Micromotives, Microstructure, and Macrobehavior: The Case of Voluntary Cooperation

Abstract

How micromotives, the microstructural features of interactions, and macrobehavior are related is a fundamental question in all social sciences. In this article we argue that laboratory experiments are a useful tool to study this question, because the experimenter can measure motivations, manipulate microstructures, and sometimes even exploit variation in the macrosocial environment. We illustrate the experimental approach with the help of four examples from the context of voluntary cooperation. The examples (from previous and new experiments) illustrate the link between individual pro-social motives and wage formation in labor markets, the inevitability of the “Tragedy of the Commons,” and two potential ways to avoid the tragedy: sorting and punishment.

Keywords:

• gift exchange
• laboratory experiments
• macrobehavior
• macrosocial background
• micromotives
• microstructure
• punishment
• reciprocity
• sorting
• voluntary cooperation

Acknowledgments

We gratefully acknowledge financial support by the University of Nottingham and the Grundlagenforschungsfonds of the University of St. Gallen. We also benefited strongly from the excellent comments by the anonymous referees, and by Vincent Buskens, Werner Raub, and Marcel van Assen.

Notes

¹See Carpenter, Harrison, and List (2005); Levitt and List (2008); Falk and Heckman (2009); Bardsley et al. (2010); and Rosser and Eckel (2010) for recent discussions of the methodology of experimental economics. The edited volume by Webster and Sell (2007) contains several articles on methodology and practicalities of experiments in sociology.

²See Hechter and Horne (2009) for a collection of important texts on the problem of social order. The problem of social order is also discussed in Fehr and Gintis (2007) and Buskens and Raub (2008). Buskens and Raub point out, for example, that already Durkheim and Weber saw contractual incompleteness as a central issue.

³In the experiment all players are only informed about the rules of the experiment and how decisions translate into payments. Hence, employers actually only learn about their employees' degree of reciprocity as the experiment progresses. At the beginning of the experiment employees can only form beliefs (drawn from their life experience) about how likely reciprocal reactions are.

⁴The experimental instructions and all further details can be found in Herrmann, Thöni, and Gächter (2008b).

⁵See Fischbacher and Gächter (2010) and the references therein.

⁶Replication studies investigated the distribution of types in different subject pools. Herrmann and Thöni (2009) find an identical fraction of Conditional Cooperators but a smaller fraction of Free Riders in experiments conducted in four cities in Russia. Kocher, Cherry, Kroll, Netzer, and Sutter (2008) conducted the experiment in Austria, Japan, and the United States and found a substantially higher share of conditional cooperators in the United States than in the other countries. Finally, Thöni, Tyran, and Wengström (2009) report results from a large-scale Internet experiment with a diverse subject pool in Denmark and also find a very high share of conditional cooperators of 70%. Other studies which use different methodologies also find preference heterogeneity in public goods games. See Kurzban and Houser (2005), Bardsley and Moffatt (2007), and Muller, Sefton, Steinberg, and Vesterlund (2008).

⁷Instructions are available in Gächter and Thöni (2004). The instructions of the new experiments were adapted accordingly and are available upon request.

⁸ Ranking is a proxy measurement because unlike in the previous example, not the full contribution vector as a function of others' contribution is measured. The reason is that sorting people into groups as a function of their degree of conditional cooperation is much less straightforward on the basis of a contribution schedule than on the basis of a particular contribution level. Moreover, free rider types clearly have an incentive to contribute nothing in Ranking, whereas people who do contribute a positive amount cannot be free rider types. Pessimistic conditional cooperator types might be behaviorally equivalent to free rider types, which is why our measurement is only a proxy tool.

⁹Tests are calculated on the basis of independent group averages. Subjects do only receive information about the contributions of the Ranking experiment of their new group members after the rematching procedure. Subjects are not informed about the outcome of the one-shot public goods game of the Ranking experiment until the very end of the whole experimental session. Thus, only within rematched groups information spillovers are possible.

¹⁰In the treatment Baseline we combine the results of two treatments: (i) a treatment where subjects simply play the 10 period public goods game with random matching, and (ii) a treatment where subjects first play the Ranking experiment and then play ten periods of the public goods game, also with random matching. We use these two treatments to check whether the presence of the Ranking treatment has an influence on contributions in the ten period public goods game. It turns out that this is not the case. The treatment averages do not differ according to a two-sample Wilcoxon rank-sum test: z = .674; p = .501.

¹¹In the treatment Sorted Strategic we have an odd number of groups (17). The category HIGH contains one group more than the category LOW.

¹²There are only a few studies that have utilized non-student subject pools to study punishment behavior (Egas and Riedl, 2008; Gächter and Herrmann, in press). These studies find limited evidence for the importance of socio-demographics for punishment behavior.