Substantial regulatory and enforcement resources are spent combatting insider trading in financial markets. In 2020 alone, the US Securities and Exchange Commission (SEC) employed approximately 1,300 staff members in its Enforcement Division and committed $550 million in resources to investigating and prosecuting illegal insider trading.
While it is clear that insider trading occurs, opinions vary significantly as to the total amount of illegal insider trading and whether it has increased or decreased over time. For example, the former US Attorney for the Southern District of New York, Preet Bharara, suggests that insider is trading is “rampant”, and is undertaken by company insiders and hedge funds. On the other hand, SEC prosecutions provide a lower bound estimate of the extent of insider trading – approximately 50 insider trading cases are prosecuted by the SEC per annum. The problem is that we can’t directly observe insider trading cases that are not detected or never brought to prosecution, leading to the longstanding puzzle of how much illegal insider trading actually occurs in financial markets and what fraction is brought to prosecution.
Related questions include what are the characteristics of stocks in which insider trading occurs most often? And what makes it more likely that a given case will be detected and brought to prosecution? These questions have remained elusive because prosecuted cases of insider trading represent a non-random subset, not the entire population. This notion is illustrated by Comerton-Forde and Putniņš (2014), who find that only one in three hundred cases of closing price manipulation is detected and prosecuted by regulators, indicating that prosecutions are only the tip of the iceberg.
Summary of approach
In our recent paper, we estimate the underlying prevalence of insider trading using structural estimation models known as “detection controlled estimation” (DCE). The models address the issue that prosecuted cases are just a non-random sample of all instances of insider trading by jointly modelling both processes: insider trading and its detection/prosecution. The model recognizes that in the absence of a prosecution case, either no insider trading occurred, or insider trading occurred but was not detected and prosecuted and estimates probabilities of these outcomes. We estimate the models using all US prosecuted insider trading cases relating to mergers and acquisitions (M&A) and quarterly earnings announcements between 1996 and 2016. M&A and earnings announcements are the two types of information most frequently present in insider trading prosecutions.
A similar approach of structural estimation (DCE models) has been effective in estimating the underlying rates of breaches of occupational health and safety (Feinstein, 1989), tax evasion (Feinstein, 1991), corporate fraud (Wang, Winton, and Yu, 2010), market manipulation (Comerton-Forde and Putniņš, 2014), and illegal activity in crypto-currencies (Foley, Karlson, and Putniņš, 2019).
Prevalence of insider trading and its detection
Using our structural estimation approach, we estimate that insider trading occurs once in every five M&A and once in every twenty quarterly earnings announcements. Our findings indicate that prosecution cases are only the tip of the iceberg, as the actual estimated underlying rate of insider trading is at least four times larger. From 1996 to 2010, we find that the probability of insider trading increases for both M&A and earnings announcements, coinciding with increases in stock liquidity. We also conclude that the likelihood of insider trading fell following the introduction of the SEC Whistleblower Program in 2010.
For both M&A and earnings announcements, we estimate that the probability of detection/prosecution of insider trading is around 15%. This estimated rate is consistent with rational crime theories that suggest no rational individual would conduct insider trading if the likelihood of detection is high (Becker, 1968). Interestingly, we see that the probability of detection/prosecution increases throughout our sample period, corresponding with increases in regulatory resources over time and detection from the Whistleblower program.
Characteristics of insider trading and its detection
Our models also allow us to identify the determinants of insider trading. The likelihood of insider trading is higher for stocks with more liquidity. Higher levels of market liquidity allow individuals to trade strategically and hide among other traders as well as earn larger profits from their information. The likelihood of insider trading is also higher for more material information as measured by market reactions to the information. When the materiality of information is greater, it becomes more attractive for individuals to illegally trade as the potential profits are larger. For M&A, our research shows that the probability of insider trading is higher when the chances for information leakage are larger, such as deals with a large number of financial or legal advisors, and for stocks with larger levels of information asymmetry (greater information advantage of the insiders). All of these results are consistent with theories of rational crime, whereby potential offenders weigh the potential profits against the probability of detection and potential penalties.
The likelihood of detection and prosecution increases when regulatory budgets are larger and when regulators have a greater focus on enforcement (when the rate of future prosecutions is higher). Abnormal trading characteristics, including excess returns and volume prior to the M&A or earnings announcement, increase the probability that a given case of insider trading is detected and prosecuted.
Our results have a number of implications. By using an estimation approach that overcomes the issue that prosecuted cases are only a non-random subset of all insider trading, we have estimated the underlying rates of insider trading, how much of it gets brought to prosecution, and where/when is it most likely to occur. These estimates can be used by regulators to focus their enforcement efforts and make more efficient use of regulatory resources. For example, by understanding where insider trading is more likely to occur (e.g., which types of material news, stock characteristics or industries), for the same level of resources, regulators can detect more insider trading violations. The increased ability of regulators in identifying insider trading can have a deterrence effect on an individual’s decision to illegally trade.
Dr. Vinay Patel is a Senior Lecturer at the University of Technology Sydney
Dr. Talis Putniņš is a Professor of Finance at the University of Technology Sydney.
The post is adapted from their paper “How much insider trading happens in stock markets?” available here.