Gene writes: [...] >In that group of "most economists" I see cited all the time I think most >would assert that you could compensate the losers out of the gains of the >winners and thus make all better off. I'm serious here. I hear that >explanation quite often for schemes that are just short of murder and >smallpox. This is a good point--I neglected this option in my earlier posts. But the question remains whether the compensation principle stands as "reasonable" practice in neocleassical welfare theory. Put it this way: the ultimate neoclassical statement on these issues is Arrow's Impossibility Theorem, which says roughly that there doesn't exist a coherent social decision rule which simultaneously satisfies a) the Pareto criterion b)non-dictatorship c)"independence from irrelevant alternatives" (more on this one in a bit) and d) universal applicability (meaning applicability no matter what the particular shape of people's preferences). Now as Gene suggests, the compensation principle gets around the Arrow impossibility result, but I don't think it's accepted as conventional wisdom in neoclassical-land that the Arrow theorem has thereby been repudiated. Here's the reason: the compensation principle clearly violates the "independence of irrelevant alternatives" assumption, because it presumes the ability to measure the relative *intensity* of preference (as captured by the level of hypothetical compensation). So is the IIA condition simply irrelevant in light of the compensation principle? Hardly. It's one thing to say *in principle* that party A could compensate party B for some change. But in practice this encounters difficulties: how would anybody go about figuring out what the called-for level of compensation is? What, for example, would keep the potentially injured party (appropriately enough, in the case of the Indian blanket and Nazi genocide examples) from insisting on an astronomical level of compensation? And on what basis could anybody insist that such demands were inappropriate (how the hell could they tell?)? So while the compensation principle sounds plausible (to a neoclassical) at first, it turns out that it is essentially impossible to apply consistently in practice, and therefore does not constitute a meaningful repudiation of Arrow's Impossibility Theorem. Therefore I stick with my original point: by any reasonable (in this sense, internally coherent) neoclassical interpretation, the examples originally advanced by Jim C. could not be considered efficient.
