Ex Interim Voting: An Experimental Study of Referendums for Public-Good Provision

Abstract

Fischer and Nicklisch [2007] present a fascinating experimental design aimed at evaluating the consequences of ex interim voting mechanisms on the efficient provision of public goods under conditions of unanimity rule, majority and no referendum (automatic provision). The authors find that (1) initial contributions are the same under both unanimity and majority rule; (2) over the course of the experiment, the average number of yes votes is essentially the same for both majority rule and unanimity, with the consequence that the public good is provided more often under majority rule; (3) over the period of the experiment, average contributions are higher under the unanimity provision, but the variance among contributions is also higher under unanimity; and (4) with respect to efficiency and group profits, the unanimity condition produces the least efficient outcome, with majority rule more efficient and automatic provision the most efficient. The authors find that players did not coordinate to produce efficient results but rather exhibited a form of path dependence in their behavior. In particular, players who voted against the good in the previous round tended to decrease their contributions, while rejection of the public good led to increased contributions. This path dependence was most pronounced under conditions of majority rule, suggesting that majority rule is the least stable voting mechanism over time. The authors express surprise that no meaningful coordination took place between the players and that unanimity voting mechanisms decreased efficiency the most. Their expectation was that majority rule would be least efficient because of unstable voting coalitions, making coordination more difficult. Yet the result may be a function of the experimental design. In the real world, majority voting mechanisms produce the problem of unstable coalitions because meaningful bargaining can take place between the players. Under the experiment's conditions, the players had full information but there is no indication that communication took place between them, limiting bargaining and negotiating opportunities that arguably create problems with systems operating under majority rule. Second, it would be interesting for the authors to explore in more detail the path dependent nature of the decision making process over repeated rounds. As noted