| Economists have learned a great deal about investor behavior over the last two decades with the availability of large discount brokerage data sets. While this has given economists a better understanding of the trading patterns that characterize individual investor behavior, less success has been achieved in understanding what drives these trading patterns. Part of the difficulty in this endeavor is that it is sometimes difficult to test alternative theories of investor behavior using only data from the field. In particular, the two trading patterns we investigate in this thesis, the disposition effect and the repurchase effect, are unlikely driven by standard rational models of trading, and alternative theories of their causes are difficult to test using only data from the field, or data from behavioral laboratory experiments.;In order to better understand the causes of the disposition effect and the repurchase effect, we use neural data, data collected from functional magnetic resonance imaging (fMRI) along with trading data to construct empirical tests of different theories. Chapter 1 uses fMRI data to test a model of realization utility, which can readily predict a disposition effect. In our experiment, we find that subjects exhibit strong disposition effects, although they are suboptimal, and the neural data strongly supports the realization utility hypothesis. While Chapter 1 is concerned with the selling behavior, we focus on systematic violations of buying behavior in Chapter 2. We propose a model of regret to explain the repurchase effect in the buy-side trading data, for which we find strong support in the neural data. Chapters 3 and 4 study whether the suboptimal trading behavior we find in the first two chapters is stable, and we explore what the source of the heterogeneity is. Specifically, in Chapter 3 we find that exogenously manipulating the display of information on the trading screen can significantly reduce the size of the disposition effect. Chapter 4 uses an approach from behavioral genetics to identify candidate genes that can help explain the cross-sectional variation in choice behavior. |
