A sufficient condition is that 0 $ cf(x) ! @(t!c)xfN(x) and the elasticity e(x) = dRnf(x)/dRn(x) is a decreasing function of x.
Proof: The traders' strategies are obviously best replies to the strategy of the city from any point in the history of the game, so only the optimality of the city's strategy needs to be proved. Beginning with x current traders, consider the subgame achieved after x!y traders depart, when y # x traders remain. By cheating a fraction D of the y current traders, the city's payoff will be g(D;y) = (t![1!D]c)f(y) + @f(y[1!D])(t!c). A necessary condition for the optimality of D = 0 is Mg(D;y)/MD # 0 at D = 0. An easy calculation verifies that this is the same as condition 6, so the latter condition is necessary for all y.
By the optimality principle of dynamic programming, it is sufficient to show that there is no subgame in which the city would do strictly better by setting D > 0 in the initial period and then adhering to its equilibrium strategy thereafter, given the strategies of the others. If f is concave, then for all y, g(D;y) is concave in D, so a sufficient condition is that for all y, Mg(D;y)/MD # 0 at
D = 0, which is again equivalent to condition 6, proving sufficiency.
The elasticity can be rewritten as e(x) = xfN(x)/f(x). Condition 6 is that e(y) $ c/[@(t!c)] for all y # x, which follows from e(x) $ c/[@(t!c)] and the hypothesis that e(A) is decreasing. Q.E.D.
Let xN be the largest solution to condition 6. The equilibrium described by proposition 4.4 suggests an interesting interpretation of the levels of trade, xN, observed during boycotts, and it explains why some merchants continued to trade and others did not. According to the theory, additional traders, beyond the number xN, would be cheated by the city and would be unable to exact retribution for their losses. Alternatively, if one thinks of the level of trade x < x* during the embargo as being determined by factors outside the model (such as existing alliances or other interests), then condition 6 implies that the minimum tax rate necessary to deter cheating is lower the lower x is. This confirms the intuition that an embargo breaker may be able to negotiate an unusually attractive deal, both because the value of trade per trader (f(x)/x) is higher when x is small and because the minimum tax rate t necessary to prevent cheating is lower for small x.
Proposition 4.4 implies that in the absence of a strong guild organization—one that can impose the embargo on its members—the guild cannot credibly threaten to reduce the city's