Document Type : Research Paper
Authors
Associate Professor of Economics, Department of Economics, Allemeh Tabataba’i University, Tehran, Iran.
Abstract
The Probability of Informed Trading (PIN) is one of the important measures of market microstructure that is generally used to estimate the level of information asymmetry. Estimating PIN can be challenging due to boundary solutions, local maxima, and Floating Point Exceptions (FPE). Additionally, the prevailing assumption of the existence of only one information layer per trading day in PIN is inconsistent with the real-world empirical evidence and exposes it to a considerable underestimation bias. In this paper, we estimate information asymmetry for 55 listed companies in the energy sector during the period from 1396:Q1 to 1402:Q1, utilizing the Multi-Layer Probability of Informed Trading (MPIN) model introduced by Ghachem and Ersan (2023). The findings indicate: First, the assumption of a single information layer is satisfied for only 2.67% of the 1,200 stock/season observations, which implies the necessity of using MPIN to estimate information asymmetry. Second, the use of PIN not only leads to significant underestimation bias, but also provides an inaccurate picture of the ranking of companies from the perspective of information asymmetry. Third, the energy sector faces an average information asymmetry of 34.4%, and estimations reveal that private information reached its peak in the summer of 2020, exceeding 49%. Fourth, the symbols "Bepeyvand" from the “electricity, gas, and steam” sector and "Shapna" from the refining sector hold the highest (64.75%) and lowest (18.9%) information asymmetry, respectively.
Introduction
The Probability of Informed Trading (PIN) is a prominent metric in market microstructure, utilized to assess the level of information asymmetry by estimating the probability of informed trading. Despite extensive international research on measuring information asymmetry and its applications across various domains, the Iranian academic landscape reveals two significant gaps. First, studies in this field remain limited in number. Second, corrective and generalized methodologies—whether in terms of estimation techniques or the underlying assumptions of the models—have not been sufficiently explored in domestic research.
The primary contribution of this study is to address this research gap and provide a more precise representation of information asymmetry levels within the stock market of companies operating in the energy sector. This investigation is structured around three central research questions: First, what differences exist between the average estimated levels of asymmetric information in the energy industry when using the PIN and MPIN models? Second, what is the magnitude of bias in the estimated asymmetric information derived from the PIN model across energy subsectors, including "chemical and petrochemical," "refining," and "electricity, gas, and steam"? Third, which companies in the energy sector exhibited the highest levels of asymmetric information, as estimated by the MPIN model, during each quarter from the first quarter of 1396 to the first quarter of 1402?
To answer these questions, asymmetric information will be estimated using the PIN and MPIN models, which are rooted in the extended methodologies proposed by Ersan and Alici (2016) and Ghachem and Ersan (2023). This approach aims to enhance the accuracy and reliability of the findings, contributing to a deeper understanding of asymmetric information dynamics in the energy sector.
Methods and Materials
To estimate information asymmetry in this study, high-frequency daily stock data from 55 companies operating in the energy sector were collected. These companies are categorized into three subsectors: "chemical and petrochemical," "refining," and "electricity, gas, and steam." The data spans the period from the first quarter of 1396 to the first quarter of 1402 and was sourced from the Tehran Stock Exchange website. The data was subsequently cleaned using Python software. The selection of these 55 companies was based on two critical criteria: the availability of high-quality, high-frequency data (with minimal non-trading days) and the inclusion of a diverse range of companies of varying sizes.
At any given moment, a vast number of bid and ask quotes exist at different price levels. Therefore, the initial step involved collecting order data, which amounted to over 30 billiard rows of data for each of the 55 symbols across the 25 quarters under study. Due to the irregular timing of trades, the bid and ask quotes were aggregated into one-second intervals. Subsequently, price data was aggregated using a weighted average, while trade volume data was summed up within each one-second interval. Following this, the traded prices and volumes were matched with the corresponding bid and ask quotes.
The second-by-second data for each day were processed using the Lee and Ready (1996) algorithm to identify the origin of each trade (i.e., whether it was buyer-initiated or seller-initiated). Finally, within the frameworks proposed by Ersan and Alici (2016) and Ghachem and Ersan (2023), the parameters of the PIN and MPIN models were estimated, respectively. These parameters were used to calculate the probability of informed trading for each quarter. The likelihood function was constructed separately for each company and each quarter, and the estimation was performed using parallel processing on a Core i9 processor in the R software environment.
Results and Discussion
In Figure (1), the average estimated values of PIN and MPIN for 1,387 quarter/stock pairs are presented, while Figure (2) illustrates the trend of PIN bias alongside the number of layers identified in the MPIN model. The findings reveal the following:
Figure (2). Average PIN bias and the number of layers identified in MPIN
Figure (1). Average values of PIN and MPIN in the energy industry from 1Q:1396 to 1Q:1402
The average PIN ranges between approximately 14% and 37%, while the average MPIN fluctuates between 21% and 51%.
The average MPIN values are consistently higher than PIN values across all quarters, with MPIN being, on average, approximately 46% higher than PIN. This observation suggests a higher probability of encountering informed traders when using the MPIN model, which aligns with theoretical expectations.
The average values of MPIN and PIN exhibit similar patterns of fluctuation over time. However, the difference between the two is not constant. Although they began the study period with relatively similar values, the gap between them gradually widened. As shown in Figure (2), the underestimation of PIN increased from approximately 5% in the first and second quarters of 1396 to around 12% in the first quarter of 1402. The highest bias was observed in the fourth quarter of 1398, where PIN underestimated information asymmetry by 18%.
These results highlight the importance of using the MPIN model for a more accurate estimation of information asymmetry, particularly in dynamic and complicated markets such as the energy sector. The increasing divergence between PIN and MPIN over time underscores the limitations of the PIN model in capturing the full extent of informed trading, especially in periods of heightened market activity or volatility.
Conclusion
The estimated PIN values in the energy industry fluctuate between 14% and 37% across different quarters, with an average of 22.9%. Among the three energy subsectors, the refining industry exhibits the lowest level of information asymmetry (20.4%), while the chemical and petrochemical sector shows the highest (23.5%). The electricity, gas, and steam subsector has an information asymmetry level of 23.1%. The estimated PIN values in the Iranian energy sector are higher than those reported by Easley et al. (2002) for the U.S. stock market (approximately 19.1%) and Hwang et al. (2013) for the South Korean stock market (20.1%), but are close to the estimates by Martins and Paulo (2014) for the Brazilian stock market (25%).
Given that the level of information asymmetry in Iran is relatively high compared to estimates from other countries and has been increasing over time, and considering that private information is more prevalent in subsectors with smaller market shares (such as electricity, gas, and steam) and smaller companies (e.g., Bepeyvand, Begilan, etc.), it is recommended that the Securities and Exchange Organization of Iran enhance its oversight of smaller companies and industries. These entities should be required to improve transparency by promptly disclosing material information that impacts current and future revenues and costs.
Additionally, analyzing the effectiveness of policies, particularly trading restrictions such as volume limits and price fluctuation limits—which are implemented in various countries to reduce information asymmetry and enhance retail investor confidence—could serve as a focus for future complementary research. Such studies would aid in designing and implementing optimal policies to address information asymmetry and improve market efficiency.
Keywords: Market Microstructure, Information Asymmetry, Multi-Layer Probability of Informed Trading (MPIN), Hierarchical Agglomerative Clustering (HAC)
JEL Classification: C13، G10، G14
Keywords
- Market Microstructure
- Information Asymmetry
- Multi-Layer Probability of Informed Trading (MPIN)
- Hierarchical Agglomerative Clustering (HAC)
Main Subjects
)