Estimation of the house money effect using hurdle models

Evidence from an experiment investigating the "house money effect" in the context of a public goods game is reconsidered. Analysis is performed within the framework of the panel hurdle model, in which subjects are assumed to be one of two types: free-riders, and potential contributors. The effect of house money is seen to be significant in the first hurdle: specifically, house money makes a subject more likely to be a potential contributor. Hence we find that the effect of house money is more than just an effect on behaviour; it has the effect of changing a subject from one type to another. This result is potentially important in the external validity debate. -- Public Good Experiment ; Hurdle Model ; double hurdle model ; Tobit ; panel data