Post-Earnings Announcement Drift - Statistics Project Example

POST-EARNINGS ANNOUNCEMENT DRIFT Introduction Typically, event studies measure the effect of specified events on security value. Traditionally, event studies like in this case try to answer questions on whether a given event matters and whether the relevance of the information is impounded mainly into aspects such as prices. This paper entails an event study in which the establishment of whether returns of large-cap stocks within an industry are higher than those of small-cap stocks. These terms originate from the term market capitalization, abbreviated as market cap. It refers to the shares of a public company that have been issued. Market capitalization entails a multiplication of the total number that a company owns by the price of each share. Usually, companies announce their earnings resulting to various anomalies that are associated with such earnings. One of these anomalies include the announcement of post-earnings announcement drift. With respect to the perspectives of the efficient market hypothesis and the PEAD, this paper aims at testing the PEAD phenomenon on a non-American market, Greek market. The paper considers the availability of 80 companies selected randomly for assessment on how PEAD affects the Greek market. The data for the 80 companies has been obtained from secondary sources especially the internet (Vaios). Considering available statistics, the Athens’ Stock Exchange a daily announcement of earnings effect on the markets. The data used in this paper considers four SUE portfolios based on events’ quoted prices as one method of testing the PEAD phenomenon and examination of whether market over and under reaction usually exist through the use of event study methodology. Finally, this paper also classifies the sample firms regarding their response or exposure to the PEAD phenomenon. Data Description In this paper, the selection of the used sample was based on the consideration of all companies listed in the ASE. Out of the 264 firms listed in the ASE, 80 were selected from which their reporting of earnings from the year 2001 to 2008 (Vaios). Among the data sets that will be considered in this case include the quarterly earnings per share, corresponding announcement dates of the quarterly returns per share, and the closing prices of the stocks (Brown and Warner, 328). Besides the random selection of the companies, all without quarterly earnings per share were excluded (Vaios). The exclusion in this case involves the dates of announcements and, therefore, the consideration of annual returns per share as this would not show the major changes that occurred after each announcement. For instance, considering that positive announcements can affect markets for as long as 40 days in the Greek market, it is clear that annual EPS may not reflect any major changes if data was presented annually (Fama, 383-417). The Greek market capitalization comprises of 61.2 billion Euros from which 81% of the firms considered in this case (Vaios). The sectors making a bigger market capitalization include the banking sector which comprises of 28.45% with those making up the smallest capitalization being the chemical industry. Illustration 1 below shows the market capitalization per industry for the duration of 8 years from the start of 2001 to the end of 2008 (Vaios). Empirical Tests In the testing of PEAD phenomenon, most event studies use the Foster et al (574-603). methodology. The methodology uses the formula below: E (Qi,t) = Qi,t-4 + Øi (Qi,t-1 – Qi,t-5) + δi where Qi,t is the quarterly EPS of firm i within period t, δi is equal to (1 – Ø)u, with u covering for the mean of seasonally differences series (Vaios). Qi is an estimate provided by the first order autocorrelation coefficient (rl) (Forster et al., 574-603). In order to test PEAD using the Foster et al. (159) time series data is required for the sample considered in this case. The time series of quarterly ESP are needed in this case as the methodology equation comprises of Qi,t which denotes a firm’s quarterly earnings over time t, t being the three months of a quarter financial year. Consecutive earnings announcement are required within the time series. However, from the date series provided for the data series in this case, some dates are missing. To cater for this missing data, a portfolio formation method is considered (Fama, 383-417). The portfolio formation methodology is presented by Asimakopolous, Lambrinoudakis, Tsangarakis, and Tsiritakis (12-19) and its relevance is in the designing of a confidence interval around a desirable earnings median and then putting the actual earnings into character after being produced in time t while considering its position within the confidence interval (Brown and Warner, 305 – 340). Suppose that an event study considers the categorization of an earning as expected or otherwise within a quart t requiring the calculation of earnings median for three of the most recent earnings within the quarter – this would be denoted as Q(t – 1) to Q(t – 3). In order to present the calculations as statistical results, the testing of the three most recent quarterly, computation of standard deviation within the same period (Vaios). And E (Qt) denotes the earning’s average for t and σ denoting the standard deviation of the three observations required for the portfolio formation (Kothari and Warner). A confidence level is created through the above equations to result to the one below: When the realized earnings are part of the above confidence interval, then the earning is classified as expected, but it would be considered unexpected. Specifically, four distinguishable possibilities from this confidence interval can be identified (Vaios). i. It is expected earning it is within the confidence interval and if it’s bigger, the increase is expected as well considering the number of observations made. ii. Given that Q is bigger than E (Q) + δ (Q) limit is therefore classified as unexpected increase iii. Like with the first point, it the earnings from the observation Q but smaller than E (Q), then this is classified as expected decrease iv. Given that Q is smaller than E (Q) + δ (Q) limit, then this is characterized as unexpected decrease. Event study methodology The study methodology involves the methods and techniques used in the study process. These methods and techniques revolves around the ways of collecting data, the type of data used, and the techniques used to analyze the collected data. A number of methods are used for similar studies, but event study methodology is the preferred study method in this case. The major reason is that the paper tries to test for the reactions around the earnings announcements as depicted in the paper. Typically, a primary research is also important besides the use of secondary data and information since data obtained from first-hand sources depict the issue as it appears during the study process. In most case, not all people requested to participate are willing to take part. This means that the number of respondents willing to offer their services and the key issue are significant. Again, the hypotheses are critical aspects of the studies since are the central facets of the study tests. Both null and alternative hypotheses are first formulated to form the testing basis. In this case, the event study aims to test if the null hypothesis could be rejected or could not be rejected (Kothari and Warner). The null hypothesis is not rejected when the mean abnormal return at t equals to zero. Typically, the mean abnormal cross-sectional return (ARt) is calculated as: Where: e = unexpected return N= sample size The null hypothesis would not be rejected if ARt is equal to zero. In this case, the null hypothesis is that ARt is equal to zero and thus the returns are not statistically significant. If the null hypothesis is not equal to zero (H0 = ARt ≠ 0), the returns are said to be statistically significant. The null hypothesis, that ARt is equal to zero, is thus rejected. Data Analysis The data analysis in this study is done on market returns and industry returns. Illustrations for each case is presented accordingly. Each form of returns is analyzed as follows: 1. The Market Returns To proceed, the sample of firms used in this study is considered. Here, the sample is divided into the listed and the unlisted firms (Kothari and Warner). The sub-samples are then analyzed with reference to the returns’ calculation of model for the market returns, which is presented as: Ai,t = Ri,t – Rm,t Where: Rm,t = the ASE general index returns 2. The Industry Returns In essence, not all firms are included in the sample are obtained from the same industry. Again, each firm has a relatively unique set of defining characteristics. Thus, General Index would be highly useful in comparing the firms within the selected sample. To analyze the data, Microsoft Excel program could be used. The ways of initiating the analysis can be show in various systematic illustrations as shown in the appendix (The Case of Post Earnings Announcement Drift in Greece). Considering the case of abnormal unexpected increase, the time series uses the formula (Vaios): Āt = Ŝ (Āt) Where: Āt = average unexpected returns for all the considered days for all sample firms In order to calculate the standard deviation of the median, the formula is given as: Again, the difference in the formula is denoted by: (Vaios), Where: = median averages of the unexpected returns Further calculation result to (Vaios): This is presented in excel as shown in illustration 6 in the appendix. Conclusion From the empirical study undertaken in this paper, unin portfolio has the highest positive CARs showing the statistical significance. This statistical significance is observed a day after announcement and remains unchanged till the end of the examination period. When considering the unin portfolio under General Index, it is observed that a bigger CAR was recorded from the announcement date as compared to non-general Index portfolio. Under this observation, it can be concluded logically that companies belonging to G.I (general index) are under tighter scrutiny and PEAD phenomenon is a rare case. Thus, provided there are exceptions in the PEAD phenomenon, it is therefore conclusive that PEAD does not affect all firms as witnessed with the non-G.I firms. Works Cited Brown S. and Warner J. Using daily stock returns – The case of event studies. Journal of Financial Economics, Vol. 13: 305 – 340. 1985. Print. Fama E. Efficient Capital Markets: A review of theory and empirical work. The Journal of Finance. Vol. 25, No 2: 383-417. 1969. Print. Foster G., Olsen C., and Shevlin T. Earnings Release, Anomalies and the Behavior of Security Returns. The Accounting Review, Vol 59., No 4: 574-603. 1984. Print. Kothari S.P. and Warner B. Econometrics of Event Studies. Handbook of Corporate Finance: Empirical Corporate Finance Vol A, Ch. 1. 2006. Print. Vaios, Plakotis. Post Earnings Announcement Drift in Greece. Academic Report. Rotterdam, 2010. Print. APPENDICES Appendix 1: Data source screenshot Screenshot url (http://www.ase.gr/content/en/ann_list.asp?sbj=&fmd=01%2F02%2F2001&tod=31%2F12%2F2008&ast=3&at=1&stp=1&submit1=Go) Appendix 2: Illustration Screenshots(source: Vaios, Plakotis. Post Earnings Announcement Drift in Greece. Academic Report) Illustration 1: the Classification of Earning Observations into four Portfolios Illustration 2: Market prices for each firm since 2001 to 2008 (after each announcement date) Illustration 3: Prices Illustration 4: Returns Illustration 5: The Calculations Illustration 6: median averages of the unexpected returns Illustration 7: A projection of the entire observation data results to Read More