I was asked to write two chapters for a Wiley book called Behavioral Finance, and for some reason I said yes.  Neither chapter is a typical review paper.  Behavioral vs. Traditional Finance will be Chapter 2 of the book, and sets the stage by viewing the conflict between the subfields in light of the philosophy of science.  Experimental Finance (written with Cornell Finance doctoral student Alyssa Anderson) is a tutorial on the role of experiments in finance, with heavy emphasis on how experiments can provide contributions beyond the economic models they test, and on the importance of controlled manipulations (which is sadly under-appreciated in economics and finance).

Both chapters are just as relevant to financial accounting researchers and finance researchers.  Chapter 2 even uses a brief history of behavioral research in financial reporting to highlight the role of sociological forces in research trends.

Teaser excerpts after the jump.

From Traditional vs. Behavioral Finance:

Anyone with a spouse, child, boss, or modicum of self-insight knows that the assumption of homo economicus is false. Behavioralists in finance seek to replace homo economicus with a more-realistic model of the financial actor. Richard Thaler, a founding father of behavioral finance, captured the conflict in a memorable NBER conference remark to traditionalist Robert Barro:  “The difference between us is that you assume people are as smart as you are, while I assume people are as dumb as I am.” Thaler’s tongue-in-cheek comparison aptly illustrates how the modest substantive differences in traditionalist and behavioralist viewpoints can be exaggerated by larger differences in framing and emphasis, bringing to mind the old quip about Britain and America being “two nations divided by a common tongue.”  (For what it’s worth, when confirming this account of the exchange, Thaler reports that Barro agreed with his statement.)

The purpose of this chapter is to guide readers through this debate over fundamental assumptions about human behavior, and indicate some directions behavioralists might pursue. The next section provides a general map of research in finance and describes in greater detail the similarities and differences between behavioral and traditional finance. The ensuing section places the disagreements between the two camps in the context of the philosophy of science: behavioralists argue, a la Thomas Kuhn, that behavioral theories are necessary to explain anomalies that cannot be accommodated by traditional theory. In return, traditionalists use a philosophy of instrumental positivism to argue that the competitive institutions in finance make deviations from homo economicus unimportant, as long as simplifying assumption is sufficient to predict how observable variables are related to one another.

Kuhn’s (1962) perspective is not in direct opposition to instrumental positivism. Yet, behavioralists tend to argue Kuhn against traditionalists, who reply with instrumental positivism.  While both arguments have substance, they also contain a rather contentious personal element.  By adopting a Kuhnian perspective, behavioralists implicitly brand their opponents as old, fading Luddites. (Kuhn famously claimed that individual scientists never change their minds; instead, fields change because the old scientists die or retire, and are replaced by a new generation of scientists who hold to the new paradigm). By emphasizing instrumental positivism, the traditionalists imply that behavioralists are arguing their case on the basis of realism rather than predictive power, and suggest that behavioralists are not even real scientists because they proffer an irrefutable theory that can be adapted ex post to accommodate almost any observation.

From Experimental Finance

Experimental economics often worry about the extra-theoretical implications of context.  Yet, recognizing that concerns about such ‘baggage’ are far more serious for a study that has no manipulations is important. Without a manipulation, any aspect of the setting could be important in driving bubble formation. Thus, the presence of context makes attributing the presence or absence of bubbles to economic factors very difficult. While concerns about ‘baggage’ are a problem, context is only one relatively obvious noneconomic factor that might drive results. The color and temperature of the room, the background and intelligence of the participants, and details of the trading interface and the noise generated by trading could all affect pricing. Moreover, economic factors not being considered by the research could also matter, such as the length of trading periods, nominal price levels, or the nature of the pricing mechanism (e.g., double auctions vs. clearinghouse markets). Quite simply, Smith et al. (1988) can only provide conjectures of why they observe bubbles and any deviation from their setting may change their results.

In contrast, imagine that Smith et al. (1988) had in fact manipulated a variable such as the amount of cash in the market. In fact, Caginalp, Porter, and Smith (2001) conducted a study like this by indicating that large cash endowments do indeed make bubbles more likely. Thus, assume that this alternative to Smith et al. would generate a similar result. In this case,  finding such a difference across treatments to be driven by the features of the two settings that are held constant would be very unlikely. As a result, there is little reason to be concerned about the fact that context and meaningful labels would detract from the inferences one can take from the study. After all, the context and labels do not drive bubbles when cash endowments are low. Thus, context and labels by themselves are unlikely to drive bubbles. Also highly unlikely is that  the presence of context and labels drives the difference between the two treatments. This would require an interaction between context and cash endowments that, at least at first glance, does not seem plausible.

The power of controlled manipulations also provides experimentalists with defenses against many common criticisms, such as the types of participants in the study and the levels of compensation. While more training and greater incentives might reduce the formation of bubbles, the level of training and incentives in this particular experiment is very unlikely to explain differences across settings in which training and incentives are identical. Thus, those reviewing experiments should be extremely cautious in criticizing experiments in which they believe participants had too little experience or were not paid enough, unless they have a specific reason to believe that experience or incentives will interact with the manipulated variable.