Center for Social Information Sciences (CSIS) Seminar

Abstract: In centralized allocation platforms participants may not fully observe other participants' types. Hence, the designer may deviate from the promised allocation mechanism without the participants being able to detect these deviations. In this paper, I develop a theory of auditability in allocation problems. I measure a mechanism's auditability by the smallest number of participants that can detect any deviation, and I find a striking contrast between the auditability of prominent allocation mechanisms. On one extreme, the Serial Dictatorship and Immediate Acceptance mechanisms are maximally transparent, in a sense that any deviation can always be detected by just two participants. On the other extreme, the Deferred Acceptance and Top Trading Cycles mechanisms are minimally transparent, in a sense that some deviations may never be detected unless there is full information about everyone's types. The findings may give an alternative explanation of the widespread usage of Immediate Acceptance mechanism despite its potentially inferior theoretical properties compared to Deferred Acceptance and Top Trading Cycles.