Empirical Analysis of a Complete Stock Market with Information Networks of Traders

Abstract

How is information incorporated into financial markets and asset prices? Network theory offers a novel approach to this important question. In a market where investors share private information through a network, specific predictions arise about the distribution of trading profits, the correlation of trades and the trading volume. A theoretical model is developed along these lines in Ozsoylev and Walden (2011). Specifically, the model introduces slow information diffusion through an information network. In the model, some information is neither completely public, nor completely private, and the network provides an additional channel through which information is incorporated into asset prices. This view, that there is “semi-public” channel through which information reaches asset prices in addition to the traditional channels, is quite different from the standard approach to asset pricing with heterogeneous information, since the standard approach assumes completely private signals. Anecdotal evidence suggests that the semi-public channel is important. The theory is consistent with several fundamental stylized properties of real-world markets: 1. Most large price movements are not related to public signals (see, e.g., Fair, 2002). 2. Price volatility is highly time-varying. 3. Trading volume is high and heterogeneous - in a predictable way. 4. Portfolio investments are heterogeneous - in a predictable way (the alternative motivation for diversity in portfolios is diverse hedging motives. However the success of the studies that take this approach to explaining portfolio heterogeneity has so far been quite limited). Since these are fundamental properties of stock markets (especially the first and the fourth property), and we still do not have a good understanding of why they arise, if it can be shown that they are driven by aggregation of information through information networks, this would lead to fundamental new insights about how markets work. It would also have potential policy implications. A high informational efficiency in a market is often valuable to society, e.g., allowing for optimal investments. If it can be shown that information networks are a source of inefficiencies, this would potentially suggest new regulations for how information should be disseminated. Our study aims at carrying out an initial evaluation of how well information networks explain these stylized properties, by testing several empirical implications of the theory.