can be inflicted on a defector, which depends on the immediacy and frequency of
punishment. Global strategies have the most power because punishment can take place
the following period and applies to everyone (Figure 1)—all else being equal, the greater
the power, the lower the continuation payoff for someone who defects when everyone
else is cooperating. Hence, a strategy with greater power reduces the incentives to defect.
Global strategies provide the largest possible threat since punishment is immediate and
indiscriminate. Targeted strategies have the least power, while reactive strategies occupy
a space in-between the two. For example, with public monitoring, the lower bound for δ
falls by about 40% when we move from a reactive strategy to a global strategy, and by
about 50% when we move from a targeted strategy to a global strategy.
The selectivity of a strategy is linked to the incentives to punish defectors. Targeted
strategies are the most selective and allow agents to punish at the lowest cost. We use the
term selectivity because it is related to the richness of information needed to support
cooperation. A more selective strategy requires a finer partition of the information set.
5 Results
We first present results on the aggregate outcome (Results 1-6) and then on the
strategies employed to sustain those outcomes (Results 7-11). The section is broken up by
treatment so that the discussion can be more focused.
Result 1. The introduction of public monitoring in the non-anonymous treatment
increased cooperation over private monitoring.
Figures 2 and 3 and Table 4 provide support for Result 1. In the non-anonymous
public monitoring average cooperation across economies is 81.5%. For an economy
k=1,..,50 we define the action aitk of an agent j=1,..,4 in period t=1,..,Tk of economy k to
be an element aitk∈{0,1}≡{Z, Y}; a cooperative action is coded as 1 and a defection is
1 Tk 4
coded as 0. Therefore, the average cooperation in an economy k is ck = 7∑∑atk and
4 t=1 i=1
17