Event Studies (Finance)

The term “event study” describes an empirical research design widely used in finance and accounting. Event studies employ a common general methodology aimed at studying the impact of specified economic or financial events on security market behavior. The occurrence of an event is used as the sampling criterion and the objective of the research is to identify information flows and market behavior both before and after the event. Although some event studies have examined the volatility of return s and patterns of trading volume surrounding events (for a review see Yadav, 1992), most studies have focused on an event and its impact on the market prices of securities. Price-based event studies were originally designed to test the semi-strong form of the efficient market hypothesis (Fama et al., 1969) with the expectation that efficiency would be reflected in a full and immediate response to the new information conveyed by the event. In the mid- 1970s a new type of price-based approach was developed (Mandelker, 1974; Dodd and Ruback, 1977) called value event studies; their main aim was not to study market efficiency but to examine the impact of events on the market value of specific companies (or groups of companies).

The scope of events studied ranges from firm-specific incidents (e.g. announcements of stock splits, or changes in dividend policy) to more general phenomena such as regulatory changes or economic shocks. Analysis occurs over “event windows” or test periods when evidence of abnormal behavior in the market is sought. Such abnormality occurs relative to behavior during an estimation or benchmark period, which is used to estimate the benchmark for the expected behavior of a parameter around the event. Abnormality can occur in the form of abnormal returns, abnormal trading volumes or changes in the levels of the volatility of returns. The research methodologies used in each case are similar, differing only in respect of evaluating criterion. Accordingly, the brief description that follows will take into account only the general price-based event studies.

Formally, abnormal return is the difference between the actual and expected return during the test period:


where ARit is the abnormal return on security i during period t, Rit the actual return on security i during period t and R*u is ^ expected return on security i during period t. Several alternatives exist to determine the expected return. The market model approach uses a regression analysis (usually OLS) to estimate the security returns as a function of the market index during the estimation period and then uses this model in conjunction with the actual market return during the test period to calculate the expected return. In this case, the classic configuration of the expected return generating model is the following (Famaet al., 1969):


where Rmt is the return on the market index for period t (systematic component of return), a is the intercept coefficient and P is the slope coefficient for security i, uit is the zero mean disturbance term for the return on security i during period t (unsystematic component of return) and T is the number of (sub-)periods during the benchmark period.

The model does not imply the acceptance of any explicit assumptions about equilibrium prices. This fact, and the specific design characteristics, which allow for an easy and powerful statistical treatment, constitute the main reasons for its wide popularity. The alternative mean adjusted method assumes that the best predictor for a security’s return is given by historic performance. This assumption implies that each security’s expected return is a constant given by its average return during the estimation period:


where Ru is the return on security i over the T(sub-)periods of the estimation interval.

The market-adjusted return method assumes that the expected market return constitutes the best predictor for each security’s market performance. The market return on the index during the test period is then the predicted return for each security:


Finally, CAPM based benchmarks define the expected return of each security as a function of its systematic risk or P and of the market price of risk, effectively the difference between the return on the market index and the return on the risk-free security:


where R# is the risk-free rate of return during period t and P is the systematic risk of the security i (previously estimated with reference to the market index).

A variant of this approach uses a control portfolio benchmark, under which the expected return of a specific security or group of securities is given by the expected return of a portfolio with the same P

The estimation of expected returns is usually considered the main source of variations in event study methodology. Other aspects of the methodology are: (1) the reference basis used to calculate the returns (logarithmic or discrete); (2) the measurement interval (the more common are monthly, weekly or daily returns); (3) treatment of disturbancies during the event window; (4) the duration of the event window; and (5) the choice of market index (where used).

To reflect the uncertain holding period preand post-event, it is usual to present the abnormal return in both periodic return form and as cumulative abnormal returns (CAR). The hypothesis normally tested then becomes whether CARs during the test period are significantly different from zero.

Some of the recent developments in event studies are: (1) the application of the methodology to the market for debt securities (Crabbe and Post, 1994); (2) the study of the likely implications of non-constant volatility on abnormal return estimates (Boehmer et al., 1991); (3) the employment of non-parametric tests of abnormal returns when the usual assumption of normally distributed returns seems problematic (Corrado, 1989); (4) and the implementation of multiple regression approaches based on the application of joint generalized least squares (GLS) techniques (Be rnard, 1987).

The volume of event study literature has grown significantly in recent years and shows every sign of continued expansion. At the theoretical level, two topics for continuing research are the control for extra market effects in the securities return generating processes; and the handling of statistical problems caused by samples of thinly traded securities. At the empirical level, the great challenge is accounting for the observed abnormal returns.

Next post:

Previous post: