1
Did the Financial Crisis Change the Influence of
Stock Analysts’ Recommendations on NYSE
Listed Banks’ Stock Returns?
A Panel Data Study Using a One-Way-Error Fixed Effects
Model
H.B. Hemmer⃰ Master’s Thesis Msc BA
Faculty of Economics and Business, Rijksuniversiteit Groningen
Abstract
This study determines whether the latest financial crisis changes the influence of stock analysts’ recommendations on the selected NYSE banks’ unadjusted and mean-adjusted daily stock returns. The influence of stock analysts’ recommendations is measured over 1- and 3-day event windows. We use a Mann–Whitney U-test and a regression that controls for four important determinants of those US banks’ daily stock returns. By analyzing both unadjusted and mean-adjusted daily stock returns, we find evidence that investors reacted more in conformity with analysts’ recommendations during the financial crisis.
Keywords: Analysts’ recommendations, financial crisis, panel data, fixed effects model JEL Classification: C23, G14, G21, G24
⃰
2
Numerous studies have examined the value of stock analysts’ recommendations. On the one hand, studies argue that they create value by providing investors pre-analyzed data that is essential for pricing stocks (Grossman and Stiglitz (1980), Womack (1996), Beaver (2002)). Investors following such recommendations become better informed and should thus be able to make better investment decisions and gain abnormal returns, which is empirically proofed by Womack (1996) and Barber, Reuven, McNichols, and Trueman (2001). On the other hand, the semi-strong version of the efficient market hypothesis states that investors cannot earn superior risk-adjusted returns using any publicly available information (Fama (1970)). Thus, following publicly available analyst recommendations should not generate risk-adjusted abnormal returns. Which restricts analysts’ recommendations value. Studies by Diefenbach (1972) and Bidwell (1977) empirically support this position.
These two opposing views on the value of analysts for investors are the main inspiration for this research. This research examines the influence of analysts’ recommendations during the latest financial crisis. A financial crisis increases uncertainty among investors, (Blanchard (2009)). Due to this uncertainty we expect that financial analysts would be most valuable to investors—and such value most detectable—during the latest financial crisis. However, there are no studies elaborating on the effect of a financial crises on the value of analysts’ recommendations. This research aims to fill this gap in the literature and presents a case study with both industrial and geographical restrictions.
We first restrict the industry analyzed. We choose the financial industry because (i) it is characterized as the catalyst of the economy (Goldsmith (1969), Hicks (1969), Levine (1997)); (ii) it is very sensitive to news due to the financial fragility hypothesis
3
on the United States since the US securities markets are well developed (Allen and Gale (1997)) and have no data limitations with regard to this research.
The following research question is the focus of this study:
Did the financial crisis change the influence of analysts’ recommendations on NYSE listed banks’ daily stock returns?
An examination of this question requires focusing on a time period that includes at least the latest financial crisis and a period were the influence of the latest financial crisis can be measured against. Moreover, to draw representative and up-to-date conclusions, we focus only on the past decade. To determine the non–financial crisis- and the financial crisis period, we use, recommended by Blanchard (2009) the volatility index of the Standard & Poor’s S&P 500 Index (VIX)(see Figure 1).
Figure 1. Standard & Poor’s S&P 500 VIX rate. Source: Yahoo! Finance1.
4
This index is a popular measure of the implied S&P 500 Index options and therefore indicates the market’s expectation of the stock market volatility over the next 30 days. Figure 1 shows two periods of relatively high volatility: the period between January 2000 and April 2003 and the latest financial crisis, between July 2007 and October 2010. The intermediate period (April 2003 to July 2007) has a relatively low VIX rate. Since we want to determine the influence of the financial crisis, we define the period July 2007 to October 2010 as the financial crisis period and the period from January 2000 to July 2007 as the non–financial crisis period. Regarding the selection of US banks, we focus on those listed on the New York Stock Exchange (NYSE). We then collect all analysts’ recommendations from January 2000 to October 2010 for these banks from Yahoo!
Finance2. The final sample comprises 64 banks, with a total of 1,338 upgrades and 1,083
downgrades.
To answer the research question, we first compare, using a Mann–Whitney U-test, the selected banks’ daily returns during the non–financial crisis on the day of a downgrade with the daily returns during the financial crisis period on day of a downgrade. We do the same test for the upgrades. To account for firm-specific risk and time period effects, we repeat the four tests, but now the average return over the 200-day pre-recommendation period (as suggested by Brown and Warner (1985)) is first subtracted from the selected banks’ stock returns. To account for any leakage (Stickel (1995), Jegadeesh and Kim (2006)) and post-event effects (Beneish (1991)) of analysts’ recommendations, the same tests, including the selected banks’ stock returns on the two days surrounding analyst upgrade/downgrade, are repeated.
Conclusions drawn on just a Mann–Whitney U-test may be sensitive to omitted variables. Therefore we make a regression were the sampled US banks' stock returns (R) over the period January 2000 to October 2010 are regressed against four sets of dummies (which are combinations of non–financial crisis/financial crisis and downgrade/upgrade
dummies), and control them by exchange rates, market returns, long-term interest rates,
5
and short-term interest rates. These control variables are considered important determinants of US banks’ stock prices (See: Choi, Elyasiani, and Kopecky (1992), Wetmore and Brick, (1994), Choi and Elyasiani, (1997), and Tai (2000). In order to make such a regression the data must be pooled. Moreover, the regressions are re-estimated using dummies that also indicate the two days surrounding a downgrade/upgrade. The research question is then answered by a Wald coefficient test that compares the financial crisis dummy variables with the non–financial crisis dummy variables.
To enhance the robustness of the regression, we account for firm-specific risk and period effects, we subtract before regressing, from all selected banks’ stock returns ( their average stock return measured over 200 days (as suggested by Brown and Warner (1985)).
Moreover, we redefine the non–financial crisis period as April 2003 to July 2007. Since this period has a very low VIX rate, it is a very calm period. We expect this change to enhance the contrast of the influence of analysts’ recommendations on US banks’ stock returns between the non–financial crisis and the financial crisis periods.
The remainder of this thesis is structured as follows. Section I reviews the literature that describes why investors follow analysts’ recommendations and the empirical evidence. Section II describes the data used for this research. Section III outlines the methodology. Section IV outlines the results of this research, followed by a conclusion and the
6
I.Literature Review
A. Do Investors Follow Analysts’ Recommendations?
Fama (1970) defines an efficient financial market as one in which security prices always fully reflect the available information. Such a market rests on the arguments that: (i) investors are rational (ii) in case that some investors are not rational, their random trades cancel each other out (iii) in case investors are irrational in similar ways, their influence on prices is mitigated by arbitrageurs. According to the semi-strong form of market efficiency, investors are not able to benefit from trades using any publicly available information. Following publicly available analysts’ recommendations would thus not beneficial. However, more recent studies, by Banz (1981) and Fama and French (1992), indicate that this assumption may be too simplistic. These authors indicate that the available information is incompletely incorporated into market’s stock prices. This shortcoming is mainly induced by average prudent investors, who lack the time, skills, or resources to analyze and correctly interpret financial statements (Beaver (2002)).
However, according to Beaver (2002), analysts are able to compensate for this deficit through the tasks they perform. According to Michaely and Womack (1999) page 658, these tasks include
(i) gathering new information on the industry or individual stock from customers, suppliers, and firm managers; (ii) analyzing these data and forming earnings estimates and recommendations; and (iii) presenting recommendations and financial models to buy-side customers in presentations and written reports.
7
number of analysts who all have a different vision of the stock market means that uninformed investors follow many different recommendations.
This multidirectional character of investors following large numbers of analysts is superseded by herding behavior among analysts. Since analysts tend toward unilateral herding due to personal prestige costs which rises as the number of people who made the same wrong decision decreases. (Scharfstein and Stein (1990), Trueman (1994),
Prendergast and Stole (1996)) And this herding behavior among analysts is, for instance, detected by De Bondt and Forbes (1999), Darrough and Russell (2002) and Jegadeesh and Kim (2010).
B. Empirical Evidence of Investors Following Analysts’ Recommendations
An early article by Brown and Rozeff (1978) compares time series models with analysts’ forecasts and concludes that analysts’ forecasts are superior to time series models. These superior analysts’ forecasts can persuade investors to follow them. The authors’
conclusion is also confirmed in more recent studies. For instance, Hawkins, Chamberlin, and Daniel (1984), using a database containing earnings estimates made by more than 70 brokerage firms for over 2,400 stocks, examine the month-to-month percentage changes in consensus estimates and find that this information can be useful in achieving positive risk-adjusted returns. Beneish (1991) also finds positive abnormal returns based on analysts’ recommendations published in the ―Heard on the Street‖ item of The Wall
Street Journal. Evidence indicates that analysts’ recommendations are associated with
significant average abnormal stock price performance on the publication dates and the two previous days.
8
and Trueman (2001) examined if it was possible for investors to profit from the publicly available recommendations of security analysts. The authors conclude that if the portfolio is rebalanced every day and if there is a timely response to analysts’ recommendations, the investor can earn 4% annual abnormal gross returns.
The reason why uninformed investors are inclined to follow analyst recommendations was mentioned above. The empirical literature shows, however evidence that informed institutional investors incorporate analysts’ recommendations into their investment decisions. Wei, Brown, and Wermers (2007) examine the relation between analysts’ recommendations and investor herding behavior, focusing on the trading activities of US mutual funds. The reason for this focus is that the authors want to determine the impact of fund trading on equity prices. More precisely, their study determines whether funds also simultaneously exhibit following behavior as a result of sell-side analysts’
recommendations. In their opinion, analyst recommendations represent public
information which is useful for institutional investors and. This result in herding behavior and therefore have a significant impact on prices in the US equity markets. One of the main results of Wei, Brown, and Wermers (2007) is that mutual funds managers clearly analyst recommendations in their investment decisions. In addition, the authors find evidence that downgrades incur stronger following among mutual funds managers than upgrades.
9 II.Data
This study focuses on all banks listed on the NYSE, Since this index contains most banks. For a complete list of those banks, we consult Bankscope3, and obtained a list of 141 banks. To obtain the precise dates of analysts’ recommendations, we use Yahoo!
Finance4 since, according to a news article published by Business Insider on September 14, 20105, Yahoo! Finance is the most highly viewed worldwide financial news website today, with more than 45.6 million unique visitors in August 2010 (comScore6).
Unfortunately, Yahoo! Finance lacks historical analysts’ recommendations from 60 banks that were gone out of business during the sample periods and from 17 living banks. Therefore these banks are removed from the sample, for a final sample of 64 banks (see Table I on the next three pages).
The analysts’ recommendations presented at Yahoo! Finance contain recommendation changes presented as upgrades and downgrades and initiated recommendations. To map the influence of analysts on stock prices as comprehensively as possible, both
recommendation changes and initiated recommendations are used to mark the event days, for a total number of 2,827. However, analysts’ scaling systems use different
terminologies and scales (strong buy–buy–hold–sell, underperform–market perform–over perform, etc.) to express their expectations. Therefore, we draw up a comprehensive rating system that encompasses all the different rating systems. This new rating system converts the analysts’ terminology into a seven-point scale from -3 to 3, where a score of -3 means that the analyst is very pessimistic about the stock outlook and a score of 3 means the analyst is very optimistic.
3
A database provided by Bureau van Dijk that contains all the available information of all the financial institutions around the world.
4 http://finance.yahoo.com/
5 http://www.businessinsider.com/bloomberg-puts-its-stories-on-yahoo-finance-2010-9
10
Table I Selected Banks’ Anysts’ Recommendations and Assets. This table shows the distribution of the four different types of recommendations along the 64 selected banks from the NYSE. Moreover, the assets in thousands of USD are given of these banks (n.a. means that Bankscope was not able to provide information on this subject.).
Assets Upgrades Downgrades Initiated Initiated Total No. Name of Bank (thousands of USD) Positive Negative
1 Bank of America Corporation 2,223,299,000.00 36 40 18 2 96
2 JP Morgan Chase & Co. 2,031,989,000.00 22 23 13 1 59
3 Citigroup Inc. 1,856,646,000.00 20 25 21 1 67
4 Wells Fargo & Company 1,243,646,000.00 30 35 17 1 83
5 Goldman Sachs Group, Inc. 848,942,000.00 29 32 20 1 82
6 Morgan Stanley 771,462,000.00 13 9 4 1 27
7 GE Capital-General Electric Cap. Corp. 623,097,000.00 11 10 6 0 27
8 Metlife, Inc. n.a. 31 29 8 0 68
9 Prudential Financial Inc 480,203,000.00 13 12 11 0 36
10 Merrill Lynch 479,195,000.00 6 14 2 0 22
11 Credit Suisse USA 374,608,688.00 7 11 4 0 22
12 PNC Financial Services Group Inc. 269,863,000.00 29 23 6 2 60
13 Bank of New York Mellon Corp. 212,224,000.00 27 22 16 1 66
14 SunTrust Banks Inc. 174,164,703.00 29 41 9 3 82
15 Capital One Financial Corporation 169,646,359.00 44 58 18 3 123
16 Sallie Mae-SLM Corporation 168,768,406.00 24 21 10 1 56
17 BB&T Corporation 165,764,219.00 31 35 15 4 85
18 State Street Corporation 157,946,000.00 35 23 13 3 74
19 Regions Financial Corporation 142,318,000.00 17 14 2 2 35
11
Assets Upgrades Downgrades Initiated Initiated Total No. Name of Bank (thousands of USD) Positive Negative
21 KeyCorp 93,287,000.00 21 22 4 2 49
22 M&T Bank Corporation 68,880,398.00 15 16 9 2 42
23 CIT Group, Inc. 60,029,102.00 0 0 3 0 3
24 Comerica Incorporated 59,249,000.00 36 38 10 3 87
25 Marshall & Ilsley Corporation 57,210,000.00 17 22 13 2 54
26 Discover Financial Services 46,020,988.00 6 5 4 2 17
27 New York Community Bancorp, Inc. 42,148,195.00 18 18 5 2 43
28 MF Global Holdings Ltd. 38,836,602.00 4 5 6 0 15
29 Synovus Financial Corp. 32,831,418.00 26 26 10 2 64
30 Jefferies Group Inc. 28,189,301.00 11 11 1 2 25
31 First Horizon National Corporation 26,068,699.00 15 19 5 0 39
32 City National Corporation 21,078,801.00 17 27 9 1 54
33 Credicorp Ltd. 20,821,100.00 4 4 0 0 8
34 Astoria Financial Corporation 20,252,199.00 14 18 2 2 36
35 Raymond James Financial Inc. 18,226,699.00 11 6 2 1 20
36 TCF Financial Corporation 17,885,199.00 19 23 9 4 55
37 Webster Financial Corp. 17,744,861.00 9 13 0 0 22
38 Cullen/Frost Bankers, Inc. 16,288,000.00 10 14 9 0 33
39 iStar Financial Inc. 15,296,700.00 2 4 12 0 18
12
Assets Upgrades Downgrades Initiated Initiated Total No. Name of Bank (thousands of USD) Positive Negative
41 Flagstar Bancorp Inc. 14,013,300.00 7 7 0 0 14
42 BancorpSouth, Inc. 13,206,923.00 4 4 2 0 10
43 Bank of Hawaii Corporation 12,414,800.00 13 12 3 0 28
44 Wilmington Trust Corporation 11,097,100.00 13 12 6 0 31
45 Doral Financial Corporation 10,232,000.00 9 16 4 1 30
46 Legg Mason Inc. 9,321,400.00 23 19 5 3 50
47 Franklin Resources, Inc. n.a. 31 36 13 1 81
48 FNB Corporation 8,709,077.00 7 10 2 2 21
49 NewAlliance Bancshares Inc. 8,434,300.00 4 5 0 0 9
50 Old National Bancorp 8,005,300.00 1 4 1 0 6
51 Provident Financial Services, Inc. n.a. 6 9 1 0 16
52 Oriental Financial Group Inc. 6,550,833.00 5 6 0 0 11
53 First Commonwealth Financial Corp. 6,446,300.00 11 9 1 1 22
54 Western Alliance Bancorporation 5,753,279.00 7 7 4 1 19
55 Community Bank System, Inc. 5,402,800.00 3 5 4 0 12
56 Capitol Bancorp Ltd. n.a. 2 1 1 0 4
57 Central Pacific Financial Corp. 4,890,484.00 2 4 1 0 7
58 BankAtlantic Bancorp 4,815,600.00 12 15 10 1 38
59 SWS Group Inc. 4,199,000.00 0 1 0 0 1
60 Banco Latinoamericano 3,879,000.00 4 2 1 0 7
61 Stifel Financial Corp 3,167,400.00 1 3 1 0 5
62 Lazard Limited 3,147,800.00 1 4 7 1 13
63 Ocwen Financial Corp 2,238,000.00 1 3 5 0 9
64 Sterling Bancorp 2,164,596.00 12 12 2 1 27
13
Implementing this new rating system results in a situation that can be categorized into one of five different types of recommendations: the two types of ordinary
recommendation changes (upgrades and downgrades) and the three types of initiated recommendations (negative, neutral, and positive). Since initiated neutral
recommendations should not lead to investor action, they are deleted from the event days. Therefore the total number of event days is reduced to 1,587 (1,837 - 250) non–financial crisis and 834 (990 - 156) financial crisis recommendation days. The event days are then converted to dummy variables for use in panel regressions. Table II below shows the distributions of the numbers of these four different types of recommendations over the non–financial crisis period and the financial crisis period.
Table II. Analysts’ Recommendations. This table outlines the distribution of the
different types of analysts’ recommendations over the non–financial crisis period (NFCP, January 2000 to July 2007) and the financial crisis period (FCP, July 2007 to October 2010).
Period FCP NFCP Total Type of Down- Up- Down- Up-
Recommendation grades grades grades grades
Upgrades 353 570 923 Downgrades 346 667 1013 Initiated positive 98 317 415 negative 37 33 70 Total 383 451 700 887 2421
This study uses daily stock prices of the selected banks’ which are obtained from
Datastream. The stock prices are converted into continuously compounded stock returns,
14
With respect to the macroeconomic variables —short-term interest rates (STI), long-term interest rates (LTI), market returns (MR), and exchange rates (FX)—this study relies on different sources of information. We follow Elyasiani and Mansur (2002) by using the daily absolute differences of the 10-year Treasury composite yield as a proxy for long-term interest rate developments. According to Choi, Elyasiani, and Kopecky (1992), we use the daily absolute differences of the 3-month US Treasury bill yield as a proxy for short-term interest rate developments. Since this study selects a group of banks listed on the NYSE, the continuously compounded daily returns of that index control for market returns. Both yields and market returns are obtained from Datastream.
With respect to the exchange rates, this study uses the relative daily changes of the broad
exchange index7, just like Choi, Elyasiani, and Kopecky (1992), obtained from the statistical releases from the Federal Reserve’s website8. Relevant figures about the macroeconomic variables are presented in Table III on the next page. Additionally, the data used in this study have the features of both cross-sectional and time series data. The 64 different banks can be seen as cross-sectional entities, while the dummies given for the financial crisis versus the non-crisis period, analysts’ recommendations, and the four macroeconomic variables can be regarded as time series data.
This so-called panel data structure has several advantages (Baltagi (2005), Brooks (2008)): First, by using panel data, it is possible to address a broader range of issues and to tackle more complex problems. Second, with the panel data, it is possible to increase the number of degrees of freedom. Since information on the simultaneous dynamic behavior of a large number of entities is employed, the statistical power of tests is improved. The additional variation resulting from combining cross-sectional and time series data also contributes to mitigate multicollinearity problems. Finally, using panel data reduces the impact of the omitted variables bias in regression results.
7 The broad exchange index is the weighted average of the foreign exchange values of the US dollar against the currencies of a large
group of major US trading partners.
15
Table III. Descriptive Statistics. This table presents the mean, median, maximum, minimum, standard deviation, kurtosis, and Jarque–Bera (JB) probability of the exchange rates (FX), market returns (MR), long-term interest rates (LTI), short-term interest rates (STI), and the selected banks’ daily returns (R). Since the variables are relative daily changes (FX, MR, R) or, in the case of both yields (LTI, STI), absolute daily changes, the values are presented as percentages. Moreover, those measures are calculated over the non-financial crisis-, non-financial crisis and total sample period. In addition the number of observations is given.
Variables Mean Med. Max. Min. Std. Dev. Kurt. JB Prob.
16
Finally, a correlation matrix (see Table IV on the next page) is constructed to check for multicollinearity between the variables. The four macroeconomic variables and the two binary variables that indicate the day of a downgrade or an upgrade are present. We do not incorporate binary variables constructed as a result of time effects (financial crisis and non–financial crisis), since those variables will bias the correlation matrix due to the fact that they are inherently correlated with each other. Moreover, their correlation with the continuous macroeconomic variables will always be partly mitigated, since they are period dependent. In addition, to highlight the differences between the non–financial crisis period, the financial crisis period and the total sample period, the correlation matrix is conducted over each of these periods. There is no evidence that any one of the variables is correlated with any other variable. This minimizes the occurrence of
17
Table IV. Correlation Matrix. The matrix shows that none of the variables have any serious correlations with any other variable, thus reducing multicollinearity. To compare the non–financial crisis period (NFCP), financial crisis period (FCP) and the total sample period, the numbers are shown next to each other.
Variables FX MR LTI
NFCP FCP Total NFCP FCP Total NFCP FCP Total
FX 1.0000 1.0000 1.0000 -0.0760 -0.3703 -0.2589 0.1778 -0.1459 0.0055 MR -0.0760 -0.3703 -0.2589 1.0000 1.0000 1.0000 0.1853 0.4072 0.3084 LTI 0.1778 -0.1459 0.0055 0.1853 0.4072 0.3084 1.0000 1.0000 1.0000 STI 0.0708 -0.0485 -0.0101 0.0867 0.1911 0.1615 0.1950 0.1651 0.1689 DOWN -0.0014 0.0035 0.0009 -0.0069 -0.0190 -0.0126 0.0005 0.0038 0.0018 UP 0.0037 0.0089 0.0059 0.0000 -0.0018 -0.0009 0.0038 0.0062 0.0047 Variables FX DOWN UP
NFCP FCP Total NFCP FCP Total NFCP FCP Total
18
III.Methodology
A. The Mann–Whitney U-Test
The hypotheses in case of an upgrade or downgrade are as follows:
(2)
(3)
where is the stock return of bank i on the day of an upgrade, UP, or a downgrade,
DOWN, during the financial crisis period c or the non–financial crisis period n.
To obtain a first answer to the research question, both hypotheses are tested with a non-parametric two-sided test. For this purpose, we use a Mann–Whitney U-test. The Mann– Whitney U-statistic is defined as
(4)
where n are the numbers of upgrades or downgrades, in the non-financial crisis n and financial crisis c periods, respectively, and is the sum of the ranks of the observations from the financial crisis sample. The Mann–Whitney U’s mean and variance are,
respectively,
(5)
and
19
In this study’s sample, the distribution of the random variable is approximated by the normal distribution and has the following identity:
(7)
The null hypothesis suggests that the distributions of both populations have the same median. With both an upgrade and a downgrade, the null hypotheses are, respectively, rejected if
(8)
(9)
According to Liu, Smith and Syed (1990) and Beneish (1991), the effect of a
recommendation on stock price is most noticeable in a 3-day event window consisting of the day before, the day of, and the day after the recommendation. However, these authors and Stickel (1995) find the most significant effect on the day of the recommendation. Due to this ambiguity, the test is performed over both the 3-day and the 1-day event window. To enhance robustness, we correct for firm-specific risk and time period effects by first subtracting the 200-day pre-recommendation average stock return, as suggested by Brown and Warner (1985), from , , , and and then repeat the Mann–Whitney U-test with these adjusted returns.
B. Regression Methodology
Conclusions drawn on just a Mann–Whitney U-test may be sensitive to omitted variables. Distilling the exact effect of the financial crisis on the influence of analysts’
recommendations on the NYSE listed banks’ daily stock returns is therefore difficult. To overcome this problem, both unadjusted and adjusted returns are regressed against dummy variables that indicate combinations of non–financial crisis/financial crisis
20
(STI), which are considered by Choi, Elyasiani, and Kopecky (1992), Wetmore and Brick (1994), Choi and Elyasiani (1997), and Tai (2000) to be important determinants of US banks’ stock returns. Due to the maturity mismatch (borrow short, lend long), banks’ gross profits are jeopardized if the short-term interest rate unexpectedly rises (Flannery (1981), Saunders and Yourougou (1990), Choi, Elyasiani, and Kopecky (1992), Van den Heuvel (2001)). According to the expectations hypothesis (Mankiw and Summers (1984)), long-term interest rates should reflect a weighted average of the present interest rate and expected short-term interest rates. Therefore, we assume that both interest rates are negatively related to banks’ stock returns. Empirical evidence on this subject (Choi, Elyasiani, and Kopecky (1992), Wetmore and Brick (1994), Benink and Wolff (2000), Tai (2000)) supports this negative relation.
Using a sample of 59 US banks divided into three groups based on asset size, Choi and Elyasiani (1997) find significant evidence that market returns positively influence banks’ stock returns. Allen and Jagtiani (1989), Choi, Elyasiani, and Kopecky (1992), and Tai (2000) also find evidence for a positive relation between the market returns and stock returns of US financial institutions.
Exchange rate fluctuations can influence a bank’s profitability and therefore, banks’ stock prices are exposed to this exchange rate risk (Tai (2000)); however, they are able to hedge themselves against these risks (Brown (2001)), although this often involves transaction costs (Koutmos and Martin (2003)). We therefore expect a positive relation between the value of the US dollar and NYSE listed banks’ stock returns.
21 The following model is therefore used:
; t = 1,… ,T
(10)
where B is the total number (64) of individual banks observed in this study; is the number of banks observed in year t, where Bt B, since the data used for this study have an unbalanced structure; it, the explanatory variable, are the stock returns of all sampled banks i on day t; a is the intercept; and is a vector of parameters to be estimated on the explanatory variables. These explanatory variables are the dummy variables indicating the combinations of financial crisis/non–financial crisis periods and
upgrades/downgrades and the four macroeconomic controlling variables. The error component (u ) is subdivided into both an unobservable bank-specific effect (it ) and a stochastic disturbance term , and is a vector of observations on each of the explanatory variables and is assumed to be independent of for all i and t.
The regression that will be estimated is9
(11)
The Mann–Whitney U-test mentioned in the previous subsection accounts for two event windows. The regression does the same by estimating the coefficients of both event windows in a separate regression.
9 Here
are the stock returns of all selected banks over the total sample period, is a constant, UP is a dummy variable that
indicates the day of an upgrade, DOWN is a dummy variable that represents the day of a downgrade; is the dummy variable that
22
To test whether the financial crisis changes the influence of analysts’ recommendations on NYSE listed banks’ daily stock returns, a two-sided Wald coefficient test, with significance levels of 1%, 5%, and 10%, is used. The following sets of hypotheses are tested by all the estimated regressions:
(12)
(13)
C. Tests for Robustness
To test whether the model is robust, we make two adjustments in the regression. Just like the Mann–Whitney U-test, we correct for firm specific risk and time period effects. For this correction we subtract from all selected banks’ stock returns ( their average stock return measured over 200 days (as suggested by Brown and Warner (1985)), before regressing.
23 IV.Results
A. Mann–Whitney U-Test
The results of the Mann–Whitney U-test are presented in Table V on the next page and show that the impact of a downgrade during the financial crisis is greater than during the non–financial crisis period, regardless of the return nuances or the event period.
However, a downgrade has the most influence on the stock return of our selected banks on the day of initiation. This finding is also confirmed by Liu, Smith, and Syed (1990) and Stickel (1995), who find the most significant results on the day of the
recommendation. During the financial crisis, our selected banks face a median positive unadjusted stock return in both event windows as a result of an upgrade. However, in the case of a downgrade, the effect of a recommendation is mitigated if it is measured over a 3-day event window. Adjusted stock returns on the day of an upgrade are only
significantly lower in times of a financial crisis if measured at a 10% significance level. Over a 3-day event window, the effect of an upgrade on the adjusted return is minimal, but significantly higher during the financial crisis.
Another interesting finding is that the absolute median returns in case of a downgrade are lower than in the case of an upgrade. This stronger reaction to downgrades relative to upgrades is also noted by Womack (1996) and Francis and Soffer (1997). An explanation for this finding may be that downgrades are less frequent.10 It may also be that analysts know that releasing a negative recommendation is potentially more costly, due to possible losses in advisory fees. In contrast, upgrades incorporate fewer potential costs for the analyst. An incorrect upgrade has less severe effects for the analyst, since the chance that other analysts also issue an upgrade is bigger than in the case of a downgrade.
10 The data used in this research confirm this. Table II shows that in the non-crisis period there were a total of 887 upgrades, versus
24
Table V. Results of the Mann–Whitney U-Test. The median of the stock returns of the selected banks on day of a upgrade or downgrade during the non-financial crisis period are compared with the stock returns with the same criteria during the financial crisis period. Both comparisons are also made over a 3-day event window and made day t for banks facing a recommendation on day t during the financial crisis period. Moreover, the same comparison is made but with the 3-day event window. In addition, the same procedure is carried out for the adjusted selected banks’ daily stock returns. For both comparisons, the recommendations are subdivided into downgrades and upgrades.
Type of
Recommendation Downgrade Upgrade Event Window 1-day 3-day 1-day 3-day Variable Adj. Adj. Adj. Adj. Median FCP -0.026 -0.016 -0.011 -0.009 0.017 0.008 0.007 0.005 NFCP -0.014 -0.007 -0.005 -0.004 0.010 0.004 0.004 0.002
Z-Statistic 5.769 2.327 4.134 3.324 3.175 1.832 3.176 2.925
25
B. Regression Analysis
Table VI on the next page shows that the signs with regard to upgrades and downgrades in both the financial crisis and non-financial crisis periods are in line with our
expectations and have statistically significant p-values. Liu, Smith, and Syed (1990) and Stickel (1995) also find highly significant positive and negative influences for an upgrade and downgrade, respectively, on the stock return on day t. The coefficients of a
26
Table VI. Regression Estimates and the Wald Test With Unadjusted Returns. This table presents the expected signs, estimated coefficients, standard errors, t-values, and p-values of the dummy variables that indicate an upgrade (UP) or downgrade (DOWN) multiplied by a dummy indicating the non-financial crisis ( ) and financial crisis ( ) period and the four macroeconomic variables (FX, MR, LTI, and STI). Moreover, the χ2 values of the Wald test, that which test the nul-hypothesis that the coefficient of the dummy variables which indicate the non-financial crisis and financial crisis are equal. The χ2 values marked with an asterisk are significant at the 1% significance level and therefore we reject the nul-hypothesis. All these measures are given for both the 1-day- and 3-day event window and calculated were the dependent variable is the unadjusted selected banks’ stock returns.
1-day Event Window 3-day Event Window
Regression Estimates Wald Test Regression Estimates Wald Test
Variables Exp. Coeff. Std. t- p- χ2 Coeff. Std. t- p- χ2 Sign. Error Value Value Error Value Value
27
The signs of market returns and short-term interest rates are consistent with the literature. The long-term interest rate is not significant. Therefore, the relation between the daily differences of the long-term interest rate and the banks’ daily stock returns may be biased. Moreover, the exchange rates have a significant opposite sign and therefore requires additional elaboration. We expect that an appreciation of the dollar against a basket of the most important currencies results in a positive effect for the banks’ stock returns. The imported data use the basket of important currencies as a base currency. Therefore the dollar appreciates when the value of the basket diminishes in terms of dollars. Hence, if the outcome of the regression is in line with our expectations, we expect a significant negative relation. However, the regression estimates the opposite relation. This outcome may be the result that the selected banks possess relatively large amounts of foreign currencies. The total value of cash flows generated in a foreign currency may also be more extensive than that generated in the home country. Both regressions have an R2 value of 0.32, which indicates that there may be evidence of omitted variables. However, the high and significant F-statistic shows that the
independent variables present in the regression all have a non-zero slope coefficient. In addition, the Durbin–Watson statistic of 2.1 suggests that the residuals in the estimated regressions are not subject to autocorrelation.
C. Tests for Robustness
We decide to check the model’s robustness only over the 1-day event window. Since the results of both the Mann–Whitney U-test and the previous regressions suggest that the effect of an analysts’ recommendation on the selected banks’ daily stock returns is strongest on the day of initiation, we make two adjustments to check the model for robustness. By the first adjustment, we subtract, before regressing, from all selected banks’ stock returns ( their average stock return measured over 200 days (as
28
stock returns of our selected banks. The highly significant F-statistic confirms that this influence differs from zero.
Table VII. Regression Estimates and Wald Test with Adjusted Returns. This table presents the expected signs, estimated coefficients, standard errors, t-values, and p-values of the dummy variables that indicate an upgrade (UP) or downgrade (DOWN) multiplied by a dummy indicating the non-financial crisis ( ) and financial crisis ( ) period and the four macroeconomic variables (FX, MR, LTI, and STI). Moreover, the χ2 values of the Wald test, that test the nul-hypothesis that the coefficient of the dummy variables which indicate the non-financial crisis and financial crisis are equal. The χ2 values marked with an asterisk are significant at the 1% significance level and therefore we reject the nul-hypothesis. All these measures are given for both the 1-day event window and are calculated were the dependent variable is the adjusted selected banks’ stock returns.
Regression Outcomes Wald Test
Variables Exp. Coeff. Std. t- p- χ2
Sign Error Value Value
29
Even though the daily stock returns of our selected banks are adjusted for their firm- and period-specific risks, the financial crisis dummy variables still have significant additional explanatory power.
30
Table VIII. Regression Estimates and the Wald Test With Unadjusted and Adjusted Returns. This table presents the expected signs, estimated coefficients, standard errors, t-values, and p-values of the dummy variables that indicate an upgrade (UP) or
downgrade (DOWN) multiplied by a dummy indicating the non-financial crisis ( ) and financial crisis ( ) period and the four macroeconomic variables (FX, MR, LTI, and STI). Moreover, the χ2 values of the Wald test, that tests the nul-hypothesis that the coefficient of the dummy variables which indicate the non-financial crisis and financial crisis are equal. The χ2 values marked with an asterisk are significant at the 1% significance level and therefore we reject the nul-hypothesis. All these measures are calculated for the 1-day event window were the dependent variable is the unadjusted- and the adjusted selected banks’ stock returns.
Unadjusted Returns Adjusted Returns
Regression Estimates Wald Test Regression Estimates Wald Test
Variables Exp. Coeff. Std. t- p- χ2 Coeff. Std. t- p- χ2 Sign. Error Value Value Error Value Value
FINANCIAL CRISIS AND INVESTORS’ BEHAVIOR
V.Conclusion
This research collects 2,421 analysts’ recommendations for 64 banks listed on the NYSE over the period January 2000 to October 2010 on Yahoo! Finance. These recommendations consist of 1,338 upgrades and 1,083 downgrades. Based on the VIX, the period of focus is divided into a non–financial crisis period (January 2000 to July 2007) and a financial crisis period (July 2007 to October 2010). We first find by using Mann–Whitney U-Test that both unadjusted and adjusted NYSE listed banks’ daily stock returns on the day of an upgrade/downgrade have significantly higher/lower values during the financial crisis. Including the two surrounding days does not reinforce the effect. Second, we regress the NYSE listed banks’ daily unadjusted- and mean adjusted returns against four sets of dummy variables (which are combinations of non–financial crisis/financial crisis dummies and downgrade/upgrade dummies) and four macroeconomic variables (foreign exchange rates, market return, and long- and short-term interest rates) that are cited as important determinants of US banks’ stock returns. With the unadjusted NYSE listed banks’ daily stock returns, we find that the financial crisis dummies have significant additional explanatory power for both event windows. However, adding the two surrounding days of the initiation of a
recommendation does not enhance the impact of the financial crisis on the influence of analysts’ recommendations on NYSE listed banks’ daily stock returns. Therefore, this event window is deleted in the subsequent analysis. Moreover, after the mean adjustment of the NYSE listed banks’ daily stock returns, the financial crisis dummy still has significant additional explanatory power. We prove that these results are robust to a change in the definition of the non–financial crisis period.
We conclude that the financial crisis significantly changes the influence of analysts’
recommendations on US banks’ daily stock returns. This means that an upgrade/downgrade during the financial crisis has more significant positive/negative influence on US banks’ daily unadjusted and mean adjusted stock returns than during a non–financial crisis period.
32
33 References
Allen, F., and Gale, D., 1997, Financial markets, intermediaries, and intertemporal smoothing,
Journal of Political Economy 105, 523–546.
Allen, L., and Jagtiani, J., 1989, Risk and market segmentation in financial intermediaries’ returns, Journal of Financial Services Research 13, 159–173.
Baltagi, B.H., 2005, Econometric Analysis of Panel Data, 3rd ed. (John Wiley and Sons, Chichester).
Banz, R.W., 1981, The relationship between return and market value of common stocks,
Journal of Financial Economics 9, 3–18.
Barber, B., Reuven, L., McNichols, M., and Trueman, B., 2001, Can investors profit from the prophets? Security analyst recommendations and stock returns, Journal of Finance 56, 531– 563.
Beaver, W.H., 2002, Perspectives on recent capital market research, Accounting Review 77, 453–474.
Beneish, M., 1991, Stock prices and the dissemination of analysts’ recommendations, Journal
of Business 64, 393–416.
Benink, H.A., and Wolff, C.C.P., 2000, Survey data and the interest rate sensitivity of US bank stock returns, Economic Notes 29, 201–213.
Bidwell, C.M., 1977, How good is institutional brokerage research? Journal of Portfolio
Management 3, 26–31.
Blanchard, O., 2009, (Nearly) Nothing to fear but fear itself. The Economist 29, 72.
34
Brown, G.W., 2001, Managing foreign exchange risk with derivatives, Journal of Financial
Economics 60, 401–448.
Brown, L.D., and Rozeff, M.S., 1978, The superiority of analysts’ forecasts as measures of expectations: Evidence from earnings, Journal of Finance 33, 1–16.
Brown, S.J., and Warner, J.B., 1985, Using daily stock returns, Journal of Financial
Economics 14, 3–31.
Calomiris, C.W., 1995, Financial fragility: Issues and policy implications, Journal of
Financial Services Research 9, 241–257.
Chamberlain, S., Howe, J.S., and Popper, H., 1997, The exchange rate exposure of U.S. and Japanese banking institutions, Journal of Banking and Finance 21, 871–892.
Choi, J.J., and Elyasiani, E., 1997, Derivative exposure and the interest rate and exchange rate risks of U.S. banks, Journal of Financial Services Research 12, 267–286.
Choi, J.J., Elyasiani, E., and Kopecky, K.J., 1992, The sensitivity of bank stock returns to market, interest and exchange rate risks, Journal of Banking and Finance 16, 983–1004.
Darrough, M.N., and Russell, T., 2002, A positive model of earnings forecasts: Top down vs. bottom up, Journal of Business 75, 127–152.
De Bondt, W.F.M., and Forbes, W.P., 1999, Herding in analyst earnings forecast: Evidence from the United Kingdom, European Financial Management 5, 143–163.
Demirguç-Kunt, A., and Huizinga, H., 1999, Determinants of commercial bank interest margins and profitability: Some international evidence, World Bank Economic Review 13, 379–408.
35
Elyasiani, E., and Mansur, I., 2002, Bank stock return sensitivities to the long-term and short-term interest rates: A multivariate GARCH approach, Managerial Finance 30, 32–55.
Fama, E.F., 1970, Efficient capital markets: A review of theory and empirical work, Journal
of Finance 25, 383–417.
Fama, E.F., and French, K.R., 1992, The cross-section of expected stock returns, Journal of
Finance 47, 427–465.
Flannery, M.J., 1981, Market interest rates and commercial bank profitability: An empirical investigation, Journal of Finance 36, 1085–1101.
Francis, J., and Soffer, L., 1997, The relative informativeness of analysts’ stock
recommendations and earnings forecast revisions, Journal of Accounting Research 35, 193– 211.
Goldsmith, R.W., 1969, Financial Structure and Development (Yale University Press: New Haven, CT).
Grossman, S.J., and Stiglitz, J.E., 1980, On the impossibility of internationally efficient markets, American Economic Review 70, 393–408.
Hawkins, E., Chamberlin, S., and Daniel, W., 1984, Earnings expectations and security prices,
Financial Analysts Journal 40, 24–74.
Hicks, J., 1969, A Theory of Economic History (Clarendon Press, Oxford).
Jegadeesh, N., and Kim, W., 2006, Value of analyst recommendations: International evidence,
Journal of Financial Markets 9, 274–309.
36
Koutmos, G., and Martin, A.D., 2003, First- and second-moment exchange rate exposure: Evidence from U.S. stock returns, Financial Review 38, 455–471.
Levine, R., 1997, Financial development and economic growth: Views and agenda, Journal of
Economic Literature 35, 688–726.
Liu, P., Smith, S.D., and Syed, A.A., 1990, Stock price reactions to the Wall Street Journal’s securities recommendations, Journal of Financial and Quantitative Analysis 25, 399–410.
Malmendier, U., and Shanthikumar, D., 2007, Are small investors naïve about incentives?
Journal of Financial Economics 85, 457–489.
Mankiw, N.G., and Summers, N.H., 1984, Do long-term interest rates overreact to short-term interest rates? Brooking Papers on Economic Activity 1, 223–247.
Michaely, R., and Womack, K.L., 1999, Conflict of interest and the credibility of underwriter analyst recommendations, Review of Financial Studies 12, 653–686.
Oppenheimer, H., and Sun, Z., 2009, Institutional trading, analyst recommendations, and stock performance, Unpublished Manuscript, College of Business Administration, University of Rhode Island.
Prendergast, C., and Stole, L., 1996, Impetuous youngsters and jaded old-timers: Acquiring a reputation for learning, Journal of Political Economy 105, 1105–1134.
Saunders, A., and Yourougou, P., 1990, Are banks special? The separation of banking from commerce and interest rate risk, Journal of Economics and Business 45, 171–182.
Scharfstein, D.S., and Stein, J.C., 1990, Herd behavior and investment, American Economic
37
Stickel, S.E., 1995, The anatomy of the performance of buy and sell recommendations,
Financial Analysts Journal 51, 25–39.
Tai, C., 2000, Time-varying market, interest rate, and exchange rate risk premia in the US commercial bank stock returns, Journal of Multinational Financial Management 10, 397–420.
Trueman, B., 1994, Analyst forecast and herd behavior, Review of Financial Studies 7, 97– 124.
Van den Heuvel, S.J., 2001, The bank capital channel of monetary policy, mimeo, University of Pennsylvania.
Wei, K.D., Brown, N.C., and Wermers, R., 2007, Analyst recommendations, mutual fund herding, and overreaction in stock prices, Working paper, University of Maryland.
Wetmore, J.L., and Brick, J.R., 1994, Commercial bank risk: Market, interest rate, and foreign exchange, Journal of Financial Research 17, 585–596.
Womack, K.L., 1996, Do brokerage analysts’ recommendations have investment value?
