SAMUELSON PRO AND CON

Fifty three years ago I would not have expected to be able to meet and talk with two presidentsJimmy Carter and George Bush, in 2002; neither would I have expected to be asked to write for Paul Samuelson's 90th. But August 2005 was the 53rd anniversary of my arrival in Cambridge Mass. I applied and

was accepted at MIT and at Harvard (also Carnegie, Stanford and Chicago), but Harvard offered a better fellowship net of tuition than MIT or the others, and in the previous May my wife had given birth to twins so I had to deal with economic survival, get my degree and get out of Cambridge; Paul's wife Marion had triplets, showing tiiat Paul was more prolific on all fronts than was I. Just after my arrival in Cambridge, I remember encountering him by accident in a supermarket out on Route 2, and we exchanged stories on die joys-for us as fathers-of multiple births; the wives were not there.

Paul doesn't know it but he was an important influence on my choosing to go into economics. At Caltech as a senior (1948-9) I was enrolled in an introductory economics course; I liked it, and went to me library to see what I could discover about the subject matter of professional economics. The first book I opened was the Foundations; amazing, economics was easy, it was just like textbook physics! I also subscribed to die Quarterly Journal of Economics, and one of the first issues had a paper by Hollis Chenery on Engineering Production Functions. So economics was also just like engineering. As I would later learn these were incredibly bad first impressions: economics was not like physics, where evidence mattered; economics pretended to be like physics, but die testing standards were far weaker, although Phil Mirowski has made a career out of taking physics-envy propositions more seriously than they deserve.

Paul also influenced my decision to go to Cambridge, and he will remember why-Paul remembers everything! While I was still living in Lawrence, Kansas, he and I had some correspondence on the "pure theory of consumers' behavior," dealing with integrability and all that, and an obscure early contribution by Antonelli. He sent me a copy of the original proofs for his article in Economica on the topic and I still have, and treasure it. I knew about such obscurities as the work of Antonelli, because I had developed a taste for the history of economic thought, thanks to my mentor, Dick Howey, at the University of Kansas, from whom I learned what deep scholarship was all about-learning all the tools you needed to solve problems as you encounter them; for Dick tiiat meant knowing mathematics, Italian, French, and German. It's called learning to learn, and they did not teach it then or now in the pretty school houses, but you can pick it up by watching a skilled scholar in action whatever his or her field. I also had seen that in spades at Caltech when I took freshman chemistry from Linus Pauling, a truly remarkable scholar and human being. Paul has had a lifelong interest in history of tiiought issues, and I admired it already in those days, although I am the last of a generation of economists that studied the history of its own thinking. Henceforth, economists were to be trained in getting things done for and to the world, or for/to whomever, and no one was interested any more in where the intellectual questions might have come from. But the style did not change: every paper started with a previous paper, almost never a problem of the world. That always bothered me and accounts for why I got interested in engineering and economics, in what would come to be called "experimental economics" and, for a while, in the ballistics of natural resource economics, and later in the methodology of science and why its constructs were all so wrong-I came to methodology not through the elegant philosophical papers, with the exception of Lakatos, but from doing experiments and thinking about what that experience had meant and how and why the discipline of testing had transformed the way I thought about economics.

Cambridge was a foreign personal experience for me-I went into the Wursthaus restaurant just off Harvard Square, circa 1952, and on the menu I found "Chili, Mexican Style (no potatoes or carrots)." Welcome to the New England cuisine-surely only some Yankee yahoo would put carrots and potatoes in chili! But the intellectual climate at Harvard and MIT was more familiar to someone from Caltech and the University of Kansas. Alvin Hansen's delightfully optimistic treatment of Keynesian economics at Harvard was much tempered by the dry wit of Gottfried Haberler, die sarcasm of Wassily Leontief (utility theory is good for teaching!), Guy Orcutt's deeply serious search for the messages hidden in all data, Alexander Gershenkron who lectured on 'ven Breetan vas ze voticshop of ze voild,' and a coterie of graduate students trying to make sense of it all for their own careers. When Fritz Machlup visited, you wondered how the two polite Austrians-he and Haberler-would resolve the issue of which one was to go through a door first. Schumpeter was no longer alive, but his ghost was lurking in me halls, with Haberler countering any macroeconomic claims that inflation ('ze monster' to Schumpeter, and Haberler would have agreed), if not too large, was good for the health, soul and spirit of the economy.

One of the best parts of being at Harvard was that you could go down river from Harvard Square, and sit in on classes at MIT, and I regularly attended Paul's classes. These were very lively, interactive, and small classes of maybe 8 students. Except for Ron Jones, I no longer remember any of my fellow students. Paul loved to dash breathlessly into die room, barehanded and bare pocketed, pick up a piece of chalk-this was the good old days of blackboards-and ask what the students wanted him to lecture about. He would get a couple or so responses and start talking and writing on one of these topics. A joke that circulated-I have no idea whether it actually happened-was that once, in one of these classes, Paul waltzed in, asked his question and there was only one suggestion. Furthermore, the set-up suggestion was esoteric enough to require a little more than an impromptu rendition from Paul's formidable memory. So Paul replied, "Well, I thought I would talk about..."

MIT had a much different feel man Harvard. The halls were light and airy, and if you passed a faculty member in the hall, he (no she's yet, I understand from a recent issue of Science that MIT is among the last to get more affirmatively active in this area, but Harvard and Larry Summers are off and running) would acknowledge or speak to you. Up river at the Littauer mausoleum the halls were dark, dingy and most if not all of the faculty would walk past you like you were a lamppost.

Paul had a fairly well deserved reputation for a sense of humor and generally MIT was better tiian Harvard in the laughs department. But trust me, no one could outdo Gottfried. Gottfried Haberler was the masterful seminar chairman whose hilarious introductions always guaranteed an audience, whether the speaker was local or a visitor like Jacob Marschak whose introduction was a gem. Harvard's Seymour Harris was a prolific publisher who supervised many PhDs in applied policy, money and macro topics-multipliers and all that. If he needed a graduate student for a new book he would post an advertisement on the first floor Bulletin Board of Littauer Center calling for some erstwhile lucky student to be funded by Harris on some AID, Bank or government grant program. Seymour was giving the seminar one day. The room was packed. Gottfried rose, walked to the podium, tall, poker faced like Buster Keaton, and gave his shortest and most memorable introduction. "Our speaker today is Professor Seymour Harris. You all know who Professor Harris is. Those of you who are not busy reading his many books and papers are busy writing them."

But Paul is hopelessly constructivist, and believes mightily that because "empirical realities of technology could vitiate Adam Smitii's comfortable views on good-thing Invisible Hands," (Colander, et al, p 311) this necessarily meant that it does so vitiate it. Could becomes is. One pencil and paper counter example and the private economy crashes and burns. Public goods and non convexities create challenging design problems, but they do not repeal the principle that information is decentralized and good solutions have to build on mis important Adam Smith principle.

It took a few dozen experiments back in the 1960s to beat out of me the nonsense that denies the proposition that people in all sorts of markets achieve ends that are not part of their intention. The human ego harbors the faith that if something works somebody has to have consciously made it possible by reason of reason. I discovered, and gradually struggled to escape my graduate education by coming to accept that under all sorts of market institutional rule system (bid/ask, sealed bid/offer, posted offer or bid) motivated humans converge to efficient equilibrium outcomes based entirely on decentralized dispersed private asymmetric information on value and cost. Hayek was dead right, but I did not get tiiere by reading him. What he lacked was a technology for trying mightily to falsify his incredibly good intuition, and learning from that process. Moreover, I learned that the subjects in my experiments have not a clue that this is what they have wrought; they deny that any model could explain their steady state outcomes, and deny that in those states they could not have done better for themselves. Their perceptions are all wrong because their dispersed information in fact defines simultaneous equations in prices and quantities that predict their convergence behavior, and by definition of equilibrium (Cournot invented it, Nash generalized it, so we call it Nash) no one person could have done better for him or herself given what every body else was doing. And you do not need a large number of buyers and sellers, a handful will dotwo is few and four is many as experimentalists like to say-although now institutions start to matter more, e.g. double auctions are better than posted offer prices when numbers are very small. No one has to be sophisticated, know any economics, understand supply and demand, or have any idea what equilibrium is. As in any real market the dispersed information that yields equilibrium is not, nor can it ever be, given to any one mind.

Some of you reading this will say to yourselves-without knowing anything about the extensive studies of the question by experimentaliststhat it must be because of the subjects! Not true. You do not have to make up the facts here. There are few results more robust to the choice of subjects than equilibrium convergence. The greatest sport of all is to do a double auction supply and demand market with game theorists. You pass out the dispersed private information on value and cost that defines the aggregate supply and demand; tell them how they can make money, and dependably one of the best of them will say: "How can we make an intelligent decision; we do not have enough information?" The answer: "Don't fret you will be able to handle it." And indeed they can, but what is missing is any curiosity about why they are able to handle it. David Hume taught us why, "When we leave our closet, and engage in the common affairs of life, (reason's) conclusions seem to vanish, like the phantoms of the night on the appearance of the morning; and 'tis difficult for us to retain even that conviction, which we had attained with difficulty ..." (Hume, 1985/1739, p 507). For the game theorists their closet analysis based on complete information is irrelevant and their presumed need for it vanishes like the phantoms of the night, as their autonomic brains take over, just like those of the sophomores, and both groups produce optimal outcomes without intending or knowing it, and have not the foggiest idea why. This is because most of the competence of the human brain is inaccessible to our self aware reasoning mind.

No wonder Reinhard Selten, who has so expertly worked on both sides of the closet door, will tell you flat out that "Game theory is for proving theorems not for playing games."

But the perceptions of the subjects are no more distorted than those of the theorists who from their closets have perpetuated the myth that equilibrium is only attainable if all agents have complete and perfect information on the "foreseen" conditions of supply and demand as Paul used to put it. The evidence is that complete information is neither necessary nor sufficient for equilibrium to be attained. I can give you economic environments in which complete information demonstrably slows, perturbs and interferes with equilibrium convergence. It's better for each to know only his/her private value and cost than for all to know everything.

Some of the new asymmetric information modelers (like jolting Joe Stiglitz) claim that the findings of information economics have enabled theory to escape its reliance on perfect information, and that this has had very negative consequences for the role of markets in economic betterment. Regrettably, the breadth of the claims made here greatly exceeds what can be supported by more soberly and cautiously expressed thoughts, and by the empirical results from dozens of experiments. This is just the rhetoric of science, not its substance.

The widely accepted and repeated proposition that equilibrium in a competitive market requires all agents to have complete information appears to derive originally from W. S. Jevons:

"A market, then, is theoretically perfect only when all traders have perfect knowledge of the conditions of supply and demand, and the consequent ratio of exchange (price) ... (Jevons, 1888/1871, p IV. 18).

On the page before he tells us his belief about the actual achievement of this state: "The theoretical conception of a perfect market is more or less completely carried out in practice. It is the work of brokers in any extensive market to organize exchange, so that every purchase shall be made with the most thorough acquaintance with the conditions of the trade. Each broker strives to gain the best knowledge of the conditions of supply and demand, and the earliest intimation of any change." (Jevons, 1988/1871, p IV.17)

Plainly this is a story telling assertion by Jevons, not a theorem; neither is it an empirical demonstration; and he has no idea what brokers know or do. It stands as a poorly articulated fallacy of constructivism. I am hard-put to think of a single conceptual "contribution" to economic thought that has done more damage-real and intellectual-to the way economists mis-mink about the world. This is because the assertion has been elevated to the status of a (folk) theorem. Moreover, the resulting frame of thought continued to infect the foundations of microeconomic theory after theorists like Akerlof, Spence and Stiglitz-as sooner or later somebody was bound to do-had turned to the task of attempting a formal analysis of information and equilibrium. Jevons' false Folk Theorem, together with the universal acceptance by economists that (1) welfare economics depended crucially on the validity of static competitive regimes, and that (2) competitive equilibrium outcomes did indeed require "perfect" or complete information, set the stage for a large industry of revisionist theoretical attacks on both the competitive model and how that model was undermined once you relaxed its claimed dependence on complete information. Spence, by the way, nowhere makes any exaggerated claims about what he was up to, and does not buy into the relevance of comparisons with what would prevail in a world of perfect information, because that is so obviously an irrelevant comparison. The rhetorical contrast among these asymmetric information theorists is pretty amazing.

There never was a theorem that I know of that stated that when everyone knows everything you get equilibrium, and when they do not you fail to get equilibrium. More than anything competition is a discovery process; it's not about making price equal to marginal cost.

But if you begin with the proposition discovered completely unintentionally by the experimentalists that complete information is not necessary for achieving efficient outcomes, and that more information can invite inefficient strategizing, what are the implications for the numerous results showing that equilibrium either does not exist, or it is inefficient, when there is asymmetric information on, say, product quality? Well, you are back to square one. There is no reason to believe in the truth value of any of these cases short of an extensive program of testing them. And without any satisfactory means of describing the dynamics of equilibrating processes-missing in both neoclassical and asymmetric information theory-the results of those tests are likely to yield more enigmatic surprises.

If human agents (like subjects) with private, and therefore inherently asymmetric information, can converge to equilibrium outcomes by dynamic processes using the standard message spaces of extant institutions that theorists can neither model or predict, why would you believe that the market failure theorems derived from asymmetric information theory apply to those agents? We are indeed back to "go," armed, I believe, with models that are still incomplete in dealing with equilibrating processes, as distinct from static equilibrium (symmetric or asymmetric) outcome specifications. We need to start over, but that requires an agonizing reappraisal that is not plainly in the interest of anyone trained in the Standard Social Science Model and its new extensions that deal in great depth with revised models of information and equilibrium, but leave untouched the treatment of ecological processes through which economic agents evolve (or fail to evolve) from their initial circumstances and knowledge to the equilibrium prediction derived from one of the new asymmetric information models. That is where all the action is, and where most of any new learning can be obtained.

It's not true, however, that no one tried. Leo Hurwicz was acutely aware of the Hayek problem, and brought powerful intellectual skills and many students and his student's students to the task of getting a fundamental theorem on Hayekian informational efficiency. Much insight followed, but the theorem proved illusive, and as I see it to get the early results you had to compromise optimality, incentive compatibility and die equilibrium tradition in ways that were not popular with the theory community. The bottom line I think is that the complete problem is just too hard. We don't have the right tools to understand how naive subjects can walk into a room with algorithmic brains that interact with the algorithmic rules of an institution selected by their forefathers and produce good results. The fact that computer simulations based on simple algorithms can demonstrate good results does not help to understand how these algorithms evolve and change in response to experience.

Many important competitive issues are raised when a subset of agents have asymmetric advance information on product quality and value characteristics. Constructivist equilibrium analysis shows that such conditions can generate market failure in the sense that the allocations are less than fully efficient. Some of these problems, however, arise because the analysis is inadequate in examining both sides of the market, and die full implications of the information content of prices. Experiments have established that constructivist inefficiency may be alleviated by one of several ecologically rational response mechanisms: competition among sellers for reputations, quality (brand) signaling, product warranties, and the aggregation of private asymmetric information into public price patterns that self correct the alleged problems.

Consequently, much of the experimental evidence, while supporting the idea that asymmetry of information has potential effects like those in the new standard models, reveals that there are countervailing responses that mitigate the proposition that asymmetric information necessarily implies market failure. Models showing tiiat there can be failure relative to perfect information regimes are not the same thing as showing that observed markets have failed in the environments to which they have become adapted. Special cases where markets (or their models) perform less than ideally do not add up to a general market failure theorem. In such "policy" analyses several things may be omitted from the formal analysis: (1) if someone has asymmetric information yielding an abnormal return, what prevents self correcting entry by competitors to exploit the advantage? Even if the entry stops short of yielding full efficiency, what is the alternative? (2) Do agents with asymmetric information reveal their position of advantage allowing others (buyers or sellers) to correct for it? (3) Analysis of the costs and benefits of intervention to correct the efficiency loss from asymmetric information; if it costs more to correct the inefficiency than it is worth, then it is not inefficient, except relative to the imagined world of perfect information; (4) Are there extra-market sources of information about reputations, even where there is no repeat trade, that serves a correctional purpose?

Gossip, in which neighbors voluntarily report to each other their experience with various physicians or repairmen, provides a powerful means by which consumers pool their information and seek out the most reliable service providers, particularly where repeat trade is non existent. Local gossip rapidly transmits experiential information on who are the best eye surgeons, the most reliable automobile repairmen, and, while they last, who are the charlatans. You only have cataract surgery once, but almost anyone in town was able to tell me that Dr. Katz was the best. And Buck, my auto repair genius is so good that he only takes new customers by referral! This is the real world of people outside the ivy covered walls who have been making the world work since long before our ancestors walked out of Africa. To my knowledge "cheap talk" among buyers as a laboratory treatment variable in these environments has not been explored. But long ago socioeconomic institutions evolved the form that they did because information is not complete. What information economics tells you-and tiiat is its really great contribution and much deserved recognition-is why there was a problem that emergent processes needed to solve.

Of course, as you would expect, experimentalists have not found that the unintended ends achieved by economic agents are always optimal. In flow supply and demand markets like those in established repetitive markets for goods and services, by and large this is true, but there are pathological environments that can cause problems for some institutions. An example would be where N is large and there is an excess supply of only one unit as in Edgeworth's market for servants but with N households and N + 1 servants. Another huge glaring class of replicable examples is in asset trading. At first Colin Camerer thought it must be an "Arizona phenomenon," but his Penn students did it too. And at Caltech they did it in spite of Charlie Plott leaning heavily on the instructional wind. Subjects, whether undergraduates, graduates, corporation executives or over the counter traders, produce big bubbles and crashes relative to fundamental value even when they "know," that is, have common information on, fundamental dividend value. Actually, they do eventually converge but it is not until the third time back in 2.5 hour sessions of trading with the same group. Common information is not sufficient to produce common expectations; the latter are achieved only through an experiential process and it takes a long time. Given common information, people come to have common expectations; but they do not get mere by thinking about it.

Where did we go wrong with all this perfect information ballyhoo? My view is that our forerunners were so enamored by the subjective utility breakthrough, the solution to Adam Smith's little diamonds and water paradox, and with the general equilibrium equations of the 1870s that we threw out the key observations and wisdom of Hume, Smith and Ferguson that the social economy tiiey observed was loaded with order that no one had consciously designed. They passed that wheredoes-it-all-come-from puzzle on to us, after articulating and documenting it very well, and challenged us to try and model it. Instead we denied it, and spent the next 90 odd years after me 1870s getting the conditions for equilibrium existence right because that was an easier problem, and by imagining that it surely must require complete information. Then we dropped the whole exercise and turned to the most partial, partial equilibrium world you could possibly imagine-game theory-and continued to work with complete information examples because they were tractable. Why look for lost keys where tiiere is so little light?

But I think the stage is now set for reappraisal, and I am optimistic about the 21st century, and its ability to build on what we know, and to forget what we know that isn't true. We now know that real people-even game theorists-can do it; they can find a multiple equation equilibrium by fumbling, stumbling repetitions.

Moreover, people in two person anonymous interactions can use trust and reciprocity to achieve cooperative outcomes in single play extensive form games. Since in many parameterizations they make more money than if they followed dominant strategy choices you cannot say they are irrational. Reciprocity is a norm that defines rights endogenously by consent-likely the ultimate origins of property rights. As Hume put it "The rules of morality ... are not conclusions of our reason." (Hume 1985/1739, p. 509)

But returning again to Paul, I love him; he just got off base here and there like all the rest of us mere humans. He was an incredibly positive influence on me, and a host of others, and I am eternally grateful. Thanks, Paul, and the best to you.

REFERENCE

References

Colander, David, Richard Holt and J. Barkley Rosser, Jr. (2004). The Changing Face of Economics. Ann Arbor: University of Michigan Press.

Jevons, W. S. (1888/1871). The Theory of Political Economy. London: Macmillan, 3rd edition; Online, Library of Economics and Liberty.

Hume, David (1985/1739). A Treatise on Human Nature. London: Penguin Books