can see why this is the case from figure (2): below some level of venality, it no longer
pays for M to skim at all, and the boss can eke out an additional 10% in the success
rate by upping the amount of money he pays to M relative to what he would be
willing to do if M were skimming. The existence of this cutoff in
0
γ —derived
analytically in the previous section—leads to a clear policy implication: state action
that serves to remove the most venal of the available middlemen can lead to a sudden
and sizable increase in the likelihood of terrorist success.
Second, the behavior of each quantity of interest is decidedly non-linear in the
budget constraint. Figure (3) clearly illustrates this dynamic. For all levels of
middleman greed, there is a constraint level below which the chance of an attack
becomes small due to the breakdown of cooperation between boss and middleman.
Essentially, there is a point at which B prefers to spend all the money on non-attack
goods rather than achieving the low success probability he can get given his budget
and M’s greed. However, once
γ exceeds that level–which will be different for
different values of the other parameters–the chance of success undergoes a
discontinuous increase, up to 80% in the case of the more venal agents. Past this
point, increasing the level of funds available to B causes a more gradual, nearly linear
increase in all the variables, until an equilibrium is reached, whereupon B no longer
desires to increase the likelihood of success. This leads to a second clear policy
implication: interdicting terrorists’ funds can provide dramatic decreases in the
likelihood of an attack, even in cases where previous reductions in available funds
seemed to have a more gradual impact.
Note that government actions which make the operational environment more
challenging, increasing
α
, should have effects similar to tightening the budget
constraint. Thus we expect important non-linearities in the returns to government
39
Convention