Game Theory and History, A. Greif & Friends (1993, 1994)

(This post refers to A. Greif, “Contract Enforceability and Economic Institutions in Early Trade: The Maghribi Traders’ Coalition”, AER 1993, and A. Greif, P. Milgrom & B. Weingast, “Coordination, Commitment and Enforcement: The Case of the Merchant Guild,” JPE 1994.)

Game theory, after a rough start, may actually be fulfilling its role as proposed by Herbert Gintis: unifier of the sciences. It goes without saying that game theoretic analysis is widespread in economics, political science (e.g., voter behavior), sociology (network games), law (antitrust), computer science (defending networks against attacks), biology (evolutionary strategies), pure philosophy (more on this in a post tomorrow!), with occasional appearances in psychology, religion (recall Aumann’s Talmud paper), physics (quantum games), etc. But history? Surely game theory, particularly the more complex recent results, has no place there? Yet Avner Greif, an economic historian at Stanford, has shown that games can play a very interesting role indeed in understanding historical events.

Consider first his Maghribi traders paper. In the 11th and 12th century, a group of Judeo-Arabic traders called the Maghribis traded across the Mediterranean. Two institutional aspects of their trade are interesting. First, they all hired agents in foreign cities to carry out their trade, and second, they generally used other Maghribi merchants as their agents. This is quite different from, for instance, Italy, where merchants tended to hire agents in foreign cities who were not themselves merchants. What explains that difference, and more generally, how can long distance traders insure that traders do not rip you off? For instance, how do I keep them from claiming they sold at a low price when actually they sold at a high one?

To a theorist, this looks like a repeated reputational game with imperfect monitoring. Greif doesn’t go the easy route and just assume there are trustworthy and untrustworthy types. Rather, he assumes that there are a set of potential agents who can be hired in each period, that agents are exogenously separated from merchants with probability p in each period, and that merchants can choose to hire and fire at any wage they choose. You probably know from economics of reputation or from the efficiency wage literature that I need to offer wages higher than the agent’s outside option to keep him from stealing; the value of the continuation game, then, is more than the value of stealing now. Imagine that I fire anyone who steals and never hire him again. How do I ensure that other firms do not then hire that same agent (perhaps the agent will say, “Look, give me a second chance and I will work at a lower wage”)? Well, an agent who has cheated one merchant will never be hired by that merchant again. This means that when he is in the unemployed pool, even if other merchants are willing to hire him, his probability of getting hired is lower, since one merchant will definitely not hire him. That means that the continuation value of the game if he doesn’t steal from me is lower. Therefore, the efficiency wage I must pay him to keep him from stealing is higher than the efficiency wage I can pay someone who hasn’t ever stolen, so I strictly prefer to hire agents who have never stolen. This allows the whole coalition to coordinate. Note that the fewer agents there are, the higher the continuation value from not stealing, and hence the lower the efficiency wage I can pay: it is optimal to keep the set of potential agents small.

What of the Italian merchants? Why do they not hire only each other? Maghribi merchants tended to be involved only in long distance trade, while Italian merchants were also involved in real estate and other pursuits. This means the outside option (continuation value after cheating if no one hires me again) is higher for Italian merchants than Maghribi merchants, which means that hiring merchants at the necessary efficiency wage will be relatively more expensive for Italians than Maghribis.

A followup by Greif, with Milgrom and Weingast, considers the problem of long distance trade from the perspective of cities. Forget about keeping your agent from ripping you off: how do you keep the city from ripping you off? For instance, Genoans in Constantinople had their district overrun by a mob at one point, with no compensation offered. Sicilians raised taxes on sales by Jews at one point after they had brought their goods for sale. You may naively think that reputation alone will be enough; I won’t rip anyone off because I want a reputation of being a safe and fair city for trade.

But again, the literature of repeated games tells us this will not work. Generally, I need to punish deviations from the efficient set of strategies, and punish those who themselves do not punish deviators. In terms of medieval trade, to keep a city from ripping me off, I need not only to punish the city by bringing it less trade, but I also need to make sure the city doesn’t make up for my lost trade by offering a special deal to some other trader. That is, I need to get information about violation against a single trader to other traders, and I need to make sure they are willing to punish the deviating city.

The merchant guild was the institution that solved this problem. Merchant guilds were able to punish their own members by, for example, keeping them from earning rents from special privileges in their own city. In the most general setting, when a guild orders a boycott, cities may be able to attract some trade, but less than the efficient amount, because only by offering a particularly good deal to the merchants who come during a boycott will entice them to come and to credibly believe the city will not steal.

This is all to say that strong guilds may be in the best interest of cities since they allow the city to solve its commitment problem. The historical record confirms many examples of cities encouraging guilds to come trade, and encouraging the strengthening of guilds. Only a reputational model like the above one can explain such city behavior; if guilds are merely extracting rents with monopoly privilege, cities would not encourage them all. Both of these papers, I think, are quite brilliant.

1993 AER (IDEAS version) and 1994 JPE (IDEAS version). Big thumbs up to Avner for having the final published versions of these papers on his website.