Estimation and Testing for Dependence in Market Microstructure Noise

Abstract

This paper proposes new test statistics for the dependence and cross and auto covariance estimators of bivariate noise processes. It derives their asymptotic distributions and provides additional tests for the statistical significance of covariance estimators. Monte Carlo simulation shows that the covariance estimators and test statistics perform better in a finite sample. Further evidence from empirical illustration suggests that the covariance estimators and proposed test statistics are capable of capturing various dependence patterns in market microstructure noise. These results can shed more light on the sign of noise autocorrelation in the presence of market microstructure frictions such as bid-ask bounces and the clustering of order flow.