Experimental Economics and Real-world Details

Lynne Kiesling

Yeah, what he said. I agree with Mike’s previous post about real-world detail and experimental economics (and thanks for mentioning the Electricity Journal article!).

The wind tunnel metaphor is apt for the use of experimental economics to test market designs and policy proposals; testing them in an environment that captures the most salient features, but not all of the features, of the real-world environment will provide valuable information on what incentives the design presents to individuals, what unintended consequences might arise, and what outcomes to expect. Importantly, experimental economics helps to identify consequences that the experimenter or policymaker could not have anticipated ex ante, due to incomplete information and/or tacit knowledge in the minds of the participants.

I also use the map metaphor to explain to people why full real-world detail is not necessary, and indeed may be a bad idea. Think about a map. Its objective is to guide you from where you are to where you are going, using salient information about landmarks, milestones etc. between here and there. If the map were in real-world scale it would be useless, because all of the detail would obscure the most salient features that would enable you to navigate from here to there. The challenge to the experimenter is to identify and incorporate the features that are most important in affecting individual behavior and the outcomes of that behavior (and the interaction of that behavior with the behavior of others).

In electric power one of the most controversial issues over which this discussion occurs is the detail embedded in the physical wires network in the experiment. Some experimenters argue that you need a pretty fully fleshed-out network, with lots of nodes and lots of opportunities for loop flow, to get a realistic depiction of the system-level effects of individual decisions in the context of a particular market design/policy environment. Other experimenters argue that you can allow for loop flow to occur but focus more cleanly on economic incentives and individual decision-making if you have a simpler network environment (such as a 4-node, 5-line Wheatstone Bridge layout). I tend to fall in the latter camp, while those with a more engineering perspective tend to fall in the former camp. My concern about the really detailed network environment is that once you incorporate that much detail, it becomes more difficult to formulate and test a clean hypothesis, and to know that there’s a real connection between the design features and the actual outcome.

3 thoughts on “Experimental Economics and Real-world Details

  1. If I may suggest, perhaps one reason that “those with a more engineering perspective tend to fall in the former camp” is that (relatively speaking) it is not that much more difficult to formulate and test a clean hypothesis.

    If you use a complex network, but really understand how that particular network works, then you get both the benefits of “complex enough to be interesting” and “simple enough (to the researcher) to be tractable.”

    At Cornell, for example, while they use a variety of networks and a platform which has the potential to do many things, a lot of their experimental economics work involves variants of a particular well-examined network. More complex than, say, the folks at GMU typically use, but since the Cornell folks know it so well it isn’t that much more difficult for them to use productively.

    (This argument is my ‘off the top of my head’ view on the stylistic differences between the experimental work done at Cornell and at GMU. Additional views welcome.)

Comments are closed.