Grant-funded Project Nr. 463/2004/A-EK/CERGE Final Report
Project title:
Biased Self-Atrribution in Bayesian Learning and Asset Pricing
Research leader:
RNDr. Eugen Kováč, 2002
Co-researcher:
Ing. Petr Zemčík, Ph.D.
Period of project:
2004-2004
Overall grant:
79 000 CZK
Project Results
In this project I analyze effects of irrational investors in financial markets and determinants of survival in the market. Irrationality means that they do not process new information correctly (using Bayesian updating). Instead, they use different statistical methods or rules of thumb to make their investment decisions. These can be caused by various congnitive errors, as overconfidence, biased self-attribution, underweighting of base rate information, etc.
The methodology I used is a theoretical model of general equilibrium. I assume a finite time simple economy with a single risky asset in unit supply, which pays one-shot payoff. The payoff is random and its distribution is not known a priori. Agents receive public signals (random draws from the same distribution) and update their beliefs. Although they may use various updating rules, it is assumed that the agents are sophisticated and are aware of their future errors. After observing the signals, agent engage in speculative trade in order to maximize the returns from the assert. In the model, I focus on long-run effects, in particular, survival of investors.
Current literature (Sandroni, Econometrica 2000) claims that investors holding inaccurate beliefs are driven out of the market. This is shown on models of Lucas trees. The main drawback of such models is that the equilibrium prices do not depend on investors' beliefs about future returns. On the other hand, in my model the equilibrium prices are determined by investors' beliefs. Drawing on the results by Sandroni (2000), I show that agents with inaccurate beliefs are driven out, in the sense that their wealth becomes arbitrarily small when the number of signals is sufficiently large. However, there is an important difference from Sandroni (2000). In his paper, the agents do not survive, because they systematicly make wrong investment decisions and henceforth have systematicly lower returns. In my model, agent bet on wrong price changes and lose because others have different (more accurate) beliefs.
Compared to the proposed project, I was able to achieve more general results focusing on a more general class of cognitive biases.
Part of the work on the project was done during my visit at the University Bonn, where I consulted and cooperated with several local faculty.
The results of the study were presented on two internal seminars: at CERGE (Charles University, Prague) and at the University of Bonn (Germany). The paper "On Survival of Speculators" can be found on author's homepage www.uni-bonn.de/~kovac/papers/fm.pdf or home.cerge-ei.cz/kovac/papers/fm. A publication in the Bonn Econ Discussion Papers Series is being prepared. Furthermore, the paper will be part of my dissertation at CERGE (I expect to submit it in fall 2005) and I plan to publish it in an international academic journal.