10th Global Conference on Business & Economics ISBN : 978-0-9830452-1-2
The relevance of Annual General Meetings in stock returns, trading volumes and volatility: Evidence from Spain
Authors:
Josep García Blandón[*]
Mònica Martínez Blasco
Lucinio González Sabaté
Facultat d’Economia, IQS, Universitat Ramon Llull
The relevance of Annual General Meetings in stock returns, trading volumes and volatility: Evidence from Spain
ABSTRACT
Although the investigation of the effects of corporate events on stock prices is a well established line of research in accounting and finance, very little attention has been devoted to one of the most important corporate events: the Annual General Meeting (AGM). The effects of AGM on stock returns will largely depend on the relevance of the information released to the market as well as on the level of efficiency of the financial market. In this paper, we have investigated the effects of AGM on stock returns, volatility and trading volumes, in the Spanish stock market. Our results indicate that AGM do not have significant effects in any of the three indicators, either on AGM days or nearby days. After the exam of the possible explanations, we conclude that no relevant information seems to be released to the market during AGM.
JEL: G21.
Key-words: event studies; Annual General Meeting; stock returns; volatility; trading volumes.
1. Introduction
The reaction of stock prices to information releases during public announcements constitutes a well established line of research in financial economics. Researchers, nevertheless, do not agree either about the real magnitude of the reaction, or about the implications for the Efficient Market Hypothesis (EMH). Regarding the last point, abnormal returns around company events have been usually interpreted as evidence against the EMH. Fama (1998) argues, however, that event studies methodology that investigates the reaction of stock prices to specific company events can not be properly used to test the EMH. The reason would be that, while this methodology assumes that any lag in the response of prices to an event is short-lived, returns should be examined over long time horizons in order to discuss about market efficiency. The author concludes that, since the literature does not clearly identify overreaction or underreaction of stock prices as the dominant phenomenon, the observed random split between over and under-reactions would not question the EMH. In the same line, Bhattacharya et al. (2000) discus about the difficulties to interpret the lack of reaction of stock prices to company events, in terms of the EMH. Four possible, and sometimes contradictory, situations were compatible with this behavior, by combining the concepts of market efficiency and the relevance of information releases: 1) the market is inefficient, and thus prices do not react to the arrival of relevant information; 2) companies do not make value-relevant corporate announcements; 3) the stock market is efficient and the news are value-relevant, but this information has been already completely anticipated by the market; and 4) insider trading prohibitions do not either exist or are not enforced and thus the superior information of insider traders has been incorporated to stock prices through their trades. The authors point it out about the importance of examining the behavior of stock prices during pre-announcement periods in order to assess the likelihood of each one of the former situations.
The effects of corporate events in the behavior of stock prices is a well establish line of research in accounting and finance. Numerous studies have investigated the reaction of stock prices to a great variety of corporate events, being earning announcements the most popular one. Some examples are the seminal paper of Beaver (1968), Aharony and Swary (1980), Ball and Kothary (1991), Abarbanell and Bernard (1992), and more recently Landsman and Maydew (2002) and Landsmand et al. (2002). Dividend announcements are another strongly investigated company event. A short list of papers dealing with the reaction of stock prices to company dividend announcements must include Watts (1973), Denis et al. (1994) and Michaely et al. (1995). Other examples of company events that have received important attention in the literature would be: stock splits (e.g. Lamoureux and Poon, 1987 and Ikemberry et al., 1996) corporate news (e.g. Battacharya et al., 2001; Chan, 2003; Frazzini, 2006 and Kothary et al., 2008) and executive compensation plans (Tehranian and Waegelein, 1985 and Gaver et al., 1992). In addition to these strongly investigated events, we can also find examples of other, somehow, more extraordinary ones, as for example, the reaction of stock prices to auditor switches (Hong, 1992) or to sudden executive death (Johnson et al., 1985). All the mentioned events have in common the release of potentially relevant information to the market. It is, therefore, quite surprising, that one of the most important company event, as it is the Annual General Meeting (AGM) has received almost no attention in the literature. During these meetings the top executives of the company address not only shareholders but the whole financial community. There are certain decisions that can only be approved on the AGM, as for example the election of the Board of Directors, and important managerial announcements, usually concerning the managers views about the company prospects, are usually made during these meetings. We have only found two previous investigations on the issue1: Brickley (1985) and Olibe (2002). The former addresses the effects of AGM on returns, without considering trading volumes or volatility, while the latter proceeds in the opposite way, investigating trading volumes and volatility, but without considering stock returns. Brickley (1985) conducted his research for a random sample of U.S. firms, reporting significantly positive abnormal returns around shareholder meeting dates. Nevertheless, the author complains that the lack of comparable investigations makes it difficult to interpret his results in the framework of previous research. More recently, Olibe (2002) investigates the effects of AGM in U.K. based companies, listed in the U.S. market. The author reports particularly high levels of volatility in stock returns on AGM days. Nevertheless, the effect of AGM on trading volumes is minimal, suggesting that U.S. investors do not generally find AGM informative.
Regarding the theoretical foundations of the expected relationship between AGM and stock prices, we can adopt the standard framework used in the literature to analyse the reaction of stock prices to any particular corporate event implying the release of potentially relevant information to the market. In particular, we can extend the explanation proposed by Kalay and Loewenstein (1985) to the reported abnormally high returns on dividend announcement dates. The authors interpreted this finding in terms of the increase in expected return and risk associated to predictable events that would likely generate new information. In such cases, the risk per unit of time on common stock would not remain constant over time but increase on the day of the event. Similarly as dividend announcements, AGM dates are also known in advance by market participants, and both situations imply the release of company information to the market.
Another theoretical approach to analyze the importance of annual general meetings in stock returns relies on information asymmetries. It could be argued that, due to the fact that AGM constitutes a media event, company executives would prefer to avoid communicating bad news during AGM, since it could have a stronger impact on the firm’s market value than if this bad news were communicated on other days. Following this line, Kothary et al. (2008) suggest that managers can time the release of bad and good company news. The rationality of this behavior would rely on the agency theory, and particularly on the existence of information asymmetry between managers and investors. One example of the management of information is provided by Frankel et al (1995) reporting that managers tend to make public good news about the company prior to the issue of new stock. Similarly, Yermack (1997) observes that managers tend to accelerate bad news or withhold good news prior to the grant of stock options, to lower the price of the stock and consequently the strike price of the option. More recently, Graham et al. (2005) conclude that financial executives managed financial reporting practices in order to influence the price of the stock. The authors conducted a survey about the factors driving reported earnings and disclosure decisions. In particular, they point out managers’ strong incentives to withhold bad news, with the hope that the situation reverse in the nearby future and thus they will never have to release this bad news. Kothary et al. (2008) explicitly mentions the recent case by the European Aeronautic Defence and Space Company (EADS) involving the new Airbus A-380, as an example of this situation2.
Another example of the behavior of managers regarding the release of company news is the tendency to release good news while the stock markets are open and bad news after the closing of the markets (Patell and Wolfson, 1982). Supporting this view, the so-called Monday effect, generally defined as returns being abnormally low on Mondays, is usually explained by the tendency to release bad news during the weekend, when the markets are close. Although no paper, to our knowledge, has investigated the withholding of bad news prior or during AGM, when the company receives an unusual attention by the media, such a behavior could be also explained in similar terms.
According with the previous discussion, both the Kalay and Loewenstain’s argument as well as the management of the release of company news could explain positive abnormal returns on AGM dates.
In this paper, we investigate the behavior of stock returns, trading volumes and returns volatility around AGM dates in the Spanish stock market. We have used the classical Brown and Warner (1985) methodology for event studies. As it has been already mentioned, we have found only two previous comparable investigations on the issue, one carried out with a random sample of U.S. companies for the period 1978-82, and the other with U.K. companies traded on the NYSE and AMEX from 1994 to 1998. With so little previous research on the issue, additional empirical evidence should be welcomed in order to draw sounder conclusions about the impact of shareholders meetings on stock prices and the possible explanations. In addition, since the transmission of information contended in company events to stock prices will largely depend on the microstructure of financial markets, evidence reported for one particular market should not be immediately translated to another. Besides, Brickley (1985) focuses only on the behavior of returns without analysing volumes or volatility, while Olibe (2002) do the opposite. Unlike both papers, we have investigated not only returns, but also returns volatility and trading volumes. This more comprehensive approach will allow a better understanding of the causes behind the behavior of stock returns during AGM. The non-significance, for instance, of abnormal returns during AGM dates will have different implications if are followed by an increase in volatility and/or trading volumes.
The remaining of the paper is as follows: next section discusses the methodology and dataset we propose to investigate the relevance of AGM on stock returns, volatility and trading volumes in the Spanish stock market. Finally in sections 3 and 4 we discuss our results and present the main conclusions.
2. Methodology and sample selection
In this section we present the methodology we propose to analyze the behavior of stock returns, returns volatility and trading volumes around AGM dates, as well as the sample and dataset used in our investigation.
2.1. Methodology
We have followed the classical Brown and Warner (1985) event studies methodology. Accordingly, abnormal returns (AR) have been computed as the difference between actual and normal returns, while normal returns are defined as expected returns without conditioning on the event occurrence.
Thus, abnormal return for stock i on day t is expressed as,
ARit = Rit – E(Rit|Xt) (1)
Where ARit is the abnormal return of stock i on day t, Rit the actual return, adjusted by dividends and stock splits, calculated in the usual way as ln((Pt+Dt)/Pt-1), where Pt and Dt are the closing price and the dividend paid on day t respectively, and E(Rit/Xt) the expected return for day t. Finally, Xt is the conditioning information set for the expected return on day t. Expected or normal returns have been computed through the market model. The event window and estimation period are given by the intervals [-5, +5] and [-90, -20], respectively, with day 0 the AGM day. Although papers on event studies tend to use wider event windows (eg Olibe (2002) uses the interval [-10, +10]), the effects, when they exists, are systematically detected nearby the event day. In addition, the aim of event studies methodology is clearly short-term. Therefore, we have investigated the effects of AGM in a five days period around AGM dates, while normal daily returns have been computed through an estimation of the market model for the seventy one days period ending twenty days before AGM. Market returns have been computed through the Indice IBEX-35, the most relevant index in the Spanish stock market, formed by the thirty five most liquid companies quoted in the Spanish Stock Exchange. Normal daily returns for each day within the event window have been estimated through ordinary least squares.
After estimating daily average abnormal returns (AAR) for each stock i, the average abnormal return on day t, has been calculated as:
(2)
Cumulative average abnormal return (CAAR) has been computed by adding AAR for different intervals through the event window, as showed by expression (3).
(3)
Our first null hypothesis states that average abnormal returns will be zero on AGM dates. We have performed two statistical tests, the parametric t-test and the nonparametric rank test, in order to decide about the rejection of the null hypothesis for each day within the event window.
The t-test is the standard procedure to test the null hypothesis in event studies. Accordingly, our null hypothesis has been tested through the t-test at the standard 5% and 1% significance levels. Brown and Warner (1985) discuss the implications of the statistical properties of daily stock returns for the event studies methodology, in particular departures from normality and the non-constantness of returns variance across days. The problem of the lack of normality of daily returns is particularly serious for small samples, since the Central Limit Theorem demonstrates that if excess returns are independent and identically distributed drawings from infinite variance distributions, the distribution of the simple mean excess return converges to normality as the number of securities increases. The implications of non-constant returns variance is also emphasized by Corrado (1989), who argues that the increase in the variance of day 0 returns distribution constitutes a major weakness in the performance of the t-test, since a variance change will significantly increase the probability of type I errors (to reject the null hypothesis when this hypothesis is true). This fact will cause significance levels in event studies to be overstated when an increase in day 0 returns variance occurs. Corrado proposes the nonparametric rank test (better known since then as Corrado test), to overcome this shortcoming. Similarly as Brown and Warner (1985), he found that doubling the day 0 returns variance causes severe misspecification for the t-test statistics, since it more than doubles the probability of a type I error. On the contrary, the rank test remains relatively unaffected by these misspecification problems. In addition, Corrado (1989) also concludes that the rank test is better specified under the null hypothesis and more powerful under the alternative hypothesis. Accordingly, regardless our relatively large number of events we have performed both, the t-test and the Corrado rank test.