Experimental Exploration into Macroeconomics

Economics, long thought beyond the reach of experimentation, began gradually to yield some of its inaccessible secrets to laboratory and field investigations over the past century. Observation and analyses of laboratory games have led economists to think about markets as social artifacts, whether evolved or designed, to achieve predictable outcomes in specified environments. The value of potential insights into properties of macroeconomic models and policies had to overcome the barrier of virtual impossibility of conducting controlled experiments at macroeconomic scale. Fortunately, micro models of macro phenomena have allowed experiments to identify the more plausible from sets of multiple or indeterminate outcomes, and assess policy alternatives and institutional designs. Without attempting a comprehensive survey, this paper summarizes some important discoveries from experimental economics, with special emphasis on macroeconomics. Challenges ahead are mentioned briefly. 1 Presented at the Silver Jubilee International Conference at the Indira Gandhi Institute of Development Studies, Mumbai, December 1-3, 2012. Revised February 25, 2013. I thank Qin Tan for her assistance. 2 Experimental Exploration into Macroeconomics 1. Economics as an experimental science The use of experimental methods to explore our understanding of economics is only a little over half-a-century old. Its use in macroeconomics originated more recently. While applications of this method to various aspects of economics have expanded rapidly, much remains to be done on its usefulness, linkages, and acceptability. I address some of these issues at the outset, before delving into what we have learned. Classical uses of experiments were confined to natural sciences such as physics, chemistry, and aspects of biology where the objects of experimentation are either inanimate, or cannot reasonably be expected to change their behavior in response to experimental treatments. This invariance makes it possible for natural scientists to identify laws of nature that retain their validity and predictive power across time and space. Universality of their laws earns for science and scientists high prestige that social sciences covet, and try to emulate. Experimental method is valued for enabling investigators to gain manipulative control over conditions or treatments under which observations are gathered. Invariance of underlying laws, combined with careful design of treatments, allow investigators to infer regularities in observations. When independently replicated with sufficient and convincing frequency, and logically linked to related phenomena, these regularities are characterized as laws by virtue of their high predictive power and contribution to better understanding. Social sciences, too, seek to identify observational regularities which can be called laws on the basis of their power to explain and predict. Their observational base had long been confined to domains which are outside the investigators’ manipulative control—often referred to as field data from ‘naturally occurring’ phenomena. There were three reasons for this restraint. First, unlike in the natural sciences, the objects of observation and analysis in the social sciences are sentient beings—ourselves. Human consciousness, including self-consciousness, means that we tend to learn, change behavior with time and experience, and are likely not only to become aware of the treatments to which we are being subjected, but also to have the capacity to expect, anticipate, and react willfully to such treatments 2 . Such reactions can make the investigator a part of the phenomenon he/she is trying to investigate. Under such conditions, identifying universal laws of behavior, or even delineating the boundaries of any local regularity we may come upon, is a far more challenging task than what the natural scientists face. 2 I do not address the deeper philosophical issue of whether human free will, and the idea of replicable laws of individual behavior, are irreconcilable with each other. Major aspects of experimental economics that I shall mention here are confined to regularities in the outcomes of social institutions (e.g., markets) that constrain the behavior of individuals who interact with one another. Even if populated with individuals whose free will cannot be captured in identifiable laws of behavior, it is entirely possible for social institutions to exhibit regular and replicable properties. It is an issue I return to in sections on the role of optimization and markets as artifacts. 3 Second, the human tendency to learn and adapt their behavior challenges the robustness of the validity of any laws relative to their discovery. If discovered “laws” in social sciences induce those who learn of the discoveries to adapt their behavior to this knowledge, they may no longer remain valid. This endogeneity risks rendering any discovered regularities into transient facts, instead of steady state functional relationships that constitute the essence of what we can call a science. Third, many social phenomena such as legislation, macroeconomic policy, and race relations, remain well-beyond manipulative control of investigators because such manipulation is infeasible, carries unacceptable risk of undesirable consequences, or is simply unethical. However, this second difficulty is shared with some natural sciences, of which astronomy, geology, and meteorology are well-known examples. All these three difficulties of extending the use of experimental method to social sciences have been at least partially addressed in the recent decades with the development of experimental methods to address economic questions. The first two can be addressed by constraining the focus of experimental investigations to properties of economic institutions. Institutional properties are, at least potentially, more stable and less subject to the abovementioned endogeneity problems associated with adaptation of human behavior to anticipations of experimental treatments and to the knowledge of institutional properties. Further, it can be argued that it is useful to make at least an approximate distinction between psychology as the study of individual behavior and economics and sociology as the study of aggregate level outcomes that result from individual behavior but may have distinct properties of their own. In this paper, I use this criterion to distinguish experimental economics from studies of individual behavior by economists, often referred to as behavioral economics, and confine my remarks to the former. Vernon Smith (1962) found that the outcomes of markets with a mere dozen or so student traders are surprisingly close to the predictions of equilibrium derived from assuming perfect competition among optimizing atomistic traders. This discovery enabled economists to have greater faith that markets with a bare handful of traders in laboratory settings can yield useful approximations and insights into the behavior of much larger markets which would be essentially infeasible to replicate in laboratory. The problem of addressing macroeconomic questions in small-scale laboratory settings has been eased by the development of microeconomic foundations of macroeconomic phenomena (Lucas 1972, Barro 1997). As with laboratory work in all fields, experimental economics results also must pass a skeptical external validity check; credibility of inferences must remain a matter of judgment by the disciplinary community. In Section 2, I take a sample of some experimental economics results over the recent decades, with additional attention to macroeconomic experiments. Since the literature has grown rapidly to become quite large, this cannot be a review, only an appetizer. Excellent collections of reviews of various aspects of experimental economics are available elsewhere and I shall not try