After examining all 14 service areas, we found that in
most of the service areas the discounted
values of both firms always increase, but with decreasing
increments, as we increase the reversion length (T) in months.
However, we could not report on the discounted values for 
.
Instead, we report on discounted values using a large
T=5000 months.
We selected to report on values where the difference in
discounted values is less than 1 U.S. dollar. That is,
U.S. dollar and 
U.S. dollar.
It can be observed in Table 4 that some service areas had a
peak values at less than . Service area A is one of
these areas, that is T=1227 months. We have only reported on the
Pareto optimal peaks and do no show suboptimal rises in values that
are smaller than all future discounted values.
In the service areas where
the model shows the Pareto optimal number of reversion
periods for a service area was infinite, this implies that
the firms actions would follow a grim strategy.
A grim strategy would mean that in those service areas collusion
between firms can never occur after one incident of the price
dropping below the trigger price. However, the model also shows
many suboptimal tuples both where the best or where the best
.
In suboptimal cases, collusive quantities
should also be chosen as the best.
In theory these outcomes
should not be selected, because they are not Pareto optimal. But,
in practise we can question if the firm is concerned with
the difference in expected discounted values from the smaller
reversion length.
In summary, the response curves demonstrate
each case where choosing collusive quantities is at least Pareto superior
to choosing Cournot quantities.