This PhD thesis demonstrates how Public Goods Game experiments can be used to design and test cooperative environments. In chapter two I propose an intergroup competition scheme (ICS) to theoretically solve the free-riding problem in the public goods game. The key feature of the ICS is a transfer payment from the group with the lowest contribution to the group with the highest contribution that is proportional to the difference in the overall contribution between the groups. The ICS is trivial to implement, requires minimal information, makes the efficient contribution a dominant strategy and is budget balanced across the groups. Consistent with the theory, the experimental results demonstrate that the ICS significantly raises contributions to almost reach optimality.
Chapter three examines the effects of in and out-group social comparisons on cooperation in team situations. Performance benchmarking, where firms compare their performance to other firms, is one channel firms can use to motivate free-riders to contribute greater effort. Three competing models are put forward to explain how comparative information might affect contribution preferences: conformity, competition, and selfish biased conditional cooperation. This study varies in-group and out-group comparative information to experimentally test the models driving behavior. Social comparisons raise cooperation with the highest level of cooperation observed when both in-group and out-group comparisons are provided. However there are differences in how in-group and out-group comparisons influence cooperation.
Chapter four compares the performance of alternative estimation approaches for Public Goods Game data. A leave-one-out cross validation was applied to test the performance of five estimation approaches. Random effects is revealed as the best estimation approach because of its un-biased and precise estimates and its ability to estimate time-invariant demographics. Surprisingly, approaches that treat the choice variable as continuous out-perform those that treat the choice variable as discrete. Correcting for censoring is shown to induce biased estimates. A finite Poisson mixture model produced relatively un-biased estimates however lacked the precision of fixed and random effects estimation.