Do Brokerage Analyst's Recommendations Have Investment Value?
Kent Womack
Journal of Finance March 1996
Executive Summary
Short answer: yes, analysts' recommendations do appear to have value. The stock tends to move in the direction of the change in recommendation both in the short run (up 3% for buy recommendations and falling by 4,7% for a sell) and in the long run (buys up 2.4% in the following month, sells down 9.1% in the following 6-months. This implies that recommendations embody valuable information for which a brokerage firm should be compensated.
With the millions of dollars spend on investment analysts, it is necessary to ask whether the analysts' recommendations have any value. This is not a new question. It has been investigated almost constantly since 1933 (Cowles). As might be guessed the answer is unclear.
"This article provides new evidence on stock price formation and on the ability of analysts to predict or influence stock prices." The basic findings are that the stock price does rise by about 3% in the three day event window around the announcement of a "buy recommendation" and falls by 4.7% in the same three-day window on the announcement of a "sell recommendation."
What is more surprising is that the author finds a drift in the direction of the recommendation. This long-run drift is statistically and economically significant. For a sell recommendation, the stock tends to be a size-adjusted return of -9.1% over the next six-month window. For a buy recommendation, there is a positive 2.4% drift that is only significant for the first month.
Data: Recommendations (and dates of the recommendations) are from First Call which is a collection of daily commentary. This data is sold at a "substantial" cost and hence is viewed as "quasi-private" information.
For the years 1989-1991 there are over 150,000 commentaries per year. This paper looks at a "small subset of [these] daily comments:" changes in stock recommendations by the 14 highest -ranking US brokerage research departments.
Womack looks only at those recommendations that either added or removed stocks to (from) the most attractive category and to (from) the least attractive category. "Only changes in recommendations, not reiterations of previous opinions (which occurs frequently) are included in the sample of 1573 recommendation changes made on 822 stocks. Some were removed if there was not Compustat and CRSP data and all foreign stocks were removed.
Sample statistics:
Not surprisingly the stocks followed by the analysts tend to be larger than average. "Only 1 percent of the collected recommendations are for stocks in the two smallest capitalization deciles."
Upgrades were much more common. "The ratio of buy to sell recommendations in this sample is about 7:1. Pratt (1993) notes that Zacks International Research estimates this ratio to be 10:1." Womack speculates that this may be due to conflicts of interest with the "brokerage firms presnt and potential investment banking relationships…and that top management [at the firms] may limit…the flow of information is an analyst issues unfavorable ratings." {Summarizer's note: there are many incidents of just such behavior). Finally, due to their relative scarcity, "sell recommendations are more visible…[and] an incorrect judgement on a sell recommendation is likely to be more costly for an analyst's reoutation."
Process: "Most additions to broker recommendation lists are deliberate, planned actions that have first been researched and proposed by an analyst, then reviewed and approved by an internal investment committee, and finally prepared for publication and dissemination. These decisions are rarely made in haste. Analysts usually have more latitude to remove firms from the recommended list without formal approval or extensive review when the reasons for recommending the stocks are no longer tenable."
Only 9% of buy recommendations and 14% of sell recommendations occur within one day of quarterly earnings announcements. Thus the announcements appear to be more price and model driven than information driven.
The average size-adjusted mean return is +3% for a upward change and -4.7% for a downward revision. (medians are +2% and -2.8%). Note the author tested industry adjusted and Fama-French 1993 model (using the value weighted market returns, returns to relative sized firms, and returns to relative P/E ratios). The results were qualitatively similar.
Using simulation techniques, the longer-term drift of the returns suggests that analysts have market-timing abilities. That is in the month following a buy recommendation, not only does the stock go up, but so too does the market as a whole. The results are less impressive for looking at specific industries where in fact the signs are reversed in the buy recommendation.
In conclusion Womack notes "The results are consistent with the expanded view of market efficiency suggested by Grossman and Stiglitz (1980): that there mist be returns to information search costs. These information search costs are often assumed to be zero when considering the efficient market hypothesis. The nontrivial magnitude of the returns reported here challenges the innocence of that assumption."
| Most Popular |
|
Sign up for Finance Professor's free newsletter! Note: We have been having difficulty with delivery via Bcentral (our main mail server). Therefore we have added two back-up mail providers for the newsletter. You can sign up through Topica or Yahoo. In all cases I will never sell, give, or trade your email address. Most recent newsletter (well at least when I remember to update it-if not check the archive) |
|
NEW!!!
Message boards for Jim's Current Students and former students
|
|
Academic Finance Conferences and Conventions |
| Park and Shop-My favorite grocery store, find out why |
| You can Find almost any book at
Amazon.com (P.S. While I am not trying to make money on this, covering
some of the costs would be nice. So if you could, when you buy a book from
Amazon,
click through here. Thanks!)
|
