David I. Laibson, Harvard University and NBER
Xavier Gabaix, MIT and NBER
The nature of the problem can be described as follows. Various measures of stock market activity exhibit puzzling "universal" features that have recently attracted much research attention. These features include the power law distributions of return, volume, number of trades, assets under management of trading institutions, and other power law relations linking them. These universal features not only present a challenge to models of market fluctuations, but their specific power-law nature also suggests new modeling directions, which include ideas from statistical physics which proved useful in understanding similar relationships that occur in the physics of critical phenomena.
We seek to understand the origins of these regularities, by exploring the following hypotheses: (i) Large returns arise when a few large institutions trade in a market of fairly small liquidity. (ii) Large returns arise when market liquidity drops, independent of the size of the institutions involved. (iii) Large returns arise because of volatility feedback mechanisms can also give rise to power law distributions, and generate short and long term correlations in trading activity.
We approach this problem by developing simple rational and boundedly rational models, and analyzing extremely large data sets. This synergy between compact tractable models and the exploration of vast data sets has been the prime method of our research group. Our empirical research will study the joint distributions of returns, volume, number of trades, and liquidity. This includes studying in detail the price impact of trade and its variations under different market conditions. The modeling will involve designing models of financial decision making, that can provide testable implications that our empirical research will study.
The potential impact of our work includes the following. First, we will understand what creates large economic fluctuations. In particular, our work will delineate between competing hypotheses that seek to explain the power-law distribution of returns, and systematically explore the role of volume and liquidity in explaining the specific exponent value. Second, more generally our work may give some insight into quantifying and understanding collective human behavior using concepts of statistical physics, behavioral finance, and psychology. We will also show that a useful way to formulate the models is to draw from research in decision making in complex environments, which itself draws from cognitive psychology.