A family of maximum likelihood estimation methods of MRR model and the corresponding empirical analysis

Madhavan et al. presented an influential model referred as MRR, which employs the GMM method to measure the information parameter and liquidity cost of each stock using the high frequency data. As the GMM method needs adding some more moment conditions when some new market characters are introduced in MRR and the new moment conditions always lead to erroneous estimation results in some empirical studies. This paper proposes a family of maximum likelihood estimation methods of MRR. Empirically, the parameters estimated by our methods are accurately approximate to the values estimated by the GMM method in original model and the MLE methods are robust when some new characters are introduced. Further, the effect of trading intensity on the information parameter is tested.