Link : Review of Financial Studies

This paper builds upon the this paper where the authors introduce a trade process model. What do the authors attempt via this study ?

They develop a framework for analyzing the information in a trading process. This is basically Bayesian learning problem where the market market is a Bayesian who updates various probabilities based on the trades that occur through out the day.

The first section of the paper talks about a trade price model, a sequential trade model :

image

The authors go on to estimate the parameters of this model via MLE. Once the parameters of the trade process are obtained, initial equilibrium quotes are generated for the market maker. Subsequently, after every trade, the market maker updates the respective probabilities of the trade process and adjusts his bid and ask process. Thus the trade process and the quote process lead to the price process under certain assumptions. The paper ends by showing the importance of trade data in the price formation process.

Broadly, the paper answers the following questions :

  • How does one formulate an asymmetric information dynamic model for a market maker behavior?

  • How can one estimate the parameters of the trading model from trades and quotes data ?

  • How does one test for model misspecification ?

  • How does one analyze the information transmission that occurs between the trade data and transaction prices?

  • How do you incorporate the effects of trade size in to price formation process ?

A lot can be learnt from this paper, not only from the microstructure modeling perspective, but also from a  general stats perspective on how to go about testing for model misspecification.