Survey_BCC_2014

Abstract

The experiment consists of a cooperative task that involves eight subjects and is repeated for an indefinite number of periods. In each period, subjects meet in pairs, where one has the option to help the other at a cost. Everyone has repeated opportunities to help and to receive help because roles alternate over time. Pairs are formed at random in every period and identities remain hidden, so the subjects interact as strangers. Cooperation requires trusting that help given to a stranger will be returned by a stranger later in the game. There are three treatments: Baseline, which serves as a control, Money and Memory. To facilitate cooperation, in all treatments subjects can see when everyone in the group helps. This type of public monitoring is theoretically sufficient to support any cooperation level, from zero to one-hundred percent (folk theorem). Here lies a social dilemma with two intertwined issues: opportunism, due to the short-run temptation to avoid helping others, and coordination, due to the existence of multiple equilibria, including the efficient outcome corresponding to full cooperation. In Money we add a fixed amount of intrinsically worthless electronic tokens, which participants can choose to exchange for help or hold as balances. Memory adds a record-keeping system based on numeric balances—positive or negative—that rise for those who help and fall for those who receive help. In this design tokens and recordkeeping are theoretically irrelevant in the sense that in all treatments there already exists a self-enforcing norm capable to support the efficient outcome. However, in Money and Memory subjects can also support the efficient outcome by conditioning help on observed balances. In this sense, tokens and record-keeping are theoretically affine: balances in Memory can be employed to replicate a pattern of monetary trade without transferring tokens, while tokens can communicate a subject’s past conduct without the need to rely on any additional public monitoring. Through this design we can uncover behavioral differences between monetary systems and systems for maintaining and sharing information about past conduct.