Experimental Economic Games (Winter 2017)

This Proseminar, co-led by four ISS Fellows, explored the interdisciplinary use of experimental economic games (EEG) and their role in enhancing our understanding of the decision-making process.

ISS Fellows: Monique Borgerhoff Mulder (Anthropology), Mark Lubell (Environmental Science and Policy), Travis Lybbert (Agricultural and Resource Economics) and Peter Richerson (Environmental Science and Policy).


ANT 298 | Thursdays, 1:00 - 3:00 p.m. | 224 Young Hall | Winter 2017 | CRN 12348 | Flyer


Jump to Proseminar Blog


The study of issues as diverse as economic development, inequality, biodiversity conservation, health service delivery and household economics depend on assumptions regarding individual preferences and strategic behavior. Increasingly, economic experiments, and especially experimental economic games (EEG) conducted in field situations, are being used to understand how humans make decisions under different scenarios.

These games are designed to uncover predispositions towards cooperativeness, to identify the role of norms of trust and reciprocity in substituting for formal institutions, to determine the extent to which such norms are built on notions of fairness and equity, and to titrate the extent to which decisions are influenced by risk and time preference. Investigators’ concerns range from probing theoretical models for the evolution of cooperation to a series of applied issues. These include, but are not limited to:

-       the potential role of local norms as opposed to state regulations in providing public goods

-       the potential for cooperation over the protection of natural resources

-       the means of regulating the use of common pool resources

-       the rendering of effective policy interventions such as in health care, and

-       the eradication of poverty.


As such, experimental economic games are playing a central role in both theoretical and applied social science. The enormous growth in the use of these methods has not come without critique, however. Issues to consider include:

-       appropriate sampling (including convenience samples)

-       the design of protocols for non (or semi) literate subjects

-       the need to provide explanations for game play that do not add unanticipated framing effects

-       the potential for deliberate framing to test culture specific models

-       the need to check levels of understanding (again without adding unplanned framing)

-       how to cope with (and/or integrate) culturally-constructed notions of chance and probability subjects bring to the experiment

-       the appropriate form of payment (cash or durables)

-       experimenter credibility, and

-       the minimization of cross-talking among untested and tested subjects.


Each meeting of this Proseminar will, using actual data sets, focus on a particular game, including Dictator Game, Ultimatum Game, Prisoner’s Dilemma, Trust Game, and Common Pool Resource, among others. Attendees will discuss anticipated inference, challenges of running the experiment in the field, pros and cons of different technologies, cultural context, and validity of interpretation.

Readings will generalize from the use of EEG to the broader understanding of preferences, individual variation in preferences, and the nature and determinants of community differences in preferences (as revealed by EEG), leading to a more general discussion of the overall utility and shortcomings of this methodology.

The Proseminar may also include lectures from visiting speakers experienced in using EEG in the field, and a ‘fieldtrip’ to an experimental games lab on campus.


Proseminar Blog

EEG Schedule

1. Introduction

January 12, 2017

Our opening class focused on general introductions to the field of experimental games based on two readings:

Cardenas and Carpenter (2008). Behavioural Development Economics: Lessons from Field Labs in the developing world. Journal of Development Studies, Vol. 44, No. 3, 311–338.

Van Lange, P. A. M. et al. (2013). The psychology of social Dilemmas: A Review. Organizational Behavior and Human Decision Processes 120: 126-141.

We have eight students enrolled, two auditors, one postdoc and a number of faculty/staff who have expressed interest in attending.

General introductions indicated a huge breadth of disciplines: economics, agricultural economics, sociology, political science, anthropology, plant pathology, and environmental sciences. General points that emerged from our discussions were:

  • Use of experimental economic games (EEG) emerged first in economics (where specific incentives were offered), growing out of an earlier tradition of experiments in psychology. They then spread to political science, and much more recently to anthropology (and a lesser extent sociology). In the 1990s the field mushroomed, with EEG increasingly played in field contexts worldwide.
  • A key objective of EEG is to understand the core principles/fundamental aspects of human behavior that occur across multiple sites (and their generalizability); more specifically to measure preferences from which better causal inferences can be derived from empirical (non game) analyses.
  • In economics, EEG were initially used by Vernon Smith as a pedagogical experiment (to teach students how market equilibria are reached).
  • Traditional design elements (see the slide below) in experimental economics are: (1) no deception (i.e. players received full information); (2) incentive-compatibility (i.e., players earn rewards, typically money, based on their performance in the game); (3) no provision of context (in other words the concepts used are abstract). Field experiments often relax the third element, but the first two remain mostly in tact today. The potential role of deception is a topic of heated debate.
  • Faculty experienced with the field noted the multitude of games that have been used, and the challenge for students to find “new frontiers” in this field, leading to discussion on the generalizability of findings, the ecological validity of results, and the extent to which different “framings” might affect results. One student suggested (interestingly) that in comparisons of game play across contexts we should perhaps not be comparing game play and decisions, but rather the effects of different treatments. For example, while it is well know that games allowing for communication and/or punishment generally produce more cooperation (or prosociality) the size of the effect of these two treatments across sites/samples/populations might be more enlightening.
  • We discussed what a game is, making a distinction between games with nature, and games entailing strategic interactions. Games against nature may also have strategic elements if other organisms are capable of strategizing. In many cases where nature is conceived of as harsh environments  human actors may strategize among themselves even if nature is non-strategic.
  • Multigenerational games were identified as an important area of intersection between EEG and cultural evolution.


2. Intergenerational Behavioral Experiments

January 19, 2017

Week 2 focused on intergenerational behavioral experiments, with emphasis on the evolution of cooperation in public goods games.  Figure 1 summarizes the basic idea.  Intergenerational experiments start with a “progenitor” generation of experimental subjects, which could be drawn from student populations, Mechanical Turk, or in the field.  The progenitor group plays the behavioral experiment, and then in some way passes down their experience the next group, with research questions oriented around the how the experience of the past generation affects behavior in the current generations.  In Figure 1, this process of “cultural transmission” takes place via written advice being provided to the next generation.  There are many possible ways of operationalizing the idea of cultural transmission.  The overall goal of intergenerational experiments is to example what type of behaviors and norms become stabilized in the population over time, for example if passing down advice increases overall levels of cooperation.  Although running intergenerational behavioral experiments entails a number of logistical challenges, most researchers agree that some of the most important findings will emerge from lineages that run for many generations.

Why is this important?

While the vast majority of experiments focus on the behavior of a group of subjects at a single point in time, human behavior in real societies evolves over many generations.  Intergenerational behavioral economics experiments try to mimic these types of processes of the lab, with hypotheses being drawn heavily from theories of cultural evolution or evolutionary models of cooperation.  The emergence of norms of trust and reciprocity, what is often called “social capital”, also accumulates over time.  In a more applied context, many collaborative policy initiatives develop and implement plans over time, and must deal with turnover over in personnel. Hence, it is important to understand how any initial cooperation developed over time

What do we know?

The literature on international generational experiments is still relatively new and sparse, but is quickly growing.  A generic but imprecise statement is that nearly all of the existing experiments have found interesting and important effects from different mechanisms of cultural transmission, often with a capacity to support cooperative outcomes.  For example, Hillis and Lubell (2015) discovered that is possible to “breed” cooperation by selecting the most cooperative advice from the first generation to pass on to the second generation (Figure 2).  Even when the second generation cannot communicate directly among themselves during the experiment (dashed line labeled NC-G2 in Figure 2),  cooperation rates are as high or higher than experiments in which participants can communicate (dotted line labeled C-G1 in Figure 2).  The cooperative advice from the past generation is a catalyst for cooperative behavior especially in early rounds of the experiment.

Where do we go from here?

There are many possible research trajectories for intergenerational behavioral experiments.  One of the most important goals is to develop experimental platforms, most likely web-based, that allows experiments to unfold over many generations—even hundreds of generations.  UC Davis researchers have already started to pilot experiments that examine the evolution of institutions, for example the levels of punishment that are suggested for free-riders in public goods games. It will also be important to experiment with different mechanisms of cultural transmission, for example selecting advice from the previous generations with the highest payoffs or allowing experienced players to participate in future generations as an analogy to “parents” or “elders” that deliver advice to inexperienced players.  As with other types of behavioral experiments, it will be important to examine the generalizability of these findings to other types of subjects (e.g.; cross-cultural comparisons), games (e.g.; trust, ultimatum, and others; see references), and substantive framings (e.g.; a generic “investment” versus contribution to climate change mitigation). 


Chaudhuri, Ananish, Sara Graziano, and Pushkar Maitra. "Social learning and norms in a public goods experiment with inter-generational advice." The Review of Economic Studies 73, no. 2 (2006): 357-380.

Hillis, V., and M. Lubell. 2015. Breeding cooperation: cultural evolution in an intergenerational public goods experiment. Ecology and Society 20(2): 8.

Schotter, A., and B. Sopher. 2003. Social learning and coordination conventions in intergenerational games: an experimental study. Journal of Political Economy 111(3):498–529.

Schotter, A., and B. Sopher. 2006. Trust and trustworthiness in games: an experimental study of intergenerational advice. Experimental Economics 9(2):123–145.

Schotter, A., and B. Sopher. 2007. Advice and behavior in intergenerational ultimatum games: an experimental approach. Games and Economic Behavior 58(2):365–393.


3. Field Experiments

January 26, 2017

This week we took the lab games to the field, and our discussion covered a myriad of issues relating to moving from lab to natural settings, as denoted in the figure we presented for week 1.  We made special emphasis on work being done by economists who are increasingly harnessing Randomized Controlled Trials or other forms of exogenous variation to help to identify causal relationships between ‘treatments’ on individual or social preferences.

For example, Jakiela et al (2015) use a girls’ scholarship program by a Dutch NGO that randomly treated primary schools in western Kenya to identify the effects of educational achievement on cooperative tendencies on an experimental economic game played by the researchers several years later. Methodologically, the study blends a “natural field experiment” (i.e., RCT) with a “lab-in-field” experiment. More specifically, the analysis uses the scholarship program (RCT) as an instrumental variable to measure the effects of education, thereby allowing the researchers to rule out the possibility that traits associated with educational success (wealth, family background, etc.) are responsible for different decisions in the “lab-in-field” experiment.

The Jakiela et al (2015) study also introduced a new variation on a familiar game type: A modified (reverse) Dictator Game. In this game, the dictator is not tasked with sharing her own windfall gain with another person. Rather she is posed with the dilemma of how much to extract from another player who had earned her experimental payout via a simple effort task (i.e., she had “earned” her payout within the experiment). While education has no effect on overall generosity in the standard Dictator Game, it has a significant effect in the modified game. Based on their identification strategy, the authors conclude that exposure to and effort in formal education settings induces individuals to have greater respect for the earned rewards of others.

The session covered other “lab-in-field” experiments as well. In a field experiment conducted in Haiti we saw how subjects randomly selected to participate in a framed public goods game actually ended up behaving more cooperatively in the real world tasks of clearing their irrigation canals than individuals who had not played the games, even after a 2-3 month interval. This result, which is part of a dissertation filed last year by an ARE PhD student, led to a discussion of the mechanisms whereby playing in an experiment might affect subsequent behavior.

In our discussion of Gine et al (2010), we saw the power of setting up a long-term experimental lab in a market place in Lima, Peru where the experimenters rented a market stall as their lab space for nearly a year. This presence enabled the authors to generate high statistical power using within individual experimental design whereby thousands of experimental results were obtained over a period of many months. This enables them to include individual fixed effects in their final regressions to control for time invariant unobservable effects at the individual level. The results captured the structure of microfinance banking, and in particular the strength of dynamic incentives that penalize poor performance as a borrower by cutting off future credit lines.

Along the way, we discussed several specific design features that frequently shape how experimentalists structure their experiments. For example, the use of the strategy method in eliciting individual responses to a full range of potential responses by others (in interactive games) or potential parameter values (in individual preference games) is a common way to create richer data and thereby greater statistical power. We also discussed the challenges of exposing subjects in an experiment to “real losses,” which is necessary to elicit preferences like loss aversion. For obvious reasons IRB is unlikely to approve research where subjects potentially lose their own money. The standard workaround is to give players money at the start of the experiment and impress on them that it is theirs to keep. By making them feel ownership over this money (so-called, “asset integration”) researchers hope they shift from “playing with house money” to treating each dollar earned or lost as their own.

Jakiela, P., Miguel, E. and V. te Velde (2014). You’ve earned it: estimating the impact of human capital on social preferences. Experimental Economics 18:385-407.

Giné, X., Jakiela, P. Karlan. D. and J. Morduch. (2010) Microfinance Games. American Economic Journal 2:60-95


4. Guest Lecturer: Alessandra Cassar 

February 2, 2017

Alessandra Cassar (professor of economics at the University of San Francisco) presented and discussed results from two sets of experiments. The first part of her talk focused on using a field experiment to test the effectiveness of the Family Independence Initiative (FII) model for poverty alleviation (Aguinaga et al. 2016). The FII program, developed by Morris Lee Miller, begins with the assumption that poor people know what they should be doing, but need help doing it. The help is provided in the form of three low-cost behavioral nudges: setting household/business financial goals, providing modest financial incentives, and forming support groups. Given the success and cost-effectiveness of the program in Oakland, CA, Cassar and colleagues wanted to determine its effectiveness in developing countries, and identify which of the three interventions – goals, groups, or financial incentives – had the biggest impact. The impact of financial incentives was particularly salient, considering the mixed results observed in previous research (e.g. crowding out effects, internal vs. external motivation, etc.). In collaboration with the Banco de las Oportunidades in Medellín, Colombia, the researchers formed 6 different groups (4 treatment, 2 control) of small business owners to test the effectiveness of each component in isolation and in combination. The business owners were all from the lowest 3 socioeconomic strata (out of 6) in Colombia. 

The results indicate that the act of simply setting goals improves household/business financial outcomes, but that the combination of all three components of the FII intervention had the greatest positive impact. The researchers have applied the same treatments to 9th grade students in Medellín, focusing on academic rather than financial goals, and though the results are preliminary, they found the same general pattern on the impact of academic performance. In the discussion after the presentation, we wondered about the long-term impact of the intervention and its positive outcomes. Did any of the households who participated in the study change income strata, which would indicate upward socioeconomic mobility?

The second presentation explored gender differences in competitiveness displayed during game play (Cassar et al. 2016). Experimental economists tend to think that part of the explanation for the wage gap is that women are less competitively inclined than men, a hypothesis supported by gender differences in experimental game play. Citing research about various experimental conditions, biological factors (i.e. hormone levels), and institutional and cultural contexts that either eliminate or reverse this gendered pattern in competitive tendencies, Cassar and colleagues suggest that experimental games might inadvertently be tailored to reveal male competitiveness, and mask female competitiveness. Perhaps women are competitive, just not in the same spheres as men. They ask if women will be more competitive if the payoffs are oriented toward child well-being, a question informed by both sociobiological and biosocial theories.

Parents in China each played four games in which they completed tasks (mathematical summations) and were paid based on either fixed rate (non-competitive) or tournament (competitive) schemes, and in the form of either cash or scholastic vouchers (ostensibly for purchasing books for their children). The four games were: fixed rate/cash, tournament/cash, choice of fixed rate or tournament/cash, choice of fixed rate or tournament/voucher. The researchers controlled for the probability of winning the tournament, risk tolerance, willingness to pay for the vouchers, and confidence. Women performed the task better than men across all treatments, and there were no statistically significant differences in preferences or beliefs. When cash was the form of payment, men preferentially chose the competitive tournament scheme; but when the payment was vouchers, men and women were equally likely to choose the tournament payment scheme. The results of this study have huge implications, because paying cash in experimental economics is the gold standard. The assumption is that cash is neutral, but this may not be the case. Unfortunately, these results cannot disentangle whether women are responding to benefits for their children, or payments that are not cash-based. Dr. Lybbert raised another potential issue if using monetary incentives affects the results we typically see for women in game play, are monetary-based measures of risk aversion (recall, the researchers controlled for risk aversion in this study using methods) subject to similar methodological issues?


Aguinaga, P., Cassar, A., Graham, J., Skora, L., & Wydick, B. (2016). Raising Achievement Among Microentrepreneurs: An Experimental Test of Goals, Incentives, and Support Groups in Medellin, Colombia.

Cassar, A., Wordofa, F., & Zhang, Y. J. (2016). Competing for the benefit of offspring eliminates the gender gap in competitiveness. Proceedings of the National Academy of Sciences, 113(19), 5201–5205.


5. Potential for Using Experimental Economic Games in Disease Management

February 9, 2017

Taking measures against a disease can sometimes create a conflict between self interest and group interest, which can give rise to a social dilemma. In recent years, there has been increasing interest in coupling game theory with epidemiology and applying it to public health and in an attempt to predict human behavior in situations where the payoff to an individual’s strategy depends on the strategy chosen by others in the population. When the ‘individual equilibrium’ associated with minimizing risk of getting a disease does not correspond to the ‘group optimum’ that minimizes the impact of the disease on the population, there might be an incentive for non-cooperation (not taking measures) due to the public good created by herd immunity (near-elimination of the risk of being infected when sufficient people take measures). 

In the context of vaccination policy, it has long been observed that the levels of voluntary vaccination uptake are always lower than the group optimum that would correspond to herd immunity, but imposing mandatory vaccination to group-optimal levels is viewed as a threat to civil rights. Therefore, several groups have tried to analyze the risks and incentives that people evaluate before making the decision to engage in vaccination, in an attempt to predict behavior and localize targets for public health policies. As an example, we discussed an online experiment (Chen et al. 2013) that uses a virtual epidemics game to study how people change their behavior in response to an epidemic. Although the experiment was criticized for lacking a solid epidemiological and game-theoretical underlying model that would clarify the existence (or not) of a cooperation problem, some of the results stimulated a fruitful discussion. The study showed that some players were willing to take the risk of not engaging in self-protection and free-riding on the protection measures taken by others; and willingness to engage in self-protection decreased when the cost was higher. Also, reported prevalence, previous experiences in the epidemic and individual risk preferences seemed to play an important role in the decision to engage in self-protection.

After this example from public health, we considered the similarities with plant disease management, where there might also be a conflict (and eventually a social dilemma) between individual decisions at the farm/field level and institutional decisions at the regional, state or national level. This dichotomy might be very relevant for some plant diseases such as citrus greening or huanglongbing (HLB), for which coordination between management at the individual and the regional level is crucial to achieve acceptable control. For the rest of the session, we discussed the particularities of the HLB pathosystem, the challenges and dilemmas associated with its management, and potential applications of the Experimental Economic Games approach. 


Bauch CT, Earn DJD. 2004. Vaccination and the theory of games. Proceedings of the National Academy of Sciences of the United States of America 101:13391-4

Bauch CT, Galvani AP, Earn DJD. 2003. Group interest versus self-interest in smallpox vaccination policy. Proceedings of the National Academy of Sciences 100:10564-7

Chen F, Griffith A, Cottrell A, Wong Y-L. 2013. Behavioral Responses to Epidemics in an Online Experiment: Using Virtual Diseases to Study Human Behavior. PLoS ONE 8:e52814

Gottwald TR, B A. 2010. Current Epidemiological Understanding of Citrus Huanglongbing. Annual Review of Phytopathology 48:119–39 


6. Behavioral Foundations of Microcredit

February 16, 2017

This week we had a vibrant discussion that centered on a paper of Michal Bauer, Jonathan Morduch and Julie Chytilova (2012) on the Behavioral Foundations of Microcredit. This work presents a further view of microcredit, as a means by which individuals exercise self-discipline in financial behavior. When individuals with present-biased preferences (Laibson, 1997) look to the future and decide to reserve some money for savings, they might find it difficult to implement their saving plan when the time comes and give into the temptation of using that money for consumption. The availability of microcredit may serve as a mechanism by which individuals agree to a public commitment to save and to feel social pressure to meet their goal.

In contrast to the typical treatment/control analysis, this work presents a series of lab experiments in the field performed to elicit discounting and risk aversion measures in rural South India and analyze whether the present-biased individuals exhibited different savings behaviors than consistent-preference individuals.

Bauer and coauthors elicit time and risk and preferences of 537 individuals but they focus their result in the subsample of the 266 women. They found that among women who borrow, those with present-biased preferences are particularly likely to be microcredit borrowers and that women with present-biased preferences are more likely than others to borrow through a microcredit organization. Why the results only hold for women with present-biased is usual in the literature (Ashraf, Karlan, and Yin, 2006), suggesting a possible spousal control motive. Bauer et al., however, show that the results are robust to controlling for a measure of women’s decision-making power within the household.

Two discount rates were elicited, a current one and a future one. For the current one, individuals were asked to choose between receiving a smaller amount earlier in time or larger amounts with three months delay. "Do you prefer Rs. 250 tomorrow or Rs. 265 three months later?, further increasing the future amount (Rs. 265, 280, 300, 330, and 375) and recording the switch point to provide an interval of the discount rate. Similarly, respondents answer to binary choices at a future time frame (one year vs. one year and three months) to provide a future discount rate. If preferences are consistent, then current and future discount rates are the same. Higher current than future discount rate are categorized as present-biased individuals (and if the reversal is true, then “Patient now, impatient later”).

We noted some points on the way the elicitation took place. First, choices were over relatively large, real stakes, as large as a week’s wage. This would incentivize participants to reveal their true preferences. However, not all the participants would receive the money. They knew that through a lottery, only 20% would be receiving the prize. There was a discussion on whether this was enough to incentivize individuals to tell the truth. 

Second, we noted that the question referred to receiving the prize tomorrow, rather than today, to lessen the extent in which answers would reflect lack of trust on the numerators and transaction costs.

Third, asking for 5 binary choices gives 6 possible bins for the discount rate; asking for more would have increased the precision of the discount rate, possibly increasing the number of individuals in the consistent category to be labeled either as weakly present biased or patient now but impatient in the future category.

To elicit risk aversion measures each participant was asked to select one out of 6 different gambles. Every gamble yielded either a high or a low pay off with a probability 0.5. In each subsequent gamble, the expected value increased jointly with the variance. These measures helped to control for attitudes towards risks, however these play a very small or null part in helping explaining the discounting factors or the savings behaviors. Intuitively, we would have expected to see some relationship between attitudes towards risks and time discounting, however when considering that we want to isolate choices between uncertain outcomes from trade-offs between certain outcomes at different points in time, we reckon there does not need to be a reason of why these primitive parameters are correlated. When analyzing time-preferences, it is imperative that individuals perceive the future outcomes as certain, when they perceive the outcomes as almost uncertain the valuations given may plunge.

To estimate how having present-biased preference affects financial behaviors, the authors rely on a linear regression of borrowing and saving on the discount factor (either the current or the future one), indicators for being present-biased or future biased, risk aversion controlled by the gambles selection, demographic and villages controls. To interpret the results is crucial to understand what does controlling for the discount rate implies for significance and direction of the individual-type indicators, depending on the mechanisms that individuals may have at hand. For instance, when conditioning on the current discount rate, (strongly) present biased women are more likely to participate in a SHG, whereas controlling for the future discount rate, the significance disappears, suggesting that sophisticated individuals adjust their participation when they acknowledge their discount rate.

Normative questions surrounding present-biased individuals, the interrelation between risk aversion and time discounting, and why do results only hold for wormen were left on the table for further discussion. Without doubt this week triggered many thoughts and inquiries on heterogeneity and deep parameters. 


Nava Ashraf, Dean Karlan, Wesley Yin. 2006. “Tying Odysseus to the Mast: Evidence From a Commitment Savings Product in the Philippines.” Q J Econ, 121 (2): 635-6

Bauer, Michal, Julie Chytilová and Jonathan Morduch. 2012. "Behavioral Foundations of Microcredit: Experimental and Survey Evidence from Rural India." American Economic Review, 102(2): 1118-39.

David Laibson. 1997. “Golden Eggs and Hyperbolic Discounting.” Q J Econ, 112 (2): 443-478.


7. The Interaction of Legal and Social Norm Enforcement

February 23, 2017

This week, we looked at a paper by Sebastian Kube and Christian Traxler on how social sanctioning in a public-goods game changes in response to the introduction of a formal legal sanctioning mechanism.  They draw on two distinct literatures, one looking at social sanctioning as a second-order public good, and one looking at the incentives created by centralized sanctions, but examine the interaction between these two.  In particular, by introducing a “strategy” empirical elicitation method, the authors are able to disentangle the direct crowd-out effects of the law on social sanctions, holding contributions constant, as well as see how contributions to the public good (and thus the need to sanction) change with the dual sanctions.  Despite a worry that introducing a relatively weak law could crowd out social  sanctions without creating sufficiently strong incentives for contribution to replace them, this careful study actually finds welfare gains.  Not only do individuals contribute more to the public good, but they pay less to punish each other, giving overall higher payoffs. 

The experiment consisted of 3 rounds: a public goods contribution (with the conventional public goods dilemma, where the Nash equilibrium is for all to free-ride, but the social optimum is for all to contribute their full endowment), a round of social sanctioning (where players can choose to punish others, at a cost to themselves), and a round of legal sanctions in one treatment arm (where with 1/8 probability, players are monitored and punished for deviations from full contribution).  In order to separately identify the effects of legal sanctions on each previous round, the authors adapt Selten’s strategy method of elicitation.  In the sanctioning round, instead of only seeing the realized contributions of their partners, players are presented with a menu of 11 ‘triples’ of potential contributions (one real and 10 hypothetical), and asked to choose sanctions for each of the three players in each triplet.  By keeping the hypothetical cases the same in both treatments, they are able to measure how a player will sanction for the same level of others’ contributions with and without the law. 

The empirical results in the paper, especially the random-effects Tobit regression, do show a drop in sanctions when the law is introduced; this only becomes statistically significant when they control for the difference in contributions between the two players, which they argue is evidence that players are motivated by inequality aversion.  They also find that the contributions of the “peripheral” players matter for sanctioning decisions; we discussed that this could be due either to framing effects or expectations about their sanctioning behavior.  They also find a significant increase in first-round contributions, and these two effects taken together increase total payoffs by around 1/3.  Thus there can be relatively large welfare gains from introducing even mild legal sanctions.  The authors explore several possible motivations that could drive the crowd-out or the crowd-in of social sanctions, but we discussed how the setup of their existing experiment does not distinguish between them.  Slight modifications, such as changing the legally prescribed level of contributions, could test the extent to which the law is acting as a focal point. 

In our discussion, we brought this framework towards reality, discussing alternate forms of sanctioning (such as ostracism), as well as what this sanctioning might look like in the real world.  Vigilantism in particular was explored, as it represents a case where socially we might desire to ‘crowd out’ social sanctions.  Within Kube and Traxler’s framework, this is one case where sanctioning is cheaper for the state than decentralized individuals.  It was mentioned that this could more generally apply when sanctioning is for low-probability, high-cost deviations from the social optimum.  We spent some time exploring possible dimensions of cost and heterogeneity, to predict where we would expect legal and social sanctions to be complements or substitutes.  The consensus was that this was an under-explored area of the literature, with several obvious immediate extensions to Kube and Traxler’s game.


Kube, S., & Traxler, C. (2011). The Interaction of Legal and Social Norm Enforcement.  Journal of Public Economic Theory, 13(5), 639-660.


8. From the Lab to the Real World

March 2, 2017

To what extent does behavior in the lab reflect behavior in the “real world”? So far, studies that have explicitly tried to determine external validity have yielded mixed results (see table below). In Week 8 of the proseminar, we discussed a recent paper (2016) by Torres-Guevara and Schlüter that explores the relationship between the ecological impact of Colombian artisanal fishermen and their behavior in the lab. The authors hypothesized that fishermen with higher public goods contributions and greater patience would have lower ecological impacts.

Using government data about fishing sites and gear used by fishermen, and perceptions about the ecological impact of different fishing sites and gear types, the authors scored the fishing impact of each fishermen. Fishermen with sufficient representation in the government catch data were invited to complete a survey and play three unframed experimental games: (1) a public goods game, (2) a standard time preference experiment framed as a future bonus, and (3) a non-standard time preference experiment based on the “marshmallow test” but using Coca-Cola instead. The authors found no relationship between game play and fishing impact. Neither public goods contributions nor time preferences predicted fishing impact, even after adding controls for education, fishing experience, socioeconomic factors, and conservation attitudes. The authors suggest that the mixed results from tests of external validity arise from differences in experimental design (testing groups of resource users vs. individual resource users), and differences in the chosen indicator(s) or real-life cooperation.

In our discussion, the class uncovered interesting conceptual and methodological issues with the paper that will help guide further investigations into external validity. Fishing entails dilemmas associated with common-pool resources, so perhaps a framed common-pool resource game would have been more appropriate; indeed, the fishermen’s responses to survey questions showed no awareness that fishing was a potential tragedy of the commons, nor (accordingly) that it represents a cooperative dilemma. We also discussed how many other factors aside from cooperative tendencies can influence ecological impact; and with so little information from the authors about the policy context of fishing activities at their site, it is difficult to gauge the accuracy and reliability of government catch data. A key challenge when interpreting the literature on external validity is defining exactly what is meant by the term – what precisely is expected to be constant between experiment and real world? A more direct test would be to determine whether there is a direct association between, for example, increasing communication in the game, and increasing communication in the real world, in other words to compare effect sizes in the experiment versus some manipulation of natural variability in the field. Finally, we found ourselves wondering whether the study was addressing concerns about external validity, internal validity, or measurement validity.

The remainder of class discussion centered on experimental game data from Nicole’s dissertation research in a Mexican fishing community. Nicole played the fishing game (Cardenas et al. 2013) with fishermen who had already completed an in-depth survey about household demographics, ecological knowledge, fishing practices, conservation values, and cooperative behavior. The goal of the fishing game is to incorporate resource dynamics into experimental games, which tend to focus more on behavioral and institutional concerns. In the fishing game, groups of 5 players must choose where to fish (location A or B), and how much to fish (0, 1, 2) each round without communicating with each other. If the fishing effort in any location exceeds 5, it shifts from high productivity to low productivity, thereby decreasing the returns to fishing. Recovery from low to high productivity requires all harvesters to choose 0 or 1 unit of fishing effort for 2 consecutive rounds. If all players behave opportunistically, both locations move to a state of low productivity within 2 rounds and get stuck there. During the next phase of the fishing game, players anonymously vote on a fishing rule, choosing between a lottery, rotation, or quota rule. A dice is rolled after each round, and if it lands on 6 players who broke the rule forfeit their payoffs for that round. Nicole added a new phase of the game, where each player had the option of contributing his earnings to increase the probability of rule enforcement.

As a framed common-pool resource game, the fishing game mimics the real world cooperative dilemma faced by fisherman. And the kinds of “real world” behaviors captured in the survey (fishing effort, rule violation, contribution to monitoring/enforcement, etc.) are analogous to behaviors observed and measured in the lab setting. Thus, Nicole’s data will allow her to overcome some of the limitations of previous studies of external validity. 


Torres-Guevara, L. E., & Schluter, A. (2016). External validity of artefactual field experiments: A study on cooperation, impatience and sustainability in an artisanal fishery in Colombia. Ecological Economics, 128, 187–201.

Cardenas, J.-C., Janssen, M., & Bousquet, F. (2013). Dynamics of Rules and Resources: Three New Field Experiments on Water, Forests and Fisheries. In Handbook on experimental economics and the environment (pp. 319–345). Edward Elgar Publishers.


9. Experimental Evidence on Information, Identity and Voting

March 9, 2017

This week, we focused on experiments dealing with the mechanisms of voting behavior, especially on the role of information and group identity, and implication of those result for the competence of voters under democracy. 

We discussed two papers. Bassi, Morton and Williams (2011) examine the role of group identity in voting. In their voting experiment, five subjects (voters) vote for two alternatives. Through somewhat complicated experimental procedures designed to examine the effects of group identity on vote choice (allocated through random color assignments to subjects and alternatives), the authors showed that identity plays a significant role in influencing voting decision. This is particularly the case under low information and none/low incentive conditions, while this effect disappears under high information or high incentive conditions. The result implies that, under certain conditions, even the very minimal level of identity – randomly assigned color – can influence the vote choice. This poses questions to the competence of citizens under democracy. By prioritizing their identity, inadequately informed or motivated voters may fail to maximize their wellbeing through voting. 

The second paper, Battaglini, Morton and Palfrey (2008), tests the implications from what is called Swing Voter’s Curse (SVC) model (Feddersen and Pesendorfer 1996). The SVC model shows how uninformed voters might act rationally in elections. Through a stylized game experiment that closely traces the logic of the SVC model, the authors find that uninformed voters do behave in the ways that is highly consistent with the prediction from SVC model. This result, in contrast to Bassi, Morton and Williams (2011), implies that voters can act logically to maximize their wellbeing even with low level of information. 

These conflicting findings inevitably raise questions. In the class discussion, we focused largely on the external validity of the findings. First, we suspected that utilities of voters may change across different political cultures. For the first paper, assigned utilities assumed that voters always get higher payoffs from voting for winning candidate. It is suggested that this utility makes more sense under clientelistic political culture than under individualistic political culture. In the clientelistic political culture, such as in Latin America, it is important for voters to vote for the winning candidate, since winners typically distribute benefits only to their supporters. However, under individualistic political culture, the policy of the winner should equally benefit all voters, not only supporters. 

Second, we suggested that the size of electorate may matter. In both papers, the size of electorate is quite small (five to twenty). The results from these voting experiments may be more relevant to voting in the small group (such as in committees), but it is not clear if they are really pertinent to elections with a much larger electorate.

Lastly, we briefly discussed the implications of “no deception” policy in behavioral economics experiments, which prohibit researchers to deceive their subjects in any way; (this is practice in economics but not psychology). Unnecessary deceptions should be avoided, but as the experimental designs more closely mimic reality, it is more difficult to avoid some level of deception. Therefore, the “no deception” rule adhered to by economists might limit the extent to which the results from behavioral economics experiments can be applied to the real world. 


Bassi, Anna, Rebecca B. Morton and Kenneth C. Williams. 2011. “The Effects of Identities, Incentives, and Information on Voting." The Journal of Politics 73(2): pp.558-571.

Battaglini, Marco, Rebecca B. Morton and Thomas R. Palfrey. 2008. “Information Aggregation and Strategic Abstention in Large Laboratory Elections.” The American Economic Review 98(2): pp.194-200.

Feddersen, Timothy J. and Wolfgang Pesendorfer. 1996. “The Swing Voter's Curse.” The American Economic Review 86(3): pp.408-424.


10. Economic Games and Forest Conservation

March 16, 2017

The ultimate goal of experimental economics is to use games that provide real incentives and create real dilemmas in the lab to provide insight into the problems we face outside in the real world. After having reviewed tests of external validity last week, we turned our attention to an experiment that tested to see if they could not capture, in game form, the impact that positive leadership has on forest conservation efforts (Kosfeld and Rustagi 2015).

In classic economic theory, it is assumed that leaders punish and enforce the public good because they are paid to do so, and because they derive reputational benefits from punishing. However, it is likely that leaders punish for other reasons, internal reasons. For example, they may have emotional commitments to groups or hold norms about proper kinds of behavior. In other words, leaders may punish because they have other-regarding preferences.

Kosfeld and Rustagi (2015) explore three internal motivations for why leaders may punish: equity, efficiency, and spite and envy. Specifically, a leader concerned with equality will punish people when there are inequitable outcomes and they generally target the person who does better. Second, leaders concerned with efficiency may punish a people when they fail to reach the Pareto optimal, or most efficient outcome. Finally, a leader who is envious and spiteful may punish people even when they are perfectly egalitarian and efficient – these leaders are very bad people, sick guys (oops!)

Kosfeld and Rustagi took advantage of a unique situation in Ethiopia to test the effect of leadership on cooperation by studying a forest conservation project that began in 2000. Historically this region of Ethiopia had very high levels of deforestation and the government implemented a project whereby the villagers formed community based forest management groups of about 30 people and every five years the government would come and inspect the forest; depending on how effective the communities were, the less money they had to pay in taxes to the government for using their land.

To test leaders’ internal motivations they ran a third-party punishment game. This game has two stages to it. First, two members of the community play an anonymous “one shot two players” public goods game. Second, the leader of the same community is able to pay some of his own money to punish the players in the game. The crux of the game is that the leader has absolutely no incentive to punish the other two people, in fact, he does worse off by doing so. The game uses what is called a strategy elicitation method where, before the leader sees how the two others played the PPG, he indicates how much he would punish under each of the possible contributions that the players could make. Using these response patterns the authors were able to construct four behavioral profiles, egalitarians, efficient egalitarians, anti-socials, and non-punishers.

What they found was that leaders who were concerned with both equality and efficiency were associated with forests that were much healthier and had many more trees per-hectare than non-punishers. The egalitarians were not significantly different from the non-punishers, and the anti-socials had significantly less trees than the non-punishers. After coming back years later they found evidence that the egalitarian leaders’ forests were doing much better than the non-punishers.

Our discussion of the article was great, just absolutely amazing, really had the best people thinking about it (oops again!). Unaddressed in the article is the lack of data on the initial state of the forests and bias this might introduce. Second, we were worried that the creation of the categories classifying leaders may have been done ex post. Finally, the students had a chance to explicate on why they thought that the article had been admitted to the prestigious AER, and as per economists, the students correctly mentioned that the article deftly dealt with concerns of endogeneity.

At the end of the class we briefly discussed my own research. After introducing the theoretical foundation of cultural multilevel selection, the challenges to forest conservation through Reduced Emissions from Deforestation and Degradation (REDD+) initiative, and the study site in Pemba, we were almost out of time. I nevertheless got some good suggestions about how to study the uncertainty inherent in REDD+, and the value of public goods to Pembans.


Kosfeld, M., and D. Rustagi. 2015. Leader punishment and cooperation in groups: Experimental field evidence from commons management in Ethiopia. The American Economic Review 105:747-783.