Improvements in estimating the probability of informed trading models

Two advances have been made in the estimation of probability of informed trading (PIN) models. First, an initial-value-setting scheme has been proposed, that sets up a grid for initial values of mixture probabilities and uses the probabilities to divide the sample so as to derive the initial values of Poisson parameters. Second, the mixture bivariate normal distribution can help approximate the compound Poisson distribution in estimating PIN models. This study implements two approaches to simulated and real data for the PIN and Adjusted PIN models and compares their performance with the literature. The new initial-value-setting scheme performs better than those of Yan and Zhang [An improved estimation method and empirical properties of the probability of informed trading. J. Banking Finance, 2012, 36(2), 454–467] and Ersan and Alıcı [An unbiased computation methodology for estimating the probability of informed trading (PIN). J. Int. Financ. Markets, Inst. Money, 2016, 43, 74–94], and using the normal distribution outperforms the Poisson distribution under certain variance specifications.