The Role of Institutions in the Revival of Trade: The Law Merchant

by Paul Wilgrom, Douglass North, and Barry Weingast:

Reputation:

Institutions help traders gain reputations in large communities where would be otherwise improbable. The system of judges before the rise of law encouraged merchants to be honest, impose sanctions on violators, become adequately informed on others' behavior, provide evidence against violators, and pay any judgments assessed against them.

The temptation to cheat in the short-run makes trust difficult to create. One way to ensure honest transactions is to make a continuing relationship an asset that a party could lose by dishonest behavior. Only the opportunity for a big pay-off would compel a merchant to surrender such a valuable bond.

A reputation system sometimes only work where it encompasses sufficiently many traders and trades as the greater volume of contacts and agreements a merchant undertakes, the higher his credit score.

But if informal arrangements based on reputations can effectively bond good behavior, then what is the role of formal institutions in helping to support honest exchange? Given the existence of so many large and specialized institutions to facilitate agreement between all kinds of businessmen and politicians, it appears something more than mere reputation is necessary. In other words, why do reputations fail to cover all transactions, and why do formal institutions emerge when they do?

With no state enforcement of contracts or an established body of commerical law, merchants had to create their own private code in the Middle Ages. Disagreements were handled by a judge, be he a local official or private merchant. Judges had only limited powers to enforce judgments, so why have them at all?

The Medieval Law Merchant:

In order to capture the gains associated with geographic specialization, a system needed to be established that lowered information costs and provided for the enforcement of agreements across space and time. There was no state to enforce contracts or protect merchants from pirates and brigands. Commercial law predated the rise of large-scale third-party enforcement.

As trading communities grew larger, it became more difficult for merchants to monitor one another's behavior. New institutions were needed to mitigate cheating. As trade grew more disperse, merchants needed greater assurance that their partners would not take advantage of wide expanses of space and reneg on deals.

Merchants moved to towns where they developed their own law. Merchant guilds arose to provide protection to local merchants against insidious ittiernants. Legal codes governing commercial transactions arose and were administered by private judges drawn from commerical ranks which by the end of the 11th century had created a uniform set of standards across large numbers of locations.

Commerical law, therefore, can be understood as coordinating the self-interested actions of merchants as well as coordinating the actions of people limitied knowledge and trust.

The code provided a means for reducing uncertainty and limited the ability of locals to cheat foreigners. As towns lacked the power to enforce standards outside of their region and states had not developed to the point where nefarious international hustlers could be caught. The development of mutually benefical trade could only occur as long as merchants obeyed the code.

Champange Fairs:

The Champange Fairs provide an interesting example that illuminates the importance of reputation. A legal system existed in the 12th and 13th centuries for judication of disputes, but nothing prevented a merchant from providing low quality goods and peacing. Even a judgment against the supplier would not matter if he never returned to the fair. Ostracism appears the logical route, but if this were the case, why would a legal system be required at all?

In a prisoner's dilemma game, assume the following: if both players go it honestly, they both make a profit. If both players cheat, no one gets anything. A trader profits through cheating an honest partner and imposes a larger loss on his honest partner. If the game is only played one, both players will cheat because the result of cheating is still better than getting cheated.

However, repeated transgressions lead players to condition their actions based on what transpired in the past, meaning they have an instrument to reward honest behavior and punish cheating. If trade is frequent, then there is a Nash equilibrium where the player adopts the Tit for Tat strategy, in which the player will go it honestly the first time and react to whatever his partner played the second time and so-on.

The central idea is that frequent trading with the same partner, or "clientization," makes it possible to find an equilibrum with efficient trading. It holds for virtually all repeated games, regardless of players. The same conclusion also holds in a community of traders in which players change partners often and cheaters do not have to face a former partner ever again, provided that information about the behavior of the traders is widely disseminated throughout the community.

Suppose there are N traders and there is some rule M that matches them at each stage. Let h, be the history of trade through date t and let M(h1,i) be the identity of the trader matched with trader i at date t + 1 at history h1. If player i plays honest at date 0 and then plays cheat at date t + 1 if two conditions hold: his partner failed to recognize his equilibrium strategy or the other player cheated the first time. A trader who cheats will eventually be punished by the next merchant he meets even if that merchant is honest and has never met the cheater.

Punishing the cheater is directly profitiable because punishment is delivered by playing Cheat. A merchant who fails to deliver a punishment, by participating in a boycott or something, is also subjected to punishment from the community.

There is no point at which a merchant can make a one time play different from equilibrium pay that raises his total payoff. It is not necessary for any pair of traders to interact frequently in order for a boycott mechanism to be effective. But that relies on the condition that members of the community are well-enough informed to know who to boycott.

The Law Merchant Enforcement System:

In the incomplete information of the Prisoner's Dilemma game where no traders meet twice and no trader's behavior can directly or indirectly influence the behavior of his future trading parners, the outcome of any Nash equilibrium is that each player plays Cheat at every opportunity. There are no incentives for honest behavior. Incentives can be restored by introducing an institution that provides full information to each trader about the other operates, but that would be costly. Efficient trade does not require that every trader know the full history of behavior of each trader. One only needs to know his own history of behavior and whether his partner had defected on the previous turn to determine his current behavior. How does one arrange that the traders are adequately well informed so that they can sanction a cheater when necessary.

Imposing sanctions and remaining aware of them can be personally costly. Institutions must keep traders adqueately informed of their responsibilities and motivate them to do their duties. One who keeps informed about who should be punished for past actions is supplying a public good by deterring others from cheating. No trader except his current partner will ever know if a trader does not check his partner's present history. A trader could avoid supplying the public good without facing any sanction. Therefore, cheated traitors must be motivated to document perfidies. If players who are cheated are unwilling to invest in informing their neighbors, the Cheater will profit from his action and all Honest trade suffers.

LM = Law merchant, the repository of information and the adjudicator of disputes. Any party can accuse any other of cheating and appeal to LM. Any dispute appealed to LM is perfectly and honestly adjudicated at cost C to the plantiff. The LM's pronouncements include the ability to award damages if the defendent is found to have cheated the plantiff. The payment of the damage is voluntary in the sense that there is no state to enforce payment. Any party can visit LM prior to finalizing a deal. At a cost of Q, the party can query for the records of previous judgments. Here is the sequence of play:

1. Players may query ther LM about their current partner at utility cost Q > 0. In response to a query, the LM reports to the players whether a party has any 'unpaid judgments,'. Whatever transpires at this stage becomes common knowledge among the LM and the two players.
2. The two players go through the Prisoners' Dilemma gameto learn the outcome.
3. Either party may make an appeal to LM at a personal cost ofC> 0, but only if he has queried the LM.
4. If either party makes an appeal, then the LM makes a judgement, J, to the plantiff if he has been honest and his trading partner cheated. Otherwise, no award made.
5. If a J awarded, the defendent may pay it at personal cost f(J), or he may refuse to pay it, at cost o.
6. Any unpaid judgments are recorded by the LM and become a part of the permanent record.

The player's utilties for the extended game are determined as the sum of payments recievied less than those made. In this case, an Honest player who is cheated and appeals recieves: - Q, - X (cheat) + J - C, of the other party pays the judgment, and - Q, - X, and -C, if he does not.

The function of f: R (the utility of paying a given judgment - it can be assumed the greater the size of the judgment, the greater the cost to the defendent. The cost of paying the judgment is never less than paying the judgment itself. This excludes the possibility that paying judgments adds to the total utiltiy of he players.

The desired behavior in such situations goes as follows:

Part 1: A trader queries the LM if he has no unpaid judgments on record, but not otherwise.
Part 2: If either player fails to queries the LM or if the query establishes at least one player has an outstanding judgment, then both players play Cheat: otherwise, both play honest.
Part 3: If both parties fuliflled part 1, and exactly one of the two players Cheated at part 2, then the victim appeals to the LM, otherwise, no appeal is filed.
Part 4: If a valid appeal is filed, the LM awards damages of J to the aggrieved party.
Part 5: The defendent pays the judgment J if and only if he has no other outstanding judgments.

1 - Q (frequency of trade)/1 - frequency of trade) > f(J) > the max amount (x - 1), f (C).

In evaluating expected payoffs, players must make certain conjectures about what others have done to forecast what will happen. We assume that the trader beleives that all others have played according to the rules in all past plays except where a detection is caught. Must assume all will obey the rules in the future.

Paying a judgment yields an expected payoff of -f(J) in the current period. In future periods, the player will spend Q to query the LM and earn a trading payoff of 1. If the player refuses to pay a judgment, then he is stuck at zero, and his payoff in every subsequent period is also zero. Therefore it pays to pay judgments if and only if f(J) < (1 - Q) frq. (1 - frq of trade). Does it pay the victim to appeal where it incurs personal cost c. Given the above, the trader can expect the judgment to be paid. So f(J) > f(C).

If there are no unpaid judgements and the LM has been queried, does it pay for the trader to play honest? If he does, it will be 1 -Q. If he cheats and adheres to the abovementioned strategy, it will be - Q + the benefit of cheating - f(J). Equilibrium requires the former is larger, or that f(J) > cheating - 1.

The cheat always maximizes payoffs for the current period, so it does pay. It makes sense for both to query LM without pay periods because they will benefit only if 1 - Q, if Q > 1. Does it pay a party with an outstanding judgment to query? No, because his partner will cheat him and o > -Q.

There is NO situation in which a one-time deviation from the LM is profitable for a trader provided that paying the judgment sufficient enough to deter cheating, the judgment must be large enough to cover the cost of the appeal, and it must not be so large that the cheater refuse to pay it, for then the injured party would get nothing to collect and find it unprofitble to appeal. If traders live at great distances from one another and if their principal asset holdings are illiquid, then wealth transfers will be costly and probably would not work.

It must be worthwhile for traders to query the LM. If traders fail to query, then they possess insufficient information to administer punishment and once again cheating will go unpunished. Traders who fail to query are constantly cheated by their trading partners.

Minimizing Transaction Costs:

There are transaction cost to this system.

Aume the following:

Q C P temp. to Cheat discount factor J
X .5 .5 50% 2 .67 1
Y .5 1 50% 3 .80 2
Z .5 3 50% 7 .90 6

The LM system is not viable if the cost of making and investigating a claim or the cost of a judgment is too high for the traders cannot be expected that others will make the same claims. They only act as deterrents. Any institution that restores incentives for Honest trading by restoring the effectiveness of decentralized enforcement must inform a player when his partner cheated in the past. If the temptation to cheat is small and trading frequent, information need not be perfect. If one can lower the cost of query, it could be more effective. LM system avoids unnecessary costs of dispute resolution and los on tranfers. It centralizes the information system so that information about any partner is in one place. The costs incurred by LM are inevitable.

Dishonest Law Merchants:
Perfect judges are a luxury no historical economist would apply to here. A trader extorted by a judge would need recourse to reestablish his reputation. How is it in the interet for the LM to behave honestly due to client incentives in the long-term relationship between himself and the trader. If a trader pays a bribe, he will be subjected to future extortion. A LM who threatens the reputation of a trader only loses business with little gain for himself.

Lets suppose LM earns 2x > 0 per contract as part of the 2Q that the parties spend to query him. Let us suppose a LM tries to extort an honest trader. The amount B is chosen by the LM. If the bribe is not paid and a query is made, the LM reports falsely. When a B is paid, the LM's payoff increases by B and the victim's is reduced.

No player who paid a bribe, cheated, and then had a judgment charged against him would ever pay it because he would only be extorted in the future. If a player bribed in the past, then the profits for paying the bribe are cheat - q - bribe. Not paying leads to zero. It is profitable whenever B is less than or equal to cheating minus q.

If a player has paid a bribe before, he will pay anything up to cheat -q. If a player hasn't paid a bribe and one is demanded, if he takes it, he will cheat, and the payoff will be Cheat - B - Q and, as a trader with an unpaid judgment, 0. His payoff in the current period is 0 if he refuse to pay and 1-q in the future.

When facing a trader who has never before paid a bribe, the LM expects that any demand for a bribe will be refused and that trader will also not query, leading to a loss of revenue.

The worse problem is preventing traders who cheated in the past from trying to pay people off.