Governing collective action in the face of observational error

We present results from a repeated public goods experiment where subjects choose by vote one of two sanctioning schemes: peer-to-peer (informal) or centralized (formal). We introduce, in some treatments, a moderate amount of noise (a 10 percent probability that a contribution is reported incorrectly) affecting either one or both sanctioning environments. We find that the institution with more accurate information is always by far the most popular, but noisy information undermines the popularity of peer-to-peer sanctions more strongly than that of centralized sanctions. This may contribute to explaining the greater reliance on centralized sanctioning institutions in complex environments.