Money Is Memory

This 1998 Journal of Economic Theory paper by Narayana Kocherlakota asks what money technologically does by comparing it against an alternative record-keeping technology the paper calls memory: costless knowledge, for every agent, of the past trading history of everyone he has ever been directly or indirectly linked to through trade. Working in a general class of economies – one broad enough to nest the overlapping-generations, turnpike, and random-matching models used elsewhere in monetary theory – in which agents cannot commit in advance to honor any proposed allocation, the paper’s main result (Proposition 2) proves that every allocation that is incentive-feasible when money is available is also incentive-feasible in the same economy when memory is available instead. Whether the reverse holds depends on the environment: in the overlapping-generations and turnpike economies the two technologies support exactly the same set of allocations, but in a Trejos-Wright-style random-matching economy memory strictly dominates money, because an allocation that pays identical continuation utility to everyone after every match can be sustained with memory but not with money, since a money-holder must be rewarded for having produced. The paper is careful to show that this equivalence rests on two specific restrictions built into its notion of memory – it must be limited to one’s own network of trading partners’ past partners, not a complete public record, and the matching process must otherwise leave agents uninformed about a partner’s history – and it demonstrates with a worked counterexample that relaxing the “limited” restriction can make some money-feasible allocations infeasible under full memory.

Narayana R. Kocherlakota | Macro Paper Warehouse

Money Is Memory