Identifying information types in probability of informed trading (PIN) models: An improved algorithm

Abstract

The multilayer probability of informed trading (MPIN) model, developed by Ersan (2016), releases the assumption of single type of information events in the original PIN model of Easley et al. (1996). Identification of the number of layers in a dataset is applied through a layer detection algorithm suggested in Ersan (2016). The algorithm is based on clustering absolute order imbalances and examination of confidence intervals for the skellam distribution. When uninformed trading intensity is assumed to be identical in the buy and sell sides, the algorithm performs extremely well. When uninformed intensities are not equal, Ersan (2016) suggests the adjustment of the data using a correction term, proxied by the minimum levels of buys and sells in the data. We improve the algorithm of Ersan (2016) in two ways. We provide accurate estimates of uninformed trading intensities used for data adjustment, and slightly modify the algorithm of determining the information layers. The improved algorithm identifies the number of layers with substantially increased precision, between 86% and 95% accuracy for the simulated data with various settings.

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

JEL Classification: C13, C38, G14, G17