Information, trade, and liquidity in financial markets

This thesis discusses the effect of information on the liquidity of financial markets and investors' trading behavior. Chapter 1 proposes a model in which liquidity is time-varying and conditionally heteroskedastic. It is demonstrated both theoretically and empirically that liquidity variation plays an important role in the dynamic cross-sectional relation among observable quantities in stock markets. The Kyle-Admati-Pfleiderer framework is extended to a setting with conditionally heteroskedastic information and multiple security trading. When information acquisition is endogenized, liquidity, trading volume, and conditional price-change variance become all stochastic. Their dynamic relation is examined by a simulated vector autoregression. It finds that liquidity shocks are persistent; a positive volume shock has a negative effect on liquidity in subsequent periods; and liquidity shocks transmit across securities. Using high frequency data, empirical results are provided that are generally consistent with the model's predictions.; Turning to the long-run effects of heterogeneous information on asset market characteristics and investors' trading behavior, Chapter 2 studies an overlapping generations model with multiple securities and heterogeneously informed agents. There are two types of multiplicity of equilibria, one due to noisy rational expectations and the other resulting from self-fulfilling prophecies. Under general conditions, there exists an equilibrium in which stock returns are highly volatile and strongly correlated, even if all underlying shocks are small and independent. When prices are partially revealing, less informed agents rationally behave like trend-followers, while better informed agents follow contrarian strategies.; Chapter 3 empirically examines the consequences of price-impact costs, a measure of illiquidity deduced from trades of possibly privately informed traders, on market efficiency. Using nonlinear price-impact functions estimated for 4,897 stocks, we implement long-short arbitrage strategies based on known anomalies, and compute the maximal fund size possible before excess returns become negative. It is found that the fund size of size and B/M strategies must remain small relative to the actual hedge fund size in order to remain profitable. However, some momentum strategies may be implemented profitably in a fairly large scale, challenging the market efficiency literature.