We consider a model of trade in a limit order market for a single asset, and examine the properties of the microstructure noise (i.e., the difference between the transaction price and the fundamental value). The asset has a common value; in addition, each trader has a private value for it. Traders randomly arrive at the market, after choosing whether to purchase information about the common value. They may either post prices or accept posted prices. If a trader's order has not executed, he randomly reenters the market, and may change his previous order. The model is thus a dynamic stochastic game with asymmetric information. We numerically solve for equilibrium in the model, and simulate market outcomes. Agents with no intrinsic benefit from trade have the highest value for information and also tend to supply liquidity. Agents' incentives to acquire information and their subsequent equilibrium trading behavior changes systematically with the underlying volatility of the asset. In equilibrium, the limit order market acts as a ``volatility multiplier'': prices are more volatile than the fundamental value of the asset. This effect increases when the fundamental volatility of the asset is higher or when there is asymmetric information across traders, due to a change in the composition of trader types that choose to provide liquidity. Further, changes in the microstructure noise are negatively correlated with changes in the fundamental value itself. This implies that asset betas estimated from high-frequency data will be incorrect. Thus, microstructure noise and other variables correlated with it may spuriously appear to explain cross-sectional asset prices. Finally, we show that the microstructure noise itself has positive autocorrelation.
