Behavioral Mechanism Design as a Benchmark for Experimental Studies

I introduce the idea of behavioral mechanism design where in addition to the usual selfish players there are noisy players who play randomly and ethical players who actively seek to maximize social welfare and are committed, up to a point, to “do their bit” to achieve that goal. I calibrate this model using data on risk aversion and giving in dictator games. I then use it to study fifteen different (out of sample) experiments including stag hunt games, ultimatum bargaining games, and public goods games with and without punishment. I show that this simple calibrated model makes sharp predictions and does a good job both qualitatively and quantitatively in explaining the data from those experiments. The theory also identifies quantitative anomalies in the data pointing the way to future improvements. I conclude that this simple calibrated model might be a good benchmark for other experiments.