A Recommendation to Rouhani and Obama

By: Ehsan Barin

His responses to the questions were very clear, and it sounds that he has a great background in his field of expertise. Born 1950 in New York City, David Kreps is a game theorist and economist and professor at the Graduate School of Business at Stanford University. His research interests include dynamic choice, human resource management, and behavioral economics. He is also best known for contributing the idea of sequential equilibrium, which is developed as a mutual work with Robert Wilson, his colleague at Stanford. His Microeconomics textbook is a source of study in Iran’s postgraduate economics schools. The following is an interview with him.

You have considerable works about “Repeated Games” and the role of “Reputation” in these games. Would you please say about the origins of emerging such an idea?

The work that I did with Robert Wilson on these topics was, of course, far the original work on repeated games and reputation.   The Folk Theorem, which came long before our work, was the seminal work on these general topics.

Wilson and I (along with Paul Milgrom and John Roberts, who worked independently at first) contributed the idea that, even with a finite horizon, reputation-like constructions would be possible if one relied on incomplete information.  (One should also see the work of Benoit and Krishna that derives “folk theorems” in finitely repeated games without incomplete information, but where the stage game has multiple equilibria.)   I think the first foray into this way of thinking was the paper of Milgrom and Roberts, entitled “Entry-deterring pricing doesn’t deter entry.”  This was only a two-period model, but it showed clearly that the interaction of timing and incomplete information would be subtle.   Wilson and I (and Milgrom and Roberts) then saw how powerful these ideas would be in the context of the chain-store paradox, and we came together to apply the same ideas to the finitely repeated prisoners’ dilemma.

Your Microeconomics text book is one of the resources in Postgraduate studies in Iran. You have discussed about the role of incentives and information economics in some chapters of your book. Is there any relation between these two (incentives and information) and repeated games?

I believe there are very strong and important connections, but they are not yet well developed in the economics literature.   There are, of course, a number of important papers dealing with incentives in repeated agency settings; I’m thinking here of Holmstrom and Milgrom’s “Linearity and Aggregation” paper, and later work, especially by Sannikov.   But I think the connections run much deeper.   Economic incentive theory deals almost exclusively with “formal” incentives---the principal and agent agree at the outset on a compensation formula that aligns the agent’s incentives with those of the principal, to some extent.   But such formal incentives are only a small piece of motivation more generally, and schemes for motivating, say, employees go far beyond such formal and formulaic incentive schemes and rely heavily on the reputations of both the employee and employer.  To take an example, several authors (Baker, Gibbons, and Murphy; Pendergast and Topel) have written about “subjective incentive contracts,” wherein the principal decides ex post how much to reward the agent, with the agent working in anticipation of this reward.   The focus has been on influence activities, but a more fundamental question is:   In such schemes (which are extremely common in real life), why does the employee trust the employer to exercise her subjective judgment in a fair and equitable manner?  This, of course, has everything to do with repeated interactions and reputation.  

One of the challenging economic issues in the late 20th century is about choosing monetary authority. Maybe we can put the start point of this challenge on an article by Professor Barro and Professor Gordon; “Rules, Discretion and Reputation in a Model of Monetary Policy” (1983). About a year before that article, you have written another one with Professor Robert Wilson; “Reputation and Imperfect Information” (1982). What is your idea about the influence of your mutual work with Professor Wilson on the Barro-Gordon works?

This is a question that must be addressed to Professors Barro and Gordon.   It is perhaps worth observing that Milgrom and Roberts completely independently came up with the same ideas as were in my paper with Wilson, so these ideas were “in the air.”

One of the most famous games is “Repeated Prisoner’s Dilemma”. How much can we rely on “Reputation” for achieving the best equilibrium in this kind of games? Or is there any experimental work to verify the efficiency of “Reputation”?

 We can certainly not rely on reputation for achieving the best equilibrium, even if we know what is meant by “best equilibrium.”   There has been work, and I think here in particular of work by Fudenberg and Levine, showing that in a one-long-run-many-short-run-players situation, the long-run player has at her disposal whichever equilibrium she most desires, as long as there is enough incomplete information.   But this leaves open the question---what are her desires?   One might say that “best” means Pareto efficient.  But in a world where we do not know the preferences of any participant---and this is of course the starting point of this class of models of reputation---not knowing preferences renders “Pareto-efficiency” into a virtual tautology.

And, I should have said at the outset, in the infinitely-repeated-game variation on reputation constructions, the folk theorem limits the range of outcomes very little.   Nothing that I know of in that branch of the theory guarantees a “best” equilibrium.

 One of the problems in labor markets is mismatching of people for jobs. Iran (as a developing country) has a large group of well-educated people which suffer from this problem. Is there any mechanism that best matches people (especially well-educated people) to their most favored jobs?

I think you want to reformulate this question as, “that achieves a good match of people to jobs,” where good encompasses both the preferences of the employees and employer.   And then, in limited and controllable situations, the answer is yes:  See, for instance, Al Roth’s work on the market for medical interns.

But I suspect you have in mind a matching mechanism for a much larger economy.   And there, the answer is a clear No.  In part, this is because of the dynamics of the situation:   “Well-educated people” (to use your phrase) make education decisions that constrain their later choices long before they know what choices they will have.   And the numbers of people involved in an economy the size of, say, Iran, make any single mechanism impractical.  Simply put, the answer is:  Yes on a very limited and confined scale.   But No as the scale grows.

In your opinion, what is the role of Dynamic Choice and Behavioral Economics in handling of a big or multinational company? I mean how is the importance of knowing about these two for CEO of such companies?

I teach every summer in an executive program called the Stanford Executive Program.   This brings to Stanford around 140 very high level executives from around the world---perhaps 25% of the group are CEOs or equivalent, while the median participant is head of a function, such as a CFO.   These people know a lot, instinctively, about dynamic choice and how people---employees, customers, political “masters”---behave.   And it is crucial to their job that they do so.

What theories (and applications!) of dynamic choice and behavioral models in economics add for these people are frameworks that help them improve their instincts when it comes to making decisions.   When I teach them (for instance) about how the Toyota system of strategic relationships with manufactures of subassemblies works, I am clarifying in their minds the process, and perhaps I even show them some details of which they hadn’t thought. But they understand already, instinctively, what is going on; it is important that they do so, although (I believe) it is useful for them to have a greater sense of clarity about these things, which is what we (economists) can add.

You are a Microeconomics expert. There are several economists who use microfoundations to support their theories firmly, but there are some other economists who pay attention mostly on statistical works and support their theories by econometrics analysis. In your opinion, which one of the two types is most important for an economic analysis? And is there any other method for experimental analysis other than Econometrics?

There is a lot of experimental work that relies on simple statistics, but I don’t think this is what you have in mind with the last part of this question. So I’ll pass on that one.

I am a very, very big fan of economic analyses that begin with rigorous theory and then test those theories empirically.   I think, in this respect, of Ariel Pakes as an exemplar.   I am a fan, but less so, of studies that “let the data speak first” or of studies whose only connection to the real world is “arm-chair empiricism,” even though most of my own work is of the latter type.   All of these forms of research contribute.  Indeed, my colleagues in economics tend to make light of the Harvard Business School’s chief “research” product, which is the case study.   But case studies also contribute to our (economic) understanding of the world, and that (after all) is the point of economics. 

 Mainstream Economics respect “Private Ownership” and use it as a base of their theories. But, there some other analyses which use “Common Ownership” or “Governmental Ownership” as a base of economic theories. What is your opinion about the ownership?

I think “ownership” in all its forms should be grist for our theories, employing in particular transaction-cost economics to understand the strengths and weaknesses in specific situations of specific patterns of ownership.   I point, in this regard, to the classic work of Grossman and Hart on the topic.

There are pressure groups in both Iran and the US, who are trying to prevent President Rouhani and President Obama from making diplomatic relations between the countries. As an expert, what would you think about the best strategy for the Presidents on both sides?

I’m not an expert on the specifics, and I certainly would not want to offer the presidents any concrete advice.  Based on my own expertise, I would offer two insights from theory blended with my observations:   (a) Actions nearly always speak louder than words.   (b) In particular, one builds credibility by being credible.

In current days the Iranian diplomats and diplomats from the so called “3+3 countries” (USA, Russia, China, and England, France, Germany) are negotiating with each other in Vienna, to catch an agreement about Iran’s nuclear program. Can we consider these negotiations as repeated game, in which “Reputation” of countries has an important role?

The situation is not a repeated game in the formal sense but, to change your phrasing a bit, I think it is obvious that the credibility of the parties to carry out their portions of any agreement is an essential element in reaching an agreement, and reputation is very often a key to obtaining credibility.   That said, reputation-management is tricky business:  X’s reputation with Y is a product of the long-term interactions between X and Y, but it also depends on how X behaved and is behaving with Z and W.   Y, in trying to assess what X will do in the future, looks closely at what X did in the past with Z and W and, at the same time, thinks hard about what X may be called upon to do in the future in re Z and W.

Initially published in Tejarat-e-Farda weekly magazine