1. I. Introduction
I chose the paper “Gender Differences in the Laboratory: Evidence from Prisoner’s Dilemma
Games” by Andreas Ormann and Lisa Tichy because it is a paper dedicated to observing how a
factor that is not included in textbook game theory may still influence the outcomes of a game.
To me, experiments such as the one described in this paper are in some ways more important than
typical economic theory because they acknowledge that people are not simple creatures who can
all be expected to behave the same way. Though this experiment, as I will go over, does not
come to any definite conclusions on the effect of gender on game play in prisoner’s dilemma
situations, hopefully, through papers like this one and ones to come we can find some decisive
evidence that can help evolve economic theory into a science based not only on theoretical
behavior but actual behavior.
II. Summary
This research was done to look for empirical regularities while running experiments on the
prisoner’s dilemma game. This comparison was done in the form of comparing the results of a
prisoner’s dilemma-type game design using subjects of different genders. The research question
that the authors were trying to answer was whether or not women exhibit more cooperative
behavior than men do. They also observed how both genders’ behaviors change over multiple
trials of the game (ie. how subjects learned over the course of multiple rounds). Previous
psychological research as well as past experiments using the ultimatum, dictator, and public good
provision design games led the authors to predict that women would act more cooperatively than
men.
A prisoner’s dilemma game has two properties. Firstly, that there is a single Nash
equilibrium as a result of both player’s dominant strategies. Second, there is another outcome
that would be better for both players. The theory that this experiment is based on predicts that
both players will pick their dominant strategy and thus end up at the non-cooperative Nash
equilibrium.

2. An important part of the prisoner’s dilemma institution is that subjects should not be allowed
to communicate directly or know each other at all. If two participants playing each other are best
friends outside of the laboratory it is much more likely that they will end up at the cooperative
outcome (both choosing option one) because they have a pre-established trust. In this particular
experiment, randomly assigning partners and then not allowing subjects to know who they were
paired with ensured anonymity.
The experimenters used three different treatments in the form of gender. Eight sessions were
done with both males and females (three-four of each gender). Two were done with only females
and two were done with only males. There were eight subjects in each session. Each subject
played against a single other subject in each game (round) for a total of seven games played in
each session. So there were a total of fifty-six games played under the mixed-gender treatment,
fourteen games with all female subject, and fourteen games played with all male subjects. All
subjects knew the payoff schedule that they were playing. They knew that they were playing a
different person each game but they did not know the gender of their opponent in any of the
rounds (except,of course, the same-gender sessions when only males or only females were
present). Subjects were paid an initial $3-4 for participating. Average earnings for the
experiment were about $10 per subject.
In a single classroom, two experimenters (one male and one female to ensure that behavior
was not influenced by the gender of the experimenter) read aloud the first part of the instructions
and distributed on paper. Once the experiment started they switched to step-by-step oral
instructions. They drew the payoff that was being used for that round on the board (payoff tables
were switched in round four of each session). Each subject had been randomly assigned an
identification letter and was given decision sheets with their letter on them. Subjects then made
their decisions, folded the sheets in a uniform way, and inserted them into a box. Once all sheets
were collected, subjects were given back the decision sheet with their randomly selected
opponent’s identification letter on it. Subjects then recorded both their and their opponent’s

3. decision on their record sheet (on which they had also recorded their own genders). At the end of
the session, the record sheets were put into envelopes with a name or word that each subject had
chosen on them. An experimenter went through the envelopes and placed each subject’s
winnings into his or her envelope. Once the experimenter left the room, subjects picked up their
envelopes. The design of this experiment was a double-blind. Neither the subjects nor the
experimenter knew who was paired with whom. This was done so that the experimenters could
not influence subjects’ behaviors by unknowingly encouraging the actions predicted and subjects
could not behave as they thought that they were expected to behave.
Over the entire experimental process the cooperation rate for female subjects was 41% and
for males it was 30%. In the first round (pooled data from both mixed-sex and same-sex
sessions), the cooperation rate for female subjects was 62% and for male subjects it was 41%.
From these data the authors concluded that female subjects are more cooperative than male
subjects. However,when comparing pooled data by round across all sessions, they found that as
multiple games were played, both overall cooperation and the difference between male and
female cooperation rates decreased. By the seventh game (round) cooperation of female subjects
had decreased from 62% in the first round to 22% and the rate for male subjects had gone from
41% to 15%. When comparing cooperation rates in mixed-gender and same-gender treatments,
the authors found that in the first round of mixed-gender sessions women cooperated 65% of the
time whereas in the first round of the all-female sessions they cooperated only 50% of the time.
Men, on the other hand cooperated only 27% of the time in the first round of mixed-sex sessions
but in single-sex session their cooperation rate was higher at 38%. In other words, when playing
against their own sex, men appear to cooperate somewhat more whereas women appear to
cooperate somewhat less.
In the first round of mixed-gender sessions the authors got the expected results. That is, in
the first round women were significantly more cooperative. However,by the ending rounds this
statistical significance went away and women and men appeared more equally cooperative. This

4. suggests that while initially men and women may perceive their environments differently (women
as part of a group, men as an individual) their decisions become continuously more similar as
their past experiences become more similar. The authors conclude that, based on these results
(and others past), it is too soon to make any strong arguments about the affect of gender
differences in the laboratory except that these differences definitely do have an effect.
III. Evaluation
Overall the procedures as they were reported would permit replication. The biggest thing that
they were lacking was how the payoffs were transferred into monetary awards for the subjects.
The payoffs in the provided payoff tables are not in dollars, clearly and other than that all that is
provided are maximum, minimum and average figures for each session and an average earnings
for the experiment as a whole. Since there are no instructions on how the payoff table got
converted into cash it is hard to precisely evaluate if the incentives were enough to establish
dominance. However,the authors did report that average earning were about $10, which was
about twice the average hourly wage of the subjects. Since each session was about fifty minutes
long I think that this incentive was appropriate. Other than that I believe that any experimenter
could implement this experiment with little difficulty with the description given.
The authors explained that in studies done with more than eight subjects per session the
transaction costs quickly escaladed. Because each subject had to have a different opponent in
every round there could only be seven rounds with a group of eight subjects. This means that for
each session they got fifty-six observations. Since they did eight mixed-gender they obtained
448 observations under this treatment, which I think is enough to justify their results. However,
they ran only two sessions, for a total of 112 observations in both of the same-sex treatments. I
find a slight issue with this for two reasons. Firstly, when looking at the cooperation rates across
all rounds for all session, only 16% (112 same-sex sessions/672 total sessions) of that data is
coming from all-female or all-male subject pools. Since only the first round of same-gender data
is reported in this paper it is hard to say how substantial this is because we do not know which—

5. if any—direction they same-sex observations may have skewed the overall results. Secondly and
I think more importantly, this means that when the authors analyze the data from the first round
of same-sex sessions they are only looking at between fifteen and seventeen observations. They
had made fifty female and forty-six male first-round observations in mixed-sex treatments and
chose to compare their results from the same-sex treatments with sessions 1, 2, 5, and 6 without
justifying why those were the mixed-sex sessions chosen. I would have rather seen all mixed-sex
observations compared with a proportional number of same-sex observations in order to justify
their conclusion that women are less cooperative in all-female treatments and men are more
cooperative in all-male treatments.
I thought that the best part of this paper was their emphasis on controlling how much both the
subjects and the experimenters knew during the games. Knowledge and communication in the
prisoner’s dilemma game is absolutely essential and I thought that the authors did an exceptional
job making sure that none of the subjects ever knew who they were playing against. They even
offered cash prizes to any subjects that could find a part of their experimental design that violated
anonymity. I did, however, see a potential confounding variable in their design having to do with
the payoff tables used. In two of the payoff tables, the defective strategy only weakly dominates
the cooperative strategy. That is, option two gives you a higher payoff when your opponent
choses option one but when your opponent chooses option two you would get the same payoff
from choosing option one or option two. In the remaining payoff schedule option two is strictly
dominant. Whatever your opponent chooses, option two will give you a higher payoff. The
strictly dominant payoff schedule was only used in four of the mixed-gender sessions. The
authors do not discuss the different schedules as a treatment that would have any effect on the
results but, personally, I think that it is an unnecessary factor. I would have rather seen the
experiment run with either all weakly or all strictly dominant payoff tables.
Even though this experiment was not enough to make any generalized conclusions about the
effect that gender differences in the laboratory may have on results it came to the very important

6. pronouncement that these differences do have an effect. This means a lot when looking at past
and future experiments because it adds a factor that should be reported when discussing the
subject pool (much like the experience level of the subjects). This is especially important because
it is known that many past experiments have been made up of mostly male subjects. If, indeed,
men and women do behave consistently differently, these past results may need to be revisited.
IV. Experimental Proposal
As a follow up to this experiment I propose to run computerized, voluntary contribution
experiments and look for empirical regularities concerning the gender of the subject. I would
especially like to look at whether or not the subjects knowing that they are in groups of same-
gender vs. mixed-gender groups has an affect on the cooperation rate. The authors of this paper
were testing the former observations that women are more apt to see themselves as a part of a
communal group and men are more likely to think as individuals but they used a design in which
there were only two players, so each individual’s decision effected only one other individual.
While theoretically the number of participants should not influence the results so long as there is
still a unique, pure-strategy Nash equilibrium, I think that, behaviorally, the group aspect of the
voluntary contributions method may induce different results when comparing men and women.
Since we know that both the number of participants and the level of marginal per capita
return effect levels of cooperation (Davis and Holt, 330), I would like to run the experiment with
average levels of both. Group size should be ten people and MPCR .5 for the group exchange
and 1 for the individual.
I would like there to be four treatments. The first is a baseline observation of twenty subjects
(ten male, ten female), randomly split into two groups in the same laboratory. Then I would like
to look at the same twenty subjects split into groups by gender but still in the same laboratory so
that none of the participants are aware that they are in groups with only people of the same
gender. I would also need to run sessions where these two treatments are reversed. That is,the
twenty subjects are first split according to gender without them knowing and then split randomly.

7. I would like the next two treatments to be groups of ten subjects at a time all of the same gender,
in the same laboratory so that they know that they are in groups with only people of the same
gender. All subjects in all sessions should be told how many periods the experiment will last for
since game theory tells us that games with infinite or unknown repetition will have different
outcomes than those of finite, known repetition. It is important that the experiment be
computerized so that there is no way for the subjects to know who is in their group.
Since past psychological research has shown that women are more cooperative than men I
would expect that the sessions where all women are grouped together without their knowledge
would have the highest cooperation rate and the sessions where men are grouped together without
their knowledge would have the lowest. However,because of the results of the evaluated paper I
would expect the sessions with all women (therefore subjects know they are grouped with only
women) to have lower cooperation rates than the sessions with all men (where subjects know that
they are grouped with only men).
V. Conclusion
Overall I thought that this experiment was extremely well done and came to an enormously
important conclusion. Even if the results were perhaps not as definitive as the authors wanted, in
that they were unable to make any sweeping claims about the effects of gender differences in the
laboratory, I think that they paved the way for more experiments to come, such as the one that I
have laid out above and at the very least have brought attention to gender as a possible
confounding factor for future experiments.

8. Work Cited
Davis, Douglas D., and Charles A. Holt. Experimental Economics.Princeton,NJ: Princeton
University Press, 1993. 325-333. Print.
Ortmann, Andreas, and Lisa K. Tichy. “Gender differences in the laboratory: evidence from
prisoner’s dilemma games”. Journal of Economic Behavior& Organization. Volume 39,
Issue 3. July 1999. Pages 327-339.

9. An Analysis of
“Gender Differences in the Laboratory:
Evidence from Prisoner’s Dilemma Games”
Andrea Woogerd
June 6, 2013
Econ 430
Final Paper