This paper experimentally examines infinitely repeated prisoners' dilemma games
with imperfect private monitoring and random termination where the probability of
termination is very low. Laboratory subjects make the cooperative action choices quite
often, and make the cooperative action choice when monitoring is accurate more often
than when it is inaccurate. Our experimental results, however, indicate that they make
the cooperative action choice much less often than the game theory predicts. The
subjects' naivete and social preferences concerning reciprocity prevent the device of
regime shift between the reward and punishment phases from functioning in implicit
collusion.
|