The US Supreme Court judgment in Basic v. Levinson1; recognised the concept of market efficiency, endorsed the fraud-on-the-market doctrine as the foundation for securities fraud cases and established the use of economic event studies as the cornerstones for expert evidence to assess loss causation and quantification of damages. With the decision in Halliburton II2, which held that defendants may rebut the presumption of reliance at the class certification stage, the importance of event studies and the ability of plaintiffs and defendants to demonstrate or reject the effect of corrective disclosures on the share price of the company has only increased.
Through the use of an event study, researchers aim to analyse the influence (or the absence) of a specific market event or news release on the price of financial securities. Assuming market efficiency the price of a security will at any time incorporate all publicly available information and new information will (almost) immediately be reflected in the form of price change. Measuring the stock price change over short periods of time can therefore be used to identify and quantify the relevance of information received during each of the periods.
While event studies have become a widely accepted method in academic research to test the hypothesis of market efficiency, concerns have been voiced about the use of event studies in securities litigation. The criticism broadly relates to two areas:
Stock price movements deviate from standard statistical assumptions
Daily stock price changes exhibit statistical properties which deviate from the assumptions used in standard regression analysis, especially the assumption of normally distributed price changes. In particular, daily stock prices show periods of low and high volatility (volatility clustering) and significantly more extreme moves than predicted by the normal distribution (fat tails).3 To address these issues advanced regression techniques and statistical methods are available which do not rely on the normality assumption and can be used to obtain robust estimations and test statistics.
Shortcomings of test design
A recently published research paper4 highlights several important issues which should be considered when relying on event studies in legal securities fraud cases:
Low statistical power of test statistics
Event studies determine the daily ‘abnormal return’ of individual stock prices after controlling for general market and industry effects. Because stock prices move continuously based on all kind of idiosyncratic information and technical trading conditions, daily returns deviate all the time from their expected value implied from the general market or industry index. Consequently, it may be difficult to identify the effect of a corrective disclosure on the stock price at a 95% statistical significance level5; unless this impact is “quite large” and peaks out enough between the daily fluctuations. The currently used approach to event studies therefore has the potential to reject the existence of a price impact caused by a corrective disclosure even if such impact was actually present6.
Solution: To fully appreciate the results of any event study it is necessary to consider the ‘statistical power’ of the test used and not only the probability of falsely confirming that a particular disclosure had a price impact when it actually did not7. This issue is often not addressed in securities litigation.
From a statistical perspective it is difficult, if not impossible, to separate the impact of ‘general information’ arriving throughout the day from the effect of a specific piece of information like a corrective disclosure. This issue has implications on the power of the statistics used as well as on the estimate of the quantum of the impact resulting from such corrective disclosure and consequently the amount of damages associated with the initial misrepresentation.
Solution: Detailed analysis of intraday price developments can be a way to isolate the effect of specific events further. At the same time a detailed comparison with the price development of shares from the same industry or geography can help to average out the impact of general news.
The implication of low test power and confounding effects is that damage estimates derived from event studies may overstate the actual loss caused by a corrective disclosure. This is because in the case that an abnormal return actually exceeds the high threshold of 95% it cannot be subsequently decomposed into the two parts to isolate solely the effect of the corrective disclosure.
Solution: Together with a detailed analysis of confounding effects the estimation bias needs to be carefully considered when quantifying damages.
When properly used event studies are powerful methods of analysing the effect of information and corporate announcements like corrective disclosures on the price of the relevant securities. However, a thorough understanding of the applied methods is necessary in order not to draw unwarranted conclusions.
Fideres is offering clients a web-based interface to perform preliminary event studies and to perform a range of tests to analyse stock price performance around disclosure dates using tailored assumptions.
Please contact us at firstname.lastname@example.org if you would like access to this online platform.
1 Basic, Inc. v. Levinson, 485 U.S. 224 (1988)↩
2 Halliburton Co. v. Erica P. John Fund, Inc., 134 S. Ct. 2398↩
3 ‘volatility clustering’ and ‘fat tails’ are technically referred to as heteroscedasticity and excess kurtosis↩
4 Brav, Heaton, Event Studies in Securities Litigation: Low Power, Confounding Effects, and Bias, March ↩
5 The 95% statistical significance level is widely accepted as ‘the standard’ in event studies used in securities litigation cases.↩
6 Rejecting an effect despite such effect actually being present is referred to as ‘Type II Error’ or ‘false negative’↩
7 Falsely confirming an effect that actually is not present is referred to as ‘Type I Error’ or ‘false positive’↩