Proposition 4.1: No Nash equilibrium of game 1 can support honest trade (Dt / 0) at the efficient level (xt / x*), regardless of the levels of c, t, T, or B.
Proof: Suppose there were such an equilibrium and consider the payoff to the city if it deviates from the equilibrium strategy and cheats a fraction D of the first-period traders. In the initial period its payoff is f(x*)(t!c[1!D]). In subsequent periods the informational assumptions of the model imply that the play of at most D traders is affected. Consequently, at least 1!D traders come to the city in each future period, and the city's payoff from treating them honestly is, in present value terms, at least @(t!c)f(x(1!D)) (for convenience define @ = B/(1!B)). So the city's total payoff from cheating a fraction D of the traders in the first period and adhering to the purported equilibrium thereafter is at least
f(x)(t!c(1!D)) + @(t!c)f(x(1!D)), (2) and this expression coincides exactly with the actual payoff when D = 0, that is, when the city adheres to the purported equilibrium. The derivative of expression 2 with respect to D at D = 0 and x = x* is
cf(x*) ! @(t!c)x*fN(x*) = cf(x*) > 0, (3) because fN(x*) = 0. This establishes that the city has a profitable deviation, that is, the specified behavior is not a Nash equilibrium. Q.E.D.
No mechanism based only on sanctions by those who are cheated can support honest trading at the efficient level, x*, because when trading is conducted at that level, the marginal trader has zero net value to the city. By cheating a few marginal traders, the city loses nothing in terms of future profits but saves a positive expense in the present period. There is no institution in which the ruler’s belief in a merchant’s retaliation enables him to commit at the efficient level of trade. The belief that the ruler will respect rights at the efficient level of trade is not self-enforcing. To support the efficient level of trading, some kind of collective action among merchants is