Three Essays on the Disposition Effect in Dynamic Settings
In this dissertation, I study the unique statistical features of multi-period returns vs. single period returns, and examine their impacts on making inference about investors' behavior from trading records. In particular, I focus on one such feature, namely the probability that investors can realize positive returns. In a simple dynamic model of a financial market, I derive a closed form representation of the dynamic relation between winning probability and holding time, and then apply the theoretical results to studies of disposition effects, referring to the tendency of investors to be reluctant to realize losses.
In essay I, I examine the holding time difference between winning positions and losing positions. Through theoretical analysis and simulations, I show that the winning probability dynamics, more convincingly than behavioral bias, accounts for the observation that investors hold the losing positions too long.
In essay II, I examine the disposition ratio difference between winning positions and losing positions (PGR-PLR). Both theoretical derivation and simulations confirm that the disposition ratio difference is not necessarily related to the selling preference of investors, but rather can be the results of investors' rational response to the winning probability dynamics.
In essay III, I examine the effects of no-trading events on the estimation of survival models. Theoretical analysis and simulations demonstrate that no-trading events in financial data can seriously bias the estimation of survival models, such as the well-employed Cox proportional hazard model, and thus distort our inference about investors' behavior.
Overall, the behavioral bias argument about the observed disposition effects does not have solid foundation. The rigorous theoretical analysis and simulations support the argument that the observed disposition effects, as well as the reversed disposition effects, are just the expression of such unique statistical features as the winning probability dynamics and the effect of no-trading events. The long existing puzzle of disposition effects can be solved within the economic framework in which market participants are rational agents.