Friedman, Harsanyi, Rawls, Boulding - or somebody else? An experimental investigation of distributive justice

This paper investigates distributive justice using a fourfold experimental design: The ignorance and the risk scenarios are combined with the self–concern and the umpire modes. We study behavioral switches between self–concern and umpire mode and investigate the goodness of ten standards of behavior. In the ignorance scenario, subjects became, on average, less inequality–averse as umpires. A within–subjects analysis shows that about one half became less inequality–averse, one quarter became more inequality–averse and one quarter remained unchanged as umpires. In the risk scenario, subjects become on average more inequality–averse in their umpire roles. A within–subjects analysis shows that about half became more inequality–averse, one quarter became less inequality–averse, and one quarter remained unchanged as umpires. As to the standards of behavior, several prominent ones (leximin, leximax, Gini, Cobb-Douglas) were not supported, while expected utility, Boulding’s hypothesis, the entropy social welfare function, and randomization preference enjoyed impressive acceptance. For the risk scenario, the tax standard of behavior joins the favorite standards of behavior. JEL classification: D31, C91, D63