perceive the differences in the incentive structure of a finitely repeated versus an
indefinitely repeated interaction, and react in the expected direction. For example, Palfrey
and Rosenthal (1994) and Dal Bo (2005) report lower cooperation for finite duration
experiments in comparison to indefinite duration experiment with a the same expected
length. Moreover, the higher the discount rate the lower the cooperation. For a recent
discussion see Normann and Wallace (2006).
In order to place our contribution within the existing literature, and given our focus
on the models in Kandori (1992) and Ellison (1994), we will discuss indefinitely repeated
experiments whose stage game is a prisoner’s dilemma (for experiments with other
games see Cason and Khan, 1999, Engle-Warnick and Slonim, 2004, 2006, Engle-
Warnick, 2007). It is helpful to classify experiments with indefinite interaction according
to two aspects, the matching protocols and the availability of information supplied about
other subjects. The protocol to match subjects within a supergame is an empirically
relevant and theoretically interesting parameter. Furthermore, all experiments we
surveyed include several supergames within a session, and hence need an additional
protocol to match the subjects after each supergame. We will come back later to this
matching across supergames and for now focus on matching within a supergame.
The most common matching protocol within a supergame is fixed matching. For
instance see Palfrey and Rosenthal (1994), Aoyagi and Frechette (2003), or Dal Bo
(2005). Under this design, which we refer to as “partner”, subjects always interact with
the same person and generally support a significant level of cooperation, sometimes full
cooperation. The present study employs instead a random matching protocol within a
supergame as, for instance, in Schwartz et al. (1999) and Duffy and Ochs (2006). In any
given period subjects still meet in pairs but after each period new pairs are randomly
formed drawing among subjects from a larger economy with N>2 people.
A comparison of fixed matching (partner) versus a random matching (stranger) in
finitely repeated games can be found in Andreoni and Croson (2002) and indefinitely
repeated games can be found in Duffy and Ochs (2006). This latter study has a random
matching treatment with private monitoring and the parameters were set in a way that full