Testing and Modelling Market Microstructure Effects with an Application to the Dow Jones Industrial Average

It is a well accepted fact that stock returns data are often contaminated by market microstructure effects, such as bid-ask spreads, liquidity ratios, turnover, and asymmetric information. This is particularly relevant when dealing with high frequency data, which are often used to compute model free measures of volatility, such as realized volatility. In this paper we suggest two test statistics. The first is used to test for the null hypothesis of no microstructure noise. If the null is rejected, we proceed to perform a test for the hypothesis that the microstructure noise variance is independent of the sampling frequency at which data are recorded. We provide empirical evidence based on the stocks included in the Dow Jones Industrial Average, for the period 1997-2002. Our findings suggest that, while the presence of microstructure induces a severe bias when estimating volatility using high frequency data, such a bias grows less than linearly in the number of intraday observatio