, Volume 43, Issue 2, pp 113-149

Divergent evidence on free riding: An experimental examination of possible explanations

Rent the article at a discount

Rent now

* Final gross prices may vary according to local VAT.

Get Access

Summary, conclusions and extensions

The most important single observation from this research is the similarity between our wide range of results and the multitude of seemingly divergent conclusions about free riding from previous experimental results. Even when defined in the restrictive manner of this paper, free riding is neither absolutely all pervasive nor always nonexistent. This ‘intermediate’ result is not the same as a general theory which states that the predictable result of public goods provision processes is always weak free riding. The extremes of strong free riding and near-Lindahl optimal behavior can and do occur.

These experiments further demonstrate that this diversity of outcome need not be attributed to inexplicable randomness. At least for the case of the voluntary contribution mechanism, there are identifiable factors which make free riding more or less likely to occur. Two such factors have been identified here: (i) the replication environment and (ii) per capita marginal return. This latter parameter can be related to although it is not equivalent to group size.

This research is not intended to ‘model’ a precise quantitative prediction about free riding. Having identified, in a general manner, these effects, the groundwork is set for many interesting questions in future research:

  1. Although the qualitative results more often suggest that ‘free riding’ behavior increases with experience, there are no statistically supported conclusions. A design focusing on the factor of experience could reveal whether this effect is sustained under more extensive replication.

  2. The influence of marginal per capita return is striking. Why is there a difference when zero contribution is a single period dominant strategy for either level (.3 or .75)? How extensive is the interaction of replication and marginal per capita return? We are currently conducting a new set of experiments focusing upon these issues.

  3. Given the power of the marginal per capita return, what happens in situations in which this parameter varies, as in a quadratic return function?

  4. Why did some individuals contribute positive amounts even in the 10th period when the learning and multiperiod gaming aspects were presumably of minimal effect? We conjecture that this represents a core of people for whom utility functions are not completely selfish or who otherwise wish to believe in ‘good guy’ fashion. However, it is possible that some individuals were still learning their single period dominant strategy or did not correctly notice the presence of a truly single period decision environment. In any case, it appears that marginal per capita return again plays a role.

  5. With the group production technology held constant, increasing group size (with a concomitant decrease in marginal per capita return) has the expected effect of increasing ‘free riding’ type behavior. Standardizing for marginal per capita return, the effect of group size becomes ambiguous, and shows evidence of reversing. This aspect of group size effects is another area for future research. Our ongoing research will also look at this question.

  6. What can be said about economies which lack condition D*? If individuals have no dominant strategy, the very concept of ‘free riding’ becomes poorly defined. Definitions and predictions must explicitly state what assumptions about expectations and what solution concepts are being employed.

In summary, we find that there is no successful general theory which states that all individuals always free ride a lot, always free ride a little, or never free ride. Under the appropriate circumstances, we find people who will do any or all of the above. This research gives some guidelines as to why and when free riding can be expected. Further work in both theory and experiments may be able to tell even more.

Isaac and Walker are Assistant Professor of Economics at the University of Arizona. The authors gratefully acknowledge the research support of the National Science Foundation (through a grant to Professor Vernon Smith) and of the University of Arizona College of Business and Public Administration (through its summer research grant program). The encouragement, advice, and assistance of Professors Vernon L. Smith, Ronald Oaxaca and Mark Walker were invaluable, although all remaining errors are entirely the responsibility of the authors.