I have always understood Scitovsky's main claim from his famous 1941 article ("A Note on Welfare Propositions in Economics") to be that Kaldor efficiency (or Hicks efficiency), used alone, is subject to reversals, such that a move from allocation-state A to allocation-state B is socially efficient and so is a subsequent move back to allocation-state A. This is obviously problematic from the perspective of social-welfare calculation. But I also understood Scitovsky to assert that an allocation-state that is Kaldor-Hicks efficient (that is, meets both the Kaldor and Hicks criteria*) is not subject to reversal. Am I wrong about this? I sometimes come across economists, usually writing in the cost-benefit analysis literature, who claim that Scitovsky reversals are a problem for the combined Kaldor-Hicks criteria. Which of us is misunderstanding Scitovsky?
*A (re)allocation is Kaldor efficient if those who gain could, in theory, completely compensate the losers without completely depleting their gains. A (re)allocation is Hicks efficient if the losers could not bribe the gainers into foregoing their gains without suffering a loss.