Résumé
We investigate experimentally the relevance of Warr's neutrality theorem (1983). In our test treatments we implement an unexpected income redistribution among group members. We compare treatments with equalizing (unequalizing) redistribution to benchmark treatments without redistribution, starting either with an equal or unequal income distribution. Our data are consistent with the neutrality theorem: redistribution (either equalizing or unequalizing) does not affect the amount of voluntarily provided public good. However at the individual level we observe that subjects tend to under-adjust with respect to the predicted Nash-adjustment. We also observe an insignificant adjustment asymmetry between the poor and the rich: after redistribution subjects who become poorer lower their contribution by a larger amount than subjects who become richer increase their contribution. Finally we also observe a general tendency for poor subjects to over-contribute significantly more than rich subjects.