cooperation was an equilibrium outcome. The study finds remarkably higher cooperation
in fixed than in random matching economies. Therefore, despite the theoretical viability
of cooperative equilibria with random matching and private monitoring, it seems that
they are empirically difficult to attain.
A key parameter when comparing cooperation rates between fixed and random
matching treatments is the expected number of encounters with any given person. This
number is higher under fixed matching than random matching for economies of equal
size and identical continuation probability. As a consequence the deck is stacked in favor
of observing higher cooperation in fixed matching. To avoid this bias, we do not use a
partner treatment. Instead we introduce a novel design that, on one hand, equalizes the
expected number of encounters with any given person across treatments, while on the
other hand, provides as much information as in the partner treatment. The new design has
random matching and public monitoring, as we provide the complete history for each
agent in the economy. A subject knows the identity of their opponent (non-anonymous)
as well as what their opponent chose when meeting other participants.1
A second novel feature of our study is to understand which one of the several
available strategies that support a given equilibrium outcome have been employed.2 This
issue has been largely unexplored in the experimental literature on supergames, as it has
mostly focused on measuring the levels of cooperation. As we will later clarify, we
develop a design where we can exploit differences in information across treatments in
order to change the strategy set and hence identify the type of strategies employed.
We also relate the choice of punishment strategy in an indefinitely repeated setting
to the literature on costly personal punishment in one-shot settings. Experimental studies
of finitely repeated social dilemmas have evidenced a surprising tendency of subjects to
engage in costly personal punishment of others, in particular defectors. Although, this
1 Schwartz et al. (1999) and Duffy and Ochs (2006) also consider treatments when subjects receive some
information about the reputation of their current opponent while preserving anonymity of the opponent.
Our non-anonymous public monitoring gives all individual histories and so provides information about
reputation and it also reveals the identity of the opponent.
2 These strategies include off-equilibrium threats that on the equilibrium path will never be employed. The
features of such threats are largely irrelevant as long as they are credible and they generate a sufficiently
low continuation payoff.