An Empirical Analysis of the Dynamic Probability of Informed Institutional Trading: Evidence from the Taiwan Futures Exchange

• Author: Pei‐Shih Weng,Miao‐Ling Chen,Wei‐Che Tsai,Ming‐Hung Wu
• DOI: http://doi.org/10.1002/fut.21830
• Date: 01 September 2017
• Published date: 01 September 2017

An Empirical Analysis of the

Dynamic Probability of Informed

Institutional Trading: Evidence from the

Taiwan Futures Exchange

Pei-Shih Weng, Ming-Hung Wu, Miao-Ling Chen, and Wei-Che Tsai *

Based upon a unique dataset of institutional transactions in the Taiwan index futures market,

we analyze the informational role of institutional investors using the “dynamic probability of

informed trading”(DPIN). Compared to “trading imbalances”(TIB) and the “volume-

synchronized probability of informed trading”(VPIN), we show that the DPIN of foreign

institutional investors outperforms the alternative measures and provides more stable effects

in measuring informed trading. Following the strategies highlighted by DPIN, foreign

institutional investors also perform better than domestic institutional investors. Our results

support the validity of DPIN and present that foreign institutional investors are more informed.

© 2016 Wiley Periodicals, Inc. Jrl Fut Mark 37:865–891, 2017

1. INTRODUCTION

The wealth of literature on the information content of institutional trading, which dates back

more than 40 years, shows that if institutional investors are informed and choose to realize

the value of their information through trading, then their transactions will improve the

informational efficiency of the market by enabling the information to be incorporated into

asset prices quicker.

1

Of the various studies, Roll (1988) demonstrated that informed traders’

risk arbitrage activities can be the primary cause of firm-specific stock price movements.

Although the study of institutional trading remains an issue of major interest to both

practitioners and academics alike, nevertheless, identifying an appropriate measure of the

informativeness of institutional trades within an empirical market microstructure is by no

Pei-Shih Weng is affiliated at National Dong Hwa University, Hualien, Taiwan. Ming-Hung Wu is Ph.D.

Candidate of Finance at National Sun Yat-sen University, Kaohsiung, Taiwan. Miao-Ling Chen is Professor

of Finance at National Sun Yat-sen University, Kaohsiung, Taiwan. Wei-Che Tsai is Associate Professor of

Finance at National Sun Yat-sen University, Kaohsiung 80424, Taiwan. We sincerely thank Professor Peter

Bossaerts, Professor Hsiang-Hui Chu, Professor Sean Foley, Dr. Taewoo Kim, Professor Chi Sheh, Professor

Albert Wang, Professor Robert Webb, and an anonymous reviewer for their insightful comments and

suggestions. We also appreciate the suggestions from participants at the 2015 Auckland Finance Meeting, the

23rd Conference on the Theories and Practices of Securities and Financial Markets, the 11th Conference of

Asia-Pacific Association of Derivatives, the 2015 International Conference of Taiwan Finance Association,

and the 23rd Annual Conference on Pacific Basin Finance, Economics, Accounting and Management.

*Correspondenceauthor, Departmentof Finance, National Sun Yat-senUniversity, No. 70, LienhaiRoad, Kaohsiung

80424, Taiwan. Tel: þ886-7-525-2000ext. 4814, Fax: þ886-7-5250136, e-mail: weiche@mail.nsysu.edu.tw

Received September 2015; Accepted October 2016

The Journal of Futures Markets, Vol. 37, No. 9, 865–891 (2017)

© 2016 Wiley Periodicals, Inc.

Published online 22 December 2016 in Wiley Online Library (wileyonlinelibrary.com).

DOI: 10.1002/fut.21830

means an easy task. One of the most common and widely accepted measurement methods is

the probability of informed trading (PIN),

2

but despite being widely used in many prior

related studies, this measure is also well recognized for one specific shortcoming—its

inability to effectively capture short-lived information.

3

In an attempt to overcome the failure of the PIN measure to capture short-lived

information, Easley, Prado, and O’Hara (2012) developed the “volume-synchronized

probability of informed trading”(VPIN) to solve the data aggregation problem in high-

frequency observations; however, Andersen and Bondarenko (2014a,b, 2015) subsequently

showed that while the VPIN measure is correlated with trading volume and concurrent

volatility, when controlling for other factors, it was found to have no incremental predictive

power on future volatility. Thus, Andersen and Bondarenko (2015) concluded that the VPIN

measure may not be suitable for capturing informed trading activity since it was confounded

by other trading-related effects.

4

Chang, Chang, and Wang (2014) proposed a new measure aimed at providing a more

intuitive explanation and more user-friendly application for the measurement of informed

trading. They extended the Avramov, Chordia, and Goyal (2006) model to construct a proxy

measure of the “dynamic probability of informed trading”(DPIN), thus enabling researchers

to more easily estimate the probability of informed trading at much finer frequencies.

5

The advantages associated with the DPIN measure make it an attractive alternative

means of directly determining the information content of all types of transactions in the

market.However, as compared to other informationmeasures, it has attractedfar less attention

in recent studies,thus leaving an obvious gap within the extantliterature. This provides us with

the motivation to examine the validityof the DPIN measure and to contribute to the literature

by providing a better understanding of the information content of institutional trading.

A unique dataset obtained from the Taiwan Futures Exchange (TAIFEX) enables us to

estimate the DPINs of transactions undertaken by foreign institutional investors versus those

of domestic institutional investors. Given the general recognition that foreign institutional

investors seem to enjoy an information advantage over domestic institutional traders in a

local market, such as the TAIFEX, the advantages of our unique dataset make it particularly

1

Examples include Chakravarty (2001); Chiyachantana, Jain, Jiang, and Wood (2004); Dasgupta, Prat, and Verardo

(2011); Kraus and Stoll (1972); Puckett and Yan (2011); Saar (2001); and Yan and Zhang (2009). In one of the

earliest works in this field, Kraus and Stoll (1972) found that block trades could affect market efficiency, whereas the

examination of the influence of informed trading on medium-size trades undertaken three decades later by

Chakravarty (2001) provided confirmation of the stealth-trading hypothesis. Saar (2001) and Chiyachantana et al.

(2004) carried out investigations into the information content of institutional trades, with Dasgupta et al. (2011)

subsequently providing a theoretical equilibrium model to confirm the association between the herding behavior of

institutional investors and both short- and long-term returns.

2

The PIN measure, which was developed by Easley, Hvidkjaer, and O’Hara (2002) and Easley, Kiefer, O’Hara, and

Paperman (1996, 1997a,b) is in general used in many fields, such as corporate finance, investment, and market

microstructure; examples include Brockman and Yan (2009); Brown, Hillegeist, and Lo (2004), Chan, Menkveld,

and Yang (2008); Duarte and Young (2008); Vega (2006); Zhao and Chung (2006).

3

In order to estimate the PIN measure, it is necessary to aggregate intraday data (that is, data occurring at

approximately 5-minute intervals within the trading day) across multiple days (Easley et al., 1997a,b), with the

resultant estimate providing a measure of informed trading over very long horizons (from 1 month to 1 quarter). The

added disadvantage of such long horizons is the increased likelihood of the actual impact of short-lived private

information becoming diluted or masked by other factors.

4

Andersen and Bondarenko (2015) found that when VPIN is constructed from an accurate classification, it behaves

in a diametrically opposite way to the “bulk volume classification”-VPIN (BVC–VPIN), with the latter exhibiting

volatility forecasting power, solely because it generates systematic classification errors that are correlated with

trading volume and return volatility.

5

The DPIN measure is well-suited to capturing information at fine frequency intervals throughout a trading day,

such as 15-minute intervals. Such frequencies are more in line with the speed at which traders react to, and digest

information in, modern financial markets.

866 Weng et al.