Using laboratory experiments in public economics.

"Argument is conclusive ... but ... it does not remove doubt, so that the mind may rest in the sure knowledge of the truth, unless it finds it by the method of experiment. For if any man who never saw fire proved by satisfactory arguments that fire burns, his hearer's mind would never be satisfied, nor would he avoid the fire until he put his hand in it that he might learn by experiment what argument taught."

--Roger Bacon (1928)

INTRODUCTION

Like other sciences, economics is a quest for truth, based on the development of theory and on the ability of that theory to explain observed activities. However, unlike many other sciences, especially the natural sciences, economics faces substantial difficulties in empirically testing the predictive power of its theories using data from the naturally occurring world. Given the dizzying array and complexity of forces that operate in the market systems studied by economists, economists can never be quite certain that they are "holding constant" the many factors that may be driving individual choices, so that they can focus on the "true" driving factors that are the object of empirical testing. Methods for achieving such identification have become increasingly sophisticated over time, especially with the use of so-called natural experiments. Even so, there are few instances in which such identification is uncontroversial.

Partly as a response to this inherent difficulty, economists have begun to emulate the methods of natural scientists, by conducting carefully controlled laboratory experiments. The methodology of experimental economics has matured significantly over the last few decades, and has yielded many insights across all fields in economics. This paper surveys the broad practice of experimental studies in one such field, public economics, attempts to identify some of the main results--and the main limitations--of these studies, and suggests areas in public economics in which experimental methods may be usefully applied in the future.

Public economics has profited significantly from the use of laboratory experiments, for several reasons. Econometric data on research questions obtained from the naturally occurring world can be unreliable, can fail to show the variation or distinctions of interest, or can fail to provide sufficient identification. Indeed, in some cases data simply cannot be assembled because the real world setting of relevance does not exist. Laboratory experiments, on the other hand, provide a controlled environment for testing predictions; that is, experiments allow one to examine the mechanisms of interest, as well as the changes in these environments and institutions, in isolation from each other. However, a laboratory experiment is only as good as its design: if the institutions and environments imposed in the lab do not parallel systems of interest in the world, the resulting experimental data can be useless or misleading. Finally, public economics often examines topics such as public goods or tax compliance where traditional models of homo economicus have fallen short. Experimental work has been used to explore, develop, modify, and test new theories in these areas, as well as to test the assumptions of these theories, in ways that field data do not allow.

We begin with a discussion of laboratory methods in economics, where we stress many of the design elements laid out in Smith (1976, 1982). We then examine several areas of experimental research in public economics. We do not pretend to be exhaustive either in our selection of areas or in our discussion of specific research contributions, but rather intend to give a flavor for the many and varied areas to which laboratory methods have been applied:

* Tax compliance--Theoretical analyses based upon expected utility theory have proven unable to explain much compliance behavior. Further, field data on tax compliance are limited in quantity and quality, for obvious reasons. Experiments represent one useful avenue by which economists can study compliance.

* Public goods--Experiments have allowed economists to understand how, whether, and why people contribute to public goods.

* Political economy--Economists and political scientists alike have used experiments to study many aspects of political economics, including voting, committees, and legislative behavior. Game theory has played a huge role in the theoretical and experimental work in this area.

* Tax incidence--Simple market experiments have been used to investigate theories of tax incidence.

* "Other" experiments--We discuss a selection of experimental papers in other areas of public economics, including behavioral responses to taxes and various macroeconomic questions.

We focus on experiments in public economics and, thus, we do not discuss a range of other areas, such as experimental analyses of industrial organization (e.g., market structure experiments, game-the-oretic models of imperfect competition), environmental economics (e.g., externalities, valuation, especially the divergence between willingness to pay and willingness to accept), risk and uncertainty, labor economics (e.g., job search, work effort, reservation wages), asymmetric information, and health economics. For comprehensive surveys of experimental methods, see Davis and Holt (1993) and Kagel and Roth (1995).

In the conclusions we suggest areas where the next decade could see laboratory experiments in public economics having a significant impact on theory and policy.

THE METHODOLOGY OF EXPERIMENTAL ECONOMICS

The use of laboratory experiments in economics began in earnest in the early 1960s with work on resource allocation under alternative forms of market organization. Growth in its applications came with the establishment of a well-defined framework for experimental work by Smith (1976, 1982), and laboratory methods are now widely accepted as a methodological approach in the analysis of theory and policy.

Laboratory experiments seem particularly well suited for the study of some aspects of public economics. Unlike theoretical work, experiments are not as constrained by the same degree of simplification required in analytical studies, which allows the impact of numerous factors not amenable to theoretical work to be examined precisely and unambiguously in a controlled environment. Unlike traditional empirical work based on naturally occurring data, experiments generate data under settings in which there is control over extraneous influences. As we discuss later, there are some obvious limitations of experimental methods. However, given the weaknesses of theoretical and econometric work, there are compelling reasons for the use of experiments. In fact, experimental work has examined a remarkably rich range of public economics issues in areas that to date have not proven fully amenable to either theoretical or empirical analyses.

Experimental economics involves the creation of a real microeconomic system in the laboratory, one that parallels the naturally occurring world that is the subject of investigation and one in which "subjects" (usually students) make decisions that yield individual financial payoffs whose magnitude depends on their decisions. The essence of such a system is control over the environment, the institutions, the incentives, and the preferences that subjects face. Of these, control over preferences is particularly crucial. As emphasized by Smith (1976), "[s]uch control can be achieved by using a reward structure to induce prescribed monetary value on actions"; that is, in experimental methods, "... it is important that one be able to state that, as between two experiments, individual values (e.g., demand or supply) either do or do not differ in a specified way." (1)

Smith (1982) identifies a set of sufficient conditions for control over preferences to be established:

* Nonsatiation: Subjects prefer more to less of the reward medium.

* Saliency: The rewards received by subjects are related to their decisions, so that subjects recognize that their actions affect their outcomes.

* Reward Dominance: The rewards are large enough to offset any subjective costs or benefits that subjects place on participation in the experiment.

* Privacy: each subject knows only his or her own payoffs, so that they do not receive any subjective value from the payoffs of other subjects.

Nearly all recent experimental studies invoke these conditions. (2)