Can investors profit from publicly available analyst recommendations? A study on analysts’ forecasting skills

1_{Student at the University of Groningen; Faculty of Economics and Business.} Email: karinhijmans@hotmail.com. Student number: S1874713.

Can investors profit from publicly available analyst recommendations?

A study on analysts’ forecasting skills

Karin Hijmans1

Master thesis Finance: EBM866B20

University of Groningen Supervisor: Dr. A. Plantinga Date: January 15th_{, 2015}

Abstract

This paper investigates the potential for investors to profit from publicly available analyst recommendations. It is expected that investors are able to profit from these recommendations because analysts tend to have more information than the investor. The models used are the CAPM and the Fama and French 3-factor model. The results indicate that negative recommendations (i.e. sell) are most likely to be correct and thus generate a positive return, whereas positive recommendations (i.e. buy) generate a negative return. Furthermore, a distinction is made between recommendations with- and recommendations without target prices.

JEL classification: G1; G11, G15

1

1. Introduction

“Wall street is the only place that people ride to in a Rolls Royce to get advice from those who take the subway.”

- Warren Buffet –

The last few years have been dificult for stock market investors all over the world. The optimistic mindset resulting from substantial stock market returns during the previous decades, made room for decreasing stock prices and uncertainty. Because of this uncertainty, it became apparent that trading stocks required more knowledge and insight. As Didier Sornette (2004) pointed out in his book, the stock market is not a “casino” of playful or foolish gamblers. Although trading stocks is still available for everybody, people started to rely more on the recommendations of professionals. But are these professionals able to predict the market and generate an abnormal return?

The theory of market efficiency implies that investors should not be able to profit from trading stocks because there should be no information asymmetry. However, there are signs that information asymmetry does occur. To reduce this information asymmetry, the Securities and Exchange Commission (SEC) introduced the Regulation Fair Disclosure (Regulation FD) in October 2000. They believed that managers provided material information to select investors, who then traded profitably at the expense of less informed investors (Gintschel and Markov, 2004). Regulation FD states that when an issuer discloses material nonpublic information to certain individuals or entities, the issuer must make a public disclosure of that information (www.sec.gov). This legislation reduces opportunities for analysts to issue profitable recommendations, since they do not have superior information compared to other analysts due to full public disclosure.

Long before Regulation FD was introduced, Cowles (1933) provided evidence that analysts are not able to beat the market with their stock market forecasts. Barber et al (2001) find that purchasing stocks with the most favorable recommendations yield annual abnormal gross returns greater than four percent before transactions costs. However, when transaction costs are accounted for, the abnormal returns become insignificant. Furthermore, even though Womack (1996) finds abnormal returns associated with analyst recommendations, he also finds that these abnormal returns are generated by stock price reactions caused by analyst recommendations. Thus, the abnormal returns are not caused by the analysts’ prediction skills but are caused by the stock price reaction following the recommendation.

2

and, Womack (1996) find significant short- and long-term abnormal returns associated with analyst recommendations.

The principal aim of this paper is to investigate whether investors are able to profit when following analyst recommendations. Therefore, the following research question is developed:

Can investors profit from publicly available analyst recommendations?

The data used for this research is provided by IEX.nl, containing stock recommendations on the stock market over a five year period. The data set includes 29,444 recommendations made by individual analysts. Also, Datastream is used to retrieve past stock prices from stocks used in this research. The Capital Asset Pricing Model and the Fama and French 3-factor model are used to estimate the ability of analysts to issue profitable recommendations. This research will contribute to existing literature because most research focuses on the reason why a certain type of (i.e. buy, hold, or sell) recommendation was issued, while this research focuses on the profitability of the issued recommendations. Also, this paper investigates the effect on profitability when target prices are disclosed in recommendations.

3

2. Literature review

As previously mentioned, the efficient market hypothesis states that all information should be publicly available to make sure that arbitrage opportunities are minimized. So why are investors able to make a profit when trading stocks? The answer is relatively simple. Because there is just too much information for investors to process, which allows for discrepancies between different investors. This is where the analysts come in. Bjerring et al. (1983) state that financial analysts make the market more efficient by passing on information to their customers. Information which would otherwise not be available for the general public, is analyzed and sorted, and should help investors to choose efficiently when composing their portfolio. In reality however, most information is not publicly available since some analysts provide material information to a select group of investors only. Since they are helping certain investors to trade at the expense of less informed investors, they are not working to make the market more efficient (Gintschel and Markov, 2004).

The introduction of Regulation FD helped to diminish these information asymmetries by demanding public disclosure of material nonpublic information given to certain investors. In contrast, due to the pressures to issue positive recommendations, the sell recommendation is likely to reflect the true opinion of an analyst about a certain stock (Ertimur et al., 2007). As Firth et al. (2013) point out, analysts are more prone to be optimistic about certain stocks if they have a business relation with a mutual fund who hold these stocks in their portfolio. If an analyst issues an unfavorable recommendation that may harm the performance of the institutional client’s portfolio, the client could take its brokerage business elsewhere which would affect the analysts’ income. To curb these conflicts of interest that affect analysts’ research, other legislation, e.g. NASD 2711, is introduced. These regulations also established stringent disclosure requirements that are intended to make research output more meaningful (Kadan et al. , 2009).

However, Kadan et al. (2009) present another effect of the introduced regulations. They find evidence that the informativeness of recommendations declined in the period following the regulations. The post-regulation period showed a migration of investment banks to a three-tier rating system, which led to a more balanced distribution of recommendations. Even though the meaning of these recommendations better reflects the intentions of the analyst, the smaller number of tiers led to a reduction of informativeness of recommendations. Because of this regulation and reduction of informativeness, it would be reasonable to believe that it is unlikely for analyst recommendations to lead to abnormal portfolio returns.

4

predict the future movements of the stock market itself. He finds that the agencies failed to demonstrate that they exhibited skill and that any successful records result from pure chance.

As mentioned in the introduction, Womack (1996) does find abnormal returns associated with analyst recommendations. He researched 1,573 recommendation changes from 1989 to 1991. He also finds that these abnormal returns are generated by stock price reactions caused by analyst recommendations. Furthermore, Barber et al (2001) researched 360,000 recommendations from 1985 to 1996. They find that purchasing stocks with the most favorable recommendations, while rebalancing the portfolio daily and responding timely to recommendation changes, yield annual abnormal gross returns greater than four percent before transactions costs. However, these abnormal returns diminish when transaction costs are accounted for, or with a delay in reacting to recommendation changes (Barber et al., 2001).

Abnormal returns associated with analyst recommendations are also found by Groth et al. (1979). Their research contained a random sample of 2,500 customers of a brokerage house, over a seven year period, ending in December 1970. They conclude that abnormal returns are associated primarily with buy, rather than sell, recommendations. Bjerring et al. (1983) find similar results. They research all recommendations of a regional Canadian money management and investment service company from September 1977 to February 1981. They find abnormal returns for stocks with a buy recommendation, even after adjusting for commissions.

When taking the discussed literature in consideration, the following hypothesis is developed:

H0: Portfolios based on analyst recommendations generate a positive abnormal return.

5

over a period from 1996 to 1999. He finds that more favorable recommendations are accompanied with higher target prices relative to current stock prices, and recommendations without target prices would have been either bad news or would not have justified the recommendations. Also, Bradshaw et al. (2012) suggest that analysts who disclose a target price are more likely to be highly skilled than their colleagues who do not provide a target price, since their valuation model is more extensive.

Even though the literature on this topic is limited, the majority focuses on the use of target prices to justify recommendations and whether there is a link between the ratio of target price to stock price and the type of recommendation. As far as we know, no literature has mentioned the link between the use of target prices and portfolio return. It would be interesting to see whether the use of target prices effects the returns generated by recommendation based portfolios. Therefore, the following hypothesis is developed:

H0: Analysts who provide a target price generate a positive abnormal return.

3. Data

3.1 Sample selection

The analyst recommendations used in this study were provided by IEX.nl, a Dutch internet site which provides useful stock and firm information for investors. The data on IEX.nl is publicly available. IEX.nl is separated in several divisions to ensure the availability of information for different groups of investors. One of these divisions is Guruwatch, which displays all recommendations on a daily basis. The data set used for this research contains all these daily recommendations provided by Guruwatch. The recommendations are made by 226 analysts over a period from august 2010 through august 2014. This period is chosen because of data availability constraints since IEX.nl does not hold data older than five years. The database contains 29,446 recommendations for 176 companies. Table 1 provides insight in the number of analysts and recommendations each year. Each database record includes, among other items, the recommendation date, the target price, the current stock price and the given recommendation. The data is matched with monthly stock returns calculated from stock prices obtained from Datastream. The recommendations were converted to a three point rating scale where -1 represents a negative (sell) recommendation, 0 a neutral (hold) recommendation, and +1 a positive (buy) recommendation. Based on this scale, three portfolios were generated for each month: a positive, a neutral, and a negative portfolio, each with 49 monthly observations.

6 Descriptive statistics

Number of analysts and recommendations per year. Total number of analysts and recommendations and the average per year.

Year Number of analysts Number of recommendations

2010 161 2,833 2011 160 8,529 2012 149 7,830 2013 122 5,851 2014 106 4,401 Total 226 29,444 Average 139.6 5,889.2

The number of recommendations in 2010, as presented in Table 1, contains only recommendations for the last five months of 2010. There seems to be a decrease in the number of recommendations from 2011 to 2013, which is probably caused by the decrease in number of analysts.

3.2 Portfolios based on analyst recommendations

We construct portfolios based on the monthly rating given by analysts. We compare the return on each portfolio with the market return. Table 2 presents an overview of the number of analyst recommendations per category.

Table 2

Summary statistics data set

Data on the analyst recommendations from august 2010 through august 2014.

Table 2 shows that the buy recommendations outnumber the hold and sell recommendations. The positive average rating of recommendations indicates that there are, on average, more stocks in the portfolios based on buy and hold recommendations, than in the portfolios based on sell recommendations. According to Mokoaleli-Mokoteli (2009), the main reason for this unequal distribution of buy and sell recommendations is that analysts are too optimistic since their conflicts of interest increases the pressure to issue buy recommendations. Graph 1 shows the composition of the different recommendations per year.

Variable Number

Number of recommendations 29,444

Number of buy recommendations(1) 16,413

Number of hold recommendations (0) 10,287

Number of sell recommendations (-1) 2,744

7 Graph 1

Composition of portfolios per year for the research period August 2010 until August 2014.

The total number of recommendations per rating category are displayed on a yearly basis. The recommendations are divided based on their type.

As can be seen in Graph 1, the number of positive recommendations decreases over the years. A possible explanation for this could be a poor market performance which enables analysts to give a positive recommendation. Another explanation is the introduction of several regulations, which diminishes an analysts’ conflict of interest.

Graph 1 also shows the decrease in the total number of recommendations. Although 2014 only contains recommendations until August, the decrease is also visible for 2012 and 2013. According to David Tailleur, head stock investigator at the Rabobank, the main reason for this decrease is a lack of courage of analysts. He also states that the economic crisis, and especially the assumed inability to solve the crisis in the near future, has a negative influence on the number of stock recommendations.

3.3 Portfolios based on whether the analyst provided a target price

To investigate the difference between recommendations with and without target prices, the data is divided in two groups. This section will describe these two groups. The first of these two groups includes the recommendations with target prices, and the second group includes the recommendations without target prices. The data set contains 3,017 recommendations without a target price and 26,427 recommendations with a target price. Both groups are sufficient to draw inferences on. Graph 2 shows the distribution of the different groups over the entire data set. As can be seen, most analysts provide a target price with their recommendations.

8 Graph 2

Composition of portfolios per year for the research period August 2010 until August 2014.

Distribution of recommendations with- and without target price over the research period from August 2010 until August 2014.

The statistics in Graph 2 exhibit a decrease in the number of recommendations with target prices. When the number of recommendations with or without target prices are presented in percentages, it becomes clear that analysts disclose a target price with their recommendation more often. Table 3 presents these percentages for each year. This increase of recommendations with target prices could be caused by the analysts’ need to justify their given recommendation.

Table 3

Table 3 shows the percentage of recommendations with and without target prices for each year researched.

% with TP % without TP 2010 82,5% 17,5% 2011 89,7% 10,3% 2012 90,0% 10,0% 2013 92,3% 7,7% 2014 90,7% 9,3%

3.4 Portfolios based on target rate

The previous cross-section of the data was based on the availability of a target price. In this section we see what happens when the data containing target prices is divided based on the given target price. Some analysts issue really positive recommendations for a particular stock, which leads to a more aggressive target price, while others are still positive, but not as positive as the aggressive analysts, and issue a recommendation containing a more moderate target price. The following formula is used to calculate the level of determination of a certain recommendation:

𝑇𝑎𝑟𝑔𝑒𝑡 𝑟𝑎𝑡𝑒 =𝑇𝑃 𝑃0 , (1) 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 2010 2011 2012 2013 2014

9

where TP represents the given target price by the analyst and P0 represents the current stock price.

The target rate is used to divide the data in four monthly portfolios. The first portfolio is an aggressive portfolio, containing all recommendations with a target rate higher than 1.3. The second portfolio is a moderate portfolio which contains all recommendations with a target rate between 1.1 and 1.3. The neutral portfolio contains all recommendations with a target rate between 0.9 and 1.1. The last portfolio, the negative portfolio, contains all recommendations with a target rate lower than 0.9. The negative portfolio could not be separated between really negative and moderate negative because the portfolios would be too small to make a proper comparison between the different portfolios. For the purpose of reliability, some outliers have been removed. Recommendations with a target rate higher than 100 are eliminated and thus are not included in the generated portfolios. Graph 3 shows the distribution of the four portfolios.

Graph 3

Distribution of portfolios based on target rate.

The graph shows the composition of the four different portfolios from August 2010 to August 2014.

As Graph 3 shows, the analysts are less positive in 2013 and 2014 compared to 2011 and 2012. The aggressive portfolio, where the recommendations are really positive, decreases rapidly from 2011 until 2014. Interesting to see is that the neutral portfolio slightly increases from 2011 to 2012. Also the 2013 neutral portfolio contains more recommendations than the neutral portfolio of 2011. This indicates that analysts tend to be less positive than the previous years, which could be due to the still apparent credit crisis.

10

4. Methodology

In order to compare the performance of the portfolios based on different recommendations, a times-series regression is executed. The research is therefore divided into two main models. The first model is the Capital Asset Pricing Model (CAPM). The second model is the Fama and French 3-factor model. Both models evaluate the performance of the portfolios with a monthly time-series regression. But first, the calculation of the portfolio returns is discussed.

4.1 Research design

We calculate the average portfolio return for each portfolio containing one particular class of recommendations. The return for each recommendation class is calculated on a continuous basis:

𝑅_𝑖,𝑡 = 𝐿𝑁( 𝑃𝑡

𝑃𝑡−1) , (2)

where

R

represents the stock price in month t. After calculating the stock returns, the returns are matched with the corresponding recommendation. Finally, we calculate the portfolio returns by using the following formula:

𝑅_𝑝,𝑡=∑(𝑅𝑖,𝑡)

𝑁

,

where Rp,t represents the return of portfolio p in month t, and N is the number of recommendations

in the given portfolio.

We first estimate the risk-adjusted performance using the CAPM. Jensen’s alpha is estimated from this model by regressing the monthly excess return earned by analysts on the market excess return (Barber and Odean, 2002). Jensen’s alpha represents the difference between the return on the actual and the benchmark portfolio, and thus provides information about the analysts’ forecasting skills. The CAPM monthly time-series regression is as follows:

11

where Rf,t represents the month t return on treasury bills having one month until maturity, αp

represents the estimated CAPM intercept (Jensen’s alpha), γ1

represents the estimated market beta,

Rm,t is the return on the global market portfolio at month t, and

ϵ

term. The values for Rf,t and Rm,t are retrieved from an online database.1 This database provides the

factors used for the Fama and French regression for most industries. Because the data used contains various companies indexed on different international stock exchanges, the values used are global factors. The global factors are annually rebalanced and include value-weighted returns from 23 countries in four regions: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Hong Kong, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Singapore, Spain, Switzerland, Sweden, United Kingdom, United States.

The second test used to evaluate portfolio performance is the three-factor model developed by Fama and French (1993). Similar to the CAPM, the Fama and French 3-factor model also generates a risk-adjusted return similar to Jensen’s alpha. However, the Fama and French model also takes firm size and book-to-market values into consideration. The monthly time-series regression is estimated as follows:

𝑅_𝑝,𝑡 – 𝑅_𝑓,𝑡 = 𝛼_𝑝+ γ₁(𝑅_𝑚,𝑡 – 𝑅_𝑓,𝑡) + γ₂𝑆𝑀𝐵_𝑡+ γ₃𝐻𝑀𝐿_𝑡+ 𝜖_𝑝𝑡 , (5)

where SMBt represents the difference between month t returns of a value-weighted portfolio of small

stocks and one of large stocks, and HMLt represents the difference between month t returns of a

value-weighted portfolio of high book-to-market stocks and one of low book-to-market stocks. The values for SMBt and HMLt are also global factors retrieved from the online data base.2

The four coefficient estimated from equations (4) and (5), αp, γ1, γ2, γ3,provide useful insights

into the nature of the firms held in each portfolio. With a value of γ1 greater (less) than one, the portfolio is, on average, riskier (less risky) than the market. While a positive value of γ2, indicates a portfolio which contains more smaller firms, a negative value indicates a portfolio constructed out of larger firms. A value of γ3 greater (less) than zero signifies a preference for high (low) book-to-market stocks in a portfolio. Finally, as mentioned before, αp provides information about the skills of the

average analyst. A positive alpha indicates that the average analyst indeed has superior skills, while a negative alpha means that the average analyst does not have superior skills.

12

5. Results

This section presents the results generated by the models discussed previously. The average monthly portfolio returns for each separate data set are presented in the Appendix by tables A1-A4.

5.1 Average portfolio returns

Table 3 presents the average portfolio returns of analyst recommendations, including the data without- and with target prices. Table 4 presents the average portfolio returns for four portfolios based on target rate.

Table 4

Average return of analyst recommendations, including portfolios with and without target price.

For the data set containing all analyst recommendations, the difference between the positive and negative portfolio is shown. To test for robustness, the data is split in two periods. The first period contains all average returns from 2010 to 2012. The second period contains all average returns from 2012 to 2014. We find the means to be insignificantly different across periods. A t-test is performed over the entire data set. The t-statistic with corresponding p-value is presented. A paired t-test has been performed to compare means between the positive- and negative portfolio. The means are proven to be significantly different.

All data With TP Without TP Difference

with-without TP Positive 0.0026 0.0029 0.0013 -0.0014 Neutral -0.0107 -0.0113 -0.0125 -0.0111 Negative -0.0247 -0.0246 -0.0198 -0.0078 Pos-Neg -0.0221 2010-2012 -0.0140 -0.0139 -0.0130 -0.0009 2012-2014 -0.0049 -0.0050 -0.0055 -0.0005 T-statistic -1.482 -1.921 -1.710 -1.710 P-value 0.145 0.061 0.094 0.094

13 Table 5

Average returns of analyst recommendations based on target rate.

The difference between the aggressive- and negative portfolio, and the difference between the moderate- and negative portfolio is shown. To test for robustness, the data set is split up in two periods, from 2010 to 2012, and from 2012 to 2014. We find the means to be insignificantly different. A t-test is performed over the entire data set. The t-statistic with corresponding p-value is presented. A paired t-test has been performed to compare means between the aggressive- and negative portfolio and between the moderate- and negative portfolio. We found the means to be insignificantly different.

The average returns for the four generated portfolios based on target rates are presented by Table 5. As mentioned before, the data set containing the recommendations with target prices are divided into four subclasses, each represented by a particular portfolio. The aggressive portfolio generates a negative average return, which is in contrast with the underlying predictions. Also, the return for the aggressive portfolio is more negative than the return for the negative portfolio. A remarkable result is the return generated by the moderate portfolio, which is zero. An explanation for this outcome is that the recommendations shifted. Because there are four categories, analysts tend to treat moderate recommendations as neutral recommendations.

5.2 Results CAPM and Fama and French 3-factor model

Table 6-8 present the output for the CAPM and the Fama and French 3-factor model. Table 6 presents the output of the portfolios including all analyst recommendations. The difference between portfolios based on recommendations with a target price and the portfolios without a target price are presented in Table 7. Lastly, Table 8 shows the output of both models for the portfolios based on target rate. 3

3 _{The data used is tested for standard CLRM diagnostics. The results show that the assumptions are satisfied.}

14 Table 6

Output CAPM and Fama and French 3-factor model of all analyst recommendations.

This table presents values for the coefficients alpha (α), the beta for market risk (γ1), the beta for market capitalization (γ2), the beta for firm size (γ3), and R². Their corresponding p-value are shown below between brackets. ***, **, and * indicate statistical significance at respectively the 1%, 5%, and the 10% levels. To test for robustness, the data set is split up in two periods, from 2010 to 2012, and from 2012 to 2014.

CAPM FamaFrench α γ1 R² α γ1 γ2 γ3 R² Positive -0.0115 0.9274 0.5510 -0.0118 0.9066 -0.0031 0.0040 0.5720 (0.0296)** (0.0000)*** (0.0256)** (0.0000)*** (0.4482) (0.2793) Neutral -0.0249 0.9417 0.6209 -0.0253 0.9221 -0.0026 0.0037 0.6393 (0.0000)*** (0.0000)*** (0.0000)*** (0.0000)*** (0.4652) (0.2555) Negative -0.0411 1.1289 0.5132 -0.0412 1.0871 -0.0004 0.0065 0.5334 (0.0000)*** (0.0000)*** (0.0000)*** (0.0000)*** (0.9443) (0.1840) Positive - Negative 0.0297 -0.2015 0.0538 0.0294 -0.1805 -0.0027 -0.0025 0.0688 (0.0000)*** (0.1089) (0.0000)*** (0.1652) (0.5155) (0.5098) 2010-2012 -0.0283 1.0086 0.7327 -0.0281 0.9969 -0.0043 0.0054 0.7637 (0.0000)*** (0.0000)*** (0.0000)*** (0.0000)*** (0.3657) (0.1515) 2012-2014 -0.0251 1.2118 0.5598 -0.0253 1.2109 -0.0021 -0.0001 0.5625 (0.0008)*** (0.0000)*** (0.0013)*** (0.0000)*** (0.6809) (0.9820)

15 Table 7

Output CAPM and Fama and French 3-factor model of the difference between analyst recommendations with target prices and without target prices.

This table presents values for the coefficients alpha (α), the beta for market risk (γ1), the beta for market capitalization (γ2), the beta for firm size (γ3), and R². Their corresponding p-value are shown below between brackets. ***, **, and * indicate statistical significance at respectively the 1%, 5%, and the 10% levels. To test for robustness, the data set is split up in two periods, from 2010 to 2012, and from 2012 to 2014.

CAPM FamaFrench α γ1 R² α γ1 γ2 γ3 R² Positive -0.0029 0.1224 0.0295 -0.0022 0.1193 0.0056 -0.0010 0.0960 (0.5048) (0.2380) (0.6040) (0.2544) (0.0993)* (0.7352) Neutral -0.0096 -0.1217 0.0084 -0.0090 -0.1242 0.0057 -0.0011 0.0283 (0.2389) (0.5316) (0.2805) (0.5373) (0.3827) (0.8467) Negative -0.0117 0.3279 0.0669 -0.0125 0.4122 -0.0065 -0.0111 0.1623 (0.1229) (0.0727)* (0.0917)* (0.0244)** (0.2624) (0.0373)** 2010-2012 -0.0124 0.1240 0.0728 -0.0130 0.1451 -0.0007 -0.0033 0.1262 (0.0054)*** (0.1571) (0.0047)*** (0.1139) (0.8424) (0.2295) 2012-2014 -0.0069 0.0540 0.0030 -0.0048 0.1053 0.0021 -0.0068 0.0993 (0.2717) (0.7662) (0.4450) (0.5628) (0.6407) (0.1354)

16 Table 8

Output CAPM and FamaFrench 3- factor model of analyst recommendations based on target rate.

This table presents the output of the CAPM and the FamaFrench 3-factor model. This table presents values for the coefficients alpha (α), the beta for market risk (γ1), the beta for market capitalization (γ2), the beta for firm size (γ3), and R². Their corresponding p-value are shown below between brackets. ***, **, and * indicate statistical significance at respectively the 1%, 5%, and the 10% levels. To test for robustness, the data set is split up in two periods, from 2010 to 2012, and from 2012 to 2014.

CAPM FamaFrench α γ1 R² α γ1 γ2 γ3 R² Aggressive -0.0354 1.0626 0.4350 -0.0347 1.0205 0.0057 0.0049 0.4532 (0.0000)*** (0.0000)*** (0.0085)*** (0.0000)*** (0.3370) (0.3616) Moderate -0.0130 0.8418 0.5369 -0.0133 0.8270 -0.0024 0.0029 0.5507 (0.0090)*** (0.0000)*** (0.0085)*** (0.0000)*** (0.5296) (0.4033) Neutral -0.0147 0.8631 0.5212 -0.0153 0.8464 -0.0053 0.0040 0.5588 (0.0056)*** (0.0000)*** (0.0036)*** (0.0000)*** (0.1838) (0.2687) Negative -0.0225 0.8583 0.2622 -0.0235 0.8208 -0.0087 0.0080 0.3249 (0.0138)** (0.0002)*** (0.0092)*** (0.0003)*** (0.2082) (0.1976) 2010-2012 -0.0207 0.9273 0.6759 -0.0205 0.9148 -0.0060 0.0068 0.7336 (0.0017)*** (0.0000)*** (0.0013)*** (0.0000)*** (0.2161) (0.0773)* 2012-2014 -0.0204 1.0543 0.4199 -0.0206 1.0549 -0.0021 -0.0003 0.4225 (0.0130)** (0.0001)*** (0.0175)** (0.0001)*** (0.7250) (0.9560)

17

6. Conclusion

The main aim of this research is to investigate whether investors are able to profit from publicly available recommendations. The literature discussed in the literature review mostly focuses on the change in recommendations or why investors issue certain recommendations. This paper contributes to the current literature since we investigate the abnormal returns generated by portfolios based on analyst recommendations and target prices. This is not investigated simultaneously before.

When observing the portfolios based on analyst recommendation types, it becomes obvious that the negative portfolios generate a significant and negative risk-adjusted return for both the CAPM and Fama and French 3-factor model. This indicates that analysts who issue a sell recommendation generally have good judgment skills, and that investors who base their portfolio selection on analyst recommendations generate a higher return when they follow negative recommendations. These findings are in line with Ertimur et al. (2007), who find that sell recommendations are likely to reflect the true opinions of an analyst because they are hesitant to issue negative recommendations. Unfortunately, this cannot be stated for the positive recommendations. The significant risk-adjusted returns found for both the CAPM and the Fama and French 3-factor model, indicate that the analysts who issue these positive recommendations do not have superior skills. A possible explanation comes from Ertimur et al. (2007), who find that analysts have an incentive to issue positive recommendations because the managers of the firms they are covering prefer favorable recommendations.

Since many analysts provide a target price with their recommendation, the recommendations with target prices are compared with recommendations without target prices to see which of both generate a higher return. The results show insignificant results. Underlying expectations that portfolios based on recommendations with target prices would perform better than the recommendations without target prices cannot be justified.

Also, the data set containing target prices has been divided based on target rate, which results in four different portfolios. The aggressive portfolio shows a negative risk-adjusted return for both the CAPM and the Fama and French 3-factor model. Remarkably, the aggressive portfolio, performs worse than the negative portfolio, which is in contrast with the underlying predictions. The negative portfolio shows a significant negative risk-adjusted return, which contributes to the earlier findings that an investor should listen to negative recommendations.

18

19

7. References

Bandyopadhyay, S., Brown, L., Richardson, G., 1995. Analysts’ use of earnings forecasts in predicting stock returns: Forecast horizon effects. International Journal of Forecasting 11, 429–445. Barber, B., Lehavy R., McNichols M., Trueman B., 2001. Can investors profit from the prophets?

Security analyst recommendations and stock returns. Journal of Finance 56 (2), 531-563. Barber, B., Lehavy R., Trueman B., 2007. Comparing the Stock Recommendation Performance of

Investment Banks and Independent Research Firms. Journal of Financial Economics. Forthcoming.

Barber, B., Odean, T., 2002. Online investors: do the slow die first?. The review of financial studies 15 (2), 455-487.

Barniv, R., Hope, O., Myring, M., Thomas, W., 2010. International evidence on analyst stock recommendations, valuations and returns. Contemporary accounting research 27 (4).

Bjerring, J., Lakonishok, J., and Vermaelen, T., 1983. Stock prices and financial analysts' recommendations. Journal of Finance 38, 187-204.

Bradshaw, M. T., 2002. The use of target prices to justify sell-side analysts’ stock recommendations. Accounting Horizons 16, 27–41.

Bradshaw, M, Brown, L, & Huang, K., 2013. Do Sell-Side Analysts Exhibit Differential Target Price Forecasting Ability? Review Of Accounting Studies 18 (4), 930-955.

Carhart, M., 1997. On Persistence in Mutual Fund Performance. Journal of Finance 52 (1), 57–82. Cowles, Alfred, 1933. Can stock market forecasters forecast? Econometrica 1 (3), 309-324.

Didier Sornette, 2004. Why Stock Markets Crash: Critical Events in Complex Financial Systems. Princeton University Press, New Jersey.

Ertimur, Y., Sunder, J., Sunder, S., 2007. Measure for Measure: The Relation between Forecast Accuracy and Recommendation Profitability of Analysts. Journal Of Accounting Research 45 (3), 567-606. Fama, E., French K., 1993. Common Risk Factors in the Returns on Stocks and Bonds, Journal of

Financial Economics 33 (1), 3–56.

Firth, M, Lin, C., Liu, P., Xuan, Y., 2013. The Client Is King: Do Mutual Fund Relationships Bias Analyst Recommendations?, Journal of Accounting Research 51 (1), 165-200.

Gintschel A, Markov S., 2004. The Effectiveness of Regulation FD. Journal Of Accounting And Economics 37(3), 293-314.

Groth, J., Lewellen, W., Schlarbaum, G., Lease, R., 1979. An analysis of brokerage house securities recommendations. Financial Analysts Journal 35 (1), 32–40.

20

Michaely, R., Womack, K, 1999. Conflict of Interest and the Credibility of Underwriter Analyst Recommendations. Review of Financial Studies 12, 653–86.

Mokoaleli-Mokoteli, T., Taffler, R., Agarwal, V., 2009. Behavioural Bias and Conflicts of Interest in Analyst Stock Recommendations. Journal Of Business Finance & Accounting, 36 (3/4), 384-418. Stickel, S., 1995. The anatomy of the performance of buy and sell recommendations. Financial Analysts

Journal 51 (5), 25–39.

21

8. Appendix

Table A.1

Average return of portfolios based on analyst recommendations

Monthly average returns generated for three portfolios; positive, neutral and negative. The yearly average returns (AR) are displayed in the last column.

22 Table A.2

Average portfolio returns based on recommendations without target prices.

Monthly average portfolio returns. The yearly average returns (AR) are displayed in the last column.

Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec AR Positive - - - -0.0296 0.0471 0.0548 -0.0238 0.0377 0.0172 2010 Neutral - - - -0.0505 0.0390 0.0451 -0.0321 0.0322 0.0068 Negative - - - -0.0727 0.0160 -0.0442 -0.0358 0.0287 -0.0216 Positive 0.0400 -0.0066 0.0146 0.0028 -0.0466 -0.0149 -0.0386 -0.0875 -0.0638 0.0730 0.0014 0.0445 -0.0068 2011 Neutral 0.0093 -0.0291 0.0250 0.0049 -0.1138 -0.0872 -0.0573 -0.1106 -0.0564 0.0578 -0.0394 0.0153 -0.0318 Negative 0.0461 -0.0346 0.0203 -0.0443 -0.0657 -0.0750 -0.2055 -0.0841 -0.0491 0.0403 -0.1384 0.0463 -0.0453 Positive 0.0543 0.0331 0.0164 -0.0304 -0.1314 0.0814 0.0435 0.0197 0.0284 0.0112 -0.0154 0.0302 0.0118 2012 Neutral 0.0605 0.0440 0.0152 -0.0344 -0.1407 0.0671 0.0334 0.0065 0.0064 -0.0419 -0.0254 -0.0053 -0.0012 Negative 0.0355 -0.0099 0.0002 -0.0306 -0.1113 -0.0138 -0.0088 -0.0634 0.0239 -0.0163 -0.0870 0.0201 -0.0218 Positive 0.0443 -0.0135 0.0043 -0.0168 0.0732 -0.1065 -0.0290 -0.0298 -0.0033 0.0381 0.0330 0.0089 0.0002 2013 Neutral 0.0267 -0.1299 0.0007 -0.0045 0.0215 -0.1892 -0.0669 -0.0318 -0.0209 0.0211 0.0089 0.0312 -0.0278 Negative -0.0349 -0.2568 -0.0195 0.0093 0.1084 -0.0313 0.0921 -0.0142 0.1852 -0.0029 0.0030 0.0248 0.0053 Positive -0.0410 0.0344 0.0197 0.0243 0.0170 -0.1108 -0.0317 0.0022 - - - - -0.0107 2014 Neutral -0.0404 0.0151 0.0237 0.0046 0.0311 0.0186 -0.0294 0.0585 - - - - 0.0102 Negative -0.0331 0.0145 0.0430 -0.0122 0.0257 -0.0528 -0.0676 -0.0388 - - - - -0.0152 Table A.3

Average portfolio returns based on recommendations including a target price.

Monthly average portfolio returns. The yearly average returns (AR) are displayed in the last column.

23 Table A.4

Average portfolio returns based on target rate.

Monthly average portfolio returns and yearly average portfolio returns for four portfolios based on target rate. The yearly average returns (AR) are displayed in the last column.