Can experimental economics be used to improve policy design? This is an interesting idea but one question, and it's one the authors acknowledge, is if in general the experiments can be made complex enough to duplicate real-world conditions and capture all of the essential aspects of the market under examination:
Experimental economics and policy design, by Steffen Huck and Jean-Robert Tyran, VoxEU: The general public and many politicians tend to be sceptical that unregulated markets are good for people’s well-being. Economists have known forever that laissez-faire yields the first best only under fairly unrealistic assumptions. The theory of industrial organisation describes when markets work, when they fail and how such failures can be remedied. Unfortunately, these theories are often difficult to relate to and to implement in practice. One reason is that theory is simple and abstract. Reality, however, rarely cooperates with the simplifying assumptions. Real-world complexity means that it is often difficult to measure exactly how efficient a particular market is or to decide if the theory can be applied to a particular case. It is often impossible to know whether the assumptions underlying the theory are met in practice, or whether other un-modelled aspects of the environment might be critical.
Experimental economics is one way to bridge the gap between simple theories and “messy” reality. Over the last decade or so, economists have used economic experiments to investigate when markets work and when they fail – at least in the laboratory. The advantage of the experimental approach is that the markets studied can be richer and more realistic than those that simple theories deal with, yet they are also more structured and “controlled” than in the field. Market experiments are richer because real people (rather than abstract optimising agents) participate in these markets, and they are “controlled” because the researcher knows the market conditions (e.g. incentives and information available to market participants), and can manipulate these conditions in a systematic manner. In addition, interaction outcomes like prices, and the quantity and quality of traded goods, can be measured without error.
For example, one market can be implemented in which firms compete for customers, and outcomes can be compared with an otherwise identical market in which firms do not compete. Experimental control enables the researcher to hold all aspects other than competition constant. Because everything else is held constant, the researcher can argue that differences in observed market outcomes, like higher product quality, must be caused by competition between firms. Experiments thus serve to fill the gap between abstract theory and complex practice, and can provide guidance to both theory and policy practice. Of course, it should be kept in mind that laboratory economies are still comparatively simple realities. Experiments enable the researcher to draw truly causal inferences, but the extent to which the implemented environment captures the essential aspects of naturally occurring markets is always open to debate and subject to scrutiny.
In CEPR Policy Insight No. 6, we argue that ‘experimental economics’ should be used to help design policy. We take as an example recent laboratory experiments that investigate economic interactions that are beset by “the trust problem” and we ask how “the market” (i.e. various forms of competition) can help to solve the trust problem. The “trust problem” (an incarnation of the moral hazard problem) that we discuss arises in markets for experience goods. Characteristic for such markets is that buyers are uncertain about the good’s quality before they buy, but experience its quality after having bought and consumed it. Experience goods cover the broad middle ground between the extremes of goods involving no quality uncertainty at all (so-called inspection goods) and goods for which quality is not fully revealed even after the consumption (credence goods). Whenever contracts for the exchange of a good are incomplete and sellers have leeway to shade its quality, about which the consumer finds out only if it is too late, the good in question is an experience good. Hence, many are.
The overall message from our experiments is clear. Free selection of trading partners without price competition drastically improves efficiency. Sellers resist the temptation to cheat customers by shading on quality because buyers would sanction them by “walking away” if they did. We go on to ask, given that this particular type of competition improved efficiency, will markets perform even better if price competition is added to free selection of trading partners? We find that adding price competition actually reduced efficiency. We believe it is useful to use experiments as a “test bed”, similar to a wind tunnel in car manufacturing. Proposed policy changes can be tried out on small scale before causing upheaval in the large national economy. Experiments can thus help to avoid costly mistakes and provide a useful indication on which policy changes might improve market efficiency.
Experiments are a long way from a perfect check on policy design, but compared to the costs of poorly designed policy – for example, it has been documented that over €10 billion were spent on ineffective labour market policies in Germany – they are a bargain.
 For the original studies, see Huck, Steffen; Ruchala, Gabriele and Tyran, Jean-Robert (2006): Competition Fosters Trust. CEPR Discussion Paper no. 6009, and Huck, Steffen; Ruchala, Gabriele and Tyran, Jean-Robert (2007): Pricing and Trust. CEPR Discussion Paper no. 6135.
 See “Relational Contracts and the Nature of Market Interactions,” Econometrica 2004 by Brown, Falk and Fehr for an application to the labour market.