• No results found

Market Reaction to Analysts’ Recommendations: Evidence from the Dutch Stock Market

N/A
N/A
Protected

Academic year: 2021

Share "Market Reaction to Analysts’ Recommendations: Evidence from the Dutch Stock Market"

Copied!
57
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Market Reaction to Analysts’ Recommendations:

Evidence from the Dutch Stock Market

L.J. Meppelink, 1257889

University of Groningen

Faculty of Economics

Thesis Economics

(2)

Abstract

This paper analyzes buy and sell recommendations for stocks listed on the Dutch AEX index from 2005 to 2006. Buy recommendations inflict a small positive, but insignificant market reaction at the publication day. Sell recommendations inflict a small negative, but insignificant market reaction at the publication day. An individual investor who follows analysts’ recommendations cannot realize short-term abnormal returns by trading according to analysts’ recommendations. With respect to the differences in forecasting skills of the investigated brokers, no statistically significant differences are found. The results indicate that the main Dutch stock market is semi-strong efficient. Lastly, brokers seem to issue far more buy recommendations than sell recommendations, which indicates a form of broker optimism.

(3)

Table of Contents

1. Introduction

3

2. Literature Review

5

3. Hypotheses

22

4.

Methodology

and

Data

24 4.1. Sample selection and descriptive statistics 25

4.2. Methodology 27

5. Empirical Results

30

5.1. Market Reaction to Analysts’ Recommendations 30

5.2. Analysis of Abnormal Returns for the Individual Investor 34

5.3. Analysis by Broker 37

6. Conclusions

47

(4)

1. Introduction

Bringing information about companies to investors is an important role of analysts. They fulfill the role of information intermediaries, as they uncover and disseminate information to the market. Their activities, reflected in the trades of their clients, may result in markets becoming more efficient. Many investors rely on the information analysts provide to them when they revise or select their portfolios. The role of analysts’ forecasts has been a topic of interest since Alfred Cowles [1933] presented his analysis of the forecasting efforts of 45 professional agencies. Cowles [1933] analyzed the forecasting efforts of 45 professional agencies which have attempted to select common stocks that should generate superior returns compared to the benchmark. Cowles [1933] showed that recommended stocks had, on average, a negative performance when compared against a benchmark. Cowles [1933] concluded that investment recommendations do not add value.

Since then a lot of empirical research has been done to add clarity to the subject of analyst value, and their impact on the market, especially for the U.S. market. A large part of the empirical research examines the market reaction to the publication of recommendations issued by brokerage houses and security analysts. This is done by, among others, Dimson and Marsh [1984], and Womack [1996]. The results indicate that brokerage firms’ recommendations do provide an investor with valuable advice. However, when considering transaction costs abnormal returns tend to become very small or even disappear. Another part of the empirical literature focuses on stock recommendations published in columns or other media. These empirical papers examine the effect of those stock recommendations on market prices. The recommendations are published in columns of the Wall Street Journal, made by prominent money managers at Barron’s Annual Roundtable, published in investment newsletters, or published in investment magazines. This is done by, among others, Barber and Loeffler [1993], and Copeland and Mayers [1982], and Wijmega [1987]. The results are mixed. Recommendations published in columns inflict a significant market reaction at the publication date. Investment newsletters do not seem to provide value to an investor. Recommendations published in investment magazines do seem to provide value to an investor. A full review of the existing literature will be given in the designated section.

(5)

brokers that publish recommendations for Dutch stocks possess different forecasting skills with respect to their published recommendations.

To accomplish these goals, an event study is carried out. The investigated events are publications of analysts’ recommendations. Consequently, abnormal returns associated with analysts’

recommendations are calculated and examined. The potential market reaction inflicted by analysts’ recommendations is analyzed by assessing possible abnormal returns before and at the publication day of a recommendation. By analyzing abnormal returns for a period after the publication date of the recommendation, for both the sample as a whole and for individual brokers, the second and third goals of this paper are covered.

This paper focuses on The Netherlands. Consequently, only recommendations which are published on AEX-listed stocks are investigated. This does not mean that the broker is Dutch or based in the Netherlands. As mentioned, the majority of empirical research has been done for the U.S. stock market. Different market settings may affect other markets in other ways. However, analysts’ recommendations for the Dutch market are not suspected to affect and provide value to the market in another way than to the U.S. market, since the Dutch stock market is highly correlated with the U.S. stock market.

Thus this paper sets out to investigate the following:

• Do analysts’ recommendations have an impact on prices of the recommended stocks at the publication date?

• Can an investor earn abnormal returns by following analysts’ recommendations? • Do differences exist in the forecasting skills of the various brokers publishing

recommendations on Dutch stocks?

The rest of the paper is organized as follows. Section 2 provides a literature review on the subject. Section 3 presents the hypotheses to be tested. Section 4 describes the data and presents the

(6)

2. Literature Review

The pioneer in the research field of analyst value is Alfred Cowles. Cowles [1933] investigated the forecasting efforts of 16 financial services companies, 20 fire insurance companies, 25 financial periodicals and an editor from The Wall Street Journal. Cowles [1933] concluded that recommended stocks had, on average, a negative performance when compared against a benchmark during the 1928-1932 period. Investing according to the recommendations of the editor from the Wall Street Journal would have done worse than investing in the benchmark for the period from 1904 to 1929.

The financial services companies, on average, generated a return 1.43 percent worse than the average return of the benchmark on an annual basis. The fire companies, on average, generated a return 1.20 percent worse than the average return of the benchmark on an annual basis. The financial periodicals generated a return 4.00 percent worse than the return of a random sample on an annual basis. In contemporary terms, it would mean that all investigated groups underperformed the market.

Furthermore, statistical tests failed to demonstrate that the analysts exhibited any forecasting skill. A decade later Cowles [1944] re-examined his earlier work. The conclusions were similar. Eleven leading financial periodicals failed to successfully predict the future course of the stock market.

Analysts’ value is closely interconnected with market efficiency. Analysts’ recommendations are based on publicly available information and should not inflict a market reaction when a semi-efficient market is assumed. To better understand the relevance of the efficient market hypothesis for analyst’ recommendations, the efficient market hypothesis is discussed before continuing with the discussion of the previous literature. The price pressure hypothesis is reported in previous empirical papers and relevant to the efficient market hypothesis, because the price pressure hypothesis deals with temporary market inefficiency. The price pressure hypothesis is discussed as well.

Efficient Market Hypothesis

According to the Efficient Market Hypothesis a market is efficient when all available information is reflected in market prices. As a consequence investors can make efficient decisions by only looking at security prices. A fully efficient market, however, is not likely to exist. For example, inside

management information should also be reflected in security prices for the market to be fully efficient. In order to empirically test the efficient market hypothesis Fama [1970] distinguishes three test forms to determine the degree of market efficiency:

- Weak form tests are tests of whether all information contained in historical prices is fully reflected in current prices;

(7)

- Strong form tests are concerned with whether all information, public or private, is fully reflected in security prices.

In a more recent review article, Fama [1991] renamed the three categories to better fit research methods. The weak form tests of efficiency are denoted as tests of return predictability. These tests investigate whether returns can be predicted from past data. Semi-strong form tests of efficiency are renamed into event studies. Event studies relate price changes to changes in publicly available information, such as financial results, or a firm’s dividend policy. Lastly, strong form tests are renamed to tests for private information.

Tests of return predictability have resulted in interesting empirical findings and even market anomalies that appear to be a violation of the efficient market hypothesis. These violations, however, do not leave investors with opportunities to generate abnormal returns. This is mainly due to the existence of transaction costs. Furthermore, the effects seem to be disappearing in the last 15 years. For a review on anomalies and market efficiency, see Schwert [2002].

Much of the efficient market hypothesis literature is concerned with the speed with which information is impounded into security prices. Event studies show that significant abnormal returns occur in periods around certain announcements. The price adjustment process, however, is rapid and happens largely within a single day. This means that new publicly available information is discounted immediately in prices, demonstrating semi-strong-form market efficiency.

Market Efficiency and the Dutch Stock Market

The Dutch stock market has been tested for its efficiency by Mourik [1988]. Mourik [1988] found that the Dutch stock exchange reacts to developments on other stock exchanges with a lag of not more than one day. Because of differences in time zones, significant lags of one period do occur but they are not a result of market inefficiency.

The momentum effect, which was found by DeBondt and Thaler [1985] on the U.S. stock market, is found by Bos [1991] and Hogenhout [1992] for the Dutch stock market as well. Bos [1991] found that a portfolio of ten 18-month losers yields a 19.8 percent higher return on an annual basis than the return produced by a portfolio of winners.

(8)

else, capable of predicting them. The findings regarding the Dutch stock market and the findings regarding the disappearance of market anomalies suggest that analysts’ recommendations published on Dutch stocks should not help an individual investor to capture abnormal returns. Because of the fact that analysts use publicly available information to devise a recommendation, as well that new information is impounded into stock prices rapidly, an investor is not likely to benefit from abnormal returns as a result of trading according to analysts’ recommendations.

Price pressure hypothesis

The price pressure hypothesis discusses whether analysts’ recommendations have an impact on security prices and whether this impact is temporary or permanent. According to the price pressure hypothesis abnormal returns should occur on the publication date of a recommendation and disappear after a short period thereafter. Among others, Liang [1999] found statistically significant abnormal returns on the publication day of a recommendation. The effect, however, is temporary and reversed within one month.

Previous literature continued

After Cowles [1944] Diefenbach [1972] began to monitor all recommendations received from the brokerage community, starting in November 1967. Diefenbach [1972] assessed the value of

recommendations to obtain an objective comparison of the usefulness of recommendations supplied by individual firms. The results of his work state that the performance of recommended stocks during 30 weeks of generally rising stock prices was better than the comparable performance of the benchmark. However, statistical significance of the results was not demonstrated. A majority of the recommended stocks of seven out of ten sources outperformed the benchmark during the 30-week period. The relative performance of recommended stocks during 50 weeks of generally declining stock prices was worse. During that 50-week period the majority of recommended stocks of only two out of ten sources outperformed the benchmark, instead of the seven out of ten sources during the period of generally rising stock prices. Because of this bad performance of recommended stocks in a bear market, Diefenbach [1972] warned to be aware of the stock-selection performance claims of analysts in a bull market. Furthermore Diefenbach’s results did not permit him to conclude that past performance and future performance are related. Lastly, the majority of the recommendations were buy

recommendations.

(9)

periods. Again, the majority of recommendations were buy recommendations. Over the 91 day holding period, following three of the six brokerage firms’ sets of recommended stocks would have resulted in statistically significant inferior performance. Following the advice of all six brokerage firms routinely and indiscriminately would have led to statistically significant inferior performance as well. Of the randomly selected sample, only one out of six sets of stocks experienced statistically significant inferior performance over the 91 holding period. Sell advice seemed to be fairly good. An investor, who sold or sold short, as the six firms advised, would have done significantly better than the market over the 91-day period.

Over the 182 day holding period, one out of six firm’s set of recommended stocks resulted in

statistically significant inferior performance, where two out of six firms’ sets of recommended stocks achieved statistically significant superior performance. Following the advice of all six brokerage firms would have led to very small, but statistically significant superior performance. The randomly selected samples exhibited the same results as the non-randomly selected samples, but their t-values were higher. Following sell recommendations would have led to statistically superior performance for all the samples.

Over the 366-day holding period, two of the six firms’ sets of recommended stocks experienced statistically significant superior performance. Moreover, following recommended stocks of all firms would have led to statistically significant superior performance as well. Overall, the randomly selected stocks exhibited statistically superior performance as well. Moreover, the overall performance of the randomly selected stocks was slightly better than the brokerage houses’ recommended stocks. Over the 366-day holding period, following sell recommendations would have led to statistically significant inferior performance. Because the patterns found in both the recommended stocks sample and the random sample were remarkably similar, Logue and Tuttle [1973] concluded that random selection is just as good as following brokerage firm advice. Depending on the investor’s goal, this may be true. However, random selection does not take risk preferences into account, whereas risk preferences are an important aspect of selecting a portfolio for any investor.

(10)

Besides Bjerring et al. [1983], Dimson and Marsh [1984], Glascock, Henderson, and Martin [1986], Womack [1996], Jegadeesh, Kim, Krische, and Lee [2004], Chakrabarti [2004], Jegadeesh and Kim [2005], Lonkani, Khanthavit, and Chunahachinda [2006], and Green [2006] all found significantly positive abnormal returns for an investor who traded according to analysts’ recommendations.

Dimson and Marsh [1984] described their findings regarding an empirical study of share return forecasts made by 35 UK stockbrokers and by the internal analysts of a large UK investment

institution during 1980 and 1981. Dimson and Marsh’ [1984] results revealed a small but potentially useful degree of forecasting ability. Furthermore, stock prices reacted rapidly to the informational content of forecasts. A large part of the informational content of the forecasts appears to be discounted in the market place within the first month. Following all the forecasts would have led to an

outperforming of the benchmark by 2.20 percent in the subsequent year.

Glascock et al. [1986] used a more theoretical point of view of assessing analyst value. If the efficient market hypothesis is correct, acting on something as routine and widely available as brokerage house’s investment advice should not be unusually profitable. Their evidence, however, suggest that one would be better off listening to the broker than to the efficient market theorists. In the 10 trading days after an aggressive buy recommendation, an investor could have realized a statistically significant abnormal return of 4.50 percent. Over the 80 trading days after the first 10 trading days, an investor could have realized another statistically significant abnormal return of 7.60 percent by following aggressive buy recommendations. In total, an investor could have realized a statistically significant abnormal return of 12.10 percent over the 90-day period by following aggressive buy

recommendations. However, Glascock et al. [1986] found no statistically significant effect on the event day. This means that the publication day price pressure effect was not found whereas analysts do identify stocks that will outperform the benchmark in the future. This suggests that brokers have real forecasting skills.

Womack [1996] examined daily stock price reactions to revisions in analysts’ recommendations of the 14 biggest U.S. brokerage houses. Womack [1996] found statistically significant positive abnormal returns for added-to-buy recommendations. For the whole sample, the three-day event return is a statistically significant 3.00 percent. The one month post-event return is a statistically significant 2.40 percent. After six months, however, the post-event return is statistically insignificant.

(11)

recommendations. Green [2006] revealed that purchasing quickly following an upgrade

recommendation resulted in an average 2-day abnormal return of 1.02 percent, after controlling for transaction costs. Selling short following downgrades produced an abnormal return of 1.50%, after controlling for transaction costs. Brav and Lehavy [2003] also found significant market reactions to the information contained in revisions of previous analysts’ recommendations at the day of the revision. Moreover, Brav and Lehavy [2003] found that revisions in analysts’ recommendations generated statistically significant abnormal returns until six months after the revision.

Jegadeesh et al. [2004] examined the relation of analysts’ recommendations with other publicly available variables that should predict future returns. They found that analysts prefer high momentum stocks and growth stocks. Furthermore, Jegadeesh et al. [2004] showed that a strategy of buying the quintile of stocks with the highest favorable analyst consensus and selling the quintile of stocks with the least favorable analyst consensus earned 2.30 percent over the next six months.

Chakrabarti [2004] examined the predictive value and market impact of analysts’ recommendations in India. Chakrabarti [2004] found that strong buy recommendations inflicted statistically significant abnormal returns over a four-month window, whereas strong sell recommendations do not inflict statistically significant abnormal returns over the same period.

With respect to the influence on the market Chakrabarti [2004] found both the strong buy

recommendations as the strong sell recommendations to have a statistically significant impact on the market in a 5-day period around the event date. Chakrabarti [2004] found that brokers are much more likely to issue a strong buy recommendation than any other kind. Like Chakrabarti [2004], Jegadeesh and Kim [2005] found that analysts who publish recommendations in the G7 countries issue far more buy recommendations than sell recommendations. Furthermore, Jegadeesh and Kim [2005] found that the frequency of sell recommendations is the lowest in the US.

Lonkani et al. [2006] investigated whether information of analysts’ recommendations on Thai stocks generates abnormal returns for investors. Lonkani et al. [2006] found that stocks strongly

(12)

despite the suggested insider use of strong buy recommendations. Lastly, the authors found higher than average trading volumes on recommendation dates.

Instead of generating abnormal returns and capturing them, Groth, Lewellen, Schlarbaum, and Lease [1979], and Barber, Lehavy, McNichols, and Trueman [2001], found abnormal returns, but an investor could not profitably capture these abnormal returns. Groth et al. [1979] found statistically significant positive abnormal returns in the month a recommendation was published. However, as Groth et al. [1979] concluded correctly, this does not mean that an individual investor could have captured these abnormal returns. Only a minority of the firm's clients may have been in a position to respond to any given investment recommendation and only those who acted quickly may have succeeded in capturing abnormal returns.Like Groth et al. [1979], Barber et al. [2001] concluded that after accounting for transaction costs none of the strategies they created generated an abnormal return that was statistically significant. Without accounting for transaction costs, however, a strategy of buying stocks that are most highly recommended generated an average annual abnormal return of 4.13 percent. A portfolio of stocks with the least favorable analysts’ consensus generated an average annual abnormal return of -4.91 percent. Buying stocks that are most highly recommended and selling short those that are least recommended generated an abnormal return of 0.75 percent per month. Following Barber et al. [2001], the inability of investors to capture abnormal returns strongly suggests that although market

inefficiencies exist, they are not easily exploitable by traders, thereby allowing the inefficiencies to persist. However, according to Barber et al. [2001] analysts’ recommendations do provide value, because, ceteris paribus, an investor would be better off investing in the most favored stocks rather than in the least favored stocks. In 2002, Barber, Lehavy, McNichols, and Trueman [2002] re-analyzed the returns of analysts’ recommendations over the 1996-2001 period. Barber et al. [2002] confirmed their previous findings by showing that the more highly recommended stocks earned greater market-adjusted returns during the 1996-1999 period than the less highly recommended stocks.

Yet, in the years 2000 and 2001 the least favorably rated stocks earned the highest returns. Barber et al. [2002] found evidence that analysts were reluctant to turn away from small growth stocks during the 2000-2001 period, while in that period those stocks significantly underperformed the market. Barber et al. [2002] argued the possibility that analyst behavior was driven by a desire to attract and retain potentially more profitable investment banking clients. In other words the years 2000 and 2001 were characterized by rising doubts on the independence of some analysts’ recommendations. Lastly, their results alert any researcher that excluding the years 2000 and 2001 from a sample period could have a significant impact on any conclusions.

(13)

performance. Furthermore, they found that the recommendations-based strategies appeared to be more effective for selection of Portuguese stocks, which can reflect a comparative home advantage. Or in other words, Portuguese analysts may have better skills in selecting Portuguese stocks rather than foreign stocks.

Stickel [1995] investigated the short-term impact of recommendations on prices. Stickel [1995] found that recommended stocks are associated with short-term price increases for buy recommendations and associated with negative price reactions for sell recommendations around the day of publication. Stocks recommended to buy displayed an average of 1.16 percent price increase over the 11 trading days centred on the date of the recommendation. Stocks recommended to sell displayed an average of -1.28 percent price decline. Furthermore, Stickel [1995] stated that larger brokerage houses have more impact on prices than smaller brokers. This could be the result of larger brokers having a stronger marketing staff. The effect however, appears to be a temporary, price pressure effect.

Besides the previously discussed empirical papers, a lot of previous research focuses on stock recommendations published in columns or other media. These empirical papers examine the effect of those stock recommendations on market prices. The recommendations are published in columns of the Wall Street Journal, made by prominent money managers at Barron’s Annual Roundtable, published in investment newsletters, or published in investment magazines. Two columns that are investigated most often are the ‘Heard on the Street’ column, and the ‘Dartboard’ column. During the 1970s and 1980s the financial service Value Line is investigated often. Value Line analyses and ranks more than 1400 stocks. Periodically, Value Line stages a contest to attract investors’ attention. Value Line expects that portfolios chosen from the stocks it ranks highest will outperform those chosen in other ways. The empirical papers concerning Value Line are discussed as well.

Davies and Canes [1978], Liu, Smith, and Syed [1990], Huth and Maris [1992], and Beneish [1991] examined the effect of the publication of analysts’ recommendations in the Wall Street Journal column ‘Heard on the Street’ (HOTS) on market prices. Their conclusions were similar. Stocks recommended to buy are associated with positive significant average abnormal stock price performance on the day of the publication in Heard on the Street whereas stocks recommended to sell display a negative

(14)

provided a short-term trader with net returns in excess of transaction costs contains the smaller firms that were the objects of negative comments in the column. Lastly, Beneish [1991] found that stocks recommended to buy are associated with positive significant average abnormal stock price

performance on the day of the publication in Heard on the Street and the two preceding trading days. Stocks recommended to sell exhibited negative significant abnormal stock price performance on the day of the publication.

Barber and Loeffler [1993], Albert and Smaby [1996], and Liang [1999] examined whether analysts’ recommendations in the ‘Dartboard’ column of the Wall Street Journal have an impact on stock prices. They concluded similar: buy recommendations inflict significant abnormal stock returns around the date of publication. However, Barber and Loeffler [1993] reported that positive abnormal stock returns around the publication date of a recommendation are at least partially reversed within 25 trading days. Liang [1999] found that the initial price effect is to be reversed within 15 trading days. These findings suggest that the initial response of the market price is partially attributable to temporary price pressure. Albert and Smaby [1996] also detected a difference between recommendations made by analysts who recommended stocks in the column more than once and to those recommended by newcomers. According to Albert and Smaby [1996] this suggests that the market uses the column to identify superior analysts. This is valid if the column does not invite back all analysts to the column, but only the ones who have proven to recommend stocks that will generate positive abnormal returns. Those analysts, who are invited back to the contest and have indeed given valuable investment advice in past editions of the column, have proven to be a superior analyst. If that is the case, the more often an analyst is invited back to the contest, the better his or her forecasting skill has been in the past.

Value Line analyses and ranks more than 1400 stocks. Periodically, Value Line stages a contest to attract investors’ attention. Value Line expects that portfolios chosen from the stocks it ranks highest will outperform those chosen in other ways. Black [1971], Kaplan and Weil [1973], Holloway [1981], and Copeland and Mayers [1982] investigated whether Value Line is a useful financial service for investors. The results are mixed. Black [1971], and Copeland and Mayers [1982] found statistically significant abnormal returns. Black [1971] found that an investor can capture significant abnormal returns by following Value Line recommendations. This implies that the returns of the recommended stocks by Value Line are in conflict with a semi-efficient market, because Value Line

(15)

abnormal returns at all; they addressed the abnormal returns generated by Value Line’s highest ranked stocks to risk differences in those stocks. Holloway [1983] re-examined the Value Line

recommendations for a larger sample. Holloway [1983] found that investing according to Value Line recommendations generated significant abnormal returns. The abnormal returns Holloway [1983] found cannot be explained by risk differences of the stocks.

Desai and Jain [1995], Wright [1994], Walker and Hatfield [1996], Sant and Zaman [1996], Desai, Liang, and Singh [2000], and Ferreira and Smith [2003] studied the effectiveness of recommendations issued by financial gurus. Where Wright [1994], Sant and Zaman [1996], Ferreira and Smith [2003], and Desai et al. [2000] found significant abnormal returns, Desai and Jain [1995] concluded that investors who invest according to the recommendations of ‘superstar’ money managers would not have benefited from the advice, because the magnitude of the abnormal returns was very small. Stocks recommended to buy generated insignificant abnormal returns of 0.33 percent, 0.21 percent, -0.38 percent, and -0.71 percent for holding periods 25 days, 250 days, 500 days, and 750 days respectively following the publication day. The market does react on the publication day, significantly. The publication day abnormal returns are 1.04 percent. On the publication day stocks recommended to sell generated significant abnormal returns of -1.16 percent. In the post-publication periods, the stocks recommended to sell generated statistically significant abnormal returns, on average, -8.12 percent for the 250 day holding period.

Like Desai and Jain [1995], Walker and Hatfield [1996] found that while analysts do identify under-priced or over-under-priced stocks, it is difficult for an investor to capitalize on investment

recommendations. Wright [1994] found that virtually all of the initial abnormal return is gradually retracted by the market over the succeeding 39 trading days after the publication day of a

recommendation. This reversion of abnormal returns is also found by Ferreira and Smith [2003], Jegadeesh [1990], and Lehman [1990].

Sant and Zaman [1996] found that recommended stocks published in a Business Week column generated significant abnormal market reactions in the publication week. However, Sant and Zaman [1996] documented that the 6-month post-recommendation abnormal return for stocks recommended to buy is negative and highly significant. Sant and Zaman [1996] stated that the results are an indication of the self-fulfilling prophecy effect of recommendations. Compared to the more closely followed stocks the self-fulfilling prophecy effect is strongest for the less closely followed stocks, which usually have lower average trading volume but the greatest increase in trading around the publication day of a recommendation.

Graham and Harvey [1997] investigated the stock selection abilities of investment newsletters.

(16)

not outperform the market. Nevertheless, investment newsletters that have published recommendations that correctly anticipated the direction of the market in previous issues may provide valuable

information about future returns. Like Graham and Harvey [1997], Jaffe and Mahoney [1999] found that stocks recommended in investment newsletters do not outperform appropriate benchmarks, where appropriate benchmarks are control firms. However, Jaffe and Mahoney [1999] found no evidence of performance persistence of investment newsletters. Kakebeeke [1999] investigated recommendations published in magazines issued by the four biggest banks in the Netherlands. Like Jaffe and Mahoney [1999] and Graham and Harvey [1997], Kakebeeke [1999] found that trading according to buy recommendations generated a return significantly lower than the return of the benchmark. Stock recommended to sell, however, generated a higher but insignificant return than the return of the benchmark. Kakebeeke [1999] concluded that investors cannot outperform the market by following analysts’ recommendations. As analysts’ recommendations are based on publicly available

information, the Dutch stock market passed the semi-strong form test of market efficiency. Kakebeeke [1999] suggested a time-lag as an explanation for the inability of recommendations published in monthly magazines to provide an investor with valuable information. In other words, the

recommended publications arrive too late. Kakebeeke [1999] found almost six times as many buy recommendations than sell recommendations. Metrick [1999] found no evidence for significant abnormal returns in the short-run as well. Yazici and Muradoglu [2001] did find a statistically significant impact on stock prices. The cumulative abnormal returns started to pick up 8 days before the publication date. On the publication day of the recommendation, abnormal returns were 2.49 percent. However, Yazici and Muradoglu [2001] concluded that published investment advice does not help small investors earn abnormal returns, as the cumulative abnormal returns between days [+1] and [+20] in event time are negative. If one could front-run the recommendations by five days, abnormal returns of more than 5 percent could be captured. The findings done by Yazici and Muradoglu [2001] seem to suggest that trading takes place prior to the official publication of the recommendation.

Wijmenga [1987], Menendez-Requejo [2005], and Kerl and Walter [2005] did find a statistically significant abnormal return in the publication week. Wijmenga [1987] evaluated the stock

(17)

variant of the efficient market hypothesis according to which the use of publicly available information cannot yield any abnormal returns.

Kerl and Walter [2005] analyzed buy and sell recommendations for stocks published by the German Personal Finance Magazine. The authors found that buy recommendations are associated with positive cumulative abnormal returns, whereas sell recommendations are associated with negative cumulative abnormal returns. For a five-day period around the event, cumulative abnormal returns of 2.58 percent for stocks recommended to buy and cumulative abnormal returns of -1.81 percent for stocks

recommended to sell are found. In addition, the trading volume increases to around 161 percent of the normal level for buy recommendations and to 285 percent for sell recommendations around the event day. Furthermore, Kerl and Walter [2005] found that smaller stocks displayed greater price reaction than larger stocks.

Menendez-requejo [2005] analyzed the return and trading volume of recommended stocks by analysts publishing in one of the most disseminated Spanish financial newspapers. Menendez-requejo [2005] found that the market reacts before the publication of the recommendations. The cumulative return for stocks recommended to buy was 1.13 percent and -2 percent for sells. Furthermore, the trading volume of a recommended stock was greater than the average trading volume.

To sum up the empirical findings discussed in this section: stock recommendations published by brokerage firms do seem to be of value to an investor, since following stock recommendations published by brokerage houses enables an investor to capture abnormal returns. Furthermore, brokerage houses are more likely to issue buy recommendations than sell recommendations. This could be referred to as broker optimism, as well as an incentive for a broker to favor issuing a buy recommendation to a sell recommendation, because of the storage costs brokers earn by holding stocks for their clients.

Further, analysts’ recommendations published in columns inflict a significant market reaction on the publication date of the column. Recommended stocks published in all the examined columns seem to lead to abnormal returns. However, the initial effects of a recommendation seem to be reversed within a short period of time. This suggests that recommendations published in columns confirm the validity of the price pressure hypothesis. Furthermore, results of, among others, Huth and Maris [1992] and Walker and Hatfield [1996] state that it is hard for an individual investor to capture abnormal returns generated by analysts’ recommendations when transaction costs are considered.

(18)

abnormal returns around the day of the announcement. However, a reaction takes place prior to the official publication date of the recommendation. This results in the inability of an individual investor to act on those recommendations. Even though some analysts tend to disclose their information publicly, it can be argued that the public disclosure probably comes only after the analysts' clients have already traded on that information. This implies that the information contained in a published recommendation may not be valuable to investors since it may have already been factored into stock prices. For a schematic overview of the empirical literature, see table 2.1.

Table 2.1

Schematic overview of the empirical literature

Table 2.1 shows the authors of the empirical paper, the year in which the paper is published, the country where the investigation took place, the composition of the sample, and the main conclusion(s) of the paper. Further, if applicable, a *, **, or a *** indicates whether the conclusions are significant at a 10%, 5%, or a 1%

significance level.

Author(s) Year Country Sample Main Conclusion(s)

Cowles 1933 USA

16 financial services companies, 20 fire insurance companies, 25 periodicals and

1 editor from WSJ

Recommended stocks had, on average, a negative performance when compared to

a benchmark.

Cowles 1944 USA 11 leading financial

periodicals

Periodicals failed to successfully predict the future course of the stock market. Diefenbach 1972 USA

Recommendations from the brokerage community from November 1967 to May 1969

Buy recommendations did better than the benchmark during a bull market, but

worse during a bear market. Kaplan and

Weil 1973 USA Value Line, 1972

Authors do not believe Value Line generates abnormal returns, differences

are due to risk. Logue and

Tuttle 1973 USA

Recommendation from 6 major NYSE brokers from

July 1970 to June 1971

Following brokerage advice does just as well as randomly selecting stocks. Davies and

Canes 1978 USA WSJ from 1970 to 1971

Stock prices adjust to analysts’ recommendations. Information is

incorporated quickly into prices. Groth, Lewellen, Schlarbaum, and Lease 1979 USA Recommendations made by one brokerage house from

1964 to 1970

Significant abnormal returns***, however, only a small fraction of clients

might have been able to capture those abnormal returns.

Holloway 1981 USA Value Line recommendations

from 1974 to 1977

After accounting for transaction costs, active trading according to Value Line recommendations does not generate

abnormal returns. Copeland and

Mayers 1982 USA

Value Line recommendations from 1965 to 1978

Significant abnormal** returns, but abnormal returns are declining in the

newer sub-samples. Bjerring, Lakonishok, and Vermaelen 1983 Canada Recommendations from a Canadian broker from 1977 to

1981

Broker was successful in outperforming the market, for both Canadian** and

US*** stocks.

Holloway 1983 USA Value Line recommendations

from 1974 to 1981

(19)

Dimson and

Marsh 1984 UK

4187 specific share return forecasts made by 35 UK stockbrokers and by the internal analysts of a large UK

investment institution during 1980 and 1981

A small but potentially useful degree of forecasting ability is found. A large part

of the information content of the forecasts, however, appears to be discounted in the market place within

the first month. Glascock,

Henderson, and Martin

1986 USA Recommendations made by

the broker E.F. Hutton

Recommended stocks performed good*. The most strongly recommended stocks

performed better than those with a weaker endorsement.

Wijmenga 1987 NL

Three weekly Dutch magazines during the period January 1978 until December

1983

Highly significant*** abnormal return in the week of publication, partly due to a

self-fulfilling prophecy effect. Liu, Smith, and

Syed 1990 USA

Daily HOTS column from September 1982 to September

1985

HOTS column appears to have an impact on stock prices on the publication

day***.

Beneish 1991 USA HOTS column 1978 to 1979

Recommended stocks generate significant abnormal stock price

performance on the day of the publication** in HOTS and the two

preceding trading days**. Huth and

Maris 1992 USA HOTS column during 1985

Statistically significant*** but economically insignificant stock price

movements. Barber and

Loeffler 1993 USA

95 Pros’ Picks and 94 Dartboard stocks, October

1988 – October 1990

The stocks selected by the professional analysts, on average, experienced a 4.06

percent abnormal return over the two-day period (publication two-day* and

subsequent day).

Wright 1994 USA WSJ: Your Money Matters

Column

Significant abnormal return on the publication day***. Virtually all of the abnormal return is gradually retracted by

the market over the succeeding 39 trading days.

Desai and Jain 1995 USA Barron’s annual roundtable,

1968 to 1991

Investors who invest according to the recommendations of ‘superstar’ money managers would not have benefited from

the advice. Significant abnormal return on the publication day is found***.

Stickel 1995 USA

Brokerage house buy and sell recommendations from the four-year period 1988 to 1991

were supplied by Zacks Investment Research

Buy recommendations are associated with short-term price increases*** whereas sell recommendations display negative price reactions around the day

of publication***.

Albert and

Smaby 1996

USA WSJ: Dartboard column

October 1988 – December 1991

Significantly positive abnormal returns on the WSJ issue date***. Securities recommended by analysts invited back

to the contest do better* than those recommended by newcomers of the

contest. Walker and

Hatfield 1996 USA

‘Market Highlights’ Section of USA Today, January 1988

to December 1990

Analysts do identify under-priced or over-priced stocks, it is difficult for an

investor to capitalize on them. Sant and

Zaman 1996 USA

Business Week column from January 1976 through

December 1988

Three day abnormal response for the positive firms is a highly significant 2.44

(20)

post-BW performance of firms with favorable reports is, however, negative

and significant***, indicating that the BW story only has a temporary effect

which reverses itself later.

Womack 1996 USA 14 biggest U.S. brokerage

houses, 1989 to 1991

Significant abnormal returns for buy*** and sell*** recommendations.

Post-event drift exists as well. Analysts appear to have market timing and stock

picking abilities.

Graham and

Harvey 1997 USA

326 newsletter asset-allocation strategies for the

1983-95 period

As a group, newsletters do not appear to possess any special information about the future direction of the market. Newsletters that have correctly

anticipated the direction of the market in previous recommendations may provide valuable information about future returns.

Jaffe and

Mahoney 1999 USA

Investment newsletters followed by the Hulbert

Financial Digest

Taken as a whole, the securities that newsletters recommend do not outperform appropriate benchmarks.

Kakebeeke 1999 NL

Recommendations of the periodicals from three out of

four biggest banks in NL, 1992-1995

Dutch market is significantly semi-strong. The portfolio with buy recommendations generates a return

significantly** lower than the benchmark. Self-fulfilling prophecy

effect of recommendations.

Liang 1999 USA

WSJ: Dartboard columnfrom

January 1990 to November 1994

Advice of financial gurus seems to lead to significant market reactions**. Initial price effect is to be reversed within 15

trading days.

Metrick 1999 USA 153 investment newsletters No evidence for significant abnormal

returns in the short-run. Desai, Liang,

and Singh 2000 USA

Recommendations made by Wall Street Journal all-star analysts from 1993 to 1996

Stocks recommended by the all-star analysts outperform** benchmarks controlled for size and industry. Barber, Lehavy, McNichols, and Trueman 2001 USA Recommendations provided by Zacks Investment Research, from 1985 to 1996

Abnormal returns*, but an investor could not profitably capture these abnormal returns, after accounting for

transaction costs. Yazici and

Muradoglu 2001 Turkey

Stock recommendations by Investor Ali during the period

December 1993 to July 1998

Significant impact on stock prices*. Cumulative Abnormal Return starts

picking up from t= -8.Published

investment advice does not help small investors earn excess returns. Barber, Lehavy, McNichols, and Trueman 2002 USA Analyst Recommendations provided by First Call from January 1996 to December

2001

The more highly recommended stocks earned greater market-adjusted returns during the 1996-1999 period* than the less highly recommended stocks.

In the years 2000 and 2001 the least favorably rated stocks earned the highest

returns. Brav and

Lehavy 2003 USA

Database provided by First Call of analysts’ target prices

issued over the period 1997 1999

Significant market reaction to the information contained in analysts’ target

prices**. On average, the one-year-ahead target price is 28 percent higher

than the current market price. Ferreira and

Smith 2003 USA

Transcripts of the Wall Street Week between December 1996 and December 1997

Significantly abnormal returns of 0.65 percent for the recommendations on the

(21)

Friday**. The gains, however, reversed over the next four trading days. Jegadeesh,

Kim, Krische, and Lee

2004 USA Data from Zacks, 1985 to

1998

Analysts from sell-side firms generally recommend ‘glamour’ stocks.

Chakrabarti 2004 India

Recommendations made by analysts from 26 firms from January 1998 to July 2003.

Analysts tend to be optimistic in their predictions, recommending buys considerably more often than sells. Their

recommendations do have investment value at least in the near term**. Clear buy recommendations appear to be the most valuable. The recommendations also seem to have some impact on stock

price. Jegadeesh and Kim 2005 G7 countries Recommendations in the G7 countries between 1993 and

2002

Stock prices react significantly to recommendation revisions on the revision day** and on the following day** in all of these countries except Italy. Largest price reactions are found

in the USA.

Kerl and

Walter 2005 Germany

Stock recommendations published by German

Personal Finance Magazines

from 1995 to 2003.

Buy (sell) recommendations earn significant abnormal returns of 2.58 percent*** (-1.81 percent***) within the five days around the publication day. Results are mainly driven by high abnormal returns for small stocks and value stocks.

Menendez-Requejo 2005 Spain

Recommendations made in the Spanish paper Cinco Dias,

from 1997 to 1999

The market reacts before the publication of the recommendations***, but no significant abnormal return starting from

the moment that the information is published are found.

Martins, Ribeiro, arreto, and Serra 2005 Portugal Recommendations made public by a Portuguese investment bank from 1999 to

2003

Results suggest that these recommendations have not been useful

for stock selection. Most of the recommendation-based strategies we have built generated negative significant performance***. The

recommendations-based strategies appear to be more effective for selection of Portuguese stocks, which can reflect a comparative

home advantage. Lonkani, Khanthavit, and Chunahachinda 2006 Thailand

I/B/E/S data of analysts’ recommendations from 1993 to 2002

Following strong buy recommendations, on average, outperforms the market**.

Existence of ‘insider’ and ‘herding’ aspects associated with the recommendations are found.

Green 2006 USA

Recommendation changes from 16 major brokerage firms during the period 1999

through 2002.

After controlling for transaction costs, purchasing quickly following upgrade recommendations results in an average two-day return of 1.02%(p-value of 0),

whereas selling short following downgrades produces a return of

1.50%(p-value of 0).

(22)

recommendations by the brokers under investigation in this paper are based on publicly available information, and should therefore not contain any information that causes a stock to be revalued. Thus, as the recommendations should not have informational content, no price reactions should be visible. The review of other empirical papers, however, has come up with different effects of analysts’

(23)

3. Hypotheses

After reviewing the relevant theory and the previous findings of section 2, the following hypotheses are constructed. When a semi-efficient market is assumed, an analyst´ recommendation should not lead to an abnormal market reaction. This assumption is the basis of the constructed hypotheses and has resulted in the construction of hypothesis 1. However, the previous findings have suggested that analysts’ recommendations may be of value to an investor. Therefore it is investigated whether an individual investor can generate abnormal returns by following analysts’ recommendations. This is elaborated in hypotheses 2 and 3. Finally, the possible difference in forecasting skill of the

investigated brokers is elaborated in hypothesis 4.

Hypothesis 1

H0: Analysts’ recommendations do not lead to an abnormal market reaction at the publication day of a recommendation.

H1: Analysts’ recommendations do lead to an abnormal market reaction at the publication day of a recommendation.

Hypothesis 2

H0: No positive abnormal returns can be captured by an individual investor who follows analysts’ buy recommendations.

H1: Positive abnormal returns can be captured by an individual investor who follows analysts’ buy recommendations.

Hypothesis 3

H0: No positive abnormal returns can be captured by an individual investor who follows analysts’ sell recommendations.

H1: Positive abnormal returns can be captured by an individual investor who follows analysts’ sell recommendations.

Hypothesis 4

(24)

H1: There is a difference in the forecasting skill of the several recommending brokers under investigation.

(25)

4. Methodology and Data

In order to investigate the hypotheses constructed in section 3 an event study is carried out. Abnormal returns associated with analysts’ recommendations are investigated. The potential market reaction inflicted by analysts’ recommendations is analyzed by assessing these abnormal returns before and at the publication day. By analyzing abnormal returns for a period after the publication date of the recommendation, it is evaluated whether an investor can earn abnormal returns by following analysts’ recommendations. This is done for both the sample as a whole and for individual brokers.

The abnormal return is the realized ex post return of the stock over the event window minus the normal return of the stock over the event window. The normal return is defined as the expected return without the event taking place. Normal returns are estimated using the market model.

The publication date of a recommendation is defined as day 0 in event time. The event window is defined as day -20 to day 20 in event time. The period from day t = -240 to t = -21 is taken as the estimation period. As a larger estimation window decreases the sampling error in

α

i and

β

i, a

relatively large estimation period is chosen. For raw returns of each recommended stock OLS parameters are estimated in the estimation period while using the AEX return as the independent variable. The AEX index is chosen as a benchmark for its relevancy to the stocks under investigation. The choice of a benchmark period is important because it can affect the outcome of the research. The benchmarking with respect to the AEX index is justified, since the average beta of the stocks in the whole sample with respect to the index is 1.00 with a maximum beta of 2.07 and a minimum beta of 0.33. Thus the stocks are not, as a group, significantly more risky than the index. Historical AEX prices are obtained from uk.finance.yahoo.com. Historical closing prices are taken to calculate daily market returns. Closing prices are adjusted by Yahoo Finance for dividends and stock splits. Daily closing prices of stocks are obtained from DataStream and used to calculate daily stock returns.

(26)

4.1. Sample selection and descriptive statistics

Data in the form of recommendations are available at MarketAdvices.com and guruwatch.nl. The recommendations published on MarketAdvices.com contain recommendations for stocks of the AEX index, the BEL 20 index, the CAC40 index, the DAX index, the Dow Jones 30 index, the FTSE 100 index, and the NASDAQ 100 index. The recommendations published on guruwatch.nl contain recommendations for AEX-listed stocks and AMX-listed stocks. This paper focuses on main Dutch stocks. Hence, only stock recommendations for stocks listed on the main Dutch stock market, the AEX index, are included in the sample. The brokers that are investigated are reported in table 4.2.

Brokers follow different reporting styles and use different terminology in their reports. This is recognized by MarketAdvices.com and guruwatch.nl. This paper follows the translation of

MarketAdvices.com and translates all recommendations and methods to a three point scale, using the categories buy, hold, and sell. As a consequence, recommendations published on guruwatch.nl are transformed to the MarketAdvices.com-scale. For a complete description of the classification see the appendix.

Only buy recommendations and sell recommendations are included in the sample, as these are most likely to cause reactions. Buy recommendations stemming from an IPO or sell recommendations stemming from a bankruptcy filing are excluded from the sample. Reiterations of a recommendation within a month of the publication of the first recommendation are removed from the sample. This means that, for example, a recommendation published by ING Financial to buy Akzo Nobel is removed from the sample when the same recommendation has been published by ING Financial to buy Akzo Nobel in the previous month. Publications regarding a stock being under review are omitted from the sample as well. Four recommendations of the stock TOMTOM made in April 2007 and one made in May are omitted, because OLS parameters could not be calculated. This is a result of the entry of TOMTOM in the AEX index since March 2nd, 2006. For TOMTOM, recommendations published after May 2007 are included into the sample. The final sample consists of 700 buy recommendations and 140 sell recommendations over the years 2005 and 2006. Table 4.1 shows the descriptive statistics of the final sample.

Table 4.1

Number of Buy and Sell Recommendations for the whole sample, per year

Year Buy Recommendations Sell Recommendations

2005 320 81

2006 380 59

(27)

recommendations is also found by, among others, Diefenbach [1972], Logue and Tuttle [1973], and Kakebeeke [1999]. Table 4.2 reports the brokers that are investigated and the number of buy recommendations issued per broker.

Table 4.2

Number of Buy Recommendation per Broker for 2005 and 2006

Broker Number of Buy Recommendations

(28)

4.2. Methodology

For recommendation i and date t the abnormal return for each day in the event period is

(4.1) mt i i it it

R

R

AR

=

α

) −

β

)

where is the return on the AEX index for day t, and where is the daily stock return for stock i on day t. Closing prices are taken to calculate daily stock returns. and

mt

R

R

it

i

α

)

β

)

i are OLS values from the estimation period.

1

)

(

) 1 (

=

it it it

S

S

R

(4.2)

− = − = − = − =

=

21 240 2 21 240

)

(

)

)(

(

t t m mt t t m mt i it i

R

R

R

μ

μ

μ

β

)

)

)

)

(4.3) m i i i

μ

β

μ

α

)

=

)

)

)

(4.4) where

− = − =

=

21 240 1

1

t t it i

R

L

μ

)

, (4.5) and where

− = − =

=

21 240 1

1

t t mt m

R

L

μ

)

. (4.6)

where is the length of the estimation window. The abnormal return observations are aggregated for the event window and across observations. For this aggregation to be valid, it is assumed that there is not any clustering. The goal of an event study is to isolate and investigate the impact of an event on a firm’s stock price. Clustering occurs when events are overlapping in time. Brown and Warner [1980] found in a simulated environment with monthly stock returns that clustering appears to have little impact on rejection frequencies when the market model is employed to calculate abnormal returns. Brown and Warner [1985] reassessed the effect of clustering for daily stock returns and found that the results are not radically altered when there is clustering in event dates, if the market model is used to calculate abnormal returns. Christie [1986] briefly reviewed three accounting studies that suffered from overlapping event windows and found no evidence that cross-sectional dependence in the data caused serious bias in standard errors. Because no clustering is assumed, the covariance terms between events are assumed zero, and allows us to calculate the variance of the aggregated sample of

cumulative abnormal returns. 1

(29)

For a sample of N recommendations, the aggregated abnormal return for date t is

=

=

Nt i it t t

AR

N

AR

1

1

t = -20,…,+20. (4.7)

The test for statistical significance of abnormal returns is based on MacKinlay [1997]. The null hypotheses that state that abnormal returns are zero are thus tested by using

2 / 1

)

var(

t t

AR

AR

(4.8) where

= = N i t i N AR 1 2 2 1 ) var(

σ

ε , (4.9)

and where 2 is an estimated parameter from the OLS procedure, and calculated as

i ε

σ

− −

=

21 240 2 1 2

)

(

2

1

mt i i it

R

R

L

i

α

β

σ

)

ε

)

)

, (4.10)

For formula 4.9 to be valid, the estimation period must be sufficiently large, because the variance of the abnormal returns consists of two components, the disturbance variance and an additional variance as a consequence of the sampling error in

α

i and

β

i. As previously mentioned in section 4, the

sampling error of

α

i and

β

i decreases as the estimation period becomes larger.

The average abnormal returns are aggregated over the event window. Cumulative daily abnormal returns are calculated as

= = = 2 1 ) 2 1, ( t t t t t t t AR CAR . (4.11)

The test for statistical significance of cumulative abnormal returns is based on MacKinlay [1997]. The null hypotheses that state that abnormal returns are zero are thus tested by using

2 / 1 ) , ( ) , (

)

var(

1 2 2 1 t t t t

CAR

CAR

(4.12)

(30)

= = 2 1 2 1 ) var( ) var( ( , ) t t t t t t AR CAR (4.13)

To test for hypothesis 1 the reaction at t = 0 is examined. The market reaction is examined by

assessing abnormal returns. As the return on stock i on day 0 is calculated by using the closing price of day t = 0 and t = -1, the effect of a recommendation on t = 0 will be incorporated into the appropriate

of t = 0. The appropriate t-statistics of abnormal returns are calculated to demonstrate their possible statistical significance. To test for hypotheses 2 and 3 abnormal returns and cumulative abnormal returns are investigated for the sample as a whole. The possible informational content of a recommendation that may cause a stock to either increase or decrease in price will be incorporated into stock prices quickly after a recommendation. This is discussed in section 2. Due to this quick

incorporation into prices, an individual investor is often too late to benefit from the possible

informational content on the publication day of a recommendation. As mentioned in section 2, Groth et al. [1979] suggested that, because of the quick incorporation of information into prices, only a minority of the firm's clients is in a position to respond in time to any given investment

recommendation. To accommodate the issues an individual investor faces, it is assumed that an investor can buy or sell a recommended stock at the closing price of day [0]. This implies that an investor can capture possible abnormal returns from day [+1], because the return on day [+1] is calculated by the difference of the closing price of day [0] and day [+1]. Hypothesis 4 is tested by comparing the different abnormal returns and cumulative abnormal returns for the brokers under investigation. Differences between brokers are determined by only examining buy recommendations, since the number of sell recommendations per broker is too small.

it

(31)

5. Empirical Results

5.1. Market Reaction to Analysts’ Recommendations

An important research question of this paper is whether analysts’ recommendations have an impact on prices of the recommended stocks at the publication date of a recommendation. As discussed in

section 2, analysts´ recommendations are based on publicly available information and should therefore not inflict a revaluation of stock prices when a semi-efficient capital market is assumed. However, as documented in section 2 of this paper, previous research has found that analysts´ recommendations may possess valuable information. Bjerring et al. [1983], Dimson and Marsh [1984], Glascock et al. [1986], Womack [1996], Jegadeesh et al. [2004], Chakrabarti [2004], Jegadeesh and Kim [2005], Lonkani et al. [2006], and Green [2006] all found significant positive abnormal returns for an investor who traded according to analysts’ recommendations. If analysts´ recommendations do possess

valuable information it will cause a stock to be revalued. If this revaluation does arise, it is expected to happen on the publication day of the recommendation, since the Dutch stock market is considered to incorporate new information quickly. As mentioned in section 4.2 the market reaction to analysts’ recommendations is examined by assessing abnormal returns. The possible statistical significance of abnormal returns is calculated for all days in the event window. Bearing in mind the possible

(32)

Table 5.1

Abnormal Returns, Cumulative Abnormal Returns, and t-statistics for all Buy and Sell Recommendations of the whole sample

Table 5.1 shows abnormal returns (ARs), cumulative abnormal returns (

CAR

s), and the appropriate

t-statistics for the event period [-20, +20].

Buy Recommendations Sell Recommendations

Day AR t-stat

CAR

t-stat Day AR t-stat

CAR

t-stat

(33)

The most striking observation is that no statistical abnormal returns are reported. For stocks

recommended to buy a small and statistically insignificant reaction takes place on the publication date of the recommendation (t = 0). The maximum value of a daily abnormal return is reported for day [0], with 0.30 percent. Except for the publication day,abnormal returns seem to fluctuate without being consequently positive or negative.

With respect to stocks recommended to sell a similar small and statistically insignificant reaction takes place at the publication date of the recommendation. The minimum value of a daily abnormal return is reported for day [+5], with -0.26 percent. Abnormal returns seem to fluctuate without being

consequently positive or negative as well, except for a clear peak on day [0] and [+5].

Cumulative abnormal returns are examined to visualize the incorporation of new relevant information into stock prices. For stocks recommended to buy, small cumulative abnormal returns start to build up from day [-2] and reach their maximum value on day [+2], with 0.37 percent. None of the reported cumulative abnormal returns are statistically significant.

(34)

Figure 5.1

Cumulative Abnormal Returns for Recommended Stocks

This figure plots cumulative abnormal returns (

CAR

s) for recommended stocks (buy and sell) for the entire

event period [-20, +20] for the entire sample.

-1,20% -1,00% -0,80% -0,60% -0,40% -0,20% 0,00% 0,20% 0,40% 0,60% -25 -20 -15 -10 -5 0 5 10 15 20 25 t CAR Buy Sell

An interesting observation is that for buy recommendations, most of the market reaction takes place at the publication date of the recommendation. Furthermore, the lack of a pre-event reaction is worth mentioning. This suggests that recommendations to be published are not disseminated prior to the official publication date of the recommendation. However, as mentioned before, the abnormal returns are insignificant, and therefore the lack of a pre-event reaction is insignificant as well.

(35)

5.2. Analysis of Abnormal Returns for the Individual Investor

In order to test the second and third null hypotheses abnormal returns and cumulative abnormal returns starting from day [+1] are investigated. As daily closing prices are used to calculate daily returns, see section 4, abnormal returns of day [+1] are calculated by using the closing price of day [0] and day [+1]. By analyzing abnormal returns and cumulative abnormal return from day [+1], it is implied that an individual investor buys or sells a stock at the closing price of day [0]. By buying or selling at the closing price of day 0, an investor captures the return realized between the closing price of day 0 and day 1, or

R

1. Table 5.2 reports the results starting from day [+1].

Table 5.2

Abnormal Returns, Cumulative Abnormal Returns, and t-statistics for Recommended Stocks starting from day [+1]

Table 5.2 shows abnormal returns (ARs), cumulative abnormal returns (

CAR

s), and the appropriate

t-statistics for the event period [+1, +20].

Buy Recommendations Sell Recommendations

Day AR t-stat

CAR

t-stat Day AR t-stat

CAR

t-stat

1 0,00% -0,002 0,00% 0,069 1 -0,16% -0,092 -0,16% -0,091 2 -0,05% -0,039 -0,05% 0,059 2 -0,11% -0,080 -0,27% -0,106 3 -0,08% -0,058 -0,13% 0,046 3 -0,04% -0,033 -0,31% -0,110 4 0,03% 0,021 -0,11% 0,049 4 0,10% 0,082 -0,20% -0,093 5 -0,04% -0,031 -0,15% 0,041 5 -0,26% -0,249 -0,46% -0,130 6 0,01% 0,010 -0,14% 0,042 6 0,06% 0,055 -0,41% -0,120 7 -0,04% -0,030 -0,17% 0,037 7 -0,07% -0,076 -0,48% -0,129 8 -0,08% -0,066 -0,25% 0,024 8 0,06% 0,052 -0,42% -0,119 9 -0,07% -0,063 -0,33% 0,014 9 0,00% 0,001 -0,42% -0,118 10 -0,14% -0,121 -0,47% -0,006 10 -0,06% -0,057 -0,48% -0,125 11 -0,07% -0,055 -0,53% -0,015 11 -0,04% -0,036 -0,51% -0,129 12 -0,02% -0,015 -0,55% -0,018 12 0,01% 0,008 -0,51% -0,126 13 0,05% 0,049 -0,50% -0,010 13 0,15% 0,126 -0,36% -0,104 14 -0,07% -0,065 -0,57% -0,020 14 0,03% 0,034 -0,32% -0,099 15 -0,07% -0,066 -0,64% -0,029 15 0,11% 0,119 -0,21% -0,084 16 -0,01% -0,011 -0,65% -0,030 16 0,01% 0,010 -0,20% -0,081 17 0,05% 0,049 -0,59% -0,023 17 0,08% 0,062 -0,12% -0,071 18 0,11% 0,093 -0,49% -0,009 18 0,06% 0,050 -0,06% -0,062 19 0,03% 0,032 -0,45% -0,004 19 0,16% 0,127 0,10% -0,041 20 0,03% 0,024 -0,43% -0,001 20 0,00% 0,003 0,10% -0,040

(36)

and selling at day [+16] would achieve statistically insignificant negative abnormal returns of -0.65 percent.

As this paper investigates the ability of an individual investor to capture abnormal returns, trading following sell recommendations is not a valid option. This is due to the fact that an individual investor does not have the possibility to sell short stocks nor is he or she likely to possess all individual stocks of the Dutch stock market at any moment in time. Aside from these practical issues and if an investor would be able to sell short stocks, small abnormal returns could be captured. However, these returns are statistically insignificant. A strategy involving selling short all stocks that have been recommended for sale at the closing price of day [0] and for example covering the short position at day [+12] would generate statistically insignificant abnormal returns of -0.51 percent. The initial effect, however, is reversed and completely disappeared at day [+19]. Figure 5.2 plots the cumulative abnormal returns for recommended stocks starting from day [+1].

After the analysis of buy and sell recommendations and their use for the individual investor, both the second and third null hypotheses cannot be rejected. For both buy and sell recommendations,

Referenties

GERELATEERDE DOCUMENTEN

In the whole sample and in all-size stocks in both stock exchanges, the highest mean return occurs on days before the Chinese Lunar New Year, with 1.063 and 1.314 in

The overreaction hypothesis predicts that the average α p over the five years of the test- period should be positive for the extreme prior losers (portfolio 1) and

Cumulative abnormal returns show a very small significant reversal (significant at the 10 per cent level) for the AMS Total Share sample of 0.6 per cent for the post event

Figure 7 Conceptual model Environmental beliefs H3 Country of origin effect H5 Energy labels H2 Purchase intention green labeled cars Gas prices H1 Performance H4

The out of sample results for the PPP model, where the benchmark model in pa- rameterisation (2) is the value weighted portfolio, with different levels of relative risk aversion

According to Chuang and Susmel (2011), the Taiwanese stock market shows evidence that individual investors trade with more overconfidence than institutional

The stock price is used as a selection criteria as stock liquidity is correlated with the stock price of a stock as is shown by Gargett (1978). 3) From the control firms that

Table 20, in the Appendix, shows the monthly effect in the volatility of the mean excess returns after running the regression of equation 4 for the period after the crisis..