait, it is more surprising that they mightbene�t from the channel of cultural alienation. Cultural alienation acts asa barrier towards children of outsiders adopting the trait. The main reason

19

why cultural leaders maintain this barrier towards their own trait and wanttheir group to be hated is because they bene�t from the education e¤ort ofthe parents belonging to their group. Cultural leaders want to cash in thehighest per capita education e¤ort from the maximum amount of people.However, direct and oblique cultural transmission are substitutes and henceper capita education e¤ort falls when the group grows. The reduction in percapita education e¤ort might not be compensated by a bigger group size.Our model shows that educational rents can lead to a bias towards smallradicalized group with strong cultural leaders.Pakistan is an example of this phenomenon. The massive in�uence of

religious leaders in Pakistan is their command over rents and not votes. In-deed, there is little control over votes: in the 1997 elections, the JUI ofFazlur Rahman was the only religious party to win seats in the 217-memberNational Assembly. However, religious leaders in Pakistan are powerful be-cause Pakistani sects are groups with forti�ed, closed ranks. The per capitale¤orts are so high that sectarian divisions are militarized. The clergy backedMusharraf�s coup and supported his policies towards India. They were, how-ever, powerful enough to defy any attempt to reform madrasas system oreliminate Islamic militancy.We see the Pakistani example as an exception only in degrees not in the

mechanism. All around the world there are leaders who send messages thatare not attractive to a majority but lock in a minority. With all the talk ofa "clash of civilizations" and the corresponding pictures on our TV screensWestern products stay in high demand in the Middle East and are actuallygathering market shares at breath-taking pace.21 As our model predictsthis "modernization" is consistent with a shrinking increasingly radicalizedminority which follows religious leaders despite the considerable harm theyin�ict on them. It is only the leader not its followers who bene�ts fromspreading cultural hatred.Similarly, only the cultural leaders bene�t from terrorism and cultural

violence which is one way to achieve alienation in other groups. How oftenit is used will depend on its cost relative to other channels. If all that isrequired is approval of an already existing terror organization and wordsof encouragement, the use of alienation could be surprisingly cheap. How

21Cocal Cola per capita product consumption (just as an example) was growing by63 percent in Egypt, 22 percent in Morocco and 275 percent in Pakistan between 1999and 2009. Source: Coca Cola Company, EURASIA & AFRICA Operating Group 2010,author´ s calculation.

20

central terrorist organizations are to the clash of civilizations might be oneof the reasons why some are longer lived than others.We argue that terror (and other forms of violence against civilians) from

both sides could be part of a collusion equilibrium played by leaders. In thisequilibrium, violence against civilians helps hold in place claims of culturalsuperiority. The combination of September 11 and Abu Graib serves inter-ested parties on both sides to fabricate stories of a barbaric outgroup whoseonly aim is to destroy and humiliate the ingroup.

References

Aigner, Dennis and Glen G. Cain (1977) "Statistical Theories of Discrim-ination in the Labor Markets" Industrial and Labor Relations Review30(2): 175-87

Bar-Tal, Daniel and Daniela Labin (2001) "The e¤ects of a major event onstereotyping: terrorist attacks in Israel and Israeli adolescents�percep-tions of Palestinians, Jordans and Arabs", European Journal of SocialPsychology 31, 265-280

Baker, WilliamD. and John R. Oneal (2001) "Patriotism or Opinion Leader-ship?: The Nature and Origins of the "Rally �Round the Flag" E¤ect",The Journal of Con�ict Resolution 45(5) 661-687

Becker, G.S. (1071) The economics of discrimination University of ChicagoPress, Chicago

Berrebi, Claude (2007) "Evidence About the Link Between Education, Povertyand Terrorism Among Palestinians", Peace Science and Public Policy,13(1)

Bisin, Alberto and Thierry Verdier (2001) "The economics of cultural trans-mission and the dynamics of preferences", Journal of Economic Theory97: 298-319

Bisin, Alberto, Giorgio Topa and Thierry Verdier (2008) "Cultural Trans-mission, Socialization and the Population Dynamics of Multiple StateTraits Distributions", International Journal of Economic Theory 5:139-154

21

Black, Dan A. (1995) "Discrimination in an Equilibrium Search Model",Journal of Labor Economics 13(2): 309-33

Bueno de Mesquita, Ethan (2005) "The Quality of Terror", American Jour-nal of Political Science 49(3): 515�530

Canetti-Nisim, Daphna, Eran Halperin, Keren Sharvit and Stevan E. Hob-foll (2009) "A New Stress-Based Model of Political Extremism. Per-sonal Exposure to Terrorism, Psychological Distress, and ExclusionistPolitical Attitudes", Journal of Con�ict Resolution 53(3): 363-389

Cohen, F., S. Solomon, M. Max�eld, T. Pyszczynski and J. Greenberg(2004) "Fatal Attraction: The E¤ects of Mortality Salience on Evalua-tions of Charismatic, Task-Oriented, and Relationship-Oriented Lead-ers", Psychological Science Volume 15 Issue 12, Pages 846 - 851

Colaresi, Michael (2004) "When Doves Cry: International Rivalry, Unrecip-rocated Cooperation, and Leadership Turnover", American Journal ofPolitical Science 48(3), 555-570

Cornell, Bradford and Ivo Welch (1996) "Culture, Information and Screen-ing Discrimination" Journal of Political Economy 104(3): 542-71

Dávila, Alberto and Marie T. Mora (2005) "Changes in the earnings ofArab men in the US between 2000 and 2002" Journal of PopulationEconomics 18:587�601

Das , E. , B.J. Bushman, M.D.Bezemer, P. Kerhof and I.E.Vermeulen (2009)"How terrorism news reports increase prejudice against outgroups: ATerror management account", Journal of Experimental Psychology 45,453-459

Freyd, Jennifer J. (2002) "In the Wake of Terrorist Attack, Hatred MayMask Fear" Analyses of Social Issues and Public Policy pp. 5�8

Glaeser, Edward L. (2005) "The Political Economy of Hatred", The Quar-terly Journal of Economics 120(1): 45-86

Greenberg, J. Pyszczynsji, T. and S. Solomon (1986) "The causes and con-sequences of a need for self-esteem: A terror management theory" inR.F.Baumeister (ed) Public Self and Private Self (pp. 189-212). New-York, Springer Verlag

22

Hauk, Esther and Maria Saez-Marti (2002) "On the cultural transmissionof corruption", Journal of Economic Theory 107, 311�335

Hitlan, Robert T. and Kimberly Carrillo and Michael A. Zárate and Shel-ley N. Aikman (2007) "Attitudes Toward Immigrant Groups and theSeptember 11 Terrorist Attacks" Peace and Con�icr: Journal of PeacePsychology 13(2), 135�152

Huntington, Samuel P. (1993) "The Clash of Civilizations?", Foreign A¤airs

Huntington, Samuel P. (1996) The Clash of Civilizations and the Remakingof World Order

Kam, Cindy D. and Donald R. Kinder (2007) "Terror and Ethnocentrism:Foundations of American Support for the War on Terrorism" The Jour-nal of Politics 69(2): 320�338

Kaushal, Neeraj, Robert Kaestner and Cordelia Reimers (2007) "LaborMarket E¤ects of Septmebr 11th on Arab and Muslim Residents inthe United States", J Human Resources XLII(2):275-308

Krueger, Alan B. and Jitka Maleckova (2002) "Education, Poverty and Ter-rorism: Is There a Causal Connection?" Journal of Economic Perspec-tives 17(4): 119-44

Lai, Brian and Dan Reiter (2005) "Rally �Round the Union Jack? PublicOpinion and the Use of Force in the United Kingdom, 1948�2001",International Studies Quarterly 49, 255�272

Landau, M.J., S. Solomon, J. Greenberg, F. Cohen, T. Pyszczynski, J.Arndt , C. Miller, D. Ogilvie and A. Cook (2004) "Deliver us fromEvil: The E¤ects of Mortality Salience and Reminders of 9/11 on Sup-port for President George W. Bush " Personality and Social PsychologyBulletin, Vol. 30, No. 9, 1136-1150

McCauley, C. (2007) "Psychological Issues in Understanding Terrorism andthe Response to Terrorism" chapter 2 of Psychology of Terrorism,edited by B. Bongar, L. Brown, L. Beutler, J. Brecehnridge and P.Zimbardo, Oxford University Press

23

Monteith, M. and Wintersm J. (2002) "Why we hate" Psychology Today35(3): 44-52

Montgomery, James, D. (2008) "Intergenerational Cultural Transmission asan Evolutionary Game", mimeo, Department of Sociology, Universityof Wisconsin-Madison

Morgan, John and Felix Vardy (2009) "Diversity in the Workplace", Amer-ican Economic Review 99(1): 472-485

Oswald, Debra L. (2005) "Understanding Anti-Arab Reactions Post-9/11:The Role of Threats, Social Categories and Personal Ideologies" Jour-nal of Applied Social Psychology 35(9): 1775-1799

Phelps, Edmund S. (1972) "The Statistical Theory of Racism and Sexism."American Economic Review 62(4): 659-61

Rabby, Faisal (2007) "The Post-9/11 Labor Markets for �Muslims�in theWest �Evidence from the USA and the UK" Working Paper Depart-ment of Economics, Rutgers University,

Rosen, Asa (1997) "An Equilibrium Search-Matching Model of Discrimina-tion", European Economic Review 41(8): 1589-1613

Sasaki, Masaru (1999) "An Equilibrium Search Model with Coworker Dis-crimination", Journal of Labor Economics 17(2): 377-407

Sen, Amartrya (2006) Identity and Violence: The Illusion of Destiny, W.W.Norton& Comany, Inc. New York

Sheridan, Lorraine P (2006) "Islamophobia Pre�and Post�September 11th,2001 ", Journal of Interpersonal Violence, Vol. 21, No. 3, 317-336

Solonon, S. Greenberg, J. and T Pyszczynski (1991) "A terror managementtheory of social behavior: the psychological functions of self-esteemand cultural worldviews" in M.Zanna (ed.) Advances in ExperimentalSocial Psychology 24, 93-159. Orlando, FL. Academic Press

Yanagizawa, David (2009) "Propaganda and Con�ict: Theory and EvidenceFrom the Rwandan Genocide", mimeo, Stockholm School of Economics

24

A Proof of Proposition 1

Given our underlying model @�2@f2

= 1 and @q�

@f2= �f1+(w1�w2)

(f1+f2)2. So

@U2leader@f2

= �2q�2(1� q�)� 1

2�22q�(2� 3q�)f1 + (w1 � w2)

(f1 + f2)2

We can evaluate the sign by looking at the sign of

q�(1� q�)� 12�2(2� 3q�)

f1 + (w1 � w2)(f1 + f2)2

After plugging in �2 and the equilibrium q� we get

3

2

(f2 � (w1 � w2)) (f1 + (w1 � w2))2

(f1 + f2)3> 0

Hence, religious leaders would always like to set the maximal f2 possible.We now turn to the question how an increase in f2 a¤ects the utility of theminority. This depends whether the increase in f2 is due to an increase inu22 or a decrease in u21. The utility of the minority is given by

U2 = V 22 � (1��2q�)�2q

� � 12(�2q

�)2

= V 22 ��2q� +

1

2(�2q

�)2

so that@U2@u21

= �@�2q�

@u21(1��2q

�) :

In words, an increase in cultural superiority (decreasing u21) harms utilityif it leads to an increase in equilibrium educational e¤ort. This is the case if

@�2q�

@u21< 0

or

� �1

�1 +�2

+�1�2

(�1 +�2)2 < 0

which is always the case.

25

For a change in u22 the calculation is

@U2@u22

= 1� @�2q�

@u22(1��2q

�) > 0:

The minority always bene�ts from an increase in u22 and su¤ers from adecrease in u21:

B Proof of Proposition 2

Note that @�2@f1

= 0 so that the change in indirect utility of the cultural leaderis given by

@U2leader@f1

=1

2�22q�(2� 3q�)@q

�

@f1(8)

which is positive for all q� < 23as @q�

@f1> 0.

The utility of the minority is given by

U2 = P22V 22 + P 21V 21 � 1

2d22

which can be written as

U2 = V22 � q(1� d2)�2 �

1

2d22 (9)

in equilibrium d2 = �2q�

U2 = V22 � (1��2q

�)�2q� � 1

2(�2q

�)2 :

Note that the change of utility with �2q� is

@U2@�2q�

= �1 + �2q� < 0

as �2q� < 1:

Now we are ready to show that cultural alienation is not in the minority�sinterest. Cultural dislike towards the minority, f1, does not a¤ect V 22 buta¤ects the minority�s utility through �2q

� = �1�2�1+�2

as follows

@�2q�

@f1=

�2 (�1 +�2)��1�2

(�1 +�2)2

@�1

@f1

=�22

(�1 +�2)2 > 0

26

which means that@U2@f1

< 0:

C Proof of Corollary 2

We �rst derive the leader�s optimal group size with wage discrimination.Note �rst that

@�2q�

@(w1 � w2)= 1� 2q�

@q�

@(w1 � w2)=

1

f1 + f2

with U2leader = 12(�2q

�)2 (1 � q�) the long term utility change with wagediscrimination is then

@U2leader@(w1 � w2)

=1

22 (�2q

�) (1� q�) (1� 2q�)� 1

f1 + f2

1

2(�2q

�)2

= [2 (1� 2q�)� q�]12(�2q

�) (1� q�)

The maximum is therefore at q� = 25:

We now isolate the e¤ect of discrimination by rewriting (w1�w2) = �f1,where � parameterizes how the fear of violence a¤ects wage discrimination.Hence, we can rewrite the leader�s maximum with respect to wage discrimi-nation as

q� =2

5=f1 + �f1f1 + f2

and the corollary follows immediately from the fact that df1d�< 0 in equilib-

rium.

27

D Proof of Proposition 3

We know that

@U2leader@f2

= �2q�2(1� q�)� 1

2�22q�(2� 3q�)f1 + (w1 � w2)

(f1 + f2)2

= �2

��1

�1 +�2

�2�1� �1

�1 +�2

��12�22

�1

�1 +�2

�2� 3 �1

�1 +�2

��1

(�1 +�2)2

=3

2(�1 +�2)

�4�22�

31

and

@U2leader@f1

=1

2�22

�1

�1 +�2

(2� 3 �1

�1 +�2

)@q�

@f1

=1

2�22

�1

�1 +�2

(2� 3 �1

�1 +�2

)

�1

�1 +�2

� �1

(�1 +�2)2

�=

1

2(�1 +�2)

�4 (2�2 ��1)�32�1

If raising f1 and f2 cost the same the leader prefers raising f2 if

@U2leader@f2

>@U2leader@f1

or3�2

1 > (2�2 ��1)�2

so that the leader is indi¤erent between the two at

3�21 +�1�2 � 2�2

2 = 0

or

�1 =��2 �

p�22 + 24�

22

6

as only positive numbers make sense we therefore have

�1 =2

3�2

28

which implies that the leader prefers raising f2 for all

�1 >2

3�2

() q� >2

5:

29