Professional Documents
Culture Documents
Bowling - Peace Economics, Peace Science and Public Policy) International Cooperation - Testing Evolution of Cooperation Theories
Bowling - Peace Economics, Peace Science and Public Policy) International Cooperation - Testing Evolution of Cooperation Theories
Jeremy Bowling*
International Cooperation: Testing Evolution
of Cooperation Theories
DOI 10.1515/peps-2016-0020
1 Introduction
Do states reciprocate the cooperative actions of other states? Are states more
likely to cooperate with other states that have reputations of being cooperators?
Are there common factors shared by states that encourage them to cooperate
towards each other? The evolution of cooperation literature seeks to answer these
questions by developing theories of cooperation when studying the relationships
between people. Starting with the assumption that general theories of interna-
tional cooperation are distinct from general theories of international conflict, this
study supplements the scholarship on international cooperation by empirically
testing evolution of cooperation theories using the international setting. Since
states are often assumed to be unitary actors in a strategic setting, states should
be found to act similarly to actors in formal models, experimental testing, and
computer simulation. Testing evolution of cooperation theories provide further
validity of the evolutionary stability of these strategies in not only human interac-
tion on the individual level but also international interaction on the state level.
*Corresponding author: Jeremy Bowling, The College of Wooster, 1189 Beall Avenue, Wooster,
OH 44691, USA, E-mail: [email protected]
2 Jeremy Bowling
An example could clarify the relationship of desired level and defection. One
actor gives aid to another. If no response is desired, then receiving no response is
not considered defection. If the intent of aid was to induce cooperation from the
recipient in response, then no response would be considered defection. Another
example is that two states have a previous or status quo level of trade, which is
a type of recurring cooperation reciprocity where each state commits to recur-
ring cooperation at the status quo level. If one state imposes trade restrictions, it
could be considered defection because it is a reduction from the desired level of
cooperation based on the status quo, in which a trade war occurs if defection is
the response to defection.
The trade war and aid examples emphasize the problem of defining defec-
tion as no action or complete cessation of cooperative action. While it simpli-
fies the game theoretic models, defection as no action is a poor representation
of defection in real world interaction, where a level of recurring cooperation
reciprocity (e.g. trade) is maintained. Defining defection as including a range
of actions does not change the theories studied. The models in this analysis use
the assumption that, at minimum, the desired response to cooperation is the
previous level of cooperation by the target of the act, and defection is a reduc-
tion from that previous level.
In the end, strategic interaction can include the acts of no action, coopera-
tion, defection, and conflict, each of which can be an initial act and a response
to the others. Although conflict can be a response to cooperation, it is not studied
here, and it is not included in the definition of defection because conflict, an
action taken by an individual that imposes a cost on at least one other individual,
can occur without defection as a response to cooperation, such as a response of
status quo cooperation reciprocity (cooperation response) with a condemnation
of aid quid-pro-quo coercion (conflict response).
(Axelrod & Hamilton, 1981). In these games, there are only two types of
actions; an actor can only cooperate or commit no action (defect). One actor
commits a cooperative act towards another, and then, after the second actor
makes an action, the first actor retaliates with the action used by the second
actor. The first actor will continue recurring cooperation if the second actor
continues recurring cooperation and defects if the second actor defects. Tit-
for-Tat has been supplemented with the strategy of Win-stay/Lose-shift, which
accounts for the presence of cooperation in response to defection (Nowak &
Sigmund, 1993). If the first actor is benefiting more from recurring coopera-
tion than defection, the first actor will continue recurring cooperation when
the second actor defects.
Theories of indirect reciprocity are developed to explain the presence of
altruism and are often based on the cooperative reputation of potential coopera-
tive partners (Nowak & Sigmund, 1998). The basic theory of indirect reciproc-
ity is that one actor will consider the cooperation history that another actor has
with third parties when deciding to act, preferring to choose cooperation with the
second actor when it has a reputation of recurring cooperation. Other theories
of indirect reciprocity include kinship and tag theory. Both are closely related,
where kinship can be considered a special and most common tag. A tag is any
immediately observable characteristic that tells two strangers that one another
are suitable candidates to commit cooperation, where kinship conveys a variety
of observable characteristics based on biological relations (Riolo, Cohen, &
Axelrod, 2001).
This study empirically tests these four strategic theories (Tit-for-Tat, Win-
stay/Lose-shift, Cooperative Reputation, and Tag/Kinship) in the interaction
between states. Doing so not only further introduces the evolution of coopera-
tion literature to international relations scholars, but it furthers the study of
cooperation by empirically testing the validity of the evolution of cooperation
theories and strategies in a setting different from computer simulation and
experiments using individual living subjects. Many of these strategies are found
to be present in the international interaction of states, which is evidence of their
validity.
In the following pages, the evolution of cooperation theories will be reviewed
and hypotheses in relation to international cooperation will be presented. The
subsequent section will describe the research methods, models, and variables
used. When discussing the results, the findings indicate that the act of coopera-
tion, whether directly or indirectly, is more likely to elicit cooperation in response
than sharing similar characteristics, concluding that, while there is little evidence
that sharing similar characteristics elicits international cooperation, cooperative
acts will elicit cooperation in return.
Testing Evolution of Cooperation Theories 5
2 International cooperation
2.1 E
volution of cooperation theories
Axelrod and Hamilton (1981) write of the early development of theories of the evo-
lution of cooperation. Early theory emphasizes kin selection and reciprocation.
Kinship theory attempts to explain the presence of altruism, but it is incomplete.
Altruism is present between people and some organisms that are not genetically
related, which leads to reciprocity theory. This, in turn, leads to the inclusion of
the strategic model of prisoner’s dilemma in the study of cooperation.
Formal modeling is ubiquitous in the evolution of cooperation literature. Not
only are these scholars attempting to explain the presence of cooperation in situ-
ations where it is more beneficial to free ride on others’ work but they are trying to
explain the evolutionary stability of strategies that foster cooperation. “A strategy
is evolutionarily stable if a population of individuals using that strategy cannot
be invaded by a rare mutant adopting a different strategy” (Axelrod & Hamilton,
1981, p. 1392). Repeated interaction within formal modeling allows for the pres-
ence of solutions to the defection problem that occurs within the single-shot pris-
oner’s dilemma, and repeated interaction is a much proper model than single
shot games for international interaction.
In multiple rounds of computer simulation tournaments, Anatol Rapoport’s
Tit-for-Tat (TfT) beat out all other strategies. The strategy is simply to cooperate
in the first round and then in subsequent rounds do what the other actor did in
the previous round. This simple strategy of direct reciprocity helped to explain
the evolutionary stability of recurring cooperation reciprocity in iterated prison-
er’s dilemma. An actor employing TfT is never the first to defect.1 If the second
actor takes advantage of the first actor’s cooperation, it loses in repeated inter-
action because defection by the second actor is punished with defection by the
first actor in subsequent rounds, where short-term gain from defection in one
round is eliminated by long-term losses from not maintaining recurring coopera-
tion reciprocity. However, the actor employing TfT will choose cooperation when
the second actor chooses cooperation, which encourages the second actor to
choose cooperation for long-term gain. If the second actor does not know that the
first actor is employing TfT, the information becomes apparent through repeated
interaction. At some point, recurring cooperation reciprocity is established due
1 Cooperating first is a viable assumption given that human cognition is driven and constituted
by social cooperation (Moll & Tomasello, 2007; Tomasello et al., 2005; Vygotsky, 1978) and if
altruistic cooperation is driven by competing for reputation (Zahavi, 1977, 1995).
6 Jeremy Bowling
to mutually beneficial gains. In other words, the second actor cannot commit
another strategy that would benefit it more than cooperation when the first actor
maintains the strategy of TfT.
In subsequent research, TfT has been supplemented by an even simpler strat-
egy called Win-stay/Lose-shift (Nowak & Sigmund, 1993). An actor repeats its pre-
vious move when doing well and changes if not. If everyone in the population is
a defector, TfT is a catalyst for cooperation due to always cooperating in the first
round and providing incentives for the continuation of cooperation, but Win-stay/
Lose-shift is better able to maintain recurring cooperation due to another’s defec-
tion not being the sole catalyst for defection as a response. Therefore, a winning
condition may encourage recurring cooperation by an actor facing defection by
another actor.
Theories of indirect reciprocity explain the occurrence and recurrence of
cooperation in situations where there is no opportunity for direct reciprocity.
While kinship was an early evolution of cooperation theory, Tag is related to
kinship in that there is some defining characteristic between actors that each
can easily recognize (Riolo et al., 2001). The level of information required for tag
is very low because, by definition, it must be easily observable. Since Tag must
be recognized by both actors when they have had no prior interaction, kinship
is more easily recognized due to similar physical characteristics. Once the tag is
recognized, the actors inherently know that they will have greater success with
cooperation reciprocity than with others that do not have the tag.
The basic model of indirect reciprocity uses third-party cooperative reputa-
tion to identify cooperators (Nowak & Sigmund, 1998). Cooperative reputation
is the observation by one actor of the cooperation a second actor has committed
towards all other actors. One actor observes the cooperative reputation of other
actors and makes the decision to commit cooperation to those that have a suf-
ficient level of reputation, excluding any direct cooperative reputation the others
have with the actor. This requires extensive knowledge of other actors’ coopera-
tive relationships. Since cooperative reputation is theorized to lead to indirect
cooperation reciprocity, actors are encouraged to develop cooperative reputa-
tion, where cooperative strategies based on indirect reciprocity also lead to social
norms of cooperation and morality (Nowak, 2006).
2.2 E
volution of cooperation theories in international
interaction
The evolution of cooperation literature has advanced beyond the TfT introduced
in Axelrod’s cooperation tournaments as a solution to the problem of defection
Testing Evolution of Cooperation Theories 7
2.2.1 T
heories of direct reciprocity
For TfT to occur under international anarchy, one state must decide to initi-
ate cooperation when it knows that the recipient state could be a potential
defector.2 Since there is ample evidence that cooperation does occur, it can be
reasonably assumed that the iteration of the TfT game is well beyond the first
stage, where a state using a TfT strategy decided to cooperate with another. At
all other stages of the game, TfT predicts a simple direct reciprocity strategy. If
both states commit cooperation in the previous period, then both states con-
tinue to commit cooperation. If one of the states defect in the previous period,
then the other state defects in the current period. Since both states gain less
from mutual defection than mutual cooperation, the threat of mutual defection
encourages the continuation of mutual cooperation or recurring cooperation
reciprocity. Also, it is unclear that the game will ever end, which discourages
defection when there is recurring cooperation reciprocity. Since states are con-
fined to finite space, the likelihood of continued interstate interaction will
be great. If states are currently using TfT in their cooperation decision, states
will simply mirror the change in cooperative action of other states, where one
state will increase/decrease its level of cooperation in response to an increase/
decrease in its level of cooperation.
2 Tit-for-Tat is simply one of many possible trigger strategies that may be present in many bilat-
eral relationships. Ward (1982) found evidence, among a variety of independently tested selected
dyads, of cooperation and conflict reciprocity, termed “action-reaction dynamic” and based on
differing levels of memory of cooperation and conflict. Hudson, Schrodt, & Whitmer (2008) find
evidence of the strategy of Tit-for-Tat in the Israeli-Palestinian relationship for both cooperation
and conflict. They used a new methodology, which they named “discrete sequence rule models”.
However, direct reciprocity has not been tested in a general analysis.
8 Jeremy Bowling
On its most basic level past the first stage in an iterative game, TfT is simply
positive direct reciprocity. If one state initiates cooperation, the other state will
initiate cooperation. If one state ceases cooperative action, the other state will
cease cooperative action. If a state has a relatively high/low cooperation towards
another state, the other state will have a relatively high/low level of cooperation
towards the first state. In other words, states will simply mirror the cooperative
action or non-action of other states. While a strong test for TfT would include
change of action, testing for positive direct reciprocity provides some additional
evidence of the theory’s validity in international cooperation.
Hypothesis 1b (Tit-for-Tat): A state will reciprocate the cooperative action towards another
state that it received from that state.
Hypothesis 2 (Win-stay/Lose-shift): When a state chooses defection from the current level of coop-
eration with another state, the other state will not defect if it currently has a winning condition.
Testing Evolution of Cooperation Theories 9
2.2.2 T
heories of indirect reciprocity
3 Standing and image-scoring are two types of indirect reciprocity strategies, based on reputa-
tion, that are being studied in the evolution of cooperation literature (Leimar & Hammerstein,
2001; Milinski, Semman, Baker, & Krambeck, 2001; Nowak & Sigmund, 1998). There is a debate
on which has more merit, evidence, and is evolutionary stable. To construct a test of these spe-
cific strategies, states would need to know information about other states that is beyond the
scope of this analysis.
10 Jeremy Bowling
Hypothesis 3a (Cooperative Reputation): States with a higher cooperative reputation are more
likely to receive cooperation from another state.
Hypothesis 3b (Cooperative Reputation): States with a higher cooperative reputation are more
likely to receive a higher level of cooperation from another state.
There are two other indirect reciprocity strategies related to the initiation and
level of cooperation of states. They are kin selection and tag. Kin selection is the
idea that genetically related organisms are more likely to initiate cooperation
towards each other. For the kin to succeed over other kinships, they have a selfish
desire to initiate cooperation and maintain recurring cooperation (Dawkins, 1976;
Hamilton, 1964; Wilson, 1975). Whereas kin selection requires genetic similari-
ties, tag is simply the idea that there is some shared trait between the donor and
recipient of a cooperative act (Riolo et al., 2001). A “tag” could be anything, but it
must be easily observable and mutually recognized as an appropriate character-
istic that signals a good candidate for cooperation. With both kin selection and
tag, all likely targets for cooperation share the characteristic(s) or tag with sender,
assuming kinship can be categorized as a special and/or stronger form of tag.
Since the decision to initiate cooperation also requires a decision of the level of
cooperation, states will not only be more likely to initiate cooperation with states
that share their tag but they will have higher levels of cooperation with tag states
than non-tag states. There are a variety of common characteristics between states
that could serve as tags, but colonial ties, plurality religion, and democracy are
of the most prominent and easily recognized characteristics of states that could
signal potential cooperators.
Hypothesis 4a (Tag): States are more likely to initiate cooperation towards states they share
a tag.
Hypothesis 4b (Tag): States are more likely to initiate higher levels of cooperation towards
states they share a tag.
3 Research methods
Since theories of strategic interaction are being tested, the unit of analysis in this
study is the directed dyad-year, including all pairs of states for the years 1946–
2006 that data is available. States are included in the observations based on
the Correlates of War (2008) specification for entering/leaving the international
system. To model a decisionmaking process that not only includes the decision
Testing Evolution of Cooperation Theories 11
to initiate cooperation but also the quality of cooperation, selection models are
used to model the effect of occurrence on the second stage models. For models
that contain a binary first stage dependent variable and a (semi-)continuous
second stage dependent variable, a Heckman selection model, with MLE in the
first stage, is used. To test models with binary dependent variables in both stages,
a probit selection model is used. Since the unit of analysis is a directed dyad-year,
the variables, when appropriate, are averaged for the year.
3.2 E
vent data operationalization
Many variables are constructed from foreign policy event data commonly used in
comparative foreign policy analysis. To construct the longest time-series dataset,
COPDAB, WEIS, and Levant WEIS are combined to create foreign policy event
intensity measures based on the Goldstein (1992) specification. These datasets
were foreign policy events based on content analysis of newspapers. The first
event dataset is the Conflict and Peace Data Bank (COPDAB), covering the period
of 1948 to 1978 (Azar, 1980). The second event dataset is an update to the World
Events Interaction Survey (WEIS), covering the period 1966 to 1978 (McClelland,
1978). Each WEIS event category is given a Goldstein (1992) score based on inten-
sity of the conflictual or cooperative event, where negative scores denote con-
flictual events and positive scores denote cooperative events. Each cooperative
event is assigned a value based on the relative costs incurred by the event initia-
tor and benefits given the recipient, where some events have a higher or lower
level of cooperation intensity than others based on these costs and benefits.
Similarly, conflictual events are assigned intensity values based on the relative
costs incurred and imposed. Since the COPDAB and WEIS-Goldstein scales are
different, it is necessary to follow Reuveny and Kang (1996, p. 299) by regressing
their overlapping periods (1966–1978) and transforming the COPDAB scale into
the Goldstein scale using the constant and beta coefficient:
Goldstein it = C0 + C1 * COPDAB it + e it .
The third event dataset is the Levant WIES dataset, which covers April 1979–March
2005 (KEDS, 2010). The Levant WEIS data is transformed into the Goldstein speci-
fication and then merged with the other two datasets. Various categories are then
dropped from the data, including two states merging into one, domestic violence,
rebellion, civil conflict, and other non-international events. To compensate for
duplicate dyad-day events, I deflate the number of events by averaging the dyad-
day Goldstein score and making the number of events equal one for the day. At
12 Jeremy Bowling
this point, the data is a list of dyadic events covering the period of 1948–2005,
where the primary measure is a Goldstein foreign policy intensity scale.
Next, the Goldstein measure is separated into two different measures based
on whether the event was a cooperative or conflictual event. The cooperation
intensity measure ranges from 0 to 8.3, in which zero, a neutral Goldstein clas-
sification, is coded as a cooperative event. The conflict intensity measure ranges
from − 10 to − 0.1. Since this study uses a full directed dyad-year time-series, 0.1 is
added to the cooperation measure, the sign of the conflict measure is flipped, and
a true zero is added to each measure to indicate no event in the dyad-year. Meas-
ures of the sender’s and target’s number of annual cooperative events are created.
Both measures of cooperation and conflict intensity are averaged for both sender
and target within the dyad-year, and dummy variables are created that indicate
there was at least one initiated cooperative event in the dyad year for both sender
and target.
The final sender/target dyad-year average cooperation intensity scale ranges
from zero to 8.4, indicating a higher intensity or level of cooperation towards
target/sender. The final sender/target dyad-year average conflict intensity scale
ranges from zero to 10, indicating a higher intensity or level of conflict towards
target/sender. The final sender/target cooperation occurrence measure is dichoto-
mous, where one indicates the presence of at least one cooperative event towards
target/sender in the dyad-year. The final sender/target number of dyad-year coop-
erative events towards target/sender ranges from zero to 134.
Two-stage models: Since the first stage models the decision to cooperate, the
dependent variable in the selection model first stage is dichotomous. An obser-
vation is coded as a zero if sender has no observed cooperative events towards
target in the year. It is coded as a one if sender had at least one cooperative event
directed towards target. This variable captures sender’s initiation of cooperation.
The presence of no action or cooperation initiation in response to no action or
cooperation initiation can provide some evidence of positive direct reciprocity.
Since the second stage models the decision of quality of cooperation, they use
some form of the cooperation intensity measure. The second stage models that
do not use sender’s cooperation intensity as their dependent variable will use a
change in intensity, as explained below.
Hypotheses 3 and 4 second stage: Hypotheses 3 and 4 are predictions for theo-
ries of indirect reciprocity. These theories predict that states will be more likely
to initiate cooperation and have higher levels of cooperation toward other states
that have certain characteristics. Therefore, sender’s cooperation intensity is
used in the second stage of the models.
simple test of direct reciprocity using a TfT strategy after some history of iterative
interaction.
After modeling positive direct reciprocity in the first stage, two iterative
conditions must be met in the second stage to test for Win-stay/Lose-shift. The
first condition is that target state must have defected, where the current level
of cooperation is lower in this period than the last. The second condition is
that sender must still be “winning” under target defection. The first condition
is easier to quantify and observe than the second. It is difficult to measure a
general winning condition because in most situations the winning condition
will be different for each state; it will depend on very specific benefits and
costs of each bilateral relationship. For this analysis, I use a measure of relative
trade to represent a general winning condition; if sender state has a positive
trade balance with target, it is more likely to cooperate when facing a defector,
assuming positive relative trade is a “winning” condition that is strong enough
to ignore defection.
If both conditions are met- defection by target state and a winning condition
by sender state- there should not be a negative relationship between the change
in sender’s cooperation intensity and the lagged interaction between target’s
defection and sender’s positive relative trade balance. Ideally there would be no
relationship or null result because, if the two conditions are met, there would be
no change, but winning trade balances in favor of sender could encourage sender
to initiate and/or increase the level of cooperation, in trade or other issues, even
when facing a defection by target. In this case, there may be a positive relation-
ship as the winning condition overrides the effect from defection. Regardless,
there should not be a negative relationship for Win-stay/Lose-shift to be a strategy
used by states.
consider target’s entire past reputation, but may consider the recent past repu-
tation. The third measure (running average) assumes that sender considers all
previous annual reputation. Each measure is lagged in their models. Ultimately,
the running average measure is presented in the tables below.
Hypothesis 4 (Tag): The theory of tag predicts that cooperation will occur when
an easily recognized characteristic, which two states hold in common, signal that
the states are good candidates for cooperation reciprocity. One problem research-
ing tag is identifying the common characteristic that is used as a tag amongst a
myriad of potential common characteristics. Under the theory, the actors should
have no problem identifying the tag characteristic that tells them that each other
are viable candidates to commit cooperation. Since regime type and religion are
often identified characteristics of states by a variety of actors and news media,
these two characteristics would make excellent tags, where actors should feel
comfortable initiating and recurring cooperation with states that share their
regime type and religion.
While regime type may also be an excellent tag, it is used as a control because
Leeds (1999) finds that joint democracies and joint autocracies are likely to coop-
erate, joint democracies are more likely to cooperate than joint autocracies,
and both are more likely to cooperate than dissimilar pairs, arguing that joint
democracies have cooperation reciprocity due to credibility to commit and joint
autocracies have cooperation reciprocity due to a greater range of agreement pos-
sibilities. Since these arguments may be what makes democracy a good tag, joint
democracy as a tag is analyzed.
Since democracy has been examined before, two other tag measures are
created. The first tag variable is a dummy indicating that sender and target share
the same plurality religion, using eight plurality religions based on the World
Religion Dataset (Maoz & Henderson, 2013). Since the state’s dominant religion
is often discussed in news as national identities, it would be a common charac-
teristic in which states can easily identify when deciding to commit cooperation.
The second tag variable is a dummy indicating that sender and target share a
colonial tie, using the ICOW Colonial History Data Set (Hensel, 2009). Colonial
tie may be tag characteristic that may resemble the stronger ties attributed to
the theory of kin selection, but a colonial tie may also be a vertical, closed-field
relationship (Barker & Wright, 1955; Berscheid 1986; Collins & Laursen, 1992;
Graziano, 1984), which can have different implications for cooperation than
tag theory. Since states with colonial ties have strong histories between them,
it is also an easily identified characteristic that may encourage states to commit
cooperation.
Testing Evolution of Cooperation Theories 17
Conflict intensity: Sender and target’s current year conflict intensity are
included as controls, using only conflictual event data. It ranges from 0.1 to 10,
with 10 indicating all-out war. A zero is included for dyad-years that have no con-
flictual event. Sender and target’s conflict intensity is predicted to be positively
correlated with sender’s initiation of a cooperative event and negatively corre-
lated with sender’s cooperation intensity. Increasing international interaction
increases the probability of both cooperation and conflict occurring, where they
both create opportunity for each other. Also, the preferred outcome of committing
conflict is cooperation in response. Conflict increases the likelihood of coopera-
tion, but it lowers the quality of cooperation that occurs.
Capital distance: Included in the models is the distance between capitals, using
Gleditsch (2004). As the distance between states becomes greater, the opportunity
for cooperation reduces due to increased costs of action. As distance increases,
the cost of providing cooperation increases. Since resources are finite, distance is
predicted to reduce both the likelihood of cooperation and cooperation intensity.
for cooperation because it creates a stable environment. Also, actors have learned
to cooperate to have an alliance and alliance may provide some stability, which
could play a factor in increasing the likelihood of cooperation in other areas.
While the presence of an alliance may increase the opportunity for cooperation,
there is not a clear relationship with the quality of cooperation. Different alli-
ances require different commitments, where enactment of some alliances would
predict higher cooperation intensity where some may predict lower levels. There-
fore, alliance is only used in the first stage of models.
Major powers: Two dummies are included. The first is coded a one if sender is a
major power, as indicated by the Correlates of War Project (2008). The second is
coded a one if target is a major power. Lastly, an interaction of the two is included.
Under a realist understanding, major powers would have a lower incentive to
cooperate with other major powers because they do not want to increase the rela-
tive balance of power in favor of the other, but since they are more active, they also
have a greater opportunity for a cooperative event to occur. If major powers are
wary of changes in the balance of power among major powers, senders that are
major powers will be less likely to initiate cooperation with other major powers,
and if they do initiate cooperation, cooperation intensity will be lower. Since
major powers are more active in international politics, joint major powers have
an increased opportunity to have a cooperative event occur between them, but
the quality of cooperation will be relatively lower. More importantly, the inclu-
sion of major powers helps control for major powers having the most incidents of
international cooperation, particularly the United States.
Democracy: Three variables are included in the analysis. The first is a dummy
that indicates sender is a democracy based on the guidelines set by Cheibub et al.
(2010). The second is a dummy indicating a one if target is a democracy. The third
is an interaction between the two, indicating that both states are a democracy.
Whether democracy is a tag or a signal of credibility to commit, joint democracies
are predicted to be more likely to have a cooperative event and higher levels of
cooperation.
Time-series: In the first stage of the Heckman selection model, I use a common
control for temporal correlation when the dependent variable is a binary indica-
tor of event presence or occurrence, which is to include a measure of the years
between the dyadic occurrence of events and two additional measures of its
square and cube (Beck, Katz, & Tucker, 1998; Carter & Signorino, 2010). In the
second stage, the lagged dependent variable is used when it is continuous, and
the cubic specification is used when the dependent variable is binary.
Testing Evolution of Cooperation Theories 19
4 Findings
Before the discussion of the findings and analysis, a couple key assumptions
should be reiterated. The first assumption is that the desired minimal level of
cooperation in response to cooperation is the level of cooperation in the previous
period. Therefore, defection has been operationalized as a reduction of coopera-
tion intensity in the current period from the previous period level. The second
key assumption is that the theories of direct and indirect reciprocity explain both
components of decisionmaking, which are the likelihood of action occurrence
and quality of action. In other words, these theories predict the decision to coop-
erate (cooperation occurrence) and the level of benefit provided and/or the level
of costs incurred in its provision (cooperation intensity).
Table 1 presents the test of the direct reciprocity strategy, Tit-for-Tat, among
directed dyads of states. The first stage of the models tests the previous year’s
target cooperation occurrence against the current year sender cooperation occur-
rence. An occurrence of cooperation is simply the presence of at least one coop-
erative event among the dyad in the given year. There is a positive and significant
correlation between the occurrence of at least one cooperative event by target
towards sender in the previous year and at least one cooperative event by sender
towards target in the current year. Even when controlling for the current-year
target occurrence, previous year occurrence still has a positive and significant
effect. This is evidence of positive direct reciprocity because states consider the
occurrence of cooperation by potential targets when deciding to commit coopera-
tion. The weaker test of TfT, Hypothesis 1b, seems to be confirmed.
While states are in a cooperative relationship, TfT direct reciprocity should
be evident in the change of cooperation intensity. If target defects in the previ-
ous period, sender will defect in the current period. Even when controlling for
the current target change in cooperation intensity, target’s previous change in
cooperation intensity continues to be positively correlated and highly signifi-
cant. This indicates that states consider changes in strategies of their partner
states when deciding how strongly to cooperate, which is evidence in support
of Hypothesis 1a. The second model indicates that target’s change in coopera-
tion intensity is positively correlated with sender’s cooperation intensity. Since
Table 1: Direct reciprocity, Tit-for-Tat.
there is evidence that states directly reciprocate the actions of their partner states,
Hypothesis 1 seems to be confirmed.
Table 2 presents the test of Hypothesis 2, which tests if sender defects under
the condition of Win-stay/Lose-shift (WSLS). The strategy of WSLS has two condi-
tions. The first is a winning condition for sender, and the second is a defection by
target. The winning condition is operationalized as a positive trade balance with
a target. A defection is a negative change in target’s intensity of cooperation in
the previous year. If both conditions are present, which is the operationalization
of the WSLS dummy variable, there should not be positive correlation between
WSLS and defection by sender.
WSLS can only be a strategy when cooperation reciprocity is occurring,
which is modeled in the first stage. WSLS is tested in the second stage of the
model. WSLS is found to be marginally, negatively significant with sender defec-
tion. When faced with defection by target, senders that have a winning condition
either do not alter their cooperation intensity any differently than states without
the WSLS condition or they do not defect. Since the result is only marginally sig-
nificant within a sample size of 85,472, it may be that WSLS has no effect. While
the negative and marginally significant beta on the WSLS coefficient supports
Hypothesis 2 because WSLS is not positively correlated with sender defection, the
interaction terms provide evidence that contradict the causal story.
First, the winning condition measure is not significant. States that have a
positive trade balance are not more or less likely than states with a negative rela-
tive trade balance to defect when target does not defect. This indicates that a
positive relative trade balance is not a very good indicator of a winning condi-
tion in support of the WSLS theory. Second, target defection is negatively and
significantly correlated with defection by sender, when the sender’s winning con-
dition is zero. If sender has a negative trade balance and target reduces its level
of cooperation, sender will either increase its level of cooperation or there will
be no change. This may indicate that states with a positive trade balance could
use defection to induce greater levels of cooperation in response, but it is not evi-
dence of direct reciprocity based on TfT or WSLS. In the end, the WSLS measure is
only marginally significant, and due to the betas of the interaction terms, there is
little evidence to suggest that WSLS is a strategy routinely used by states.
is presented in the Model 1 of Table 3 because all three had positive and significant
results.4 The previous year target cooperative reputation is found to be positively
significant in correlation with both the current year sender cooperation occur-
rence and sender cooperation intensity. This suggests that target’s cooperative
reputation matters not only for sender’s decision to cooperate but also the deci-
sion of cooperation quality. Since these findings are consistent across cooperative
reputation measures, Hypotheses 3a and 3b seem to be confirmed.
Models 2 and 3 of Table 3 have the results of the models based on tag. A
colonial tie is operationalized as a dichotomous variable, where a one indicates
the dyad has or did have a colonial relationship. In the first stage of the models,
colonial tie is positively and significantly correlated with sender cooperation
occurrence. Senders that have a colonial tie with a target state are more likely to
initiate cooperation with that target than other targets. Hypothesis 4a seems to
be confirmed. However, a colonial tie is not a significant indicator of cooperation
quality. While having a colonial tie may make it more likely for states to initiate
cooperation towards each other, it does not factor into the decision of the quality
of cooperation. Therefore, Hypothesis 4b is not confirmed, and colonial tie may
not be a very good tag. States with a colonial tie initiate cooperation towards each
other for reasons other than being suitable candidates for cooperation.
Model 3 presents the test of sharing a plurality religion as a tag. In the first
stage of the models, sharing a plurality religion is positively and significantly cor-
related with sender’s decision to cooperate. When the dyad shares a plurality reli-
gion, sender state is more likely to cooperate with target. Hypothesis 4a seems to
be confirmed. However, sharing a plurality religion is negatively correlated with
cooperation intensity. Hypothesis 4b is not confirmed, and sharing a plurality
religion may not be a very good tag. While states that share a plurality religion
are more likely to initiate cooperation, they do so at a level that is significantly
lower than with targets they do not share a plurality religion. Therefore, states
that share a plurality religion also initiate cooperation towards each other for
reasons other than being suitable candidates for cooperation.
Table 4 contains the model with all the explanatory variables. The hypotheses
based on positive direct reciprocity and cooperative reputation are supported in
4 Table 7 has the models containing third party reputation measures based on weaker
information requirements.
Table 3: Indirect reciprocity.
(70.37) (3.54)
Colonial tie 0.176c 0.007
(12.75) (0.34)
Religion tag 0.097c − 0.079c
(20.78) (8.96)
Sender’s conflict intensity 0.132c − 0.054c 0.154c − 0.049c 0.155c − 0.051c
(53.46) (15.53) (63.40) (13.88) (63.65) (14.41)
Target’s conflict intensity 0.128c − 0.045c 0.143c − 0.041c 0.143c − 0.043c
(51.54) (12.86) (58.26) (11.50) (58.54) (12.03)
Capital distance − 0.000c 0.000 − 0.000c − 0.000 − 0.000c − 0.000
(75.45) (1.53) (80.32) (0.18) (79.73) (0.83)
Power asymmetry − 0.565c − 0.057 − 0.601c − 0.077a − 0.601c − 0.079b
(38.07) (1.86) (41.80) (2.52) (41.74) (2.59)
Alliances 0.383c 0.462c 0.419c
(59.47) (73.57) (63.17)
Sender major power 0.581c 0.029 0.594c 0.049b 0.621c 0.031
(66.38) (1.84) (68.53) (3.12) (72.56) (1.94)
Target major power − 0.026a − 0.211c 0.595c − 0.131c 0.621c − 0.151c
(2.13) (10.11) (68.30) (8.21) (72.22) (9.43)
Major power interaction − 0.016 − 0.230c − 0.242c − 0.260c − 0.261c − 0.251c
(0.37) (5.23) (5.85) (5.90) (6.31) (5.70)
Sender democracy − 0.151c − 0.059c − 0.189c − 0.067c − 0.190c − 0.061c
Table 3 (continued)
the first stage of the model. Of the variables in the model that could be considered
a tag, including colonial tie, plurality religion, and democracy, only plurality reli-
gion has a positively significant correlation with cooperation occurrence. When
controlling for target’s cooperation occurrence and cooperative reputation, tags
become a less significant indicator of cooperation occurrence. Overall, it seems
that it is reputation and not tags that matters for cooperation occurrence. In other
words, sender states will initiate cooperation with target states that are either
Testing Evolution of Cooperation Theories 27
Variable Δ Probability Δ
Variable Δ Predicted Δ
measures. The effect is an increase of 0.096, which is only about 3.39% of the
mean.
Therefore to induce interstate cooperation reciprocity, states must be
cooperators. Target’s cooperation towards sender has a strong positive effect
on sender’s decision to cooperate and the quality of cooperation. Target’s
cooperative reputation has weaker effects on sender’s decision to cooperate
and the quality of cooperation, but it still matters. Sharing some of the most
salient characteristics, such as religion, colonial ties, and democracy, does
not matter near as much, if at all, for interstate cooperation. The findings for
joint democracy are problematic for previous studies. While some models find
consistent results with previous studies, in which joint democracy increases
the likelihood and quality of cooperation, it is not found to be robust in the
combined model. This provides further evidence that direct and indirect coop-
erative action more consistently affects interstate cooperation than sharing
state characteristics.
While the control variables mostly behave as predicted, some do not have
consistent effects. First, the distance between capitals is predicted to reduce the
likelihood and level of cooperation. Capital distance is found to consistently
reduce the likelihood of cooperation, but while it has the predicted direction of
effect, capital distance is not a robust predictor of cooperation intensity. Having
an alliance between the states in the dyad is predicted and found to increase the
likelihood of cooperation. As the relative power difference increases between
the states in the dyad, senders are predicted and found to be less likely to ini-
tiate cooperative events, but while asymmetric power is predicted to reduce
sender’s cooperation intensity, asymmetric power has a less consistent effect
on the quality of cooperation. Of all the control variables, the least consistent
is major power status. Major powers were predicted to be less likely to initiate
30 Jeremy Bowling
cooperation and have a lower cooperation intensity with other major powers.
When sender is a major power, sender is found to be more likely to initiate
cooperation when target is not a major power, and while there is some evidence
that senders have a higher level of cooperation with non-major powers, it is not
as robust across all models. Joint major power status has no consistent effect
across models, which may be due to a combination of realist reduction in like-
lihood and increased opportunity due to increased interaction. Lastly, sender
and target conflict intensity is predicted and consistently found to increase the
likelihood of a cooperation occurrence and reduce cooperation intensity.
5 Conclusion
While all five strategies tended to behave as hypothesized in various models,
there are inconsistent results for some of the evolution of cooperation strat-
egies, particularly those based on tag. States seem to behave as individu-
als behave in the evolution of cooperation literature, but direct cooperative
action is the major cause of reciprocity. There is evidence for theories of direct
reciprocity, particularly the strategy of Tit-for-Tat. Target’s change in coopera-
tion intensity is predicted and found to be positively correlated with sender’s
change in cooperation intensity. Win-Stay/Lose-Shift predicts that sender will
not defect when target defects if sender has a winning condition. The WSLS
conditions are predicted and found to not be negatively correlated with sender
defection, but there is little evidence to support its causal story. While the
indirect reciprocity theory of Tag was not robust, cooperative reputation was
found to have a consistent effect. Cooperative reputation was predicted and
found to be positively correlated with both cooperation occurrence and coop-
eration intensity. The tags of colonial ties, sharing plurality religion, and joint
democracy are predicted and found to be positively correlated with coopera-
tion occurrence, but colonial ties and joint democracy was not robust. None
of the tag measures had consistent and predicted results in correlation with
cooperation intensity.
These results have significant implications for international coopera-
tion. The strongest strategy to elicit cooperation is to simply be a cooperator.
States that initiate cooperation with others are more likely to receive coop-
eration in return, where the strongest likelihood and best quality of coopera-
tion are returned by those states that are the direct recipients of the initiator’s
Testing Evolution of Cooperation Theories 31
cooperative acts. Overall, states should be aware that cooperation will beget
cooperation and defection will beget defection, but there is some evidence that,
if a state has a positive trade balance, it may induce greater cooperation by
defection.
While the combined effect of Tit-for-Tat and cooperative reputation can
increase the likelihood of reciprocal cooperation occurrence by 35%, tags do not
increase the likelihood or quality of cooperation. There is some evidence that
sharing a plurality religion makes states more likely to initiate cooperation, but
the effect is weak and not likely due to the characteristic signaling a best coopera-
tor. Plurality religion has a negative effect on the quality of cooperation. While
democracy, religion, and colonial ties may influence international cooperation, it
is not due to the theory of tag.
In the future, this work should be expanded to include a more in-depth exam-
ination of new theories of indirect reciprocity based on image scoring (Nowak &
Sigmund, 1998) and standing strategy (Sugden, 1986). The future studies of direct
reciprocity should examine the relationships between certain specific coopera-
tive acts to parse the magnitude of effect of the reciprocity between types of acts.
In other words, does military aid make target states more cooperative than eco-
nomic aid? Further analysis of the relationship between state characteristics and
international cooperation need to be examined. What is the theoretical relation-
ship between regime type and international cooperation, controlling for direct
and indirect reciprocity? What is the relationship between colonial tie and inter-
national cooperation? How do vertical, horizontal, closed-field, and open-field
relationships affect international cooperation? Lastly, why are states more likely
to initiate cooperation towards states that have the same religion but have a
reduced quality of cooperation?
Table 7: Other third-party cooperative reputation measures.
References
Axelrod, R. (1984). The evolution of cooperation. New York, NY: Basic Books.
Axelrod, R., & Hamilton, W. D. (1981). The evolution of cooperation. Science, 211, 1390–1396.
Azar, E. E. (1980). The Conflict and Peace Data Bank (COPDAD) project. Journal of Conflict
Resolution, 24(1), 143–152.
Barbieri, K., & Keshk, O. (2012). Correlates of War Project Trade Data Set Codebook, Version 3.0.
http://correlatesofwar.org.
Barbieri, K., Keshk, O., & Pollins, B. (2009). TRADING DATA: Evaluating our assumptions and
coding rules. Conflict Management and Peace Science, 26(5), 471–491.
Barker, R. G., & Wright, H. F. (1955). Midwest and its children. New York, NY: Harper.
Beck, N., Katz, J. N., & Tucker, R. (1998). Taking time seriously in binary time-series cross-
section analysis. American Journal of Political Science, 42(4), 1260–1288.
Berscheid, E. (1986). Emotional experience in close relationships: Some implications for child
development. In W. W. Hartup & Z. Rubin (Eds.), Relationships and development. Hillsdale,
NJ: Erlbaum.
Bowling, J. (2015). Determinants of international cooperation: Reciprocity, opportunity, and
cooperative norms. PhD diss., University of Missouri-Columbia.
Carter, D. B., & Signorino, C. S. (2010). Back to the future: Modeling time dependence in binary
data. Political Analysis, 18(3), 271–292.
Cheibub, J. A., Gandhi, J., & Vreeland, J. R. (2010). Democracy and dictatorship revisited. Public
Choice, 143(1–2), 67–101.
Collins, W. A., & Laursen, B. (1992). Conflict and relationships during adolescence. In C. U.
Shantz & W. W. Hartup (Eds.), Conflict in child and adolescent development. New York, NY:
Cambridge University Press.
Correlates of War Project. (2008). State System Membership List, v2008.1. http://correlatesofwar.org.
Dawkins, R. (1976). The selfish gene. Oxford: Oxford University Press.
Gibler, D. M. (2009). International military alliances, 1648–2008. Washington, D.C.: CQ Press.
Gleditsch, K. S. (2002). Expanded trade and GDP data. Journal of Conflict Resolution, 46,
712–724.
Gleditsch, K. S. (2004). Kristian Skrede Gleditsch//Data, etc, Accessed September 9, 2014,
http://privatewww.essex.ac.uk/∼ksg/data-5.html.
Goldstein, J. (1992). A conflict-cooperation scale for WEIS International Events Data. Journal of
Conflict Resolution, 36, 369–385.
Graziano, W. G. (1984). Lost in thought at the choice point: Cognition, context, and equity.
In J. C. Masters & W. P. Smith (Eds.), Social comparison, social justice, and relative
deprivation: Theoretical, empirical, and policy perspectives. Hillsdale, NJ: Erlbaum.
Hamilton, W. D. (1964). The genetical evolution of social behavior. Journal of Theoretical
Biology, 7, 1–16.
Hensel, P. R. (2009). ICOW Colonial History Data Set, version 0.4. Accessed September 9, 2014.
http://www.icow.org/colhist.html.
Hudson, V. M., Schrodt, P. A., & Whitmer, R. D. (2008). Discrete sequence rule models as a
social science methodology: An exploratory analysis of foreign policy rule enactment
within Palestinian-Israeli data. Foreign Policy Analysis, 4(2), 105–126.
Kansas Event Data System (KEDS) Project Website. (2010). Levant Data Sets: Levant WEIS data.
Accessed November 5, 2011. http://web.ku.edu/∼keds/data.dir/levant.html.
34 Jeremy Bowling
Keohane, R. O. (1984). After hegemony: Cooperation and discord in the world political economy.
Princeton: Princeton University Press, 2005.
Leeds, B. A. (1999). Domestic political institutions, credible commitments, and international
cooperation. American Journal of Political Science, 43(4), 979–1002.
Leimar, O., & Hammerstein, P. (2001). Evolution of cooperation through indirect reciprocity.
Biological Sciences, 268(1468), 745–753.
Luo, Y. (2002). Multinational enterprises in emerging markets. Hendon: Copenhagen Business
School Press.
Maoz, Z., & Henderson, E. A. (2013). The world religion dataset, 1945–2010: Logic, estimates,
and trends. International Interactions, 39(3), 265–291.
McClelland, C. (1978). World event interaction survey (Computer file). Los Angeles: University of
Southern California.
Milinski, M., Semman, D., Baker, T. C. M., & Krambeck, H. (2001). Cooperation through
indirect reciprocity: Image scoring or standing strategy? Biological Sciences, 268(1484),
2495–2501.
Moll, H., & Tomasello, M. (2007). Cooperation and human cognition: The Vygotskian
intelligence hypothesis. Philosophical Transactions: Biological Sciences, 362(1480),
639–648.
Nowak, M. A. (2006). Five rules for the evolution of cooperation. Science, 314, 1560–1563.
Nowak, M. A., & Sigmund, K. (1993). Letters to nature: A strategy of win-stay, lose-shift that
outperforms tit-for-tat in the Prisoner’s Dilemma game. Nature, 364, 56–58.
Nowak, M. A., & Sigmund, K. (1998). Letters to nature: Evolution of indirect reciprocity by image
scoring. Nature, 393, 573–577.
Reuveny, R., & Kang, H. (1996). International conflict and cooperation: Splicing COPDAB and
WIES series. International Studies Quarterly, 40, 281–305.
Riolo, R. L., Cohen, M. D., & Axelrod, R. (2001). Letters to nature: Evolution of cooperation
without reciprocity. Nature, 414, 441–443.
Rummel, R. J. (1971). A status field theory of international relations. Dimensionality of Nations
Project, Research Report 55, University of Hawaii.
Sigelman, L., & Zeng, L. (1999). Analyzing censored and sample-selected data with Tobit and
Heckit models. Political Analysis, 8(2), 167–182.
Singer, J. D. (1987). Reconstructing the correlates of war dataset on material capabilities of
States, 1816–1985. International Interactions, 14, 115–132.
Singer, J. D., & Small, M. (1966). Formal alliances, 1815–1939. Journal of Peace Research, 3, 1–31.
Singer, J. D., Bremer, S., & Stuckey, J. (1972). Capability distribution, uncertainty, and major
power war, 1820–1965. In B. Russett (Ed.), Peace, war, and numbers (pp. 19–48). Beverly
Hills: Sage.
Small, M., & Singer, J. D. (1969). Formal alliances, 1815–1965: An extension of the basic data.
Journal of Peace Research, 6, 257–282.
Sugden, R. (1986). The economics of rights, co-operation and welfare. Oxford: Basil Blackwell.
Stevens, J. R., Cushman, F. A., & Hauser, M. D. (2005). Evolving the psychological mechanisms
for cooperation. Annual Review of Ecology, Evolution, and Systematics, 36, 499–518.
Tomasello, M., Carpenter, M., Call, J., Behne, T., & Moll, H. (2005). Understanding and sharing
intentions: The ontogeny and phylogeny of cultural cognition. Behavioral and Brain
Sciences, 28(5), 675–735.
Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes.
Cambridge, MA: Harvard University Press.
Testing Evolution of Cooperation Theories 35
Ward, M. D. (1982). Cooperation and conflict in foreign policy behavior: Reaction and memory.
International Studies Quarterly, 26(1), 87–126.
Wilson, E. O. (1975). Sociobiology. Cambridge, MA: Harvard University Press.
Zahavi, A. (1977). Reliability in communication systems and the evolution of altruism. In
B. Stonehouse & C. M. Perrins (Eds.), Evolutionary Ecology (pp. 253–259). London:
Macmillan Press.
Zahavi, A. (1995). Altruism as a handicap-the limitations of kin selection and reciprocity. Journal
of Avian Biology, 26, 1–3.