Behavioural and Experimental Economics

Practice Class Week 1

Introduction
In the practice session, you will discuss these papers in light of what you learn in Topic 1.

If you wish, you can read the full papers by finding them on Google Scholar (which would be good
practice for you!). However, the synopses give you the main details of the paper and should be
sufficient for the practice session.

Questions for the Practice Class

Both papers

• What are the research question(s)? Why are they important?

• What treatments are included in the experiment
• What type of experiment is described? You could consider the Harrison and List taxonomy
from Topic 1.2. Why do you think the authors chose this type of experiment?
• How was the sample selected? Is the sample WEIRD (see Topic 1.3)?
• How was the sample size selected?
• How are the participants incentivised in the experiment?
• Is the experiment internally and externally valid? How could external validity be improved?
• How do the authors draw conclusions from their results? Could you draw any additional
conclusions or recommendations?

Drouvelis et al. (2018)

• What do you think are the key benefits and drawbacks of studying this unusual population?
• Can you think of other experiments you would like to do with these participants?

Jamison et al. (2008)

• After reading the synopsis, do you believe deception in experimental economics is “worth
the risk”?
• If you were taking part in this experiment, how would you have felt when you found out you
had been deceived? Do you think it would have influenced your future behaviour?

1
PAPER 1
Drouvelis, M., Malaeb, B., Vlassopoulos, M., & Wahba, J. (2021). Cooperation in a fragmented
society: Experimental evidence on Syrian refugees and natives in Lebanon. Journal of Economic
Behavior & Organization, 187, 176-191.

Abstract
Lebanon is the country with the highest density of refugees in the world, raising the question of
whether the host and refugee populations can cooperate harmoniously. We conduct a lab-in-the-
field experiment in Lebanon studying intra- and inter-group behavior of Syrian refugees and
Lebanese nationals in a repeated public goods game without and with punishment. We randomly
assign participants to Lebanese-only, Syrian-only, or mixed sessions. We find that randomly formed
pairs in homogeneous sessions, on average, contribute and punish significantly more than those in
mixed sessions, suggesting in-group cooperation is stronger. These patterns are driven by Lebanese
participants. Further analysis indicates that behavior in mixed groups is more strongly conditioned
on expectations about the partner's cooperation than in homogeneous groups.

Paper synopsis
Background and context

• In 2018, Syrians were the largest forcibly displaced population – with 6.7 million Syrian
refugees. Lebanon is one of the world’s largest embracers of refugees. Almost 1 million
Syrian refugees were hosted in Lebanon in 2018.
• There is some evidence of strained relationships between Lebanese nationals and Syrian
refugees. This raises the question of whether the two populations can co-exist and co-
operate.
• The provision of public goods (recall micro!) relies on co-operation between people.
Previous studies showed ethnic diversity can harm cooperation and public good provision
and people tend to favour members of their own social group (their “in-group”). But not
much is known about how these dynamics play out with refugees and host communities.
• This paper aims to shed empirical light on the intra- and inter-group cooperation of Syrian
refugees and Lebanese communities.

Experimental design and procedures

• The authors use a linear social dilemma game without and with punishment opportunities.
• Subjects are randomly assigned to a two-person group.
• Each subject is endowed with 10 tokens. They decide how many to keep and how many to
contribute to the public good. Keeping a token means adding 1 point to their own payoff.
Contributing it increases everyone’s payoff by 0.75 points (so it costs them 0.25 points but
benefits the other player by 0.75 points), i.e. 𝜋𝑖 = 10 − 𝑔𝑖 + 0.75 ⋅ (𝑔𝑖 + 𝑔𝑗 ) where 𝑔𝑖 is an
individual i’s contribution, 𝜋𝑖 is the payoff, and 𝑔𝑗≠𝑖 is the other player’s contribution.
• In a second stage, it is possible to punish other players by assigning up to 5 punishment
points after seeing what the other player contributed. Each point costs the punisher 1 token,
but costs the punished player 3 tokens, from their final payoffs.
• For every participant, one out of the six rounds was randomly drawn for real payment. All
payments were paid in cash at the end of the experiment.

2
 • The game theory prediction is that all subjects free-ride when there is no punishment, and
no subject ever uses punishment (because it is costly for them).
• Participants are randomised into one of three possibilities: a pair with only Syrians, a pair
with only Lebanese, or a mixed pair. They stayed in the same pair throughout the game.
• Participants were recruited in the Aley region in Lebanon. The experiment was advertised
and participants signed up by phone or online.
• 312 participants took part: 78 in Syrians only pairs, 70 in Lebanese only pairs, and 164 in
mixed pairs.
• To make participants aware of whether they were with Syrians or Lebanese people the
researchers used an ice-breaker where the accents of the people would reveal which group
they belonged to.

Results

• A balance check was done to make sure Syrians in the Syrian only treatment were not
different on observable characteristics than Syrians in the mixed groups, and the same was
done for Lebanese participants. This was largely successful.
• Fig. 1. Average contributions over rounds across treatments in the public goods game
without and with punishment.

• RESULT 1. In the public goods game without and with punishment, average contributions are
significantly higher in the homogeneous groups compared to the mixed groups. We find the
same treatment differences when we examine beliefs about partners’ contributions. These
patterns are driven by the behavior of Lebanese participants.
• RESULT 2. In the public goods game with punishment, average punishment assigned is
significantly lower in mixed groups compared to homogeneous groups. We also observe the
use of antisocial punishment. These patterns are driven by the behavior of Lebanese
participants.

3
 •
• RESULT 3. Average earnings are significantly lower in mixed groups compared to
homogeneous groups in the public goods game with punishment, when we only consider
earnings from the contribution stage. If we consider total earnings including costs of
punishment, there are no differences between homogeneous and mixed groups.

Discussion/Conclusion
• Mixing of the two groups leads to lower contributions to the public good, making individuals
worse off.
• Sanctions (punishment) are not able to redress this lack of cooperation.
• It appears that the host community shows less reciprocity and cooperation towards the
refugees.
• In many countries hosting large numbers of refugees, this implies a need to help both
refugee and host communities to alleviate social and economic pressures.
• Interventions aimed at increasing inter-group trust and cooperation would help reduce
biases and increase public good provision, and lead to improved well-being of refugees and
their hosting communities.

4
PAPER 2
Jamison, J., Karlan, D., & Schechter, L. (2008). To deceive or not to deceive: The effect of deception
on behavior in future laboratory experiments. Journal of Economic Behavior & Organization, 68(3-4),
477-488.
Abstract
Experimental economists believe (and enforce the idea) that researchers should not employ
deception in the design of experiments. This rule exists in order to protect a public good: the ability
of other researchers to conduct experiments and to have participants trust their instructions to be
an accurate representation of the game being played. Yet other social sciences, particularly
psychology, do not maintain such a rule. We examine whether such a public goods problem exists by
purposefully deceiving some participants in one study, informing them of this fact, and then
examining whether these deceived participants behave differently in a subsequent study. We find
significant differences in the selection of individuals who return to play after being deceived as well
as (to a lesser exten) the behaviour in the subsequent games, thus providing qualified support for
the proscription of deception. We discuss policy implications for the maintenance of separate
participant pools.

Paper synopsis
Background and context

• Economists prefer to avoid deception because they want the participants to continue to
trust the experimental instructions in future. But this is costly, because it rules out some
research questions that could only be answered with deception.
• Some research in psychology suggests participants exposed to deception have different
expectations, suspicions, and future behaviour.
• Deception can come in many forms but often it relates to who the other players are, e.g.
pretending they are humans but they’re really computer algorithms, or pretending there are
many but really there are only a few others playing, etc. In this study they focus on only one
type of deception (see below).
Experimental procedures
• Participants were members of the student participant pool for experiments at the University
of California at Berkeley.
• 261 subjects played across 10 one-hour sessions.
• 5 sessions had deception, 5 did not have deception.
• Without deception: subjects were assigned to roles, matched into pairs, played a game, and
then got paid according to their play.
• With deception: subjects were assigned to roles, matched with a computer algorithm (but
told it was another player), played a game, and then got paid according to their play.
Deception was revealed at the end when they got their payoff.
• Next (after 2-3 weeks) they were invited back to a new experiment, with different
researchers. 142 participants returned (so 119 did not). At the end, they were told the two
experiments were linked.
• In the first phase, they played a trust game: $20 endowment was given to player 1, who
chose to give some (or all, or none) of it to player 2. Money transferred was tripled in value.
Then player 2 chose to give some (or all or none) of it back to player 1.
• In the second phase, they played three different games, all common in
behavioural/experimental economics. You will meet some of this games later in our course!

5
Results

• Successful randomisation checks were conducted

• The main interesting row in table 1 shows that women who were not deceived were
significantly more likely to return than females who were deceived.
• Those recipients who passed less than half of their maximum possible transfer back to player
A were also less likely to return if they had been deceived.
• However, males who were deceived were more likely to return than males who were not
deceived. This is hard to interpret.

• Those who received a low amount back AND were deceived were much less likely to return
to the lab than those who received a low amount back and were not deceived.
• “Summarizing Table 2, being deceived made individuals less likely to return only when they
were unlucky insofar as having received a low payout.”
• Participants also behaved differently in some of the games in the later experiments if they
had been deceived. In particular, they seem to be more inconsistent in a risk aversion task
(suggesting they take it less seriously).
• Many of the effects do not stand up to “multiple hypothesis testing” corrections like
Bonferroni tests.

6
Discussion/Conclusions
• Mixed evidence of whether deception really matters in experimental economics.
• It seems to have some effect, but arguably small and not more than other sources of bias in
real life behaviour.
• Nonetheless, because some systematic effects were found, it seems sensible to control for
deception as far as possible.
• Interesting point: what if something is not deception but subjects might think it is? E.g. if a
fair coin is tossed 10 times it COULD come up with 10 heads. If a subject thinks this is
deception, what should the experimenter do about it?
• Also, what if real life involves deception? Does an experiment lose external validity if it rules
out deception altogether, when in real life it does exist?