Are tax-financed contributions to a public good completely crowded-out? Experimental evidence

Abstract

We report the results of a laboratory experiment on crowd-out in a voluntary contribution mechanism public good game. In our setting, a standard argument states that a tax should not be effective in raising contributions, because agents respond by reducing voluntary contributions by the amount of the tax. Our experimental design focuses in on this intuition by abstracting away from several potential confounds. We use a specification for the payoff function in which there is a dominant strategy for own-earnings maximizing agents, located interior to and in the upper half of the strategy space. The dominant strategy ensures that changes in contributions are attributable to the tax directly, rather than second-order effects due to responses to out-of-equilibrium play by other agents. The dominant strategy is made more transparent by the use of a novel graphical decision interface. We find that individuals robustly choose at or above the own-earnings dominant strategy level. Even with the controls of the design, crowd-out is incomplete, but the degree of crowd-out is higher than in previous studies. Analysis of individual-level decisions provides evidence of different player types. Behavior of subjects not choosing the dominant or Pareto-efficient contributions is well-organized by a model of warm-glow giving with a logit decision error.