[Note for bibliographic reference: Melberg, Hans O. (1998), Is ethnic conflict the
outcome of individually rational actions?, www.oocities.org/hmelberg/papers/980522.htm]
Is ethnic conflict
the outcome of
individually rational actions?
A case study in the application of the tools of rational choice and game theory to
ethnic violence
by
Hans O. Melberg
The University of Oslo
May 1998
Table of contents
1. Introduction
- 1.1. General introduction
- 1.2. Detailed introduction
- 1.3. Arguments and Structure
2. Ethnic violence
- 2.1. Importance
- 2.2. Definition
- 2.3. Theories of ethnic violence
3. Three big rational choice explanations: Grand Theory
- 3.1. Selective incentives
- 3.2. Reputation
- 3.3. Including and excluding beliefs and preferences
4. Rational and Emotional as Complementary
- 4.1. Multiple motivations
- 4.2. Multiple phenomena to be explained
5. Two small rational explanations: Mechanisms
- 5.1. Laws vs. mechanisms
- 5.2. Collapse
- 5.3. Window of opportunity and Credible promises
6. Conclusion
7. References
8. Appendicies
- 8.1. Appendix 1: The formal solution of the KRMW model
- 8.2. Appendix 2: A critical note on the logic behind the KRMW results using the Chain
Store Paradox as an example
- 8.3. Appendix 3: Policy implications and the future
Is ethnic conflict the outcome of individually rational actions?
1. Introduction
1.1. General introduction
When Durkheim (1912) wrote his famous study of suicide he chose a topic
which appeared to be far out of reach for his approach. Suicide was viewed as a problem of
micro-psychology, not as a promising area for a subject concerned about the general laws
of society. Yet, Durkheim chose the latter and by so doing gave fuel to a debate that has
lasted ever since.
In this paper, I want to take a critical look at a similar project: The use of rational
choice explanations for behaviour which seems to be obviously irrational. The behaviour
under question is ethnic violence. This kind of behaviour appears irrational for at least
two reasons; one flawed and one worthy of investigation. First, the very evilness of the
phenomenon may make people believe that it is irrational - how can war, violence and
ethnic cleansing be rational? Yet, this is surely flawed reasoning. Individual rationality
as it will be defined in this paper, does not exclude bad outcomes. In fact, the well
known Prisoners Dilemma illustrates precisely that rational individuals may end up in an
inferior social equilibrium. hence, the true reason why ethnic violence appears
irrational, is the basic problem of collective action. If individuals are rational, how
can we explain that they risk their lives fighting for the larger group? It is this
question I have judged worthy of closer examination.
1.2. Detailed introduction
The basic problem of collective action is formalized in the game below
(the one-shot, two player, complete information game in figure 1):
Figure 1 The Prisoners' Dilemma
Player 2
(C)ooperate (D)efect
Player 1
Cooperate 1 , 1 b , a
Defect a , b 0 , 0
It is a well known result that as long as a>1>0>b rational and selfish players
will end up receiving 0 each. The short explanation is that whatever Player 2 does, it is
best for Player 1 to play D. If Player 2 cooperates, Player 1 can get a by defecting (and
a>1 which is what you get if you cooperate when the other also cooperates). If Player 2
defects, Player 1 should also defect since 0>b. Technically speaking the strategy C is
strictly dominated by D for both players and elimination of strictly dominated strategies
in this case leaves only one pair of strategies (D, D) which then must be the Nash
equilibrium (Roughly defined as a situation in which no player wants to change his
strategy).
A more general way of illustrating the problem, is by using so called Schelling diagrams.
On the horizontal axis we measure the number of people who play cooperation (n) and on the
vertical axis we measure utility. Sine the line representing Defect is always above the
line Cooperate regardless of the size of n (i.e. the number of cooperators) this indicates
that it is always most profitable to defect. Later we will see how assumptions about the
motivational structure may change the situation - for instance, if people have a desire to
do what the majority does.
Figure 2 The Generalized Multi-person Prisoners Dilemma
The relevance of this to ethnic violence should be clear. Ethnic violence requires actors
who are willing to risk their lives ("cooperate"), and it seems that this action
cannot be consistent with the assumption of rational and selfishness players (who would
"defect"). Despite the theoretical prediction that cooperation will not take
place, ethnic violence is an undeniable empirical phenomena and this calls for an
explanation. Is it because we are irrational or unselfish? Or is it because the game above
does not capture the true situation and its payoffs?
1.3. Arguments and Structure
It should be obvious that I cannot survey the whole literature on ethnic
violence or try to answer all the questions about this topic. My focus will be on the
extent to which we can use the tools of rational choice and game theory to help us in our
understanding of ethnic violence. To answer this question I selected a few big and small
rational choice arguments for closer examination. My conclusion, and this is my arguement,
was that many of the rational choice arguments were flawed as rational choice
arguments i.e. when they were put under closer and more rigorous scrutiny they
appeared unrealistic and/or not to give the conclusion they were supposed to. Thus, my
main criticism in this paper is not that the theories ignore emotions (which they do, see
Petersen 1997 and Cunningham 1997), but that they are simply not good and realistic
rational choice explanations.
To present this argument I first discuss ethnic violence - its importance, its definition
and the various theories of ethnic violence. In the next chapter, I critically examine
three big rational choice explanations of how the problem of collective action is solved:
using selective incentives, by introducing uncertainty, and by playing with the concepts
of beliefs and preferences. Having partly rejected these rational explanations of
collective action, I go on to examine whether there is a conflict between explanations
that focus on rational vs. emotional forces, using Schelling diagrams to explain how
rational and non-rational motivations may interact to create collective action. The next
section examines whether the main use of game and rational choice theory is to aid the
construction of small scale mechanisms that help to explain some aspects of ethnic
violence, as opposed to producing grand theories of ethnic violence.
2. Ethnic violence
2.1. Importance
According to Gurr and Harff (1994:xiii) eighteen out of the twenty-two hot
wars in 1993 involved ethnic conflict. Moreover, 25 million refugees were fleeing from
communal conflict and 4 million had died as a result of these conflicts (200 000 in 1993).
Altogether 233 significant ethnic groups were mobilized during the 1980s (Väyrynen
1991:59) and ethnic nationalism has been claimed to be the main cause behind the collapse
of the Ottoman Empire, Austria-Hungary and the Soviet Union. Surely we are facing a
phenomenon worth investigating?
Against the grim view that ethnic conflict is widespread, Fearon and Laitin (1996) ask us
not to exaggerate the occurrence of violence between ethnic groups. For instance, using
the number of languages in a country as a proxy for the number of ethnic groups, they
calculated the number of potential conflicts in 36 African countries. They then examined
the number of actual conflict and found that less than one percent of the groups that
potentially could be in conflict actually experienced violent conflict (Fearon and Laitin
1996:717).
In addition to presenting a factual picture, the statistics above suggest that a good
grand theory of inter-ethnic relations should both predict low levels of
enter ethnic violence as well as suggesting how the non-violent equilibrium might break
down. Thus, Fearon and Laitin do not think ethnic violence is unimportant, only that a
good explanation should not over-predict violence.
2.2. Definition
When is violence ethnic and what counts as violence? On the first
question, we need to examine the motive for the violence. If a Hindu kills a Muslim in
India, this should not automatically qualify for the label ethnic violence since the
motive for the killing may have been non-ethnic (economic gain and so on). Only when a
person is selected because he or she belonged to that group do we have a
clear example of ethnic violence. One may then doubt the numbers indicating the importance
of ethnicity given in the beginning of this paper. Maybe ethnicity is just a fig-leaf for
economic interests in conflicts?
As for the question of what counts as violence, we could make several distinctions. In the
mentioned article by Fearon and Laitin, they distinguished between four types of violent
ethnic conflict: regular violence, irredentism, rebellion and civil war. Other
commentators have distinguished between two types of violence: high and low (Fearon 1998).
High is defined as secessionist wars and fights over control over government. Low is
characteristic of riots and pogroms.
One might say that this is simply a question of words, but this is to miss the
methodological problems involved. When we are trying to explain the level of ethnic
violence, we have to find a number which measures ethnic violence. The problem involved in
finding this number is that violence takes many forms and cannot simply be added to create
an aggregate measure of violence going from 0 to 100 against which we can run a regression
with different explanatory variables. In short, it is difficult to add the number of
rapes, the number of killings and the number of beatings by one group against another to
create a nice one-dimensional dependent variable.
2.3. Theories of ethnic violence
Explanations of ethnic violence is commonly divided into two categories:
Rational (or instrumental) vs. Emotional (or primordial, psychological) (Gurr and Harff
1994:78). As an example of the emotional theory, consider the following quotations from
Walker Connor:
"the national bond is subconscious and emotional rather than conscious and
rational in its inspiration" (Connor 1994:xiii).
"Explanation of behaviour in terms of pressure groups, elite ambitions, and rational
choice theory hint not at all the passions that motivate Kurdish, Tamil, and Tigre
guerrillas or Basque, Corsican, Irish, and Palestinian terrorists. Nor at the passions
leading to the massacre of Bengalis by Assamesc or Punjabis by Sikhs. In short, these
explanations are a poor guide to ethnonationally inspired behaviour." (Connor
1994:74)
As a representative of the rationalist school, we may cite Russell Hardin who writes that:
"In our time there is massive mobilization of groups, especially ethnic groups,
for ostensibly group-level purposes. Individuals have identified with groups so strongly
that they seem to forgo their personal interests while seeking their group's interests.
The common understanding of the logic of collective action suggests that these
group-oriented actions cannot be rational in the sense of being individually
self-interested. Instead, seemingly, they must be irrational (people are foolish or crazy)
or extra-rational (people are motivated by moral or group commitments). The latter
inference is, however, wrong - because the former suggestion is wrong. There may be
foolishness, craziness, morality, and extra-rational group identification at work for many
participants in violent ethnic conflict. But these are given their field of play by the
individually rational tendencies to group identification." (Hardin 1995:180)
Faced with this disagreement, we have at least three options. First, we may claim that one
of the two approaches is correct. Second, we may claim that the theories are mutually
compatible so that both are correct. Third, we may argue that none is correct or useful as
a general theory. In the following three chapter I will discuss some arguments that fit
under these three general answers with the aim of evaluating the usefulness of game and
rational choice theory in the field of ethnic violence.
3. Three big rational choice explanations: Grand Theory
In this section, I want to give three examples of how one may rationally solve the
collective action problem described in the introduction.
3.1. Selective incentives
In his book The Logic of Collective Action (1961/1975:51,60-64),
Mancur Olson suggests that leaders may use selective incentives to make it profitable for
individuals to join the group. For instance, organizations intended to protect the rights
of car owners in Norway (NAF, KNA, MA and so on), often offer favourable car-rescue
services, magazines, discounts on various items and so on in order to attract members. By
so doing they make it profitable to join and the situation can no longer be described as a
Prisoners' Dilemma.
What selective benefits can be offered in the context of ethnic conflict. Russell Hardin
argues that there are two main incentives to identify with a group: "Individuals may
find identification with their group beneficial because those who identify strongly may
gain access to positions under the control of the group and because the group provides a
relatively secure and comfortable environment." (Hardin 1995:70)
While this certainly sounds plausible, closer scrutiny reveals empirical problems.
Empirically speaking the vast majority of people will usually not receive a
government position or become economically much better off just because their ethnic group
is in power. Hence, multiplied by the probability of gaining these positions the overall
motivating force of these variables seems weak; I am not likely to be "much better
off" (Hardin 1995: 217) just because my ethnic group is in power.
3.2. Reputation
In a path-breaking paper from 1982 Kreps, Milgrom Roberts and Wilson
(KMRW) suggested that a small amount of uncertainty may make cooperation the rational
strategy in a repeated Prisoners Dilemma game. Once again we should note that KMRW do not
"solve" the Prisoners' Dilemma, but simply argue that it does not accurately
describe the real world. For instance, a common feature of ethnic conflict is uncertainty
about the type of players in the game: Are they just a career seekers or are they really
fervent nationalists? To describe the real world we thus introduce uncertainty in a game
of many rounds. This means we now have a finite and dynamic game of incomplete
information, rather than a static game of complete and perfect information.
The essence of the argument is that a small amount of uncertainty (for instance that
player 1 is unsure whether player 2 really is rational), can generate cooperation. I have
derived the formal result in appendix 1. Informally the argument is that by defecting
Player 1 convinces Player 2 that he is irrational. This, in turn, makes it profitable both
for Player 1 and 2 to play trigger strategies for some rounds before it becomes more
advantageous to defect. The precise result is that it is best to cooperate as long as n
(the round you are in) is larger than the following expression:
n > [1/p] * [2 (a - b) + b p]
This is still too abstract to prove whether it is a true explanation of how ethnic groups
overcome the free rider obstacle. For instance, the KRMW result is weakened if we need
many players - not just two - to create the collective good, as typically is the case in
ethnic conflict. Moreover, the analysis is empirically questionable since it predicts more
defection than we would expect. For instance, if a=2, b= -1 and p=0.05 we get n > 119
i.e. defection in the last 119 rounds. Thus it can hardly be said, as one author has, that
Kreps et al. show that cooperation "only breaks down in the last few rounds"
(Lyons 1992: 123). 199 is far from a few rounds. In sum, while I was initially very
optimistic about the idea that KRMW could give us a convincing account of how groups may
form, a closer investigation weakened these beliefs.
3.3. Including and excluding beliefs and preferences
Two easy ways of making participation in ethnic violence seem rational, is
by postulating that people have beliefs or preferences that make the act rational. For
instance, I may believe that the opponent is about to kill me so it becomes rational to
attack him. Or, I may have a strong preference in favour of violence, or - more plausible
- to live in an ethnically homogeneous neighborhood. In this section, I will criticize
these theories. First, I think that it is counter-intuitive to label actions based on
clearly irrational belief for rational. Second, the rational choice approach becomes
useless if we end up in the position that 'ethnic violence is rational because people have
a preference for ethnic violence.'
Hardin admits that the beliefs that often inspire ethnic violence are "not
convincing, even patently not so in the sense that it would not stand serious scrutiny
..." (Hardin 1995:62). The myths about your own group superiority, about old
historical events and about your own ancestry (being of pure blood) are very often simply
false. How does this affect whether we call the action based on these beliefs rational?
According to Jon Elster the following requirements should be satisfied before we have a
complete rational choice explanation:
"Ideally, a fully satisfactory rational-choice explanation of an action would have
the following structure. It would show that the action is the (unique) best way of
satisfying the full set of the agent's desires, given the (uniquely) best beliefs the
agent could form, relatively to the (uniquely determined) optimal amount of evidence. We
may refer to this as the optimality part of the explanation. In addition the
explanation would show that the action was caused (in the right way) by the desires and
beliefs, and the beliefs caused (in the right way) by consideration of the evidence. We
may refer to this as the causal part of the explanation." (Elster, 1985,
p. 71)
In constrast to this consider Hardin's argument that "It would be odd, however, to
conclude that the action was irrational when taken if it was fully rational given
the available knowledge " (Hardin 1995:16, my emphasis). It seems like we have
to make a choice; Either we demand that the collection of evidence and the construction of
beliefs be rational before we call the action rational (as Elster claims) or we agree with
Hardin that we are justified in labeling an action rational even if it is based on
irrational beliefs.
What are the arguments in favour of each approach? In favour of Elster, I would use ask
the reader to consider the following question: Would you would say it is rational to buy a
house without first trying to collect some information on the quality of the house? I do
not think it conforms to our intuition to label such this act rational. To use a more
extreme example, a person who believes he is God may try to walk on water, but we would
hardly call the action rational. Against this one may argue, as Hardin does, that to
require rational construction of beliefs before we label an action rational would make
"virtually everyone's actions [...] always irrational" (Hardin 1995:62).
Where does this leave us? Instead of saying that an action either is or is not rational, I
think we should speak of degrees of rationality. For instance, when Hardin argues that
everything becomes irrational when we demand rational beliefs, we could counter this by
pointing out that his position leads to an equally absurd conclusion: that almost all
actions are viewed as rational. To place no demands on beliefs makes if too easy and not
very interesting to claim that an action is rational.
When faced with the problem that far more people voted than his rational choice model
predicted, Downs (in)famously "solved" the paradox by assuming that people had a
desire to vote (Green and Shapiro 1994:ch. 4, esp. p. 51). The explanation then reads:
People vote because they enjouy voting. The rational choice approach to ethnic conflict
has also tried this avenue, for instance arguing that people identify within their ethnic
group because they desire to belong to groups. Also cooperation in collective action
problems can be explained in this way by pointing out that people find it pleasurable to
cooperate.
Not surprisingly this way of turning the irrational into something rational has been
criticized. First one could argue that it amounts to a tautology, as opposed to an
explanation. Second, one say that it is methodologically wrong to include non-quantifiable
psychic benefits since this makes it difficult to falsify the theory.
4. Rational and Emotional as Complementary
In this section, I want to examine two possible ways in which one could say that the
rational and the emotional approaches to ethnic violence do not conflict.
4.1. Multiple motivations
Some causes are mutually dependent in the sense that you need both in
order to get the consequence. For instance, it is meaningless to ask whether the man or
the woman was the most important "cause" of a baby. Both are needed and without
one there would be no baby. In one sense it is the same with collective violence: To reach
the critical level in which violence gets started on a mass scale we probably need people
with many different types of motivation.
To make the idea somewhat more rigorous, imagine a population of 20% fanatic, emotional
and irrational nationalists; 30% who have nationalist preferences, but who are more
rational about acting upon these preferences; 20% who are driven by a norm that tells them
to cooperate if a majority does so; 20% who are selfish, rational and who seek only
personal monetary gains; and finally 10% who will think it is morally wrong to use
violence under any circumstances. Corresponding to these different motivations one might
imagine a threshold level beyond which each group will take part in violent protests. If
we simplify and assume that we are facing a population of 10 individuals we may have the
following threshold sequence (The following analysis is inspired by Kuran 1991, but the
numbers and the motivational structure are my own):
A = {0, 0, 20, 20, 20, 50, 50, 70, 70, 101}
In this structure the different motivations interact in a way that makes violence a
possible outcome. The two fanatics will start and when the three rational nationalists
realize that 20% are fighting, they will join in. This brings the total participation up
to 50% at which point the two conformists also will join, making the total 70% which is
beyond the level required before the careerist will find it profitable to join. At 90%
involvement the process will stop and assuming this is beyond the critical level required
to make violence happen we now have mass scale violence.
The above example illustrate how it might be meaningless to ask whether violence is mainly
driven by emotional or rational factors. Without the two emotional fanatics the spiral
could not get going, but without the rational nationalists it could not get to the point
where the conformists are brought into the game.
4.2. Multiple phenomena to be explained
The second sense in which the emotional and the rational approach can be
complementary, is by explaining different phenomena in the causal chain leading up to
violence. For instance, one may try to explain the formation of preferences and beliefs
using non-rational forces, while the choice of action for given preferences may best be
handled by rational choice models. Another example is to explain the actions of leaders
using rational choice models, while the general population is considered to be easily
manipulable by emotional and nationalistic appeals.
Once again we need to be more concrete before we can evaluate these arguments, for
instance by using Yugoslavia as an example. According to one observer the events taking
place in former Yugoslavia over the last 10-12 years have been the consequences of
purposeful and rational strategies applied by the ruling elites threatened by changes in
the economic and political structure that sustained them. (Barth Eide 1991:55).
More specifically, after the collapse of the Soviet Union, the revolutions in Eastern
Europe and the failing economy in Yugoslavia, Slobodan Milosovic had to divert attention
and remove political opponents in order to maintain his power. One way of doing so, is to
create a situation in which it is viewed as legitimate to use extreme measures (so you can
eliminate opponents. This is called securitizing the situation by Barth Eide). In this
situation you also divert attention since economic difficulties is perceived as less
important than the supposed danger facing the nation. How do you create this situation?
According to Barth Eide it is a question of partly using information that creates a tense
atmospheres, partly a question of engaging in some provocative actions such as sending
Serbs into non-Serb strongholds to demonstrate and stir violence. In sum, securitizing the
nation may be an instrument for achieving or enhancing the status of undisputed leader
(Barth Eide 1991:53).
It should be evident that we are here facing a phenomenon that need to be explained at two
levels. First, the initiation of the conflict itself can be seen as the result of rational
choice by the political leaders. However, we also need to explain why their emotive
appeals work so well among the population. This part of the explanation seems to involve
emotions and psychology more than rational belief formation.
5. Two small rational explanations: Mechanisms
5.1. Laws vs. mechanisms
How should we approach large questions such as Why ethnic violence?
Traditionally there has been two large methodological approaches to such questions in the
social sciences. Either you were a grand theorists who tried to establish scientific laws
that determined the relationships in general or you specialized in one case
study of ethnic violence (the resulting accounts are technically known as nomothetical and
ideographic). Recently, however, some social scientists have argued that it is more
fruitful to focus on a third option, loosely named mechanisms (see Hedström and Swedberg
1998 for a collection of these views).
A mechanism is defined in this context as frequently occurring and easily recognizable
causal patterns than are triggered under generally unknown conditions or with
indeterminate consequences. (Elster 1998:45, italics removed). For instance, sour grapes
is one mechanism which makes people dislike what they cannot have. Another mechanism is
the grass is greener which implies that you want what you cannot have. In a particular
situation we do not know which mechanism will come into effect, but ex post
we are able to see which mechanism that dominated and thus explain an event by reference
to the mechanism. An example could be whether a man who was forced to live in London would
eventually come to think that London is the best city to live in, or whether he would
develop a strong desire to live in a city he was not allowed to live in. This should
convey the idea of what a mechanism is. (The source of the example is also Jon Elster).
Why should we focus on mechanisms? The basic argument is that it is practically impossible
to establish deterministic laws. There is not enough information to test our grand
theories and laws with sufficient statistical power. Given this imperfection, we are
better off not even trying to establish the most reliable laws. The strategy with the
highest payoff in terms of understanding is rather to focus on middle range theories and
mechanisms. This is not the time or place to evaluate the correctness of this
methodological argument. I will, however, discuss two mechanisms that have been proposed
in the study of ethnic violence. The argument is that this will illustrate the usefulness
of focusing on mechanisms and game theory in the study of ethnic violence.
5.2. Collapse, uncertainty and beliefs about victory
To describe one mechanism that sets the stage for ethnic violence, Laitin
writes: ... at the time of (Soviet) state collapse, it is very difficult for leaders
trying to rule from the decaying center, or leaders of ethnically distinct regions within
the state (e.g. Estonia), to properly calculate the balance of power between the center
and the region. Bad calculations on both sides (especially if they exaggerate the weakness
of their adversaries) may yield the expected probabilities of victory by planners of each
side to be greater than .5, a situation that might lead both parties to calculate positive
returns to war (Laitin 1998:327-328).
This might be one plausible link between state collapse and ethnic violence, but closer
scutiny, once again, reveals a more complicatd picture. First of all, there is a small
mathematical error in the focus on 0.5 as the tipping point beyond which war is more
likely. Technically speaking using a (cooperative) rational choice approach, war is most
likely when the expected net gain from war is grater than the expected net gain from
alternative strategies. (Where expected net gain is calculated by multiplying the
probability of all the possible outcomes - positive and negative - with their respective
probabilities). Depending on the payoffs and the probability distribution when using
alternative strategies, and the possible asymmetry of the size of benefits and losses from
winning vs. loosing, a war could be rational for all kinds of probabilities. There is no
logically necessary reason why the probability for victory has to be larger than 0.5
before a war gets going.
Second, and more importantly, we could question whether the relationship is as simple as
Laitin suggests. In fact, he himself uses an argument only two pages later that in my
opinion relies on the directly opposite mechanism. The argument is that the conflicting
signals from the Russians authorities about their support for the Russian diaspora meant
that neither titulars nor Russians had sufficient surity (about Russia's probable
reaction) to aggress upon the other (Laitin 1998:330). Andrei Kozyrev in The Ministry of
Foreign Affairs said one thing (moderate), while the Military establishment sent hard line
signals. Mixed signals in this sense also could increase the likelihood of conflict since
both parties would put emphasis on the signals they wanted to hear, thus exaggerating the
probability of victory. In short, in a situation of uncertainty - as Laitin himself
suggests when discussing his first mechanism - war is more likely to happen. The net
effect of the two opposing mechanism is uncertain and we cannot a priori tell
whether uncertainy increases or reduces the probability of war.
5.3. Window of opportunity and Credible promises
When a state collapses this creates two kinds of temporary opportunities.
First the population within a region may think that they temporarily have an
opportunity to gain independence. However, the minority within this region may also
believe that they only have a temporary possibility of fighting independence. (As soon as
the new government gets consolidated it becomes too late for the new minority to
realistically work for independence.) Finally, it is very difficult for the new
authorities to promise the minority that as long as you agree to our independence you will
be treated well. The problem is, as we know from game theory, that the promise is not
credible. As soon as you have consolidated your power you can go back on the promise of
minority rights since by that time you do not need their support.
Is this really important? In other words, is this how politicians think? Consider the
debate about the so called zero-option in Latvia and Estonia before independence. The
zero-option simply meant giving citizenship to everybody who lived in the country,
including the Russophone population. The Popular Fronts in both countries endorsed this
principle, and many Russians voted in favour of Estonian and Latvian independence in the
referendum in the belief that they would become full citizens of these new countries.
After independence, however, the political climate changed and neither Estonia nor Latvia
implemented the zero option. One leader in the Popular Front who later became the leader
of the Latvian Way faction in the Saeima (the parliament), Andrej Pantelejevs explained
this as follows:
I will say this very candidly and somewhat cynically. Theoretically, the ethnic Latvians
could have fought for their independence with honest means. That would have led to
confrontations with weapons in hand. The alternative course, therefore, looked much more
sensible: To penetrate with legal means the power structures which existed at that time.
In order to do that we needed votes, and, as you know, at that time all inhabitants of
Latvia had the right to vote. So, we fully consciously said that our goal was the
so-called zero option. Yes, this was a conscious lie in order to avoid human casualties
(cited in Antane and Tsilevich 1997:91).
Although this quotation does prove that politicians think strategically in a way one may
analyze using game-theory, it does not create the link between the failure to make
credible commitments and violence. In fact, it seems like the promise was believed by many
Russians and that this contributed to the peaceful transformation to independence.
6. Conclusion
Is ethnic violence the outcome of individually rational actions and is game and
rational choice theory useful in the study of ethnic conflict? The arguments that I have
considered - three big rational choice theories and two small mechanisms - all had serious
wekneses when examined closely. In the argument about selective incentives the author
failed to calculate the size of the gains (government position) with the probability of
receiveing the gains to get the expected payoff. The KMRW model turned out to underpredict
empirical cooperation when we used actual numbers in the formula. The argument which
excluded rationality conditions on beliefs seems counter-intuitive and the inclusion of
psychic preferences creates a tautology and not an explanation.
The two smaller mechanisms did not fare much better; There is no a priori
simple connection between uncertainty and the probability of violence and incredible
promises appeared to be believed in Estonia and Latvia thus falsifying the game theoretic
prediction.
All this, of course, do not prove that all rational choice models of ethnic violence is
useless. In chapter four, I showed how rational choice models and emotional mechanisms
could complement each other. Moreover, there may still be rational choice models that fare
better than those I have considered. Finally, even if choice is dominated by non-rational
mechanisms we may often gain more in terms of reliable understanding by
focusing on the rational element. We may also derive better policy implications and
predictions using the as-if rational assumption since alternative theories are often
non-quantifiable and difficult to study scientifically (Melberg 1998).
In conclusion, unlike Durkheim I do not claim that my attempt to do the impossible was
successful - but I do not rule out future advances.
7. Literature
Anatane, Aina and Boris Tsilevitch (1997): Nation Building and Ethnic Integration in
Latvia, ch. 4 in Pål Kolstø (ed.) Nation Building and Ethnic Integration ,
Unpublished manuscript, University of Oslo.
Barth Eide, Espen (1997): Conflict entrepreneurship: On the art of waging civil war, PRIO
Report 4: 41-69
Connor, Walker (1994): Ethnonationalism: The Quest for Understanding. New
Jersey: Princeton University Press.
Cunningham, Frank (1997): Critical note on Russell Hardin, One for All: The Logic of Group
Conflict, Canadian Journal of Philosophy 27:571-594.
Durkheim, Emile (1912): Le suicide : etude de sociologie . Paris: Felix
Alcan. (I did not read the French text)
Elster, Jon (1985): "The nature and scope of rational-choice explanations," pp.
60-72 in Ernest LePore and Brian P. McLaughlin (eds.): Actions and Events:
Perspectives on the philosophy of Donald Davidson . Oxford: Blackwell.
Elster, Jon (1998): A Plea for Mechanisms, kap. 3 in Peter Hedström and Richard Swedberg
(eds.): Social Mechanisms . Cambridge: Cambridge University Press.
Fearon, James D. (1998): Paper on Ethnic violence presented at the ASN conference in New
York in April.
Fearon, James D. and David D. Laitin (1996): Explaining Interethnic Cooperation, American
Political Science Review 4:715-735.
Gurr, Ted Robert and Barbara Harff (1994): Ethnic Conflict in World Politics
. Boulder, CO.: Westview Press.
Hardin, Russell (1995): One for All: The Logic of Group Conflict . New
Jersey: Princeton University Press.
Haregreaves Heap, Shaun, Martin Hollis, Bruce Lyons, Robert Sugden, Albert Weale (1992): The
Theory of Choice: A Critical Guide . Oxford: Blackwell.
Kuran, Timur (1991): Now Out of Never: The Element of Surprise in the East European
Revolutions of 1989, World Politics 44:7-48.
Laitin, David (1998): Identity in Formation: The Russian-Speaking Populations in the
Near Abroad . Ithaca and London: Cornell University Press (forthcoming, my
quotations are based on the uncorrected proof version).
Melberg, Hans Olav (1998): Et forsvar for antagelsen om rasjonalitet, Impuls
, forthcoming. See also Papers by Hans O. Melberg at Internet URL:
www.oocities.org/hmelberg/papers/papers.htm (the review of Hardin on the 2nd of April
1998 and of Connor on the 10th of December 1997)
Olson, Mancur (1965/1971): The Logic of Collective Action . Cambridge Mass.:
Harvard University Press.
Petersen, Roger (1997): Ethnic conflict, Social Science and William Butler Yeats: A
Commentary on Russell Hardin's One for All: The Logic of group Conflict, European
Journal of Sociology 28:311-323.
Shapiro, Ian and Donald P. Green (1994): Pathologies of Rational Choice Theory: A
Critique of Applications in Political Science . London/New Haven: Yale University
Press.
Schelling, Thomas C. (1978): Micromotives and Macrobehaviour . New York,
N.Y.: W. W. Norton & Company.
Väyrynen, Tarja (1998): Ethnic Communiality and Conflict Resolution, Cooperation
and Conflict 33:59-80.
Williams, Mark (1990-93), Unpublished lecturenotes, Oxford University.
8. Appendicies
8.1. Appendix 1: The formal solution of the KRMW model
Consider the following formal argument. In the game already described in figure 2 (but now
repeated and with uncertainty), Player 2 believes with probability p that Player 1 is
irrational. Being irrational is here defined as using a trigger strategy i.e. to play
whatever strategy your opponent played in the last round (if he defected you defect in the
next round and so on). Knowing that Player 2 is uncertain about Player 1's type, Player 1
faces the following choice:
a) Defect (thereby revealing that he is rational)
b) Use a tit-for-tat strategy, starting by cooperating (essentially this is to pretend
that he is irrational)
To find the condition for choosing the trigger strategy, one could reason as follows. The
expected payoff for Player 2 if he plays trigger (starting to cooperate) is:
E P2 (trigger) = p n + (1 - p) b
In plain words: Player 2 gets 1 with probability p in n rounds if Player 1 is irrational
(they will simply get 1 in each round), but he also risks (with probability 1 - p) getting
b if Player 1 really is rational and plays D in the first round.
The question is then whether Player 1 should defect or play trigger (and cooperate in the
first round). To answer this I reason as follows. Player 1 will at least get the same as
Player 2 by pretending to be rational. However, Player 1 can exploit the fact that he
knows that he really is rational. So while pretending to be irrational (playing trigger)
Player 1 can, if it is profitable, switch to the rational strategy of defecting. More
formally we have the following relationship:
E P1 + a > E P2 + b
Player 1 can always get a instead of 1 by switching to "defect" after
cooperating some time. This switching, in turn, means that Player 2 wille receive b in
that round (which is a negative payoff). Hence, Player 1 will always get b - a more than
Player 2 when they both play trigger strategies. This implies that:
E P1 > E P2 + (b - a) or (by substitution):
E P1 > p n + (1 - p) b + (b - a)
Having worked out the payoff from the trigger strategy, we may then compare this to the
playoff from the strategy of defecting. The highest possible payoff from defecting is a.
So a sufficient condition for the profitability of playing trigger instead of defecting
is:
p n + (1 - p) b + (b - a) > a
Solving this we find:
n > [1/p] * [2 (a - b) + b p]
8.2. Appendix 2: A critical note on the logic behind the KRMW results
using the Chain Store Paradox as an example
I would question the reasoning behind the KRMW solution. If it is
rational to pretend to be irrational, surely the other player will recognize that Player 1
is probably only pretending. If Player 2 realizes this, he will try to defect one period
earlier than Player 1's critical number. But then Player 1 should realize that Player 2 is
not fooled by his faking and defect one period before this and so on until we get
defection in the first period again. Now, KRMW claims to have solved this problem using
Bayes rule. Player 2, they argue, know that Player 1 is probably faking being irrational,
but still there is a slight chance that he really is irrational and the fact that he plays
"Cooperate" reinforces Player 2's belief that Player 1 is irrational. After
observing an irrational action Player 1 updates his belief on the probability of Player 1
being irrational according to Bayes rule. It is this step which I find problematic: Being
rational I do not increase my belief that you are irrational when you do something that in
fact is in your best interests.
To avoid vague words and confusion we may use the Chain Store Paradox as a formal example
(the analysis is originally grom Kreps and Wilson, but my presentation is based on
lectures by Mark Williams (Jesus College, Oxford University) though adapted for the
present purpose). In brief, the situation is as follows: An enemy does not know whether
you are tough (T) or rational (R). Tough agents always fight when faced with an opponent;
P (F | T) = 1, rational agents only fight when it is profitable, say P (R | F) = x where
0<x<1. Assume that the probability that you meet a tough opponent is d i.e. P (T) =
d, and the probability of meeting a rational opponent then becomes P(R) = 1 - d. The
question is now, how should the fact that you observe a firm fighting affect your belief
that it is tough - i.e. we want P(T | F). The "rational" answer is given by
Bayes formula:
In general: P(A | B) = P (AB) / P(B)
In our situation this can be expanded to:
P(T | F) = [P(F | T) P(T)] / [P(F |T) P(T) + P(F | R) P(R)]
Inserting the information we have we get:
P(T | F) = d / [d + (1 - d) x]
Now ask yourself about the probability that this player is tough if you observe him
fighting even one more time. The answer is given by starting from the updated probability
that the player is tough (after fighting we have P(T) = d / [d + (1 - d) x] for our
player) and inserting this into Bayes formula. We get:
P(T | F2) = d / [d + (1 - d) x2]
Or, in general:
P(T | Fn) = d / [d + (1 - d) xn]
The conclusion from the last general formula is that a little bit of fighting greatly
increases your reputation for being tough (since the relationship is non-linear). My
point, however, is that the act of fighting cannot be viewed as a sign of irrationality
when the other agent also knows that a little bit of fighting would increase your
reputation for being tough and hence be in your interst. This problem opens up a web of
complications that are yet to be solved.
8.3. Appendix 3: Policy implications and the future
(After writing this sub-chapter, I decided that it was not relevant for
my main argument. This is why it is placed in an appendix).
In his book Micromotives and Macrobehaviour (1978), Thomas Schelling gives a
wonderful example which for some time motivated me to think about ethnic violence. He
describes how he always ended up sitting next to people like himself (skin colour,
accents, income, dress) when he went to a baseball game (Schelling 1978:40). Somehow
people were segregated although they never expressed a wish to be so. The reason, he
learned later, was that each ticket agent got a block of tickets with seats next to each
other. This is surely a practical arrangement since many people want to buy two tickets to
sit next to their friend when they watched the game. Yet it also ensured that people from
the same neighbourhood who bought tickets from the same place sat together. To me the
story suggested that segregation could come about unintentionally and that that it could
be solved quite easily. One could simply give the ticket offices a batch of random seats
tickets, not a block of seats next to each other.
As I read more about ethnic violence, my initial hope of discovering equally simple and
easy solutions evaporated. Indeed, I came to realize that even the above solution could
have the effect of increasing ethnic tension. For instance, Walker Connor convincingly
describes how increased interaction between different groups may in fact increase tension.
Moreover, as I discovered the middle range approach - the mechanisms - my hope of
discovering some useful and non-obvious policy prescriptions suffered another blow. Doing
research on mechanisms allows you to explain ex post, but it does not enable you to
predict. As already mentioned we do not know in advance whether a man put in London will
start to like London or whether he will hate it. Since the variables are connected with so
many and contradictory links, it may seem impossible to give reliable policy presriptions
or predictions about the future.
With the above doubts about the very weak realibility of predictions in the background, I
will discuss some of the advice presented in the litterature.
In Hardins account groups could form because, amongs other reasons, the leaders could
promise government positions and favourable policies. Accordingly one way of reducing the
atractiveness of being active in ethnic groups, would be to reduce the role of government.
As Hardin writes: not the least among the reasons to prefer smaller government, is to
reduce the value of getting control of it (Hardin 1995:217). Another way of achieving the
same thing, is to increase the possibilities for ecnomic growth in the private sector.
Economic growth would both reduce the need for leaders to rely on nationalism and conflict
to maintain power, as well as making people less disposed to nationalistic appeals.
Finally, Hardin seems to predict that urbanization reduced the potential for ethnic
conflict since it reduces the ignorance that he believes constitute a precondition for
ethnic violence.
Laitins proposals, in contrast to Hardin, are more instututional and in some ways
increases the power of the central government. He has at least four concrete suggestions
(Laitin 1998:344-345). First, the career prospect for elites must be such that it is more
profitable to stay within the country than to press for independence (Laitin 1998:344).
Second, there should be bureaucratic hurdles which makes it difficult to seccede as well
as making it profitable to stay within the country. For instance there could be a
transferpayment system that increases your share the more people you have in your group.
Third, one must protect those who do not want to identify with a particular group from
ethnic vigilantes i.e. self-appointed people who punish those who are not natiuonalistic
enough. Fourt and finally, it is important to have a well paied and rotating police force
that can stop vigilante terror, as well as giving the vigilantes the option of being
reintegrated into society (so he is not forced to be a vigilante foreover).
Those believing that the the ethnic bond is subconscious and emotional, tend to argue that
the best policy to prevent violence is the Wilsonian principle of national
self-determinantion. If all groups had their own state, the argument goes, ethnic violence
would be greatly reduced. Laitin attacks this argument for implicitly assuming that the
number of nations or ethnic groups is constant. Without this assumption the process might
never end (there is always a minority within the minority), and besides it is impossible
to create ethnically pure states in areas where the population is geographically mixed.
[Note for bibliographic reference: Melberg, Hans O. (1998), Is ethnic conflict the
outcome of individually rational actions?, www.oocities.org/hmelberg/papers/980522.htm]