DP10788 The Dynamic Free Rider Problem: A Laboratory Study
|Author(s):||Marco Battaglini, Salvatore Nunnari, Thomas R Palfrey|
|Publication Date:||August 2015|
|Keyword(s):||durable public goods, experiments, voluntary contribution mechanism|
|JEL(s):||C72, C73, C78, C92, H41|
|Programme Areas:||Public Economics|
|Link to this Page:||cepr.org/active/publications/discussion_papers/dp.php?dpno=10788|
Most public goods are durable and have a significant dynamic component. In this paper, we report the results from a laboratory experiment designed explicitly to study the dynamics of free riding behavior in the accumulation of a durable public good that provides a stream of discounted benefits over a potentially infinite horizon. This dynamic free-rider problem differs from static ones in fundamental ways and implies several economically important predictions that are absent in static frameworks. We consider two cases: economies with reversibility (RIE), where the agents? voluntary contributions to the public good can be positive or negative; and economies with irreversibility (IIE), where contributions are non negative. For both economies, we characterize the unique Markov perfect equilibrium. The evidence supports the main predictions from the theory: behavior is generally consistent with stationary, forward-looking behavior; both in RIE and IIE the accumulation path is inefficiently slow and the public good under-provided; and RIE induces significantly higher public good contributions than IIE. A number of interesting deviations from the theoretical predictions are observed: both in RIE and in IIE we have over-investment in the early rounds of the game; in RIE over-investment is followed by periods in which negative contributions correct the stock, bringing it back to the predicted steady state; in IIE over-investment tends to decline approaching zero. To test the Markovian assumption, we compare the predictions of the Markov equilibrium with the prediction of the most efficient subgame perfect equilibrium and propose a novel experimental methodology that relies on the comparison between the behavior in the dynamic game and the behavior in a one-period reduced-form version of the dynamic game.