"Sell-Side" Security Analysis

Thomas E. Berghage

     Given the almost equal probability of a gain or loss on a random draw from the stock market that we identified in Chapter 2, and the inability of traditional financial measures to discriminate between the two groups that we demonstrated in Chapter 3, the question becomes, “how well do human analysts do at overcoming these problems?”  Can financial analysts add value over and above a random draw from the market or a simple market index?  To answer these questions we embarked upon a major study of analysts published research reports and recommendations, and their subsequent performance.

     The analysts whose work we evaluated were by and large using traditional methods of financial analysis and were being paid handsomely to be right.  They were and are Wall Street’s best and brightest.  They have, for the most part, learned their trade at the best universities in the country and are completely knowledgeable regarding current financial analysis dogma.  In most cases they carry the designation of Chartered Financial Analyst and are members of the professional Association For Investment Management And Research (AIMR).

     The reports we evaluated came from just about all of the major Wall Street firms and were listed in Barron’s Magazine during the period 1992 through 1999.  The listing in Barron’s generally included a few cryptic remarks about the company and the analyst’s recommendation.  Most, but not all of these reports, have positive recommendations.  It is not unreasonable to assume that there is some relationship between the companies being reported on and the firms that the analysts work for, but the presence or absence of that relationship is never indicated in Barron’s.  It would be useful if Barron’s would list when a brokerage firm has an investment banking or market making relationship with the companies covered by their research reports, but that might be a bit too much to hope for.

          Many of the research reports include estimates and forecasts of corporate revenue and earnings growth, and often times there is an indication by the analyst of what he/she thinks the change in revenue and earnings will mean for investors. This is despite the fact that, as documented in Chapter 3, there is little or no meaningful/useful relationship between revenue and earnings growth, and stock performance. 

Evaluation Study

     The comprehensive evaluation study undertaken here is to our knowledge the most complete and intensive study of analyst performance ever attempted.  It covers the analysis work and published reports produced by over 200 firms and published in Barron’s over 8 years.  It represents 5,000 recommendations by a number of different analysts working in a host of different research environments.  For the most part the reports were done by some of the biggest and most prestigious firms on Wall Street.  I am sure the evaluation done at these firms covered more companies than those published in Barron’s, but we limited our evaluation to those reports that the firms provided to the general public though the “Research Reports” section of Barron’s Magazine. 

     As one would expect, most of these reports were positive and in most cases were “Buy” recommendations.  Few firms spend the time, effort and money to write and publish a research report unless they are going to recommend purchase of the company’s stock.  You have to understand that these reports are generally not done to inform the public, but are instead “marketing” pieces to be used by retail brokers, investment banking departments and the companies themselves.  In fact, in some cases the company being reported on has paid some or all of the cost of producing the report.  Now there are exceptions to this rule, but lets not kid our-selves, brokerage firms and investment banking organizations are in business to make money and they do so by selling stocks.

     It is interesting and informative to look at the distribution of the stocks in the study that were recommended as “Buys” or “Strong Buy’s.”  We took this data and plotted in Figure 4-1 a frequency distribution for various levels of return.  The shape of the distribution looks very similar to that of the total market distribution shown in Figure 2-2 with one major difference.  There is a shift of the mean value to the left indicating that the analyst’s “Buy” recommendations did worse than a random selection from the market as a whole.

Figure 4-1

Distribution Of The “Buy” and “Strong Buy”

Recommendations Made By Analysts (Years 1992-1999)

     There is a 12-point difference between the mean of this distribution and the one for the market as a whole shown in Figure 2-1, and remember, these are the best they had to offer, their “Buy” recommendations.  One would have expected there to be a mean shift between these two distributions, but it should have been in the other direction.  The “Buy” recommendations should have had a significantly higher mean return.

Recommendations

     The brokerage/research firms use different terminology and classification systems for their recommendations.  Among the terms used are: Strong Buy, Buy, Purchase, Accumulate, Add, Long-Term Buy, Positive, Outperform, Neutral, Core Holding, Market Perform, Hold, Negative, Sell, and Reduce.

To reduce the complexity of the analysis task we have combined a number of the above terms into a seven category rating scale.  The seven categories and the terms they include are: 

Strong Buy (strong buy)

Buy (buy, purchase)

Accumulate (accumulate, add, long-term buy)

Positive (positive, outperform)

Neutral (neutral, core holding, market perform, hold)

Negative (negative)

Sell (sell, reduce)

     As indicated earlier the distribution of reports among these categories is far from equal.  Analysts generally use the top four categories for their recommendations.  In fact, 85.2% of the 5,000 reports evaluated fell into these four categories.  Some of the firms tried to hedge their bets by suggesting that their “Buy” recommendations were for “High Risk” portfolios, but the fact remains that most of the reports written are positive in nature.  Less than 3% of the reports were negative and only about 1.3% actually recommended a “Sell”.  The number of reports and there relative proportion of the total are shown in Table 4-1. 

Table 4-1

Recommendations By Wall Street Analysts

     Now you will recall from Chapter 2 that a random selection from the Ford Investor Service, Inc. data base during the five-year period 1996 through 2000 and held for one year had a 48.4% chance of losing money and the mean twelve-month return for the random selections was 13.33%.  Now it is assumed that the analysts that produced the reports published in Barron’s and that were the subject of this study were attempting to produce returns better than that of a random selection or a simple market index.  Working for some of the biggest and best firms on Wall Street and having almost unlimited access to any information they needed, one would think that they should have no trouble beating a market index or a random selection.  The facts indicate, however, that the probability of losing money by using the analysts “Buy” and “Strong Buy” recommendations was higher than either a random selection or the use of a simple market index.  Investors that acted on these “Buy” recommendations and held them for a year had a 60.3% chance of losing money on their investment and the mean return on their investment was –2.22%. 

     This is unbelievable!  How could Wall Street’s best and brightest produce such horrible returns?  If these were their best recommendations, I would hate to see their worst.  The reasons for such poor performance are complex and I will attempt to deal with some of them in Chapter’s 8 and 9.  In the mean time, let’s look at the rest of the results of this study, and explore them a little more in depth.  First, let’s check to see if there has been any change or trend in the use of these recommendation categories over time.  It would be important to know if analysts were becoming any more objective in their stock recommendations.  To do this we have tabulated and plotted the recommendation categories used each year from 1992 through 1999.  We have combined a few categories, “strong buy” and “buy”, “Accumulate” and “Positive”, and “negative” and “sell” to make the graph a little less cluttered and more understandable.

Figure 4-2

Recommendation Categories Used

By Analyst, 1992-1999

     There appears to be little significant change in the relative use of recommendation categories.  The great majority of recommendations are still “strong buy” and “buy.”   Many will suggest to you that due the recent turmoil in the markets and the focus on analyst’s performance this will change.  That in the future we will see more “sell” recommendations.  This might true on a temporary basis, but there is no long-term incentive for analysts to write negative reports.  The investment banking community attracts new corporate clients by promising that they will provide research coverage.  Although they do not promise that the research reports will be favorable, they are very hesitant to publish a poor report for fear of losing the corporate client. The “sell” side corporations and analysts are just human and they are not going to bit the hand that is feeding them.

     Now that we have documented the obvious fact that analysts that work for brokerage companies that sell stock tend to recommend the purchase of the company’s stock they follow, lets look at how good their recommendations are.  To accomplish this we tabulated the recommendations and calculated the probability of loss and the mean returns for each of the seven recommendation categories (Table 4-2).

Table 4-2

Probability Of Loss And Mean Return

For Analysts Recommendations

     The first thing that jumps out at you from Table 4-2 is that there appears to be no substantial difference in the probability of losing money among the seven recommendation categories.  This observation has been verified by statistical test and indicates that as a profession the analysts recommendations are ineffectual, and that regardless of what the analysts indicate as appropriate investment action the probability of loss is about the same.  If an investor acts on the reports in the “Research Reports” section of Barron’s and holds the position for one year there is, on average, a 60% chance of losing money regardless of what the Analysts have recommended. The investor may be better off just randomly selecting a stock from the Wall Street Journal or buying an S&P500 Index Fund. 

     Some would question our use of a one-year time horizon and would suggest that most of these research reports are focused on traders and that a shorter time frame would be more appropriate.  To respond to these concerns we looked at the returns for three different time frames: 3 months, 6 months, and 12 months.  The results are shown in Table 4-3.

Table 4-3

Mean Returns For Different Time Horizons

     Regardless of the time frame used the mean returns were all negative and significantly below the average return of the S&P500 over the same time periods. Acting on the analyst’s recommendations was equally disastrous for both traders and investors.  There is just absolutely no value to be derived from reading this section of Barron’s or acting on the advice given. 

     Some would suggest that this analysis is a bit harsh and that it is lumping all of the analysts and all of the firms sponsoring the research into the same basket.  Kind of like, one size fits all.  To respond to this criticism we broke out the thirty-three individual firms that published the greatest number of reports in Barron’s over the eight year time period of this study.  First we looked at the distribution of recommendation categories used by these thirty-three firms.  The thought being, that some firms might be more objective and even handed in their use of recommendations.  Well put that thought away.  It appears that they all play the same game.  They only publish positive reports and positive recommendations in Barron’s.  There was one exception to this rule, and that was for the firm, Swiss American.  Although forty-seven percent of their reports were still “buy” and “strong buy” recommendations, they balanced these recommendations with 39.2% “neutral” and 13.7% “sell” recommendations.  This is the most evenhanded use of the recommendation categories we found.  

     Twelve of the firms were especially aggressive in their use of the “buy” and “strong buy” categories.  We have highlighted those firms that had 70% or more of their reports falling into these two categories.  The use of these categories in-and-of-itself is not bad if the firm can deliver and provide above average returns to the investor’s that act on their advice.         

Table 4-4

Recommendations Used By Various

Research Organizations

     In Table 4-5 we have tabulated and displayed for the same thirty-three firms the probability of loss on an investment held over three different investment time horizons. We have highlighted those firms that had a 70% or greater probability of loss on their recommendations.  In other words, if you acted on the “buy” recommendations coming out of these firms, you had less than a 30% chance of turning a profit over the three time horizons evaluated.  Those are not very good odds, and the firms that are highlighted need to take a long hard look at their analysis process or at least what they are releasing to the public.

Four firms, Dillon Read, George K. Baum, Needham, and Rauscher Pierce, are highlighted in both Table 4-4 and 4-5 indicating that they make a lot of “buy” recommendations and that their recommendations have produced significantly higher probabilities of loss. You have to wonder about the corporate cultures at these firms and the objectivity of their research efforts.

Table 4-5

Probability Of Loss For “Buy” Recommendations

By Various Research Organizaitons

     The results of this study have been very disturbing.  The heavy use of “buy” recommendations by analysts and their research firms without the investment performance to support them is very distressing.  The total inability of analysts to discriminate among good and poor performing stocks with their recommendation categories is a real indictment of their analysis techniques.  We are assuming that the analysts are honest and well intending, but just lack the analysis procedures or technology to make the needed discriminations.  As we demonstrated in Chapter 3, the traditional financial analysis methods fail to discriminate and it is apparent by this study that the analysts employing these analysis techniques have failed to overcome the shortcomings of their analysis tools. 

     There are some that would suggest that we are a bit naive in our first assumption regarding the honesty and well intentions of analysts.  They would suggest that most “sell-side” analysts are biased and lack objectivity; that their entire analysis process is designed to support the corporate goal of selling stocks.  We are certain that this “sell-side” bias exists, but poor analysis performance is not confined to the “sell-side” of the Street.  As we will see in the next chapter, “buy-side” analysts, those that are building portfolios for institutions, mutual funds, and individual investors, have almost an equally poor performance record.

Return To Home Page

Return To Article Index