New fairness notions in align with the merit principle are proposed for designing exchange rules. We show that, for an obviously strategy-proof, efficient and individually rational rule, an upper bound of fairness attainable is that, if two agents possess objects considered the best by all others, then at least one receives her favorite object. Notably, it is not possible to guarantee them both receiving favorites. Our results thus indicate an unambiguous trade-off between incentives and fairness in the design of exchange rules.