A methodological approach to the computational problems in the estimation of adjusted PIN model

Abstract

It is well documented that computational problems may lead to large biases in the estimation of probability of informed trading (PIN) models. While effective remedial solutions have been suggested for the case of original PIN model (Easley et al., 1996), computational problems for its most broadly applied extension, the adjusted PIN model of Duarte and Young (2009) , are yet to be addressed. Given its larger parameter set, estimates of the AdjPIN model are more likely to suffer from computational problems. We address these computational problems by developing an estimation method comprising a) a logarithmic factorization of the likelihood function, and b) an algorithm to strategically generate initial parameter sets. We show that the introduced method outperforms existing best-practices and strongly suggest its use in further studies.

Keywords: Multilayer probability of informed trading, MPIN, layer detection algorithm, cluster analysis, information asymmetry, private information.

JEL Classification: C13, C38, G14, G17