Grief (1993)

Jump to navigation Jump to search

Theory Questions:

What is the research question?

Before the modern era, traders faced a problem. Exporting goods abroad required labor for delivery. If the trader *himself* supplied this labor, he would limit his trading (and other) opportunities by spend long period of time traveling. However if the trader hired someone, that agent could steal, sabotage or ransom his goods.

The research question of this paper is: How did traders overcome this? The author describes how 11th Century Maghribi traders overcame this problem. He uses historical evidence and a game theory model to explain its rationale.

What is the author's hypothesis?

The author describes a "coalition" of cooperative traders who shared information about their experiences with individual agents. Members of the coalition would punish agents who deviated together. This would help overcome the imperfect monitoring problem that traders and courts faced when agents left their supervision to deliver goods.

Agents would earn a wage and premium whose discounted value was generally greater than the value of cheating.

How does the author test the hypothesis?

The author tests the hypothesis by deriving a theoretical model, and showing that the strategies employed by the traders (according to the archival/historical records) approximate the equilibrium strategy of his model.

The theoretical model in question is essentially an infinitely repeated relational contract game in which a reputation system (through a wage premium) incentivizes agents to tell the truth as a best strategy.

Note that intergenerational transfers of reputation and wealth -- as was practiced by the agents in real life -- make the model's infinite horizon more believable.

How does the author rule out alternative hypotheses?

  • Based on the lack of historical evidence, the author says that a court-enforced legal system was not used to overcome the commitment/agency problem. The author also notes that asymmetric information problems would hinder the court's ability to enforce contracts. For example: An 11th Century court has no way of verifying whether a storm destroyed goods or if the goods were stolen. On the rare occasions in which the courts were indeed used, they were evidently expensive and time consuming. The author speculates that this would be prohibit their use. Jewish law also (according to the author) restricts the ability to sue agents.
  • The author also notes that a model based on "types" in which "bad types" carry their cheating around is not supported by historical evidence (pg 532, left).

How might these tests be run if one had quantitative evidence?

A key, underpinning aspect of the model is its infinite horizon assumption. Without it, the agents have less of an incentive to be honest.

Suppose that an exogenous factor results in some agents having better opportunities than being an honest trader. If the agent knows he will leave the trading business in advance, he would have no more reason to maintain his reputation and should begin shirking until he leaves the industry.

If we had data about entries and exits to the agency business, we may be able to learn something about whether reputation is indeed driving good behavior.

What problems might arise in this quantitative analysis?

Unfortunately the data mentioned above probably does not exist. Even if it did, one might wonder why an agent (and we researchers) would know about his improved outside options without his employers ever knowing about it.

The types of events that would lead an agent to change careers (starting a new business, marrying into a rich family, etc) are often observable in a social context and are likely to be known by their employers (the merchants).

If the merchants knew that the infinite horizon was about to end, they would respond by no longer hiring the agent -- since the agent no longer has incentives for good behavior.