A PIN per day shows what news convey: the intraday probability of informed trading

Abstract

This paper develops a new intraday estimation procedure for the sequential microstructure trading model initially proposed by Easley et al. (Rev Financ Stud 10:805-835, 1997a). Using a full year of intraday trading data for the top 100 German stocks, we demonstrate how the new estimation procedure eliminates or significantly reduces the shortcomings of the original approach in recent, high-frequency trading environments. We slice a trading day in buckets of several minutes' length to obtain one estimate of the composite variable probability of informed trading (PIN) per day. This approach makes PIN applicable in short horizon event studies. Convergence rates are above 95 % even for the most liquid stocks and the model's underlying assumptions of independence for the arrival of traders and information events are fulfilled to a much higher degree than in the original approach. An empirical application in an event study type setting demonstrates how official announcements stipulated in German insider trading legislation significantly reduce information asymmetry upon public disclosure.

Probability of informed trading during the COVID-19 pandemic: the case of the Romanian stock market

“…These values are similar to those reported by Kang ( 2010 ) and Aslan et al ( 2011 ) for more advanced countries. Compared with other studies such as Easley et al ( 1996 ) or Pöppe et al ( 2016 ), we fail to identify a clear pattern suggesting that PIN is strongly decreasing from low to high volume stocks.…”

“…4 The capital markets usually respond to news almost instantly (Aslam et al 2020 ; Espinoza-Mendez and Arias 2021 ); therefore, the computation of PIN over long periods represents a shortcoming (Pöppe et al 2016 ). Few articles in the existing literature analyse PIN over short periods (Tay et al 2009 ; Pöppe et al 2016 ). The disadvantage of the estimation for a longer period is acknowledged by Easley et al ( 2008 ).…”

“…The trade direction was computed using the Lee and Ready ( 1991 ) algorithm. 7 We exclude from the analysis the trades conducted in the first and last 15 min of the trading sessions as recommended by Pöppe et al ( 2016 ). 8 To estimate the weekly 9 PIN, we use the approach proposed by Yan and Zhang ( 2012 ) 10 and Gan et al ( 2015 ) 11 respectively.…”

Informed trading around biotech M&As

“…We provide supportive evidence that the reformulated likelihood function introduced by Lin and Ke (2011) and the YZ algorithm can (generally not significantly) mitigate the downward bias of the EHO PIN estimation method. In addition, by elaborating on the sampling procedure proposed by Kryzanowski and Lazrak (2007) and subsequently used by others (Chang et al , 2014; Pöppe et al , 2016), we demonstrate that EHO PIN estimates are sensitive to the number of observations used in their estimation.…”

“… 3. Pöppe et al . (2016) also introduce the use of multi-minute slices of a trading day to obtain a daily PIN in an event study type setting.…”