タイトル：Testable implications of fair allocations
We investigate testable implications of fair allocations via the revealed preference approach. For an observed data which consists of finitely many allocations for a given group of agents, we seek what conditions must be satisfied by the data in order to rationalise that the data is consistent with some fairness concepts. Specifically, we focus on two fairness criteria on allocations, the envy-freeness and the equal division lower boundedness (EDLB). We derive two sets of equivalent conditions for the rationalisation, and the results are fully utilised to
obtain graphical representations of how non-rationalisable allocation data sets look like, and to provide clear intuition on testable implications of the envy-freeness and the EDLB. We also deal with testable implications of efficient- envy-free allocations (allocations that are efficient and envy-free) and efficient-EDLB allocations (allocations that are efficient and satisfy EDLB) and we show that efficiency puts no further observable restrictions to either the envy-freeness nor the EDLB when both the number of agents and observations are two. We then provide the Broners’ power index to show the strength of the testable implications of the fair
allocations. Finally, we apply our results to simple bargaining experiments.