Impact of analysts’ recommendations on stock prices:
comparison between a non-crisis period and a crisis
period.
-This research examines the abnormal returns with respect to buy and sell recommendations and compares them in the period before and during the financial crisis of 2007-2008 (respectively the non-crisis period and the crisis period). In both periods recommendations
generate an initial return in the direction predicted by the analyst. In the non-crisis period, where the market is assumingly stable, recommendations support the price-pressure hypothesis whereas sell recommendations support the information hypothesis. In the crisis
period these sell and buy recommendations generate the opposite effect, so buy
recommendations support the information hypothesis where sell recommendations support the price-pressure hypothesis. The abnormal returns suggest the content of the sell analysts’ recommendations become less economically valuable during the financial crisis whereas buy
recommendations become more valuable -
Sabine Huijbers
Student number 10283722 University of Amsterdam,
Faculty of Economics and Business Supervisor: Maximilian Hoyer 14th of July 2014
Table of Content
1. Introduction 1
2. Literature review 4
2.1 Initial abnormal return 4
2.2 The price pressure hypothesis or the information hypothesis 5 2.3 Different stock-price impact of buy and sell recommendations 6
3. Data and Methodology 8
4. 4.1 Results 12
4.2 Robustness test 17
5. Conclusion 18
1. Introduction
As long as stocks exist stock analysts have been there to help investors make investment decisions. Because of the increasing use of online resources and the all-day stock market news, stock analysts get more and more exposure. The stock analysts’ job is to advise
investors and make recommendations on stocks using financial data on stocks and companies. Many traders rely on their expertise and investors are prepared to pay for these advisory services. The investors and analysts claim the information is valuable, which is in contrast with the strong form of the noted efficient market hypothesis1. This hypothesis predicts that second hand information can not influence the stock price because all information (private and public) that analysts examining are already reflected in the price. It suggests stock recommendations have no value and no abnormal returns 2 can be earned by studying or analyzing it (Wijmenga, 1990, p. 1). However, the majority of prior research report
significant initial abnormal stock price responses in the direction of the forecast after analysts revise their recommendation. Although this is inconsistent with the strong EMH, the findings can be in line with the traditionally accepted semi-strong version of the EMH (Albert & Smaby,1996, p 61). The semi-strong version of the EMH holds that while prices reflect all publicly available information they react instantaneously to new unknown information, suggesting analyst recommendations contain new unknown information. Whether this stock-price effect is temporary or permanent was and is still an interesting issue in finance research. To understand whether the stock-price impact of recommendations is temporary or permanent two hypothesis have to be examined: the information hypothesis and the price pressure
hypothesis. The information hypothesis poses that analyst recommendations are treated as
new relevant information, which was not yet reflected in the price. For example an analyst publishes a buy recommendation on a stock based on new information. Investors will believe the stock is undervalued and continue to purchase the stock. This results in an abnormal stock price increase and ultimately the information contained in the recommendation is reflected in the price (Davies and Canes, 1978, p. 44). The price pressure hypothesis holds that analyst recommendation will be self-fulfilling prophecies meaning investors believe analysts have new information when they don’t. Investors who believe an analyst has new positive information about a stock will buy this stock. Similar with the information hypothesis this buying pressure will causes abnormal returns. However when the recommendation is based on
1 From here on I refer to efficient market hypothesis as EMH 2
Abnormal return or excess return is actual return adjusted for market risk
no real economically valuable information these effect are only temporary. The direct excess return due to overreaction of investors will eventually fall again, because the stock price on the long term is eventually determined by the firm’s cash flows. Several prior studies have come up with evidence in favor of the information hypothesis, where other support the price pressure hypothesis.
Another interesting issue that did not reached a consensus yet is the possible different stock price impact of buy and sell recommendations. Davies and Canes (1978) argue that recommendations to buy may have a smaller initial impact on stock prices than buy
recommendations because of tax liability reasons and the possible reluctance to short selling. In addition, Liden (2004) suggests buy and sell recommendations have different longer-term stock price effects due to structurally difference between the two recommendations (p. 4). In short, he initiate that analyst who recommend to sell the stock have done more research compared to recommendation to buy and therefore sell recommendations create more economic value for investors.
The topics mentioned above raise the following questions: Do recommendations have initial impact on stock prices? If yes, is this stock-price impact temporary or permanent? Do buy and sell recommendations have different impact on stock prices? These three questions have been already investigated more than once. However with this paper I will add new insight in the existing literature and emphasize the stock price effect of revised buy and sell recommendations in a crisis period and compare the results with the results of a non-crisis period. In this paper the distinction is made between the crisis period as the period during the financial crisis of 2007-2008 and the non-crisis period as the assumingly stable period before the financial crisis.
During the financial crisis some investors were exposed to huge losses and markets became more volatile, meaning stock prices can drastically change in a short period of time. Stock market volatility makes investing more risky and therefore can make investors more nervous. When market conditions are unstable and stock patterns are unpredictable, investors can put more weight on extra economically valuable information like analyst
recommendations. Contrary, the need for external information might be lower in stable markets because investors are more confident and returns are more easily to predict. This suggests that the market will respond more intensively in the crisis period and therefore the initial abnormal returns in the crisis period are of a greater magnitude compared to the abnormal return in the non-crisis period. Depending on the amount of new relevant
information included in the recommendation abnormal returns might be permanent or the price pressure effect will occur.
The content of this thesis is set up as follows. In section 2 I will present the existing literature and their empirical findings about the stock price effect of recommendations. Section 3 describes the data and the methodology used for this research. Section 4 presents the results and a robustness test. And finally, section 5 contains the conclusion and provides suggestion for further research.
2. Literature review
Prior research examine the immediate effect on stock prices as well as the influence in months before and after the recommendation. Although the majority found a positive relationship between abnormal return and the direction forecasted by the analyst, no
unanimousanswer is given with respect to the magnitude of the excess return, the quickness of the price-adjustment, information leaking, the longer-term pattern, the different impact between sell and buy recommendations etc. The different results might probably be due to the fact studies examine different recommendations in different countries and different periods and use different models or time-events. Because it is difficult or maybe impossible to measure the stock price impact of each condition separately this is only a suggestion. This section provides an overview of the existing literature of the initial abnormal return and the post-publication pattern of this return divided in studies supporting the price pressure hypothesis or the information hypothesis. Lastly this section present literature about the different stock-price impact of buy and sell recommendations.
2.1 Initial abnormal return
Davies and Canes (1978) are one of the first who present evidence that stock prices do adjust to analysts’ recommendations. They examine the effect of the column “Heard on the Street” in the Wall Street Journal where Wall Street analysts published their recently revised sell or buy recommendations. The result of their study indicates an event-day return (t=0) of 0,923% in case of a buy recommendation and -2,374% in case of a sell. After day 2 abnormal returns are not significant anymore, so the initial price adjustment seems to persist till 1 day after the day of recommendation. Lui et al.(1990) and Beneish (1991) examine the impact of the same column. They both found significant abnormal returns on the publication date and small significant abnormal returns two days before the publication date. According to Beneish (1991) the stock price reaction one and two days before publication is most likely due to information leaking (p. 396).
Stickel (1995) uses data from brokerage house recommendations supplied by Zacks Investment Research3, where they divide recommendations in a 5-point scale: 1=strong buy 2=buy 3=hold 4=sell and 5=strong sell. In his study buy recommendations are defined as all
3 Research company obtains the recommendations and forecasts from written reports brokerage houses issues to
the customers and potential customers.
upward revisions to rank 1 and 2 and sell recommendations are defined as all downward revisions to rank 4 and 5. Consistent with Davies and Canes (1978) the results of this paper indicate a positive abnormal return associated with a buy recommendation and a negative return associated with a sell recommendation. Assessing the short window he looks at the mean abnormal return period between t=-5 and t=5, where t=0 is defined as the day of the recommendation, because there was uncertainty about the day of recommendation and early dissemination was possible (p. 30). Also Elton Gruber and Grossman’s (1986) paper uses this 5-point scale to divide recommendations collected by an investment group. They found a large excess return in the month of the positive recommendation announcement and a negative, although insignificant, return for sell recommendations.
Compared to other events previously studied Womack (1996) presents initial abnormal returns that are quite large. He investigates the effect on stock prices by examining
recommendations (added-to-buy, removed-from-buy, added-to-sell and removed-from-sell) from the fourteen major U.S. brokerage firms. The study shows that new buy (sell)
recommendations were associated with a positive +3% (negative -4,7%) three-day
recommendation period return. Consistently Asquith et al. (2005) also observed abnormal returns looking at a three-day event return. Unlike the majority of prior research they oriented not only in the U.S. market but examine 1126 complete analyst reports covering companies, industries and commodities worldwide (p. 263).
Additionally, one of the more recent studies about this topic is from Ryan and Taffler (2006). They observe an average abnormal return for new buy recommendation of 2,06% and an average abnormal return for new sell recommendations of -3,3% in the month of the recommendation change analyzing the revised recommendations of the six leading London-based brokerage houses.
2.2 The price pressure hypothesis or the information hypothesis
Next to examining the short-term abnormal return, Womack (1996) also looks at the long-term return after a recommendation. His conclusion was that the abnormal return immediate after a recommendation was permanent and not quickly mean-reverting, suggesting the analysts’ information had to be economically valuable (p. 139). These empirical findings are consistent with the information hypothesis. He found significant abnormal returns for buy recommendations after 1 month (+2,4%) and for added-to-sell after 6 months (-9,1%). Also Moshirian et al (2009) present a price drift in the direction of
the forecast by the analyst considering a sample of recommendations out of emerging markets. The study of Moshirian et al. (2009) shows an excess return for a 2-day event of 3,46% (-3,78%), for a one-month event of 4,03% (-10,74 %) and for a six-month event of 10,80% (-7,16%) associated with a strong buy and buy (strong sell and sell) recommendation.
Partly consistent with the price pressure and partly consistent with the information hypothesis are the results of Kerl and Walter’s study (2007). Their paper analyzes buy recommendations of stocks in Germany published by magazines between 1995 and 2003. Evidence show an initial abnormal return of 2,58% and a reversal effect of -1,04% looking at the period after the initial return trough day 20. As a consequent we observe an abnormal return of 1,54% due to the permanent information effect.
Also Barber and Loeffler (1993) found a partial price reversal over a 25-day period after an initial abnormal return associated with a positive recommendation. They examine the ‘Dartboard’ column in the Wall Street Journal, where investment analysts recommend on randomly selected stocks. Barber and Loeffler (1993) find on average a significant abnormal return of 4,06% looking at a two day period and from day 2 through day 25 they observe an negative abnormal return of 2,08%. Despite this partial price effect they do not reject the information hypothesis, because the reversal is not complete and after 25 days there still remains a positive average abnormal return. In addition to this study Liang (1999) extended the observed period from 25 days to 125 days by looking at the same column. After
measuring a significant two-day abnormal return of 3,52%, results show that there is a strong mean reversion during the first 15 trading days and the reversal effect is complete 53 days after the publication date (p. 124). Furthermore, when following the analyst’
recommendations investors earn an abnormal loss of 3,8%. Therefore the information hypothesis is fully rejected by Liang and it seems the initial abnormal return is due to overreaction of investors.
2.3 Different stock-price impact of buy and sell recommendations
The study of Davies and Canes (1978) shows that the abnormal return pattern is different for buy and sell recommendations. The days before publication the abnormal price increase associated with a buy was relatively large, whereas with a sell the abnormal price decrease occurred mostly on the publication day itself and not before. Clients of the analyst, who received the analyst report one to two weeks earlier, could possibly be reluctant to (short) selling a stock and are more likely to act on a buy recommendation ( p. 52). The reluctance to
short selling might arise because in contrast to holding a stock it can result in limited losses and in missing out potential dividend- and interest income. Additionally the sale of a stock can result in increased tax liability for the investor. As a result sell recommendations to sell may have a smaller impact on stock prices than recommendations to buy (p. 45). Consistent with the reasoning of Davies and Canes, Stickel (1984) found a greater magnitude stock price reaction with respect to upgrade recommendation compared to a downward when
investigating the Value Line Investment Survey (VLIS). The VLIS was published each week and included financial information on common stocks and ranked these stocks into 5 ranks, where a change in rank may be interpreted as an upgrade or downgrade recommendation.
Prior research on this issue does not always reflect the above-mentioned impact. Liu et al (1990) show that buy and sell recommendations have symmetric impact on the stock price looking at a period 10 days after publication. Nevertheless, completely in contrast to the possible reluctance of (short) selling, Ryan and Taffler (2006) show a greater magnitude of the abnormal return associated with sell than buy recommendations in the month of the recommendation change. Also Womack (1996) argues that the market react more strongly to added-to-sell recommendations compared to added-to-buy recommendations in the three-day event period, as well as in the post-recommendation period of one- and six month periods. Similarly to Womack’s post-recommendation periods Liden (2004) also examines the
different longer-term impact of sell and buy recommendation. Although most of the literature has examined U.S. stocks, he analyzes stock price reactions to stock recommendations
published in six large and well-known Swedish newspaper and business magazines in the period 1955-2000. His paper examines the abnormal returns of 20 and 125 trading days after the publication date. After findings of a positive abnormal return (0,79%) on publication date Liden (2004) found evidence of a reversed stock price effect 20 days after publication with respect to buy recommendations. In contrast, sell recommendations result in a permanent abnormal return after a negative abnormal return on publication date of -,150%. Thus
evidence of buy recommendations support the price-pressure hypothesis where the results of sell recommendations support the information hypothesis. In addition the post-publication drift of the stock prices looking at period of 125 days after publication have been analyzed in order to further support or reject one of the hypothesis (p. 21). Investing in a positive
recommended stock for 125 days would result on average of a abnormal loss of 3,8%, whereas following a sell recommendation will lead on average to a 11% abnormal return.
3. Data & Methodology
I. Data
All data, including recommendations, historical stock prices and beta’s, required for this study is retrieved from the website www.finance.yahoo.com. In the following test I analyse 96 changes in recommendations for S&P 500 companies in the non-crisis and the crisis period. The sample of recommendations was chosen randomly out of all revised recommendations of S&P 500 firms given by research firms. An overview of these recommendations is presented in table 1.
The revised recommendations in both periods had to meet the following two requirements. Firstly, the change in recommendation must not be followed by a recommendation by the same research firm up to 6 months after the initial revised recommendation is made.
Secondly, I examine only added-to-buy and added-to-sell recommendations. Research firms often divide recommendations into the simple stock rating system of “Buy”, “Hold” and “Sell” or they use a five-point scale such as 1= Strong buy, 2=Buy, 3= Hold, 4=Underperform and 5=Sell. In this last five-point scale added-to-buy recommendations are defined as changes from 3,4 or 5 to a 1 or 2 whereas added-to-sell recommendations are defined as changes from 1,2,3 to a 4 or 5. Womack (1996) and Moshirian (2009) suggest that these types of
recommendations are among the most valuable recommendations for investors, because they contain the most prominent new items. For simplicity in this thesis I talk about added-to-buy (added-to-sell) recommendations when I refer to buy (sell) recommendations.
Table 1: Overview of the recommendations given by research firms
Research firm #buy non-crisis #buy crisis #sell non-crisis #sell crisis # total
AG Edwards 2 2
AmTech Research 1 1
Banc of America Sec 3 1 3 7
Bear Stearns 1 1 2
Bernstein 1 1 1 3
Caris & Company 1 1
Citigroup 1 2 1 4 8 Credit Suisse 3 1 2 6 Deutsche Securities 1 1 3 1 6 Edward Jones 1 1 Friedman Billings 1 1 2 FTN Midwest 1 1 Goldman Sachs 4 1 5 Jefferies & Co 1 1 2 JP Morgan 1 2 3 6 KeyBanc Capital Mkts / McDonald 1 1 Lehman Brothers 2 1 3 Longbow 1 2 1 4 Matrix Research 1 5 6 Morgan Stanley 2 2 Oppenheimer 1 1 Piper Jaffray 2 2 Prudential 1 1 RBC Capital Mkts 1 1 Robert W. Baird 1 1 2 Roth Capital 1 1 Sandler O'Neill 1 1 Soleil 1 1 2 Stanford Research 1 1 Sterne Agee 2 2 Stifel Nicolaus 2 2 Susquehanna Financial 1 1 UBS 2 1 1 1 5 Wachovia 3 2 5 total 23 24 24 25 96
In this study I have to classify each recommendation into the crisis period of the non-crisis period. October 12, 2007 is defined as the beginning of the crisis because the S&P 500 total
return index started to decline after this point achieving a new all-time intraday high4 the day before. The bankruptcy of Lehman Brothers in September 2008 and the failure of the
Treasury bailout plan on September 29 result further in a drastically decline of the S&P 500. We define March 9, 2009 as the end of the crisis, because the S&P 500 index bottomed and after that day started to increase again. Figure 1 shows the pattern of the S&P 500 index between 2005 and 2010. The period before the crisis from April 12, 2006 till October 12, 2007 is defined as the non-crisis period and is approximately the same length as the crisis period.
Figure 1: S&P 500 index
Because the abnormal return six months after the change of recommendation is examined , the possibility exists that the recommendation is made in the non-crisis period whereas the
abnormal return is (partly) measured in the crisis period. Because I want to test if there is a significance difference between the two periods clear distinction is needed. Therefore I
exclude the changes in recommendations issued between April 12, 2007 and October 12, 2007 (6 months before the start of the crisis) and between September 9, 2008 and March 9, 2009 (the last 6 months of the crisis).
4
New intraday highmeans a security reached a new high relative to all other prices during a trading session
II. Methodology
In order to examine the recommendation stock-price effect, the average abnormal returns for the three-day event, one-month event and the six-month event are calculated as the actual return adjusted for market risk. The calculation is as follows:
ARit = Rit – (βi Rmt) where
ARit = abnormal return of stock i on day t (where day t=0 is the publication date) Rit= actual return of stock i on day t
βi =beta of stock i
Rmt = actual market return on day t (where the S&P 500 index is used as a proxy)
The three-day event window is analyzed in order to measure the average initial abnormal return, because results of prior research suggests that the initial abnormal stock price performance is spread into multiple days (before and after publication date) instead of only the publication day (t=0). Liu, Lloyd and Beneish found significant abnormal returns on t=-1 suggesting this is due to information leaking. Therefore the observed period for abnormal returns starts at t=-1. Since only closing stock prices5 are examined in this research, the closing stock price of day t=-2 is used in order to capture the stock-price effect of the day prior the publication day. To capture the entire initial market reaction I also have to include the day after publication in the three-day event window (t=1) as the majority of prior research showed significant abnormal returns on that day.
The longer-term event windows are examined to check if the average initial abnormal return is a permanent or temporary effect. Again I take into account information leaking one day before publication day. Approximately consistent with the studies of Womack and Liang, I analyze the one-month and six-month period after publication. The purpose of examining the one-month event is to see if the initial abnormal return is due to investors’ overreaction to analyst information or not. To further support the relevant hypothesis the six-month event will be analyzed.
5
Adjusted for dividends and splits
4.1 Results
This study is divided into three parts. First the comparison between buy recommendations is made between the non-crisis period and the crisis period and second the comparison of sell recommendations are made. Lastly, the different stock-price impact will be examined between sell and buy recommendations for each period separately.
I. Comparison buy recommendations
The initial effect and the pattern of the abnormal return in the non-crisis period for buy
recommendations is completely different compared to the recommendations made in the crisis period. As presented in table 2 the average three-day return for buy recommendations in the non-crisis is 2,38% and is significantly different than zero, whereas the initial abnormal return in the crisis period is roughly one-third (0,80%) though insignificant. For buy
recommendations made in the non-crisis period results show an mean-reversal effect supporting the price-pressure effect. Thus the relatively large initial significant abnormal return is due to buying pressure and decreases after one month to a insignificant return of 1,39%. Although also insignificant, the average abnormal return diminishes and results show it can become slightly negative (-0,1%) after 6 months. Since the results of the one and six-month windows are not significant they should be interpreted with caution. In contrast with the results associated with the non-crisis period, evidence of buy recommendations made in the crisis period support the information hypothesis. After a small three-day event return, results show an significant abnormal return of 4,02% after one-month. Although not
significant, after six-months the abnormal stock price impact increased further to 8,17%. The return drifts into the direction of the analyst prediction and suggests the analyst’
recommendations contain new information. Hence, buy recommendations can provide economic value to long-term investors during the financial crisis where during the non-crisis period buy recommendations do not. Because most of the results are not significant the above statements must be interpret with caution.
Table 2: descriptive statistics of the abnormal returns associated with buy recommendations in the crisis and non-crisis period
*=significant at 10% **=significant at 5%, ***=significant at 1%
The descriptive statistics are all statistics of the sample, so the sample mean, standard deviation and the variance. The t-tests are all two tailed tests. The t-test value of the single abnormal returns associated with buy and sell recommendations is the test statistics for the null hypothesis of the mean being equal to zero and is calculated as
follows: t= X�
s/√n where 𝑋𝑋�= the sample mean, s= the sample standard deviation and n=the sample size. The t-test value of the difference of the average abnormal returns between two periods is the test-statistic for the null
hypothesis of the difference of the means being equal to zero and is calculated as follows:
t=
X�1−X�2�s12 n1+n2s22
where 𝑋𝑋�= the sample mean, 𝑠𝑠2= the sample variance and n=the sample size.
II. Comparison sell recommendations
Results in table 3 suggest the pattern of the abnormal return associated with sell recommendations made in the non-crisis period is different compared to sell
recommendations made in the crisis period. However the initial stock-price effect is
approximately similar. The initial significant abnormal return for sell recommendations made in the non-crisis and crisis-period is respectively -3,01% and -2,78%. With respect to the non-crisis period, this return stays approximately the same after one month and results show a significant abnormal return of -2,84%. After six month the return will decrease further in the direction predicted by the analyst to -7,14%. These results support the information hypothesis. Again in contrast with these results, evidence of the stock-price impact with respect to sell recommendations made in the crisis period are in favor of the price-pressure effect. After a significant initial negative abnormal return of the three-day event, return increased to a positive return of 3,84% after one month and remains approximately the same after six months. Hence in contrast to what analyst predict, abnormal returns continue to increase even after a complete mean-reversal. The abnormal returns of the one-month and six-month event should be looked at with caution, because they are not significant. However results show that the one-month and six-month returns are significantly different in the two periods and thus you should not ignore the fact that the pattern of the stock-price effect after the three-day event is significantly different.
Table 3: descriptive statistics of the abnormal returns associated with sell recommendations in the crisis and non-crisis period
*=significant at 10% **=significant at 5%, ***=significant at 1%
three-day event
one-month event
six-month event
sell non crisis (n1=24)
mean
-0,0301
-0,0284
-0,0714
standard deviation
0,0316
0,0576
0,1692
variance
0,0010
0,0033
0,0286
t-test
-4,68***
-2,41**
-2,07**
sell crisis (n
2=25)
mean
-0,0278
0,0384
0,0331
standard deviation
0,0564
0,1421
0,1954
variance
0,0032
0,0202
0,0382
t-test
-2,46**
1,35
0,85
difference sell
non-crisis and crisis
mean
-0,0024
-0,0668
-0,1046
t-test
-0,18
-2,17**
-2,00 *
The descriptive statistics are all statistics of the sample, so the sample mean, standard deviation and the variance. The t-tests are all two tailed tests. The t-test value of the single abnormal returns associated with buy
and sell recommendations is the test statistics for the null hypothesis of the mean being equal to zero and is calculated as follows: t= X�
s/√n where 𝑋𝑋�= the sample mean, s= the sample standard deviation and n=the sample size. The t-test value of the difference of the average abnormal returns between two periods is the test-statistic
for the null hypothesis of the difference of the means being equal to zero and is calculated as follows: t= X�1−X�2
�s12
n1+n2s22
where 𝑋𝑋�= the sample mean, 𝑠𝑠2= the sample variance and n=the sample size.
III. Comparison different stock-price effect of buy and sell recommendations
Table 4 presents the difference of the average abnormal returns with buy and sell
recommendations for the non-crisis and crisis period. A negative difference indicates that the sell recommendations generate a greater negative abnormal return than the positive abnormal return associated with the buy recommendations. Even though the difference of the stock-price effect is small (-0,63%) and insignificant, the initial market response is more intensively
to sell recommendations in the non-crisis period compared to buy recommendations.
Additionally the negative difference of the abnormal returns in the non-crisis period continues to decrease up to the one-month event. Hence the average abnormal return associated with a sell recommendation is higher compared to the return of the buy recommendation after three days and one month. This is due to the fact that sell and buy recommendations in the non-crisis period generate respectively a permanent and a mean-reversal effect. Because the difference of the average abnormal returns are not significant it is difficult to interpret and make overarching statements with respect to outcomes for the non-crisis period presented in table 4. Consistent with the non-crisis period the average initial abnormal return for sell recommendations made in the financial crisis is higher compared to buy recommendations, although again the difference is insignificant. Where the difference is notated such as ‘-’, the calculation is not useful because the sell and buy generated both positive or negative
abnormal returns.
Table 4: the difference of the mean abnormal return associated with buy and sell recommendations in the non-crisis and crisis period.
*=significant at 10% **=significant at 5%, ***=significant at 1%
three-day event
one-month event
six-month event
difference buy/sell
non-crisis
mean
-0,0063
-0,0144
-t-test
-0,59
-0,93
-difference buy/sell
crisis
mean
-0,0198
-
-t-test
-1,50
-
-The mean is the sample mean of the difference between abnormal return associated with buy and sell recommendations. The t-tests are all two tailed tests. The t-test value of the difference is the test statistic for the
null hypothesis of the difference of the means being equal to zero and is calculated as follows:
t=
X�1−X�2�s12 n1+n2s22
4.2 Robustness test
As a robustness check I explore whether the specification of not including the revised recommendations made in the periods 6 months before the start of the crisis and the last 6 months of the crisis has influence on the results. In table 5 the abnormal returns are presented with respect to buy and sell recommendations, where non-crisis recommendations are made between April 12, 2006 and October 12, 2007 and crisis recommendations are made between October 12, 2007 and March 9, 2009. The abnormal returns with respect to buy
recommendations are different when excluding the above mentioned specification. The buy recommendations in the non crisis period support the information hypothesis whereas in the result section they support the price-pressure hypothesis. Additionally the buy
recommendations made in de crisis period generate abnormal returns approximately twice as large for all three events compared to the abnormal returns observed with the specification. Although the results are different, they are supporting the same hypothesis. Also abnormal returns associated with sell recommendations in both periods support the same hypothesis as presented in the result section. The abnormal returns in table 5 present that the specification do not have much influence on the results with respect to sell recommendations.
Table 5: the average abnormal returns associated with buy and sell recommendations made between October 12, 2007 –March 9, 2007 (non crisis) and October 12, 2007-
March 9, 2009 (crisis)
three-day event
one-month event
six-month event
buy non crisis (n=23)
0,0228
0,0263
0,0390
buy crisis (n=24)
0,0192
0,0705
0,1884
sell non crisis (n=24)
-0,0287
-0,0335
-0,0788
5. Conclusion
Results of buy recommendations made in both periods show positive initial abnormal returns. Although the direct stock-price effect is not significantly different, the abnormal return with respect to buy recommendations made in the non-crisis period is roughly three times the abnormal return with respect to the crisis period indicating the market reacts more strongly to buy recommendations made in the non-crisis. This suggests that the need for positive analyst information decreased in times of financial crisis. The initial abnormal return during the non-crisis might due to investors’ initial overreaction to analysts’ positive information because the one- and six months results with respect to buy recommendations made in the non-crisis period support the price-pressure hypothesis. However the results of buy recommendations made in the crisis-period support the information hypothesis. Consequently only result of buy recommendations made by analyst during the financial crisis provides economic value to long-term investors. On the contrary sell analysts can forecast the direction of the excess stock price only in times when the market is assumingly stable. For both periods we observe a approximately similar significant negative abnormal return at first with respect to sell recommendations, however the results of one and six months supports the information hypothesis in the non-crisis period and the price-pressure hypothesis for the crisis period. Therefore the price-reversal effect during the financial crisis suggests (naive) investors are more likely to temporary overreact directly upon the sell recommendations.
Hence, in the period before the financial crisis results associated with buy recommendations support the price-pressure hypothesis whereas results associated with sell recommendations supports the information hypothesis. This suggests buy and sell recommendations generate different patterns of abnormal returns, although the difference between the two
recommendations in the non-crisis period is insignificant. Results found for the non-crisis period match the results of Liden (2004) who examined recommendations on the Swedish stock market. He suggests the different information content of the analysts sell or buy recommendations is due to the different nature of the recommendations. Sell
recommendations are based on more (new) economic value compared to buy
recommendations. However the results of my study show the opposite effect during the financial crisis. Abnormal returns of buy recommendations support the information hypothesis whereas results of sell recommendations support the price-pressure hypothesis suggesting the information content changed. The most important suggestion of this study is
that during the financial crisis analysts’ information become less economically valuable with respect to sell recommendations and more valuable with respect to buy recommendations. One reason for that could be that the nature of the recommendations changed. During the financial crisis buy analysts possibly expand their research and therefore the chance that analyst find new information increases. On the other side sell analysts could base their predictions on the mutual fear of a drop down in the market instead of new information. However evidence of this reasoning is not clear in this research. Therefore let this work be an inspiration to further investigate the content of analyst recommendations during ‘bad’ times.
Reference list
Albert, R.L. and Smaby, T.R. (1996). Market response to analyst recommendations in the “Dartboard” column: The information and price pressure effects, Review of financial
economics, 5(1), 59-74.
Asquith, P., Mikhail, M.B., & Au, A.S. (2005). Information content of equity analyst reports.
Journal of Financial Economics, 75, 245-282.
Barber, B.M., & Loeffler, D. (1993). The “Dartboard” column: second-hand information and price pressure. The Journal of Financial and Quantitative analysis, 28(2), 273-284.
Beneish, M.D. (1991). Stock prices and the dissemination of analysts’ recommendation. The
Journal of Business, 64(3), 393-416.
Davies, P. L., & Canes, M. (1978). Stock prices and the publication of second-hand information. Journal of Business, 51(1), 43-56.
Elton, E.J., Gruber, M.J., & Grossman, S. (1986). Discrete expectational data and portfolio performance. The Journal of Finance, 41(3), 699-713.
Kerl, A.G., & Walter, A. (2007). Market responses to buy recommendations issued by personal finance magazines: Effects of information, price-pressure, and company characteristics. Review of Finance, 11, 117-141.
Liang, B. (1999). Price pressure: Evidence from the “Dartboard” column, The Journal of
Business,72 (1), 119-134.
Liden, E.R. (2004). Swedish stock recommendations: Information content or price pressure? Retrieved from Göteborg University, department of economics, website:
Liu, P., Smith, S.D., & Syed, A.A. (1990). Stock price reactions to The Wall Street journal’s securities recommendations. The Journal of Financial and Quantitative Analysis, 25(3), 399-410.
Moshirian, F., Ng, D., & Wu, E. (2009). The value of stock analysts’ recommendations: Evidence from emerging markets, International Review of Financial Analysis, 18, 74-83.
Ryan, P., & Taffler, R.J. (2006). Do brokerage houses add value? The market impact of UK sell-side analyst recommendation changes. The British Accounting Review, 38, 371-386.
Stickel, S.E. (1995). The anatomy of the performance of buy and sell recommendations.
Financial Analysts Journal, 51(5), 25-39.
Stickel, S.E. (1984). The effect of value line investment survey rank changes on common stock prices. Journal of Financial Economics, 14, 121-143.
Wijmenga R.Th (1990). The performance of published Dutch stock recommendations.
Journal of Banking and Finance, 14, 559-581.
Womack, K.L. (1996). Do brokerage analysts’ recommendations have investment value? The
Journal of Finance, 51(1), 137-167.
Yahoo Finance. Recommendations, historical stock prices and beta’s. Retrieved June 23, 2014 from http://finance.yahoo.com/
