- Do prices increase when demand increases?
- How many traders are required for a market to be “competitive”?
- Does a sense of fairness play a role in bargaining? If so, can people be paid to be less fair?
- Is it more efficient to auction complementary products, such as pairs of vases, sequentially or simultaneously?
Questions such as these are difficult to answer completely using field data. Often it is difficult to find naturally-occurring environments simple enough to cleanly test economic predictions, and often information not easily obtained in the field is required to draw conclusions. However, these questions are ideally suited to economic experimentation.
Economic experiments ask human subjects respond to incentives in a controlled laboratory environment which represents the essential economic features of a theoretical or naturally-occurring economic problem of interest. They are distinguished from other social science laboratory techniques, such as those used in psychology, by monetary incentives which depend on decisions made by experiment participants; those who respond better to the incentives provided are paid more money, usually in cash, for their participation.
In some experiments, it is easy to implement monetary incentives. In bargaining problems, for instance, the experimenter can simply provide a cash prize over which subjects must bargain, according to certain rules.
However, most experimental incentives are structured using a technique called induced value. Each subject is given redemption schedules, indicating how much real money the experimenter will give him or her for each unit of laboratory good he or she holds at the end of the experiment.
For instance, a subject with one unit of Y may be offered $1.50 for the first unit of X; $1.00 for the second unit of X; and $0.75 for the first unit of Y. This subject may trade her Y for an X with another subject, who values Y more highly than X, so that each will be paid more at the end of the experiment. These values are “induced” because they are given to subjects who value the goods only through the money the experimenter is willing to pay.
These performance-dependent incentives link results in the lab to the situations being modeled outside the lab. Experimentalists argue that laboratory tests provide information about naturally-occuring economic behavior. because economic behavior is governed by basic principles which apply both inside and outside the lab. These axioms of rationality are the basis of economic theory, which makes no distinction between, for instance, maximizing utility inside and outside the laboratory. The cash incentives ensure that better decisions inside the laboratory result in more utility outside the laboratory (money earned in the experiment can be spent outside the laboratory). The external validity of any particular result depends on how well the experimental design captures the incentives present in the naturally-occuring situation being modeled.
Advantages of the Experimental Approach
Laboratory analysis can complement traditional field data modeling by testing economic theories in several ways. First, laboratory experimentation offers a degree of control which allows economists to generate tests of alternative policies at low cost, or to give theories their “best chance” by testing them in environments which exactly satisfy their assumptions. Second, laboratory analysis facilitates replication; it is often difficult to find an independent field sample with which to accomplish this critical step in the scientific method. Finally, experimentation allows economists to have information they could not know in the field. For instance, many economic theories suggest that people respond to their beliefs in particular ways.
Beliefs are difficult or impossible to measure in the field, but they can be naturally integrated into an experimental design.
Types of Experiments
There are many ways to categorize experiments, but one of the more useful is by which models they are attempting to test. (The taxonomy also has the advantage of distinguishing experiments which test models from those which do not.) Most experiments can be characterized by one or more of these objectives:
- Testing theoretical predictions. Economic theories make predictions which can be tested in the laboratory. For instance, general equilibrium theory predicts market prices at the intersection of supply and demand. Experiments have shown this prediction is accurate with a wide variety of trading institutions.
- Testing robustness of theories. Economic theories often include very strong assumptions, and experiments can test the sensitivity of their predictions to weakening of those assumptions. For instance, general equilibrium theory predicts market prices at the intersection of supply and demand when there are a very large number of agents. Experiments have determined the “very large” number of agents at which competitive outcomes reliably obtain is three.
- Testing assumptions. Rather than testing the predictions of a theory, models are sometimes examined by testing their assumptions. This technique is frequently used by behavioral economists who seek to replace standard rationality assumptions with more descriptively accurate, yet mathematically tractable, models.
- Identifying stylized facts. Because replication is straightforward in the laboratory, experiments are often used to identify patterns in behavior which may or may not be consistent with theory. For instance, in the dictator game (in which two people are anonymously paired and one, the dictator, is asked to allocate $10 between the two) dictators often give money to their paired partner, although they could have kept the money for themselves. New models are being developed to integrate such stylized facts, and the real economic norms they imply, into economic theory.
- Comparing institutional designs. Theory does not always provide guidance when deciding among institutions. The alternative institutions, or policies, can be implemented in laboratory and the outcomes compared on the basis of efficiency or other desiderata. This technique has been used to make design decisions in the auction the FCC uses to sell spectrum permits and the auctions used to sell transferable pollution permits by the EPA and by the South Coast Air Quality Management Board in Southern California.
Conspicuously absent from this list of applications is measurement and calibration. Experiments have not, in general, been used to successfully measure the level of efficiency attained in natural applications of institutions, or to determine how much people value goods or ideals. Obstacles to these kinds of research include focal and framing considerations which are difficult to translate from the naturally-occurring situation to the lab. Furthermore, the “economic principles” argument for external validity does not extend well to magnitudes. However, as economists continue to grapple with ways to measure values, especially non market values, laboratory techniques may be an important part of new methodologies.
Over 2000 papers have been published in economics journals using experimental methods, most since 1990.
- Davis, Douglas and Charles Holt. 1993. “Experimental Economics“, Princeton University Press.This book provides an excellent introduction to experimental methodology and to a basic set of experimental results.
- Kagel, John and Alvin Roth, ed. 1995. “Handbook of Experimental Economics“, Princeton University Press.This book is a collection of detailed surveys of major areas of research.
If you are interested in participating in an experiment, click here.