by Filip Palda
The flip side of the cheating problem in socialism is the lying or “adverse selection” problem in capitalism. If potential firm managers are either good or bad, but telling them apart is difficult, bad prospects will lie to become a part of the firm. According to Roger Myerson, “a potential advantage of socialism … is that a socialist state’s monopoly of capital can facilitate honest communication, as bad managers cannot gain from imitating good managers if neither type gets any profits from entrepreneurial management” (2008, 600). He argues that, “society does not need to pay anything more than the manager’s normal cost of time, and the manager has no incentive to lie about his type to get the project funded because he gets no special benefit from managing it” (2008, 599).
Myerson’s view of the socialist calculation debate can be disputed. Makowski and Ostroy (2001) use a subtle argument about who gets what in capitalism to suggest that perfect competition with “full appropriation” of the social contribution everyone’s production makes is an incentive compatible system.
Getting deeper into this debate is not our objective. The important thing to understand is that the old questions of how people use information to coordinate their actions is very much alive in present research on game theory and mechanism design. Coordination is really where everything “is at” in game theory, just as this is the preoccupation of classical economics, where strategic interactions between individuals do not figure.
The apprenticeship of game theory
GAME THEORY IS an abstract vision of how people formulate strategies against others contesting some valuable resource. It fits into the logic of economics because it postulates a rational individual maximizing some objective subject to constraints. Yet it differs from most other branches of economics in that the constraints are not material, but mental. Material constraints do not allow for strategic interaction. As discussed, the consumer has no control over prices or income. These are givens that emerge from the competitive jostling of masses of people looking to buy or sell. In their numbers they are either anonymous to each other, or have their behavior prescribed by the rules governing the exchange of property. In the interplay of supply and demand, no one exerts control over how resources are to be divided.
The situation changes when people confront each other over control of a resource and have either no interest in striking a bargain with each other, or no reliable way of entering into one. In that case, personal intentions rather than impersonal parameters such as prices become the constraint against which every person must make his or her calculations of how to make the best of the situation, though of course behind intentions there may be a backing force. The subject matter of game theory is thus hostility and antagonism, though elements of common interest may sometimes be present. Antagonism is rife because either the rules governing the transfer of property are not well defined, or because the owners of property interact in a non-competitive setting where buyers do not feel the pressure of other potential bidders to force them to reveal their true valuation of a product.
Hostility is present because the resource in question is up for grabs. For some thinkers, the view that everyone is out to “get you” is a paranoid and unrealistic stance that relegates game theory to irrelevancy. They believe that through trial and error the masses strive to replace unpredictable and potentially catastrophic game playing between individuals with rules. An ordered society allows people to disengage from the fabricated constraints of hostility and focus their attention on dealing with the immutable constraints imposed by limited physical resources. This critique of game theory misses the point that a theory which explains behavior that never takes place is not useless. The US and Russia did not drop atomic bombs on each other but the military expense of that era was a reaction to game theoretic forces. In the same vein, even if a society finds a fix for the problems that lead to gaming, the structure of the games in question allow us to understand whether these fixes are the best possible.
Becoming a game theorist is a life-long quest, but mastering its essentials is within reach of almost anyone. To figure out how people will interact in a strategic encounter only three pieces of information are needed. One must know what the prize in question is, how the game is played, and how a resolution, or equilibrium, is reached. All critiques of game theory must center on questioning the validity of any one of the assumptions we make about these three necessary pieces. David Kreps’s simple yet brilliant exposition of game theory exposes these three pillars to a critique in Chapter 5 of his 1990 book.
Starting from this base you can obviously come up with any sort of game your imagination may conceive, but game theorists have come up with only a handful that they believe are at the basis of human strategic interaction. In this vein, Myerson explained that, “… the task for economic theorists in the generations after Nash has been to identify the game models that yield the most useful insights into economic problems. The ultimate goal of this work will be to build a canon of some dozens of game models, such that a student who has worked through the analysis of these canonical examples should be well prepared to understand the subtleties of competitive forces in the widest variety of real social situations” (1999, 1080). These “canonical” games are really just stories told with a high degree of precision and consistency.
The consistency and logic of game theory are its strengths. Yet by setting itself the lofty task of understanding interactions between individuals, it runs up against the “curse of dimensionality”. Originally used to express problems in dynamic programming, this term now describes situations in which a wealth of categorizing and assumptions leads to an explosion of possible outcomes such that no solution or equilibrium is evident. Game theory’s most basic lesson may be that as we shift focus from crowds (supply and demand) to individuals, general insights cede their place to a proliferation of stories and possible outcomes known as “multiple equilibria.’
References
Alchian, Armen A. and Harold Demsetz. 1972. “Production, Information Costs and Economic Organization.” American Economic Review, volume 62: 777-785.
d’Aspremont, Claude and Louis-André Gérard-Varet. 1995. “Collective choice mechanisms and Individual Incentives.” First published in French as “Théorie des jeux et analyse économique 50 ans après” in Revue d’Economie Politique (Special Issue), 1995, volume 4: 529–733. Published in English in Christian Schmidt, ed., 2002. Game Theory and Economic Analysis: A Quiet Revolution in Economics. Routledge. The edition cited is the 2004 e-library PDF published by Taylor & Francis e-Library and as such no page numbers can be referred to in the present text.
Groves, Theodore and John Ledyard. 1977. “Optimal Allocation of Public Goods: A Solution to the ‘Free Rider’ Problem.” Econometrica, volume 45: 783-809.
Harsanyi, John, 1967. “Games with Incomplete Information Played by ‘Bayesian’ Players: Part I. The Basic Model.” Management Science, volume 3: 159-182.
Harsanyi, John C. 1995. “Games with Incomplete Information.” American Economic Review, volume 85: 291-303.
Hayek, Friedrich A. 1945. “The Use of Knowledge in Society.” American Economic Review, volume 35: 519-530.
Kreps, David M. 1990. Game Theory and Economic Modelling. Oxford University Press.
Makowski, Louis and Joseph M. Ostroy. 1992. “General Equilibrium and Market Socialism: Clarifying the Logic of Competitive Markets.” UCLA Economics Working Papers 67.
Makowski, Louis and Joseph M. Ostroy. 2001. “Perfect Competition and the Creativity of the Market.” Journal of Economic Literature, volume 39: 479-535.
Maskin, Eric S. 2008. “Mechanism Design: How to Implement Social Goals.” American Economic Review, volume 98: 567-576.
Myerson, Roger B. 1999. “Nash Equilibrium and the History of the Economic Theory.” Journal of Economic Literature, volume 37: 1067-1082.
Myerson, Roger B. 2008. “Perspectives on Mechanism Design in Economic Theory.” American Economic Review, volume 98: 586-603.
Myer
son, Roger B. 2009. “Learning from Schelling’s Strategy of Conflict.” Journal of Economic Literature, volume 47: 1109-1125.
Myerson, Roger B. and Mark A. Satterthwaite, Mark. 1983. “Efficient Mechanisms for Bilateral Trading.” Journal of Economic Theory, volume 54: 265–81.
Nash, John F., Jr. 1950. “Equilibrium Points in N-Person Games.” Proceedings of the National Academy of Sciences of the United States of America, volume 36: 48-49.
Nash, John F., Jr. 1951. “Noncooperative Games.” Annals of Mathematics, Second Series, volume. 54: 286-295.
Schelling, Thomas C. 1960. The Strategy of Conflict. Harvard University Press.
Selten, Reinhard. 1975. “Reexamination of the Perfectness Concept for Equilibrium Points in Extensive Games.” International Journal of Game Theory, volume 4: 25-55.
Spence, Michael. 1973. “Job Market Signalling.” Quarterly Journal of Economics, volume 87: 355-374.
Tideman, T. Nicolaus and Gordon Tullock. 1976. “A New and Superior Process for Making Social Choices.” Journal of Political Economy, volume 84: 1145-1159.
Vickrey, William. 1961. “Counterspeculation, Auctions, and Competitive Sealed Tenders.” Journal of Finance, volume 16: 8-37.
Von Neumann, John. 1928/1959: “On the Theory of Games of Strategy.” English translation of von Neumann’s 1928 German article in A.W. Tucker and R.D. Luce, eds. 1959, Contributions to the Theory of Games. Princeton University Press, 13-42.
Von Neumann, John and Oskar Morgenstern. 1953. The Theory of Games and Economic Behavior, Third edition. Princeton University Press.
CONTROL 8
THERE EXIST SPECIALIZED BUILDINGS THAT float in shallow pools of mercury. The mercury absorbs tremors from the earth so that delicate scientific instruments within the building can conduct experiments without interference. The measurements sought are ones that might show a link between a stimulus and a response. If you have set up a machine that vibrates a chemical mixture with a certain frequency and amplitude in order to see how this affects the chemical reactions within the mixture, you do not want vibrations from the earth added. If vibrations from the earth can subtly join with vibrations from the machine, you will not know what really caused the strength of the chemical reaction. Was it the machine, was it the earth, or was it both? By removing possible earth tremors from the stimulus part of the experiment, the mercury pool allows the scientist to establish a clear link between a cause and an effect.
And that is what science is about. Science is not about people in lab coats, peering through thick-rimmed spectacles and scribbling on clipboards. It is about separating the signal from the noise, the wheat from the chaff, so that clear linkages can be established. You create an area—a laboratory—where you control the environment so that no vibration, or noise, or electric shock, or speck of dust can interfere in the link you are trying to measure between a cause and an effect. The key word is control.
Along with physicists and other scientists, economists strive to show a link between cause and effect. Was foreign aid to Senegal responsible for the improvement in the farm yields of poor villages? Is the low salary of minority workers in a company the result of racial discrimination? Does the minimum wage hurt the employment opportunities of the young? These are questions about cause and effect.
Yet as you might have noticed from the examples, human events do not unfold in a controlled laboratory setting. In a laboratory all other confounding influences can be eliminated by controlling the environment in which the experiment takes place.
The economist enjoys no such control. Nature is the laboratory and it does not care about concepts of control. The economist must accept that perhaps it was fine weather instead of foreign aid that improved the yield on Senegalese farms. Perhaps the company pays minorities less because they have different educational qualifications from non-minorities. Perhaps youth employment fell not because of a rise in the minimum wage but because the economy weakened. The “perhapses” gather, making it harder to see how one thing is linked to another. What is an economist without a laboratory to do?
The Dark Ages of econometrics
ALTHOUGH THEY LACK a laboratory which would allow them to exclude outside interference from causal experiments, what economists do have is an abundance of imagination with which to analyze the numbers the economic environment generates. They might not be in control of the “data generating process”, but they certainly are in control of how they can interpret the data. That is one freedom experimental physicists rarely have. In physics, once the laboratory experiment is set, the numbers generated generally speak for themselves.
An experiment allows for only one interpretation of events. In contrast, for economists, interpretation was until recently the main method they had of making sense of the data. Nobel Prize winner Christopher Sims explains that, “Because economics is not an experimental science, economists face difficult problems of inference. The same data generally are subject to multiple interpretations. It is not that we learn nothing from data, but that we have at best the ability to use data to narrow the range of substantive disagreement. We are always combining the objective information in the data with judgment, opinion and/or prejudice to reach conclusions” (2010, 60). To understand what this means, we have to go back to the 1930s and the dawn of “model building” in economics.
A “model” in economics is some simple statement of the relation between important quantities, also known as variables. During the 1930s, John Maynard Keynes developed a theoretical model of the whole economy. He connected all the important macroeconomic dots such as national income, interest rates, investment, and inflation in his 1936 book The General Theory of Employment, Interest, and Money. A year later in 1937, his protégé John Hicks worked Keynes’ verbal descriptions into mathematical equations in which consumption by everyone depended on national income, investment on interest rates, and vice versa, and so on. There were few real-world numbers in The General Theory. It was mostly, well, theory.
At the same time that Keynes was building his model, Dutch economist, physicist, and statistician, Jan Tinbergen, was confronting the problem of how to precisely measure the relation between these variables by using real-world data and their evolution through time, so-called “time-series”. His problem was that to relate, for example, investment to interest rates, it was not enough to simply see how they moved together through time and claim he had found a causal link. Other factors could be influencing investment, such as investor confidence in the future, or more technically, the anticipated marginal value product of capital. Randomness could also contribute to the final effect. You had to account for these other forces or you might end up thinking you had found a causal link where none existed. Accounting for these forces meant isolating each of their individual effects on investment so that you did not attribute to interest rates an effect that was really due to anticipations of returns to investment.
The method for isolating these effects and distinguishing them from random forces that could influence the dependent variable is a tool called linear regression. Think of linear regression as a very directed means of searching through a spreadsheet or database for the independent effect that each of many possible causal or “independent” variables has on a single target or “dependent” variable. The types of effect to look for are not left to the discretion of the computer, but rather are specified by the researcher. If you think that the overall level of investment in the economy should move in some constant but unknown relationship to interest rates and other possible causal variables, regression will only search for proportional, or “linear” links in the data.
The word “regression” was coined by British scientist Francis Galton in the 19th century. He wanted to see the relation between the height of parents and the height of their children. Using the linear statistical technique which he called regression he found that taller than average parents tended to have smaller children than they, and small parents tended to have taller children. In other words, chil
d height “regressed” to the mean of the population, which is a good thing, because otherwise the world would be covered with giants and ant people.
A physical analogy that might help explain what regression does, concerns the famous chord that opens the Beatles’ song, A Hard Day’s Night. For decades people thought it was a combination of bass and guitar. Then Canadian mathematician Keith Devlin used Fourier analysis to decompose the waves that combined to make up the chord, while at the same time filtering any randomness or “noise” from the recording that could obscure his search. The logic is similar to regression analysis. By restricting his search using the method of Fourier analysis to sift the “data” of sound waves that make up A Hard Day’s Night, Devlin was able to discover that a piano was also striking notes and contributing to the final sound.
Regression is a powerful statistical tool for isolating the effect of all causal variables on some target variable of interest but there is a catch. Regression is an apt tool provided all causal variables are fed into the regression formula. If you are missing important variables that move with, or are “correlated” with the ones you include, then the effect of these “omitted variables” may get folded into the ones included, thereby giving the latter a possibly false importance in the causality one is trying to establish.