Non-Parametric Tests of the Tragedy of the Commons

Abstract

The "tragedy of the commons" occurs when agents over-exploit a common resource. Although often applied, the canonical behavioral model of the tragedy has rarely been tested directly. We derive a non-parametric test of behavior consistent with the tragedy-of-the-commons model. Our approach allows for an arbitrarily concave, differentiable production function of total inputs and for heterogeneous agents with arbitrarily convex, differentiable costs of supplying inputs. We develop three additional tests to allow for sampling error, to establish metrics defining distance from the data to the model, and to account for measurement error in the data. Applying our approach to panel data of Norwegian commercial fishers, we find evidence rejecting the tragedy-of-the-commons model. Significantly, we find rejection rates of the model increase after property rights reforms limited access to the resource for a segment of the fleet. Results from parametric difference-in-difference regressions are consistent with these findings.