Gamson's Law versus non-cooperative bargaining theory

We compare Gamson's Law, a popular empirical model of legislative bargaining, with two non-cooperative bargaining models in three players divide the dollar games in which no player has enough votes to form a winning coalition on their own. Both of the game theoretic models better organize the comparative static data resulting from changes in nominal bargaining power than does Gamson's Law. We also identify deviations from the point predictions of the non-cooperative bargaining models. Namely, proposer power is not nearly as strong as predicted under the Baron-Ferejohn model, and a significant number of bargaining rounds tend to take more than two steps under demand bargaining and more than one stage under Baron-Ferejohn, counter to the models' predictions. Regressions using the experimental data provide results similar to the field data, but fail to do so once one accounts for predictions regarding coalition composition under Gamson's Law.