How do experiments inform collective action research?

Maroš Servátka MGSM Experimental Economics Laboratory Macquarie Graduate School of Management 99 Talavera Rd., North Ryde NSW 2113, Australia [email protected]

(Prepared for the volume on Research Priorities in New Institutional Economics)

February 22, 2018

Experimental economics and institutional economics One might wonder why experimental and institutional economists get along so well. If you have ever attempted to design an economics experiment involving personal or impersonal exchange, the answer is obvious – you need an institution that determines how interactions take place (Smith, 1982). Experiments clearly demonstrate that institutions matter. The rules determine the incentives that economic agents face and the information they have at the time of making their decisions. “Institutions matter, because incentives and information matter,” (Smith, 1994). Economists examine naturally occurring phenomena using happenstance data and construct theories to explain them. Scientific progress requires that we test the validity of proposed theories and evaluate their predictive power. The complexity of economic systems means that many variables change simultaneously and affect behavior of economic agents in an uncontrolled manner, making it hard to attribute causality, leaving us with multiple plausible explanations. Although sophisticated econometric techniques have improved our ability to analyze happenstance data, they often rely on particular statistical assumptions. Experiments, on the other hand, enable us to tackle causality by introducing a truly exogenous ceteris paribus change to the policy variable and observe and measure the subsequent change in behavior. The practical aspect of economics could be compared to engineering science. Economists are not only interested in understanding how economic systems function, but often make efforts to improve efficiency of government interventions or business solutions. The experimental methodology allows a researcher interested in institutional economics to directly compare institutions, analyze their welfare properties, and asses the causal impact on observed

behavior by creating counterfactuals, all while keeping the underlying environment (i.e., the number of economic agents and their characteristics, such as endowment, utility/technology or knowledge) constant. In a similar fashion as engineers test-bed prototypes in wind tunnels, experiments are a fantastic tool for test-bedding new economic institutions designed to alleviate inefficiencies such as free-riding (see surveys by Ledyard, 1995; Chaudhuri, 2011) or over-extraction of resources. Furthermore, they allow the researcher to identify vital components of social and economic exchange (e.g., trust and trustworthiness or norms) that promote cooperation, or study factors/mechanisms that influence the strength and prevalence of these individual components. In what follows I discuss three areas of experimental research in the context of collective action and illustrate how experiments have informed and continue informing us about human behavior in a way that happenstance field data do not. As experiments are a research methodology rather than a subfield of economics, it is challenging to define priorities for experimental research in studying institutions, much like it would be for econometrics; their suitability depends on the research question and interests of the researcher. From the perspective of institutional analysis, I argue that experiments have a powerful empirical bite in the three discussed areas. To inspire future research, within each area I selectively provide examples of studies that have made an important contribution to our understanding of how institutions impact cooperation and that demonstrate the strengths of experimental methodology. Finally, I also add a few examples from my own work.

Test-bedding of new institutions Voluntary contributions to a public good or extraction of a common pool resource are notorious examples of collective actions where cooperation can yield a significant improvement in social welfare but where following individual incentives results in inefficient outcomes. Economists have proposed a plethora of theoretical solutions to curbing opportunism and increasing cooperation, many of which have never been implemented in practice. While there could be multiple reasons why the take up rate is relatively low (e.g., the implementation might involve significant costs), one of the main problems is that ex ante it is not obvious whether these proposed institutions will shape behavior in a way predicted by theory. (If it has never been done, how do you know what works?) Practitioners are thus often reluctant to adopt new solutions because the risk might be too high. This is where experiments can (and do) help. It is my opinion that for a researcher to make a significant contribution in this area, one not only needs to propose a new institution or a new mechanism, but also demonstrate that the observed behavior indeed corresponds to the theoretical predictions. Without such testbedding we would not know whether and how individuals respond to incentives embedded in these institutions, understand their allocative and distributive properties, recognize their behavioral limitations, and would not be able to fine-tune relevant parameters to improve performance. While already quite prominent, with the current trend of economics becoming a

truly behavioral science and economists being increasingly more involved in proposing (and testing!) practical solutions in business and policy, I anticipate that this area of research will gain even more ground in the upcoming years. Let me illustrate test-bedding using a recent paper by Cason and Zubrickas (2017) that introduces a non-coercive mechanism for raising funds for threshold public projects, provided only if voluntary contributions are at least as large as the costs (i.e., reach the threshold). The mechanism is based on refund bonuses, meaning that each contributor has his contribution refunded and also receives a bonus proportional to his contribution if the threshold is not met. In equilibrium, the public good is always provided as long as the sum of refund bonuses does not exceed the net value of the public good. Cason and Zubrickas subject their mechanism to an experimental test to empirically verify whether refund bonuses induce higher contributions. The results support the equilibrium predictions and demonstrate that the participants respond to the incentives embedded in the mechanism. This includes a crucial prediction that refund bonuses that are too generous will lead contributors to converge on a total contribution marginally below the threshold, highlighting an important limitation of the mechanism that has considerable practical potential in crowdfunding. If you are considering employing the mechanism, you would want to know these results in advance.

From anecdotal evidence to causal inference Some solutions to the collective action problem emerge endogenously, without relying on external authorities for implementation and execution. Two canonical examples of such solutions are covenants with and without sanctions (Ostrom et al, 1992) and decentralized punishment (Fehr and Gächter, 2002). For example, in Ostrom et al. (1992), the difficulty lies in assessing whether villagers do not deplete the fish stock because of an agreement among those fisherman that have access to the nearby lake (even if we know that such agreement exists!) or because of other plausible explanations, for example, repeated interaction, individual reputation, or social norms that may regulate their behavior in addition to the agreement. It is also possible that such solutions emerge only in societies that are more cooperative (suggesting that selection might play a role) and thus the fishermen could cooperate even without the agreement. The lack of counterfactuals therefore makes it hard to determine whether agreements or willingness to altruistically punish free-riders drive the observed behavior. Fehr and Gachter designed a public goods experiment to test whether individuals are willing to punish free-riders and how much free riding occurs with and without punishment. Since punishment is costly, theoretically a rational selfish individual will never punish in a one shot game and free-riding is the dominant strategy with or without the opportunity to punish. In contrast, Fehr and Gächter observed that contributions to public goods were much higher with punishment than without, providing direct evidence that punishment reduces free-riding. Removing such “appear-to-be-effective” solutions from their natural environment and placing them in a controlled laboratory setting, all while preserving the essential elements of

the economic problem, allows for the researcher to draw causal conclusions regarding their impact on behavior. Studying other existing solutions that have proven successful in mitigating the collective action problem in everyday life (e.g., specific sanctions and rewards), learning about their properties, applicability, and limitations is a fruitful avenue for future institutional research. Discriminating among multiple explanations When one observes cooperation in the field, there often exist multiple behavioral reasons why people could be cooperative: for example, they are conditional cooperators, have unconditional other-regarding preferences (e.g. altruism or inequality aversion) or are reciprocal, guilt-averse, follow a particular norm and choose an action that is socially acceptable, are trusting and/or want to maintain their image of being trustworthy. Solely observing that a particular institution increases cooperation therefore does not necessarily tell us about the underlying motivation or the transmission mechanism. Distinguishing between such explanations, however, is crucial for a deeper understanding of cooperation and is necessary for correctly formulating policy implications. From the perspective of future research in this area it is important to note that the above list is by no means exhaustive as there is an abundance of unexplored reasons why people cooperate, for example linked to psychological incentives or sociological motives. How can one distinguish whether, say, cooperation observed when two players move in sequence is due to trust of the first mover and/or unconditional other-regarding preferences and whether the second mover is reciprocating and/or acting out of, say, altruism? To provide an answer, Cox (2004) designs a triadic experiment employing the investment game and two diagnostic treatments manipulating the motivation that the first and second movers have for investing and returning money, respectively. Cox finds that the first movers exhibit both trust and other-regarding preferences whereas the second movers’ decisions are driven by reciprocity and also other-regarding preferences. His experimental results therefore indicate that attributing cooperation solely to trust and reciprocity would be an incorrect conclusion as both can be inflated by people caring for others. Being able to pinpoint the main factors that drive cooperative behavior is a comparative advantage of experiments and allows for theory to evolve by constructing models that can maintain consistency with the empirical evidence. These models can subsequently be applied to new scenarios and tested further. In a similar fashion, Cox et al. (2016) separate out individual components of trustworthiness: other-regarding preferences, responsiveness to vulnerability, deal-responsiveness (reacting to actions that allow for a mutual improvement, i.e. deals), and gift-responsiveness (which is different from reciprocity in that it relies on reacting to actions that allow the second mover to obtain an improvement by adopting actions that benefit the first mover). They find that otherregarding preferences and responsiveness to vulnerability are the main drivers of trustworthy behavior. Such identification has theoretical implications for modelling trust and trustworthiness as it makes it possible to define both terms on observables and without the presence of other-regarding preferences. This is important for studying trust and trustworthiness in the field where in many economic and social situations no other-regarding

preferences are involved in trust acts. Defining trust and trustworthiness on observables, such as the maximum feasible payoff available to the player, makes it easier for researchers to formulate testable hypotheses regarding both types of behavior and subjecting them to empirical tests. Trust is ubiquitous to all economic interactions involving transaction costs (Arrow, 1974) and without its presence many welfare increasing transactions would not take place. Experimental research in this area has therefore studied both formal mechanisms that rely on enforcement or intervention from an external party, such as escrow (Bracht and Feltovich, 2008) or satisfaction guaranteed (Andreoni, 2005), and informal mechanisms, such as promises (Charness and Dufwenberg, 2006), informal agreements (Dufwenberg et al, 2017), or group identity (Morita and Servátka, 2013) and compared their relative effectiveness by means of a unified experimental design (Servátka et al, 2011). Trust and cooperation can also be sustained in the presence of reciprocal preferences of economic agents (see Fehr and Gächter, 2000; Camerer, 2003; Sobel, 2005; Fehr and Schmidt, 2006; Chaudhuri, 2008 for surveys). Does reciprocity govern certain types of economic behavior? How sensitive is it to intentions that are usually unobservable in the field? For example, it might be hard to ascertain whether I am kind towards you because of the goodness of my heart or because of strategic reasons. Similarly, I could hurt you accidentally or because it benefits me in some way. Laboratory experiments unambiguously identify intentions to be an important driving factor of reciprocal behavior (e.g., Blount, 1995; Falk et al, 2008,). Acts of commission, which actively impose kindness or harm, are found to reveal intent to a greater degree and therefore lead to a stronger reciprocal response than acts of omission, which represent failures to act kindly or to prevent harm (Cox et al, 2009; Cox et al, 2017). Experiments not only enable us to measure the impact of unobservables by, say, allowing for the presence of intentions in one treatment and removing them in the other (e.g., Charness, 2004), they also make it possible to manipulate the nature and intensity of such unobservables to learn that people do not discriminate between kind selfless and kind but self-serving motivations (Woods and Servátka, 2018). Consider another example in which it is hard to identify what drives the observed behavior in the field. You arrive to Bratislava, Slovakia to attend the Ronald Coase Institute workshop and at the welcome dinner observe me buying a beer for a third person. You decide to buy a beer for me. What can we conclude about your motivation? It could be that you wanted to reward me for my reputation of being generous, suggesting that indirect reciprocity might be in place. At the same time, it is possible that your behavior was motivated by social influence. Since this is your first time to Bratislava, you do not know what the norms are and how you should behave. Me buying a beer for a third person constitutes a signal of what is socially appropriate. Therefore, it could be that rather than rewarding me for being generous, you conformed to the social norm. At the same time, you might have extended the courtesy due to the mere fact that you were dealing with me and not some random stranger. Identification can increase the cognitive attention given to a particular individual and thus result in different behavior towards him. Last but not least, you could be an altruist who gets

pleasure from being kind to other people and your motivation has nothing to do with my previous action. To summarize, we observe that virtue seems to be contagious, but how can we find out what the transmission mechanism is? In Servátka (2009) I present a laboratory experiment designed to behaviorally discriminate between indirect reciprocity, social influence, identification, and altruism through a series of treatments that generate within-experiment reputation and manipulate information that the decision-maker has about the recipient. By comparing behavior in these treatments I observe that the reputation effect has a stronger impact on actions than the social influence and identification. Such separation of motives is a central step in trying to understand how impulses towards selfish or generous behavior arise and clearly deserves more research. Related topics in this area include, for example, the importance of information about the other party and its effect on behavior, changes in beliefs and changes in customs and social norms. Importantly, it is also possible to study transmission mechanisms using field experiments, as exemplified by Mujcic and Leibbrandt (2017) study on indirect reciprocity.

Experiments not a panacea One has to keep in mind that experiments also have their limitations. My goal is not to elaborate in detail on objections to experimentation; such objections and responses to them are nicely summarized in e.g. Falk and Fehr (2003), Charness and Kuhn (2010) or Fréchete and Schotter (2015). Rather, I would like to share my thoughts on a situation that I frequently encounter. Students who have just discovered the advantages of using experimental methods often approach me because they would like to run an experiment, thinking it will solve all their problems. Suppose the big question you are interested in is: “Why are some societies able to achieve a higher level of cooperation than others?” and to provide an answer you intend to compare country A with country B. Ignoring the fact that this (commonly asked) research question is not specific enough, how much light can experiments shed on this issue? To the extent argued above, experiments can help us identify individual factors promoting or hindering cooperation in the two countries, contribute to our understanding of motivations behind cooperation and the transmission mechanisms through which they operate, measure the strength of social norms (Krupka and Weber, 2013) and informal institutions (Jakiela, 2011) or the impact of culture (Gächter et al, 2010, van Hoorn, 2012), and study the evolution of institutions (Kimbrough et al, 2008) or their adoption (Gürerk et al, 2006). They can also be helpful in implementing policy changes in a way so we could draw conclusions about their effectiveness and learn about the drivers of observed differences in behavior. At the same time, one has to keep in mind that a possible selection bias, different historical perspectives, and the quality of formal and informal institutions likely play a role in how cooperative societies A and B are. The effects of such factors are not going to be automatically identified in, say, a simple comparative public goods experiment conducted

with members of the two societies. Therefore, in order to draw conclusions regarding the reasons why we observe more cooperation in some societies than in others, one has to have a thorough understanding of the organization and functioning of these societies and control for potential confounding factors in a way similar to empirical studies that use happenstance data. Ideally, the researcher will collect proposed theories and conjectures that explain why the societies differ in their cooperation rates, identify their implications for the given environment and in a series of experiments test them one by one to assess their individual impact. While the main advantage of experiments is the controlled variation, it would be incorrect to think of them as a substitute for the econometric analysis of happenstance data, field experiments/randomized control trials, survey data, and qualitative methods, as each of these methods has its strengths and weaknesses. Experiments are not a hammer that a researcher can use to tackle any question. Instead, their appropriate use is better motivated by the anecdote about a drunk man looking for keys under the lamppost. A passerby offers to help him but asks whether he is sure he lost the keys under the lamppost. “No,” the drunk man replies, “but this is the only place where I can see.” As Duncan James, an experimental economist and a good friend of mine, once said, experiments could be compared instead to a (movable) torchlight – a tool that allows us to search in the dark. But just because we have a torchlight, it does not automatically mean we will find the keys; we also need to know where to look for them. Acknowledgement: I wish to thank Claude Menard, Mary Shirley, and Rado Vadovič for their constructive comments and suggestions.

References Andreoni, J. (2005) "Trust, Reciprocity, and Contract Enforcement: Experiments on Satisfaction Guaranteed," University of California San Diego working paper. Arrow, Kenneth J. (1974) The Limits of Organization. Norton, New York. Blount, S. (1995) “When Social Outcomes Aren't Fair: The Effect of Causal Attributions on Preferences,” Organizational Behavior and Human Decision Processes, 63:2, 131-44. Bracht, J. and N. Feltovich (2008) “Efficiency in the Trust Game: An Experimental Study of Precommitment,” International Journal of Game Theory 37:39–72 Camerer, C. (2003) Behavioral game theory: experiments on strategic interaction. Princeton: Princeton University Press. Cason, T. and R. Zubrickas (2017) “Enhancing Fundraising with Refund Bonuses," Games and Economic Behavior, 101, 218-233. Charness, G. (2004) “Attribution and Reciprocity in an Experimental Labor Market,” Journal of Labor Economics, 22:3, 665-88.

Charness, G. and M. Dufwenberg (2006) “Promises and Partnership,” Econometrica, 74(6), 1579-1601. Charness, G. and P. Kuhn (2011) “Lab Labor: What Can Labor Economists Learn in the Lab?” Handbook of Labor Economics, 4a, 229-330. Chaudhuri, A. (2008) Experiments in Economics: Playing fair with money. Routledge. Chaudhuri, A. (2011) “Sustaining cooperation in laboratory public goods experiments: A selective survey of the literature,” Experimental Economics, 14 (1), 47-83. Cox, J. (2004) “How to Identify Trust and Reciprocity,” Games and Economic Behavior, 46, 260-281. Cox, J. C., Ostrom, E., Walker, J. M., Castillo, J., Coleman, E., Holahan, R., Schoon, M., & Steed, B. (2009). Trust in private and common property experiments. Southern Economic Journal, 75, 957–975. Cox, J. C., Kershbamer, R., and D. Neururer (2016) "What is Trustworthiness and What Drives It?" Games and Economic Behavior, 98, 197–218. Cox, J. C., Servátka, M, and Vadovič, R. (2017) “Status Quo Effects in Fairness Games: Reciprocal Responses to Acts of Commission vs. Acts of Omission,” Experimental Economics, 20, 1-18. Dufwenberg, M., Servátka, M. & Vadovič, R. (2017) “Honesty and Informal Agreements,” Games and Economic Behavior, 102, 269-285. Falk, A. and E. Fehr (2003) “Why labour market experiments?” Labour Economics, 10, 399– 406 Falk, A., Fehr, E., and Fischbacher, U. (2008) “Testing theories of fairness - Intentions matter.” Games and Economic Behavior, 62:1, 287-303. Fehr, E. and Gächter, S. (2000) “Fairness and Retaliation: The Economics of Reciprocity,” Journal of Economic Perspectives, 14:3, 159-81. Fehr, E. and S. Gächter (2002) “Cooperation and Punishment in Public Goods Experiments,” American Economic Review. 90, 341-369 Fehr, E. and Schmidt, K. M. (2006) “The Economics of Fairness, Reciprocity and Altruism Experimental Evidence and New Theories.” In S. Kolm & J. M. Ythier (Eds.) Handbook of the Economics of Giving, Altruism and Reciprocity. Elsevier. Fréchete, G. and A. Schotter (Eds.) (2015) Handbook of Experimental Economic Methodology, Oxford University Press. Gächter, S., Herrmann, B. and C. Thöni (2010) “Culture and Cooperation,” Philosophical Transactions of the Royal Society B – Biological Sciences 365(1553), 2651-2661. Gürerk, Ö; Irlenbusch, B.; Rockenbach, B. (2006) “The Competitive Advantage of Sanctioning Institutions,” Science, 312, 108‐111.

Jakiela, P. (2014) “Using Economic Experiments to Measure Informal Institutions,” In S. Galiani and I. Sened (eds.) Institutions, Property Rights, and Economic Growth: the Legacy of Douglass North, Cambridge University Press, Kimbrough, E., Smith, V. and Wilson, B. (2008) “Historical Property Rights, Sociality, and the Emergence of Impersonal Exchange in Long Distance Trade,” American Economic Review, 98(3), 1009-39. Krupka, E. and R. Weber (2013) “Identifying social norms using coordination games: Why does dictator game sharing vary?” Journal of the European Economic Association 11 (3), 495-524 Ledyard, O. (1995) “Public Goods: Some Experimental Results,” In J. Kagel and A. Roth (Eds) Handbook of Experimental Economics, Princeton University Press, Princeton, NJ. Morita, H. and M. Servátka (2013) “Group Identity and Relation-Specific Investment: An Experimental Investigation," European Economic Review, 58, 95-109. Mujcic, R. and A. Leibbrandt (2017) “Indirect Reciprocity and Prosocial Behaviour: Evidence from a Natural Field Experiment,” Economic Journal, forthcoming. Ostrom, E., Walker, J., and R. Gardner (1992) “Covenants With and Without a Sword: SelfGovernance is Possible,” American Political Science Review, 86, 404-417 Servátka, M. (2009) “Separating Reputation, Social Influence, and Identification Effects in a Dictator Game,” European Economic Review, 53(2), 197-209. Servátka, M., Tucker, S. and R. Vadovič (2011) "Words Speak Louder Than Money," Journal of Economic Psychology, 32(5), 2011, 700–709. Sobel, J. (2005) “Interdependent Preferences and Reciprocity,” Journal of Economic Literature, 43, 392-436. Smith, V. (1982) "Microeconomic Systems as an Experimental Science," American Economic Review, 923-955. Smith, V. (1994) "Economics in the Laboratory," Journal of Economic Perspectives, 8(1): 113-131. Van Hoorn, A. (2012) “Cross-cultural experiments are more useful when explanans and explanandum are separated,” Proceedings of the National Academy of Sciences, 109, E1329E1329. Woods, D. and M. Servátka (2018) “Nice to You, Even Nicer to Me: Does Self-Serving Generosity Diminish Reciprocal Behavior?" Experimental Economics, forthcoming.