Media events and stock returns : the case study of the effect of E3 and Gamescom on stock returns from 2009-2015 for five companies

(1)

Name: Quinten Duin Student number: 10658300 Supervisor: Yumei Wang

Media events and stock returns:

The case study of the effect of E3 and Gamescom on

stock returns from 2009-2015 for five companies.

Abstract:

This paper examines the effect of media events on stock returns from in the videogame industry for five companies from 2009 till 2015. I use event study methodology to study the events E3 and Gamescom effect on stock returns in the videogame industry from 2009-2015. In this study both the market-model and the Fama-French three factor model are used to calculate abnormal returns. The results show positive significant abnormal returns for the event Gamescom using both models. Further results show significant positive abnormal returns for the E3 announcement using the three factor model and significant negative abnormal returns for the Gamescom announcement using both models. The negative abnormal returns are in contrast with findings from earlier papers.

(2)

Statement of Originality

This document is written by Student Quinten Duin who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1. Introduction

The videogame industry is one of the largest entertainment industries. In the USA alone there are 155 million people playing videogames who spend a total of $22,41 billion on games. Video game publishers use different sources to announce games and distribute information. One of the ways is by releasing information at a large media event. One of these events is the E3. The electronic entertainment expo (E3) is the largest business-to-business video game expo organized by the entertainment software association. In 2015 the E3 had over 300 exhibitors, over 52,000 professionals attending and the live stream coverage reached 21 million unique viewers. Another media event is the Gamescom held in Cologne Germany. The Gamescom is the largest open-public video game expo with 345.000 consumers and 33,200 professionals attending the 2015 edition. It is interesting to examine the effect of the effect of these events on reach on stocks. Lately this effect of media on stock prices picked up interest from multiple researchers. This study contributes to this research by examining the effect of the E3 and Gamescom on the stock prices for five companies in the video game industry from 2009 till 2015

Previous research in this field is done by Fang and Peress (2009), who determine the cross-sectional relation between media coverage and stock returns. Their results show higher returns for stocks without media coverage, even after controlling for risk factors. Further research from Ferguson et al. (2015) studies the effect of both tone and volume of media coverage. They find a relation with stock returns for both tone and volume. Their results show higher returns after positive media content and lower returns after negative content. However, the effect of volume is stronger than the effect of tone. Another study in this field is done by Ikenberry and Ramnath (2002). Their study examines the effect of self-selected news events on stock returns. They study these events using event study methodology and show positive abnormal returns after these events. However, they argue that stocks underreact to these self-selected news events. Antweiler and Frank (2005) show contrary results. Their results show positive abnormal returns followed by negative abnormal returns after news stories. This shift implies an overreaction of stocks to news stories.

Prior research into pre-announcements is done by Sorescu et al. (2007) using event study methodology. Their results show positive abnormal returns after product pre-announcements. However, the short-run results are only significant if the pre-announcement contains specific information about the product. In a related study Koku et al. (1997) showed similar results. They likewise find significant positive abnormal returns after a product pre-announcement. Lastly, the study from Mishra and Bhabra (2001) shows corresponding results. Their event study results show positive significant abnormal results for credible product pre-announcements. This is similar to the results from Sorescu et al. (2007).

This thesis conducts an event study using cumulative abnormal returns to examine the effect of the event and the pre-announcement on stock returns. The results show significant positive abnormal returns for the event Gamescom. In addition, the results show significant negative abnormal returns for the pre-announcement of Gamescom using both the market model and the Fama-French three factor model. For the pre-announcement of E3 the results show significant positive abnormal returns using the Fama-French three factor model.

The remainder of the paper is ordered in the following way. Part 2 examines the existing literature. Part 3 explains the research method. Part 4 shows and discusses the results and part 5 concludes.

(4)

2. Literature review

2.1 Media coverage and stock returns

Mass media have a large effect on the spreading of information to the public. Because of the large reach, spreading company news through mass media could influence stocks. This effect of media on stocks has grabbed several researchers’ attention. Earlier researches related to mine include Fang and Peress (2009), who use the degree of media coverage to determine the effect of media on stock returns. Fang and Peress separate stocks in portfolios of no, low and high media-coverage. The medium of the number of news articles is used to divide the stocks into high and low groups (Fang & Peress, 2009, p. 2032). The event period in this thesis is considered to be a period with high media coverage. The results from Fang and Peress (2009) show that stocks without media coverage gain over 0,65-1% higher monthly returns than stocks with high media coverage. Similar to my study, Fang and Peress control their results for various risk factors like size and book-to-market value from the Fama-French three factor model (2009, p. 2034). They show that their results hold after controlling for these risk-factors. The results Fang and Peress find are stronger among small stocks (2009, p. 2049). The distinction with this thesis is that Fang & Peress examine the cross-sectional difference between stocks with and without media coverage, whereas this thesis compares stocks around an event with high media coverage to a time without an event.

A second related paper is from Ikenberry and Ramnath (2002). This paper examines the underreaction to self-selected news events. In this paper they consider stock splits to be a self-selected news event. Ikenberry and Ramnath calculate abnormal returns using two different methods. First, they use a buy-and-hold strategy comparing returns of splitting firms to a single control firm (Ikenberry & Ramnath, 2002, p. 495). Secondly, they calculate abnormal returns using a calendar-time approach (Ikenberry & Ramnath, 2002, p. 519). For the calendar-time approach they use both the ordinary least squares (OLS) and the weighted least squares (WLS) method. These two regressions are based on control variables size and book-to-market value from the Fama-French three factor model, and on the momentum factor as suggested by Carhart (Ikenberry & Ramnath, 2002, p. 519). Their results show significant positive abnormal returns using both methods. The buy-and-hold method shows significant abnormal returns of 9% in the year following the self-selected news event. However, they argue that the return evidence implies that markets underreact to self-selected news events (Ikenberry & Ramnath, 2002, p. 522). On the contrary, Antweiler and Frank (2005) find that stocks typically overreact to corporate news stories in the US. Similar to my paper, this paper uses event study methodology. Antweiler & Frank (2005) examine the reaction of stocks to corporate news stories. Their empirical analysis uses abnormal returns. The results show that the returns after an event shift from positive to negative. This reversal regularly exceeds the positive returns. This indicates the process of overreaction (Antweiler & Frank, 2005, p. 20). However, they argue that the results are sensitive to the assumed event window. Their results show that it could take many days or weeks before the stock market assimilates all the reported news (Antweiler & Frank, 2005, p. 21).

More recent research on news and stock returns is done by Ferguson et al. (2015), who examine both the tone and volume of news on stock returns. Their results show that both tone and volume of news media content affects stock returns (Ferguson et al., 2015, p. 29). However, they find stronger results for volume than tone. Ferguson et al. show that positive media content predicts higher returns, and negative content predicts lower returns (2015, p. 29). Their paper shows that using a news-based trading strategy yields significant daily abnormal returns of 14.2 to 19 basis points (Ferguson et al., 2015, p. 29). Similar to this

(5)

thesis, they use the factors size and book-to-market value from the Fama-French three factor model and momentum from Carhart to regress the abnormal returns.

Further research is done by Chan (2003). He compares returns after public news to stocks with similar returns without public news. He finds that there is a strong post-event drift after bad news (Chan, 2003, p. 33). His results show reversal for stocks with extreme returns without news headlines, while news stocks do not reverse (Chan, 2003, p. 18). Similar to my study, Chan uses the Fama-French three factor model to control for risk factors. The results still hold after controlling for these risk factors. The found results are especially seen for small stocks (Chan, 2003, p. 33). This is in line with the results from Fang and Peress (2014).

In a similar research Barber and Odean examine attention grabbing stocks. The research from Barber and Odean (2008) shows that individual investors have trouble choosing stocks to buy, and that individual investors are more likely to buy stocks that grab attention. These are stocks with for example news attention. Their results show that attention-grabbing events are followed by extreme one-day returns (Barber & Odean, 2008, p. 24). The individual investors do not face search problems when selling stocks. They show that attention-grabbing stocks bought significantly underperform compared to the stocks sold. They not only show that buying these stocks is unbeneficial, but also that they influence subsequent stock returns (Barber & Odean, 2008, p. 25)

In addition, there are multiple papers on media and stock returns written by Tetlock. First, research from Tetlock (2007) shows that high media pessimism predicts downward pressure on market prices. Tetlock finds that these changes are spread throughout the trading day, instead of grouped after the moment of release (2007). This effect on stock prices is particularly large and slow to reverse for small stocks (Tetlock, 2007, p. 1166). This is consistent with the results from Fang & Peress (2009) and Chan (2003). Additional work from Tetlock et al. (2008) shows a small delay in the reaction of stock prices to information enclosed in “negative words”. They show that negative words are useful predictors of returns and earnings. However, they find an underreaction of market prices to negative words in news stories (Tetlock et al., 2008, p. 1465). Lastly, Tetlock (2011) examines the stock response to stale information. His results show that market reactions to news-events with a large amount of old information will partially reverse in the next week (Tetlock, 2011, p. 33). On the other hand, he finds that market reactions reverse less or even continue for news-events with new information. Moreover, individual investors’ overreaction to stale information results in a short change in stock prices (Tetlock, 2011, p. 34). The results from Tetlock his papers can be used to explain possible findings of my study.

A research in to the causal impact of media on trading and prices is done by Peress (2014). Peress designates newspaper strikes as a period without news. His results show lower market returns on days before a newspaper strike, but no change in returns in the days after a strike (Peress, 2014, p. 22). He finds that on strike days the trading volume falls and the dispersion of stock returns and intraday volatility are reduced (Peress, 2014, p. 28). However, the aggregate returns are unaffected. His results show that these findings help stock prices incorporate news from the previous day (Peress, 2014, p. 29).

Previous research done by Klibanoff et al. (1998) shows that prices react more to changes in fundamentals after front-page news of the New York Times compared to weeks without front-page news. This is in line with their hypothesis that news events lead some investors to react more quickly (Klibanoff et al., 1998).

Overall the most papers tend to find a relation between media coverage and stock returns. However, not all the papers find the same results. Some papers find stronger results for smaller stocks while this is not the case for others. Also some papers find under reaction to news and others do not. In general, it can be concluded that there is a relation, but its reaction depends on certain factors.

(6)

2.2 Pre-announcements and stock returns

To examine the full effect of these events, the initial announcement should be examined. This announcement could have a so-called “announcement effect”. This is the effect now on news that something in the future will happen. The pre-announcement is the announcement of a future announcement. Companies use the E3 and Gamescom to announce new information. Therefore, the announcement of participating in these events can be considered as an announcement that they will announce something at those events: a pre-announcement.

A research related to mine is from Sorescu et al. (2007). In their paper the effect of new product pre-announcements on stock returns is examined. This is similar to my study where I examine the effect of the event’s pre-announcement on stock returns. They use cumulative abnormal returns to examine the effect of the pre-announcement on stock returns, and use the Fama-French three factor model to calculate those abnormal returns (Sorescu et al., 2007, p. 474). Their results show significant positive abnormal returns in the long-run and significant abnormal returns in the short-run (Sorescu et al., 2007, p. 482). However, Sorescu et al. show that the short-term abnormal returns are only significant if the pre-announcement contains specific information about the pre-announced product (2007, p. 482).

A research related to the one from Sorescu et al. (2007) is done by Koku et al. (1997). This paper separates the announcement and the pre-announcement of products. Their results show that on average only the product pre-announcement has a significant positive effect on stock prices (Koku et al., 1997, p. 184). Therefore, the pre-announcement should be analyzed independent from the announcement itself. However, the results vary substantially between industries.

Mishra and Bhabra (2001) also examined product pre-announcements. Their paper uses event study methodology to calculate abnormal returns after a product pre-announcement. Similar to my study, they calculate abnormal returns using the market model. The results show positive returns on the stock market at pre-announcements. However, they only find this for “credible” pre-announcements containing evidence (Mishra & Bhahbra, 2001, p. 85). The pre-announcement “bluffs” are ignored by the stock market (Mishra & Bhabra, 2001). This is similar to the result from Sorescu et al. (2007).

Overall, the findings from previous researchers show a positive effect of announcements on stock prices. However, the results are related to the credibility of the pre-announcements.

(7)

2.3 Fama-French model

The capital asset pricing model (CAPM) is generally seen as the start of the literature about return and risk. This model uses the market premium and the risk-free rate to explain asset returns. Several researchers started to study extensions of CAPM. Fama and French are one of the most prominent researchers in this literature. Their research showed that the factors size and book-to-market equity capture cross-sectional variation in stock returns (Fama & French, 1992, p. 450). As extension to their work, Fama and French (1993) proposed the factors SMB and HML to represent the risk factors size and book-to-market. Their results show that adding these factors to the excess market return explains the cross-section of average returns (Fama & French, 1993, p. 54). This results in the following formula:

_,

SMB is used as risk factor for returns related to size. HML is the factor for returns related to the book-to-market ratio (Fama & French, 1993). The factors for the three-factor model are calculated using 6 value-weight portfolios (Fama & French, 1993, p. 9). These 6 portfolios are formed on size/book-to-market value. Hereby the Small minus Big (SMB) factor is calculated by the difference between 3 small portfolios and 3 big portfolios.

1

3

1

3

The High minus Low (HML) factor is the difference between the average return of value portfolios and growth portfolios and is calculated in the following way:

1

2

1

2

Fama and French argue that using value weight portfolios to calculate SMB and HML results in a minimized variance. Moreover, using value-weighted portfolios result in factors

consistent with realistic investment portfolios (Fama & French, 1993, p. 10).

(8)

3. Methodology 3.1 Event study

This paper is based on event study methodology. The event study is an empirical tool to measure the effect of an event. The most common approach is based on the effect of events on stock returns. This approach examines significant abnormal returns around events (De Jong, 2007, p. 1). Several studies related to mine use event study methodology to examine abnormal returns. Examples are from Antweiler & Frank (2005), Sorescu et al. (2007) and Koku et al. (1997). These papers are more thoroughly examined in part 2. De Jong (2007) describes three steps in performing an event study. The first step is to identify the event. The event and announcement dates are described in part 3.2. My paper uses the methodology from Brown and Warner (1985) to determine the event window. Their study uses an event window from 5 days before till 5 days after the event.

However, not all announcements and events are anticipated. Previous research into unanticipated events is performed by Barrett et al. (1987). This paper examines the response of stock prices to completely unanticipated events. Their results show that the response of stock prices for unanticipated events is only significant for one trading day (Barrett et al., 1987). In a related paper from Brooks et al. (2003) they find that unexpected events have an immediate price effect for overnight events. The price reaction takes longer when the market is open (Brooks et al., 2003). Based on the findings of both papers, an event window from day 0 till day +1 is examined.

The second step introduced by De Jong is to specify the model for normal stock returns. There are multiple models to calculate normal returns. Examples of the models are the market model, CAPM and the Fama-French three factor model. Part 3.3 examines the used models. The last introduced step by De Jong (2007) is to calculate abnormal returns. Abnormal returns are a common approach to examine effects on stock returns. Almost all literature examined in part 2 use abnormal returns in their study. The abnormal return is the difference between the real return and the normal return, where the normal return is the expected return in an event period if there had been no event.

, , ,

(9)

3.2 Data

In 2007 and 2008 the E3 was a small business summit. From 2009 the ESA decided to organize a greatly expanded media event. The expo grew from 5,000 attendees in 2008 to 41,000 in 2009. In the same year the Bundesverband Interaktive Unterhaltungssoftware (BIU) decided to organize the largest open-public video game expo. Therefore, I chose to examine the data from 2009 till 2015 (last edition). For the data I use the stocks of five companies in the gaming industry. These companies are: Nintendo, Ubisoft, Activision-Blizzard, Electronic Arts (EA) and Take-Two interactive. The stocks from Nintendo and Ubisoft are listed on the Paris stock market, and the other three companies are listed on the NASDAQ. The historical data is retrieved from Yahoo finance. The stock prices are used to calculate the stock returns in the following way:

1 0

0

Further, the data for the Fama-French 3-factor model is retrieved from the Kenneth French database. This database contains the data for the Fama-French factors, as well as the risk-free rate and the market return. The factors for the three-factor model are calculated using 6 value-weight portfolios (Fama & French, 1993). These 6 portfolios are formed on size/book-to-market value. Hereby the Small minus Big (SMB) factor is calculated by the difference between 3 small portfolios and 3 big portfolios. The High minus Low (HML) is the difference between the average return of value portfolios and growth portfolios. The formulas used to calculate these factors can be found in part 2.3. For the market return, the Kenneth French database uses returns from firms listed on the NASDAQ, AMEX and NYSE. The Kenneth French database uses the one-month Treasury bill rate as the risk-free rate. The France government bond rates retrieved from investing.com are used as risk-free rate for the two companies listed on the Paris stock market. The CAC40 is the market index for the Paris stock market. The CAC40 represents the 40 most significant firms on the Paris stock market. This market index is used to calculate the market return for the two companies listed on the Paris stock market. Unfortunately, the data for the European Fama-French factors is only available on monthly basis. Therefore, the companies Nintendo and Ubisoft cannot be used in the Fama-French three factor model. The Kenneth French library uses data from ‘The center for research in security prices’ (CRSP) to calculate their factors.

To examine the events, the companies’ participations are observed. The following

tables are the participating tables for both events.

Table 3.2.1 Participations E3

Year Date Activision-Blizzard EA Nintendo Take-Two Ubisoft

2009 02/06 - 04/06 1 1 1 1 1 2010 15/06 - 17/06 1 1 1 1 1 2011 07/06 - 09/06 1 1 1 1 1 2012 05/06 - 07/06 1 1 1 1 1 2013 11/06 - 13/06 1 1 1 1 1 2014 10/06 - 12/06 1 1 1 1 1 2015 16/06 - 18/06 1 1 1 1 1

(10)

Table 3.2.2 Participations Gamescom

Year Date Activision-Blizzard EA Nintendo Take-Two Ubisoft

2009 19/08 ‐ 23/08 1 1 0 1 1 2010 18/08 ‐ 22/08 1 1 1 1 1 2011 17/08 ‐ 21/08 1 1 1 1 1 2012 15/08 ‐ 19/08 1 1 0 1 1 2013 21/08 ‐ 25/08 1 1 1 1 1 2014 13/08 ‐ 17/08 1 1 1 1 1 2015 05/08 ‐ 09/08 1 1 1 1 1

*The value is “1” if participating, “0”if not

Additionally, this thesis examines the participation announcement. In this announcement the companies announce their attendance to the events. News articles publishing the exhibitors list are used as announcement. The date of publication will be used as event date. The following are the publication dates of the news articles.

Table 3.2.3

Participation announcement dates

Year E3 Gamescom 2009 02-02-2009 X 2010 18-01-2010 X 2011 21-03-2011 16-06-2011 2012 12-12-2011 12-06-2012 2013 X X 2014 X 26-06-2014 2015 27-11-2014 X

3.3 Abnormal return testing

This thesis is based on the event study methodology from De Jong (2007) and Brown & Warner (1985) explained in part 3.1. Similar to related papers examined in part 2, this thesis will use abnormal stock returns to test the effect of an event. Abnormal stock returns are the “difference between actual returns in time of an event and the expected returns if there had been no event” (De Jong, 2007, p. 3). That is, actual return minus the normal return. One model to calculate these abnormal returns is the OLS market model. The market model uses the market portfolio and its beta to explain returns. This results in the following abnormal returns:

, , ,

Whereby , is the abnormal return for stock i at day t, , is the return on the market index and are the OLS estimates of the regression on the estimation period. Another model is the capital asset pricing model (CAPM), which uses the market premium and the risk-free rate. In CAPM, the excess return and normal return are calculated in the following ways (De Jong, 2007, p. 6

_, .

(11)

The paper from Brown and Warner (1980) finds that the market model performs well under various conditions. Brown and Warner show that choosing a more complicated model is not necessarily beneficial to calculate abnormal returns (Brown & Warner, 1980, p. 249). Therefore, I choose the market model as one of my models to calculate the abnormal returns. However, several studies related to mine use a more extensive model to calculate abnormal returns. For example, earlier papers about media coverage and stock returns from both Fang & Peress (2009) and Chan (2003) also use the Fama-French three factor model to calculate abnormal returns. Therefore, I choose to also use the Fama-French three factor model. This model is more thoroughly examined in part 2.3. Fama and French (1993) add two extra factors ‘Small minus Big’ (SMB) and ‘High minus Low’ (HML) to the market model. SMB is used as risk factor for returns related to size. HML is the factor for returns related to the book-to-market ratio (Fama & French, 1993). Fama and French show that using these factors in event studies does better at isolating the firm-specific component of stock returns (1993, p. 54). The Fama-French three factor model for abnormal returns follows as below:

, , , ,

The above factors are estimated by regressing them on the estimation period. This period is the time prior to the event period. Based on the paper of Brown and Warner (1985) I choose 5 days before till 5 days after the event as event period. I define the day of the event as day ‘0’. An explaining graph is shown in part 3.1. Next, the estimated factors are used to predict the returns in time of a period (normal returns). The abnormal returns are calculated using the actual returns and the predicted normal returns. To test the significance of the abnormal returns, most researchers use the cumulative abnormal returns (CAR). This is the sum of abnormal returns, which follows below:

, ⋯ ,

The testing of abnormal returns is mostly done with a simple t-test. For this test we need to assume that the abnormal returns are independently and identically distributed (De Jong, 2007). This means I assume the returns are normally distributed. This allows me to use the following test statistic:

0,1

However, these individually analyzed abnormal returns are not very informative because stock returns can be caused by information unrelated to the event (De Jong, 2007, p. 7). Therefore, the cross-sectional abnormal returns are used. The cumulative average abnormal returns are typically used to test the cross-sectional returns, which follows as below (De Jong, 2007, p. 8):

(12)

Hereby, N is the number of firms. In this thesis N=5 using the market model and N=3 using the Fama-French three factor model. The cumulative abnormal average return can be tested using a T-test. The test statistic is as follows:

√ 0,1

Whereby s is the standard deviation, calculated as follows:

1

To test the significance, the following hypotheses are used. The null hypothesis is that the cumulative average abnormal return is equal to zero.

H0: CAARE3 = 0 H1: CAARE3 ≠ 0

H0: CAARGC = 0 H1: CAARGC ≠ 0

H0: CAARAnnE3 = 0 H1: CAARAnnE3 ≠ 0

(13)

4. Results

4.1 Overall model

To see how well the control variable of the market return explains the model, regressing the market model in the period 2009-2015 for all five companies yields the results in table 4.1.1.

Table 4.1.1

Regression market model

This table examines the OLS regression based on the formula for the market model: Ri = α + βRm + ɛ

p-value is in parentheses. The coefficients for Activision, EA and Take-Two are multiplied with the factor 100.

Activision EA Nintendo Take-Two Ubisoft

Market return 0.913*** (0.000) 1.093*** (0.000) 0.4677275*** (0.000) 1.071*** (0.000) 0.817*** (0.000) Constant 0.0004362 (0.289) 0.0000356 (0.941) -0.003194 (0.571) 0.0003879 (0.512) -0.0000325 (0.956) R-squared 0.3245 0.3371 0.0825 0.2422 0.1965 Observations 2015 2015 1976 2015 1976 *: significant at 10% level **: significant at 5% level ***: significant at 1% level

As seen in the table above, the coefficients in the market model are significant for all five companies at the 1% significance level. This means that the return on the market portfolio contributes to the returns of all the companies. However, the coefficients for the companies Activision, EA and Take-Two are small compared to the coefficients of Nintendo and Ubisoft. This difference might be caused by the use of a different database. The companies Activision, EA and Take-Two use market return data retrieved from the Kenneth-French database, while the CAC40 is used for both Nintendo and Ubisoft.

Following, the Fama-French 3-factor model is constructed by adding SMB and HML to the market model for the three companies listed on the US stock market. Regressing the excess return on the control variables for market premium, SMB and HML yields the results in table 4.1.2:

(14)

Table 4.1.2

Regression Fama-French model

This table examines the OLS regression based on the formula for the market Fama-French three factor model examined in part 2.3. p-value is in parentheses. The coefficients for Activision, EA and Take-Two are multiplied with the factor 100.

Activision EA Take-Two Market premium 0.785*** (0.000) 1.173*** (0.000) 1.085*** (0.000) SMB 0.095 (0.742) 0.114 (0.150) 0.442*** (0.000) HML 0.798** (0.045) -0.407*** (0.000) -0.208** (0.036) Constant 0.0008636 (0.513) -0.00096** (0.046) -0.000605 (0.305) R-squared 0.4210 0.3455 0.2537 Observations 2015 2015 2015 *: significant at 10% level **: significant at 5% level ***: significant at 1% level

The results in table 4.1.2 show significant coefficients for the company Take-Two. Both coefficients of market premium and SMB are significant at the 1% level, while HML is significant at 5%. Further, the results show significance of the market premium and HML for both companies EA and Activision. However, the SMB factor is not significant for these two companies. Additionally, the Fama-French three factor model has increased the total regression R2 for all companies.

4.2 Abnormal return calculation

First, the abnormal returns for both events are calculated. The expected return during event period is calculated using the regression on the estimation period for both the market model and the 3-factor model. The abnormal returns are calculated using the difference between the real returns and the normal returns, as examined in part 3.3. The cumulative abnormal returns are calculated by summing the abnormal returns. Testing the cumulative abnormal returns using a t-test yields the results in tables 4.2.1 and 4.2.2.

(15)

Table 4.2.1

cumulative abnormal returns E3

This table examines the cumulative average abnormal returns for the events E3 and Gamescom. The CAR is calculated using the methodology examined in part 3.3. The calculations in stata are based on the file from Kim (2011). Critical t-values are 1.645, 1.96 and 2,576 for respectively significance at the 10%, 5% and 1% level.

Market 3-factor Market 3-factor Market Market 3-factor Market

CAR 0.038 0.107 -0.163 -0.099 0.163 -0.539** -0.487* 0.044 t-value 0.211 0.602 -0.784 -0.478 0.682 -2.127 -1.927 0.172 Event-days 91 91 91 91 91 *: significant at 10% level **: significant at 5% level ***: significant at 1% level Table 4.2.2

cumulative abnormal returns Gamescom

CAR 0.030 0.103 0.185 0.259 0.179 0.007 0.085 0.057 t-value 0.172 0.585 0.892 1.251 0.877 0.027 0.335 0.204 Event-days 91 91 62 91 91 *: significant at 10% level **: significant at 5% level ***: significant at 1% level

The results from table 4.2.1 show positive cumulative abnormal returns for the companies Ubisoft, Nintendo and Activision at the E3. None of these results are however significant. Using the Fama-French model results in higher cumulative abnormal returns and increased t-value for Activision. However, the findings are still not significant. Both EA and Take-Two show negative abnormal returns. The results for EA are insignificant. The results from the market model for Take-Two show significant cumulative abnormal returns at the 5% level. After controlling for risk factors using the three-factor model, the results show significant abnormal returns at the 10% significance level. These stronger results for negative returns are in line with the findings from Fang & Peress (2009) who find lower returns for stocks with media-coverage. The findings also correspond with the findings from Barber & Odean (2008) who find that attention-grabbing stocks underperform to stocks which are chosen based on attention. This finding of negative abnormal returns could also be explained by negative media attention. Based on the findings from Tetlock (2007) and (2008) where he finds that both media pessimism and “negative words” lead do downward pressure on stock prices.

Further, all the results show positive cumulative abnormal returns for the Gamescom. However, none of the results are significant. The fact that most results are insignificant could be explained by earlier findings from Ikenberry and Ramnath (2002) that stock prices under

(16)

react to self-selected news events. Both E3 and Gamescom are self selected events by the companies, and therefore fit in the results from Ikenberry and Ramnath.

As previous studies show that the pre-announcement influences stock returns, in this thesis the announcement effect is investigated. First, the cumulative abnormal returns are calculated for both announcements using a period from day -5 till day +5 as indicated in part 3.1. This follows as below:

Table 4.2.3

cumulative abnormal returns Announcement E3 [-5:+5]

CAR 0.022 0.042 0.017 0.035 -0.117 -0.026 -0.004 -0.066 t-value 0.159 0.307 0.110 0.218 -0.649 -0.130 -0.019 -0.336 Event-days 55 55 55 55 55 *: significant at 10% level **: significant at 5% level ***: significant at 1% level Table 4.2.4

cumulative abnormal returns Announcement Gamescom [-5:+5]

CAR -0.049 -0.025 -0.166 -0.143 -0.091 -0.090 -0.069 -0.112 t-value -0.459 -0.232 -1.335 -1.151 -0.789 -0.588 -0.453 -0.716 Event-days 33 33 33 33 33 *: significant at 10% level **: significant at 5% level ***: significant at 1% level

The results in table 4.2.3 and 4.2.4 show that the announcement of the E3 results in positive cumulative abnormal returns for both EA and Activision. The three other companies show negative abnormal returns for this announcement. The cumulative abnormal returns for the announcement of the Gamescom are negative for all companies. However, the results in tables 4.2.3 and 4.2.4 show the announcement of E3 event and Gamescom event are insignificant. This is contrary to the results from Koku et al. (1997) and Sorescu et al. (2007) who both find significant positive abnormal returns. A problem might be that it is not known in advance when these announcements are done. Barret et al. (1987) did research in unanticipated announcements and find that the stock price change after these announcements is only significant for one trading day. Therefore, the event period is changed to a period from day 0 till day +1 to see if the results would be significant. To do this, it is assumed that the returns

(17)

are normally distributed; otherwise the test would not hold. This shows the following results in table 4.2.5 and table 4.2.6:

Table 4.2.5

cumulative abnormal returns Announcement E3 [0:+1]

CAR -0.002 -0.009 -0.028 -0.031 0.136* -0.135 -0.127 -0.003 t-value -0.043 -0.148 -0.411 -0.448 1.719 -1.601 -1.526 -0.033 Event-days 10 10 10 10 10 *: significant at 10% level **: significant at 5% level ***: significant at 1% level Table 4.2.6

cumulative abnormal returns Announcement Gamescom [0:+1]

Activision EA Nintendo Take-Two Ubisoft

Market 3-factor Market 3-factor Market Market 3-factor Market CAR -0.037 -0.027 -0.090* -0.079 -0.033 0.002 -0.009 -0.072 t-value -0.824 -0.600 -1.711 -1.504 -0.540 0.023 -0.136 -0.856 Event-days 6 6 6 6 6 *: significant at 10% level **: significant at 5% level ***: significant at 1% level

The different chosen time-period shows different results. The cumulative abnormal returns for the E3 announcement shifted from positive to negative for the companies Activision and EA. However, Nintendo gets contrary results, which shows positive abnormal returns at the 10% significance level. Further, the results from table 4.2.6 show significant negative abnormal returns using the market model for EA. However, a lot of the results show negative cumulative returns. This is in contrast with the findings from Koku et al. (1997) and Sorescu et al. (2007) who both found that pre-announcements have a significant positive effect.

(18)

4.3 Event effect

In part 4.2 the cumulative abnormal returns are tested individually around the event and announcement. However, individually tested abnormal returns are not very informative because stock price changes could be the result of information unrelated to the event (De Jong, 2007, p. 7). Therefore, the total effect of the event is tested using the cumulative average abnormal returns. Using the average should cancel out the changes unrelated to the event (De Jong, 2007, p. 7). The testing is based on the calculations explained in part 3.3, which follows in the results below:

Table 4.3.1

cumulative average abnormal returns Event

This table examines the cumulative average abnormal returns for the events E3 and Gamescom. The CAAR is calculated using the methodology examined in part 3.3. The calculations in stata are based on the file from Kim (2011). Critical t-values are 1.645, 1.96 and 2,576 for respectively significance at the 10%, 5% and 1% level.

E3 Gamescom

Market 3-factor Market 3-factor CAAR -0.091 -0.159 0.092** 0.149*** t-value -0.739 -0.918 2.431 2.700 Firms 5 3 5 3 *: significant at 10% level **: significant at 5% level ***: significant at 1% level

The results in table 4.3.1 show insignificant negative CAAR values for the E3 event. However, the results for the event Gamescom are positive and significant. The findings show that there is an effect of media events on stock returns. This result is in line with the results from Ikenberry and Ramnath (2002) that self-selected news events generate significant positive abnormal returns. The findings also correspond with the results from Ferguson et al. (2015), who show that media content affects stock returns. Based on the findings from Ferguson et al. (2015), the positive returns could be the result of positive media content at the event. The results are in contrast with the finding from Chan (2003) that there is a stronger reaction to bad news. However, this could be caused by the chosen event window. As Antweiler and Frank (2005) showed, the results are very sensitive to the chosen event window. The insignificant results for the event E3 might be caused by contradicting abnormal returns for the individual companies. Some companies generate positive abnormal returns, while others generate negative returns. Therefore, the positive and negative abnormal returns could cancel each other out. The results in table 4.3.1 are also in contrast with the findings from Fang & Peress (2009). Their result, showing lower returns for stocks with high media coverage, does not correspond with the findings in Table 4.3.1.

(19)

4.4 Announcement effect

Similar to the event, the participation announcement should not be examined individually. Therefore, the announcement is also tested with the cumulative average abnormal returns. First, the returns are tested based on the event window from day -5 till day +5 as specified by Brown and Warner (1985).

Table 4.4.1

cumulative average abnormal returns Announcement [-5:+5]

This table examines the cumulative average abnormal returns for the participation announcement of E3 and Gamescom. The CAAR is calculated using the methodology examined in part 3.3. The calculations in stata are based on the file from Kim (2011). Critical t-values are 1.645, 1.96 and 2,576 for respectively significance at the 10%, 5% and 1% level.

E3 Gamescom

Market 3-factor Market 3-factor

CAAR -0.034 0.024* -0.101*** -0.079** t-value -1.295 1.715 -5.33 -2.287 Firms 5 3 5 3 *: significant at 10% level **: significant at 5% level ***: significant at 1% level

The results from table 4.4.1 show significant positive abnormal returns at the 10% level for the E3 announcement using the 3-factor model. Further, the results show significant negative abnormal returns for the Gamescom announcement. The results therefore show that there is an effect of the announcement on stock returns. However, the results for Gamescom are in contrast with the findings from both Sorescu et al. (2007) and Koku et al. (1997). Both studies find positive significant abnormal returns around a pre-announcement. The difference could be caused by additional factors like reliability and amount of information used by both studies. Further, the announcements are also tested on an event period from day 0 till day +1 based on the findings from Barrett et al. (1987).

Table 4.4.2

cumulative average abnormal returns Announcement [0:+1]

This table examines the cumulative average abnormal returns for the participation announcement of E3 and Gamescom. The CAAR is calculated using the methodology examined in part 3.3. The calculations in stata are based on the file from Kim (2011). Critical t-values are 1.645, 1.96 and 2,576 for respectively significance at the 10%, 5% and 1% level.

E3 Gamescom

Market 3-factor Market 3-factor

CAAR -0.006 -0.056 -0.046*** -0.038** t-value -0.147 -1.519 -2.888 -1.819 Firms 5 3 5 3 *: significant at 10% level **: significant at 5% level ***: significant at 1% level

(20)

The results from table 4.4.2 show significant negative abnormal returns for the Gamescom announcement. This result is similar to the findings for the event period from day -5 till day +5. Further, the announcement of E3 using the Fama-French three factor model results in insignificant negative abnormal returns. The t-values decreased compared to the results in table 4.4.1, resulting in a lower significance. These findings are not in line with the results from Barrett et al. (1987). This could be explained by the fact that the announcement might be somewhat anticipated. The results from Sorescu et al. (1997) and Mishra & Bhabra (2001), that pre-announcements only show significant results if specific information is included in the announcement, could explain the found insignificant results. The announcement only contains information about the participation and not about what is announced at the event itself. Further, the Gamescom announcement dates are very close to the E3 event dates. Therefore, the found results for the Gamescom announcement could be influenced by the E3 event.

5. Conclusion

This thesis examines the effect of the media events E3 and Gamescom on stock returns of five companies from 2009 till 2015. Both event and the initial participation announcement effect have been studied. The results show significant positive abnormal returns for the event Gamescom and insignificant results for the E3. The findings for positive abnormal returns are in line with the findings from Ikenberry and Ramnath (2002) that self-selected news events yield positive abnormal returns. The results also correspond with the results from Ferguson et al. (2015). However, the results are in contrast with the results from Fang & Peress (2009) who show lower returns for stocks with high media coverage.

The initial announcements are tested using an event period from day [-5:+5] as done by Brown & Warner (1985). The results show significant negative abnormal returns for the Gamescom announcement. After changing the event period to [0:+1], based on the results of Barrett et al. (1987), still shows negative abnormal returns for the Gamescom announcement. The negative abnormal returns are in contrast with the findings from Koku et al. (1997) and Sorescu et al. (2007).

There are some implications for this research. The biggest implication is that normal distribution for the returns is assumed. This allowed me to use the t-test to test for significance. However, returns are often not normally distributed. In this case another test should be used to test the significance. Another implication is cross-sectional dependence. This thesis assumes there is no correlation between the abnormal returns of two events. However, with several events in the same calendar period there may be cross-sectional correlation (De Jong, 2007, p. 13). This could lead into biased upward t-statistics (De Jong, 2007). Another problem indicated by De Jong is serial correlation (2007, pp. 14-15). Just like cross-sectional dependence could this lead to upward biased t-statistics.

There are some recommendations for further research. First, the companies participating in these events are each other’s competitors. It could be considered to include the effect competitors have on each stock returns. Additionally, different event windows should be studied. As the results from Antweiler and Frank (2015) show, the abnormal stock returns are very sensitive to the chosen event window. Further recommendation is based on the fact that there is an implication for using the market model. This implication is the calendar time effect. There are multiple papers who find that certain days have higher returns. To tackle this problem a dummy-variable for the day of the week could be added to the model.

(21)

6. Bibliography

Antweiler, W. & Frank, M. Z. (2005) The Market Impact of Corporate News Stories. Working paper, University of British Columbia.

Barber, B., & Odean, T. (2008), All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors. The review of financial studies, 21 (2): 785-818.

Barrett, W. B., Heuson, A. J., Kolb, R. W. and Schropp, G. H. (1987), The Adjustment of Stock Prices To Completely Unanticipated Events. The Financial Review, 22: 345–354. Brooks, R., Patel, A., & Su, T. (2003). How the Equity Market Responds to Unanticipated Events. The Journal of Business, 76(1), 109-133.

Brown, S. J., & Warner, J. B. (1980). Measuring security price performance.Journal of

financial economics, 8(3), 205-258.

Brown, S., & Warner, J. (1985), Using daily stock returns: The case of event studies. The

Journal of financial economics, 14: 3-31

Chan, W. (2003), Stock price reaction to news and no-news: drift and reversal after headlines.

The journal of financial economics, 70: 223-260

De Jong, F. (2007). Event studies methodology. Lecture Notes. E3-Expo. (n.d.) Retrieved from https://www.e3expo.com/

Fama, E.F., & French, K.R. (1992). The cross-section of expected stock returns. The Journal

of Finance, 47(2), 427-465.

Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of financial economics, 33(1), 3-56.

Fang, L.,& Peress, J. (2009), Media Coverage and the Cross-section of Stock Returns. The

Journal of Finance, 64: 2023–2052.

Ferguson, N.J., Lam, H. Y. T., Philip, D.,& Guo, J. M. (2015), Media Content and Stock Returns : The Predictive Power of Press. Multinational Finance Journal, 19 (1): 1-31

Gamescom-cologne. (n.d.) Retrieved from http://www.gamescom-cologne.com

Ikenberry, D. L.,& Ramnath, S. (2002), Underreaction to Self-Selected News Events: The Case of Stock Splits. The Review of financial studies, 15 (2): 489-526

Kim, W. (2011) Event Study Using STATA. Lecture Notes.

Klibanoff, P., Lamont, O. and Wizman, T. A. (1998), Investor Reaction to Salient News in Closed-End Country Funds. The Journal of Finance, 53: 673–699

(22)

Koku, P. S., Jagpal, H.S.,& Viswanath, P.V. (1997), The effect of new product announcements and preannouncements on stockprice. Journal of Market Focused

Management , 2(2): 183–199.

Miller, J. S. (2005). Effects of preannouncements on analyst and stock price reactions to earnings news. Review of Quantitative Finance and Accounting, 24(3), 251–275

Mishra, D. P.,& Bhabra, H. S. (2001), Assessing the economic worth of new product pre‐announcement signals: theory and empirical evidence, Journal of Product & Brand

Management, 10 (2) : 75 - 93

Peress, J. (2014), The Media and the Diffusion of Information in Financial Markets: Evidence from Newspaper Strikes. The Journal of Finance, 69: 2007–2043

Sorescu, A., Shankar, V.,& Kushwaha, T. (2007), New Product Preannouncements and Shareholder Value: Don’t Make Promises You Can’t Keep, Journal of Marketing Research, 44(3), 468-489.

Tetlock, P. C. (2007), Giving Content to Investor Sentiment: The Role of Media in the Stock Market. The Journal of Finance, 62: 1139–1168

Tetlock, P. C., Saar-Tsechansky, M. and Macskassy, S. (2008), More Than Words: Quantifying Language to Measure Firms' Fundamentals. The Journal of Finance, 63: 1437– 1467

Tetlock, P. C. (2011), All the news that’s fit to reprint: do investors react to stale information?