New Classical and Bayesian Estimators for Classifying Investor Intentions from Trade Data

We propose methods for estimating the effective bid-ask spread and classifying trades without access to quotes compatible with the conceptual framework for uncovering trading intentions outlined in Easley, Lopez de Prado, and O'Hara (2016). Our state-space approach accommodates informational asymmetries and unbalanced and auto-correlated order flow. The resulting parameter estimates possess a structural model interpretation, providing alternatives to existing semi-parametric bid-ask spread estimators. An analysis of trading in CME's gold futures contract surrounding the UK's June 2016 Brexit referendum, indicates the spread estimates and trade classifications reflect information flow dynamics during this period of financial market uncertainty and information asymmetry