Economics Lab

Alessandra Cassar, Dan Friedman

• Publisher: Routledge
• Year: 2004
• ISBN:
  9781134347728
• Edition: 1
• Topic: Business
• Subtopic: Business General
• Index: Business

Laboratory experiments with human subjects now provide crucial data in most fields of economics and there has been a tremendous upsurge in interest in this relatively new field of economics. This textbook introduces the student to the world of experimental economics. Contributors including Reinhard Selten and Axel Leijonhufvud that s

Part I
Introductions

An intensive course in experimental economics

Daniel Friedman

For two weeks, 18–29 June 2001, twenty economics graduate students from around the world gathered to learn how to run economics experiments. Despite the distraction of a stunning setting—a cliff-top hotel overlooking Trento and the Adige valley—the students made remarkable progress. Students sorted themselves into eight groups and on the last day, each group presented the results of an original pilot experiment. After returning home, many students continue to run laboratory experiments and to show others how to do it.

The structure of the summer school contributed to its success. Morning lectures began with an overview of the history and purposes of economics experiments, and then alternated between presentation of laboratory methods and surveys of applications. Methods lectures covered experimental control, emphasizing induced value theory; design, including the proper use of randomization and disposition of focus and nuisance variables; data analysis, including qualitative summaries as well as hypothesis tests; issues concerning human subjects and laboratory facilities; and project management. Applications topics included the mysterious efficiency of double auction markets; the successes and failures of institutional design, including spectrum auctions and California electricity markets; the successes and failures of game theory and learning theory in predicting bargaining behavior; and the promises and pitfalls of behavioral economics.

Afternoons usually featured guest lecturers. Distinguished guest lecturer and 1994 Nobel laureate Reinhard Selten lectured on his new theory of impulse balance equilibrium and laboratory applications, and also lectured on his recent theory of imitation equilibrium and applications to oligopoly. Guest lecturers Massimo Egidi, Steffen Huck, and Rosemarie Nagel surveyed laboratory discoveries in their fields: social learning, oligopoly, and coordination games. Program Director Axel Leijonhufvud lectured briefly on adaptive economic processes, and Peter Howitt gave a talk on themes for the next summer school that inspired one student project a year ahead of schedule!

Most important, the student groups met several times a day to hammer out a research question, design an experiment, and run a pilot session. The groups had scheduled afternoon meetings as well as impromptu meetings over meals, during coffee breaks, and late at night, in balconies, lobbies, and eventually in the CEEL facilities at the University of Trento. The groups worked with Teaching Assistant Alessandra Cassar, and often consulted with the summer school director and guest lecturers. Our students’ areas of applications went beyond what is covered in this book, to include two projects inspired by macroeconomic questions, and one in public good. Altogether, it was an intense learning experience for everyone!

This volume is intended to capture the essence of that summer school and to make it available to economists everywhere. We have written up most of the lectures, and edited the student project papers. We have tried not to homogenize everything as in a normal textbook, however. An intensive course works better when there is more than one voice, and we have tried to preserve the informal flavor of lobby discussions by sprinkling the text with comments in boxes.

Several monographs and textbooks on experimental economics appeared in the early and mid-1990s; we draw on and acknowledge these excellent books in subsequent chapters. The present volume makes four sorts of contributions:

• surveys of applications that have progressed rapidly in the last few years;
• streamlined and unified presentation of methods;
• original material by the distinguished guest lecturer and other contributors; and
• seven examples of early project development by our student groups.

This volume will serve as a helpful reference book for experimental economists, but it is primarily intended as a self-contained introduction to economists who want to develop a laboratory experiment but are not sure how. It can serve as a primary or secondary text in a formal course, or as the backbone of a do-it-yourself course.

1 The Trento Summer School
Adaptive economic dynamics

Axel Leijonhufvud

This Summer School in Experimental Economics is the second in a series. The first, on Computable Economics, was directed by “Vela” Velupillai. Next year, the third one, on Adaptive Economic Processes, will be run by Peter Howirt. And we hope to go on to Behavioral and Institutional Economics, for example. It may not be obvious what they have in common.

They are all part of our ongoing program in Adaptive Economic Dynamics—as we call it “for want of a better name.” Although we—and some other colleagues—have made common cause in these efforts, the chances are that no two of us would explain what we are about in exactly the same way. What follows, therefore, is my own perspective on the matter.

The economic theory of recent decades has been built on the basis of the optimality of individual decisions and the equilibrium of markets. This “neoclassical” economics is often criticized, but it has many achievements to its credit. Indeed, it embodies most of what economists know and the tools of what they know how to do. If you are to become an economist you had better learn it!

Yet, neoclassical economics is the subject of constant criticisms from within and from without. But the notion that one might somehow abandon it, in favor of one or another alternative, founders on the enormity of the prospective cognitive loss. Those “schools” that have defined themselves largely in opposition to neoclassical economics have remained marginal.

We had better accept, therefore, that for now and for the foreseeable future, neoclassical economics is the core of our subject. Instead of looking for an alternative theory to replace it, we should try to imagine an economic theory that might transcend its limitations. Easier said than done! To get a start on it, it may help to compare how optimality and equilibrium are understood in modern theory with how they were understood in neoclassical economics many decades ago.

The architecture of modernity: choice, optimization, equilibrium

A brief summary of how a modern neoclassical model is built up may run as follows:

• All behavior is conceptualized as choice.
• Choice is formalized as constrained optimization.
• The solution to a choice-problem is a plan.
• If this plan is to explain observed behavior:
  • the agent must know his opportunity set in all relevant dimensions;
  • this information must be objectively true (as observed by the economist);
  • the agent must be able to calculate the optimum.
• When this behavior description is applied to all agents, it follows that the system as a whole must be in equilibrium.¹ If observed actions are to be interpreted as realizations of optimal plans, the state of the system must be such that all plans are consistent with one another.
• In modern general equilibrium theory, this conception is transposed into a temporal framework where opportunity sets have to be defined over all future periods. Uncertainty about the future can be represented only in the form of objectively knowable stochastic distributions. If at any given date, the probability distribution has to be defined over k possible states for the next period, the dimension of an agent’s opportunity set will be: n goods×t periods×k^t contingencies, where n and k are arbitrarily large and t is usually taken to be infinite.
• In modern macroeconomics based on intertemporal general equilibrium (IGE) theory, the economy is represented as following an intertemporal equilibrium time-path. Such a trajectory is basically ballistic, not adaptive. There is no sequencing of decisions. The information required for each individual optimization problem to have a solution includes the equilibrium prices for all future (contingent) markets, which means, in effect, that everyone’s choices have to be reconciled before anyone’s choice can be made.
• In modern theory, the escape from this logical impasse is sought in the postulate of rational expectations. Past experience in a closed system of stationary stochastic processes enable agents to forecast the required future prices.
• An equilibrium time-path is a sequence of states where no one learns anything they did not know to begin with.² The construction, on the other hand, raises the question of how these rational expectations were learned once upon a time. This has also become a front-line question in recent years given greater urgency by the finding that IGE models very often have multiple equilibria.

Complaints, complaints, …

A list of the more common complaints directed against constructions of this kind would include the following:

• the conception of “rationality” attributed to the individual agent in standard choice theory;
• the treatment of firms and other organizations (including governments) as if they were individual decision-makers;
• “situational determinism,” the practice of assuming that individuals or firms are always perfectly adapted to their external environment, so that inquiry into internal structure or functioning becomes otiose (cf. Latsis, 1972);
• the interpretation of the economy’s motion as always in equilibrium;
• the treatment of time as simply the (n+1)th dimension of the commodity space;
• the treatment of uncertainty as probabilistic and based on stable underlying frequency distributions;
• the lack of room for fundamental novelty—innovation, emergence, evolution;
• the explanation of institutions (money, firms) as market imperfections or market failures; and
• the theory’s isolation from neighboring disciplines (sociology, cognitive psychology, etc.).

All the properties of the model that were stressed in the brief summary above, and the corresponding complaints, stem directly from the commitment to constrained optimization as the (exclusive) way to represent how people make decisions.

An older tradition

In the older neoclassical literature, the optimality conditions for an individual agent were commonly understood as a state that the agent would attain by some entirely unspecified or at best sketchily described process of trial and error. Similarly, equilibria were understood as rest states of processes of market interaction. Thus, both individual optima and collective equilibria were understood as point attractors of dynamical systems. Static theory dealt with what economists thought they knew about the properties of these attractors. Applied economic theory was largely couched in terms of comparative statics, all of which rested on the presumption of underlying adaptive dynamics that would carry the system from a historically given state to a new point attractor.

Statics comprised almost all of formalized economic theory. Economists knew little in substance about the dynamics of either individual adaptation or market interaction and the mathematics of such processes were on the whole beyond what they (or not so long ago, anyone) could do. Adaptive dynamics was the unfinished business of neoclassical theory. And so it largely remains. Intertemporal equilibrium is a generalization of the earlier statics that does not tackle the dynamic issues.

Two traditions

Elsewhere (Leijonhufvud, 1998), I have summarized the contrasts between these two ways of apprehending the core of economic theory in the form of a table, which is convenient to reproduce here. I have labeled the two traditions “classical” as opposed to “modern” in order to emphasize that the focus on process as opposed to the properties of optima and equilibria has its roots in the “magnificent dynamics” of the British Classical school.

In the “classic” camp, I would put not only Ricardo and Marx, but also Marshall and Keynes. Examples of “moderns” would be Arrow and Debreu and later Lucas, Sargent, and Prescott. The great teachers of my own generation, Hicks and Samuelson, would move back and forth between the two as the problems they dealt with would dictate.

see table

Marshall’s laws of motion

Let us take Marshall as an example of the road not taken. The core of Marshall, I want to argue, is what we today would call an agent-based model:

• Marshall started from individual demand-price and supply-price schedules, p^d(q) and p^s(q)—not from demand and supply functions, q^d(p) and q^s(p), as Walras did. That is why he drew his diagrams (correctly) with quantity on the horizontal, price on the vertical axis, as we still do (incorrectly) today.
• These demand- or supply-price functions are not based on underlying optimization experiments. The demand-price is obviously not the “optimal” price that the consumer would pay for a given quantity. So, Marshall does not build from maximization.
• Instead, the demand- and supply-price schedules give rise to simple decisionrules that become the “laws of motion” of the agents:
  • For the consumer: if demand-price exceeds market-price, increase consumption; in the opposite case, cut back.
  • For the producer: if supply-price exceeds the price realized,³ reduce output; in the opposite case, expand.
    And we should imagine a similar rule for the price setter.
  • For the middleman: if inventory turnover slows down, reduce prices to both customer...

Table of contents

• Cover Page
• Economics Lab
• Title Page
• Copyright Page
• Illustrations
• Contributors
• Acknowledgments
• Part I Introductions
• Part II Laboratory methods
• Part III Applications
• Part IV Student projects