The consensus in the financial literature is that individual analyst recommendation changes do not significantly impact a stock’s price after the influence of company-specific news is considered. The authors present evidence that supports the opposite view. They find that 12 percent of recommendation changes are influential even after filtering out company effects.
The authors’ goal is to offer insight into the effect of analyst recommendations on stock prices as well as other issues related to analyst recommendation changes. Although they are not the first to delve into the asymmetrical effect of analyst recommendation changes, they are the first to apply a methodology that distills the stock price effect of individual analyst recommendation changes from the typical influence of a group of analysts. This approach represents the first step in permitting an assessment of an individual analyst’s ability to materially influence stock prices.
The authors gather analyst recommendation changes for 1994–2006 from Thomson Financial I/B/E/S and cross-reference the data with the First Call database to ensure the dates for analyst recommendations are accurate. This condition protects against improper alignment between any recommendation change and the associated stock price movement. The main sample comprises 154,134 recommendation changes. The authors cleanse the data by filtering out the influence of any analyst recommendation changes that contain fundamentally the same information as company-initiated news, such as earnings guidance releases, to isolate the stock price impact that is solely related to analyst motivations.
In the methodology, the authors consider analyst recommendations as changed when the current recommendations differ from the previous recommendations based on the following rating categories: “sell,” “underperform,” “hold,” “buy,” and “strong buy.” The current rating is subtracted from the previous rating, which results in a calculation that ranges from –4 to +4, with minuses being downgrades and pluses being upgrades. The impact of a recommendation change is cataloged as influential based on statistically abnormal results immediately after the recommendation change when compared with the previous three months for either of the following two criteria: (1) the stock’s cumulative buy-and-hold abnormal return (CAR) being in excess of the market return, contingent on it being in the same direction as the recommendation change, or (2) trading activity in the stock increasing.
The authors first conduct various statistical analyses on the sample of recommendation changes to set the stage for later granular analyses. Interestingly, the recommendation levels exhibit a strong positive sentiment with only a small portion of sell or underperform ratings. The transition matrix exhibits a strong central tendency around the hold rating in which more than 76 percent of recommendation changes are in the –1 to +1 categories, with –1 representing a one-level downgrade and +1 representing a one-level upgrade from a hold rating. Histograms of the CAR for one-level upgrades and downgrades provide initial support for the authors’ assertion that individual analyst’s ratings can be influential. Although a majority of the CARs fall between –0.5 and +0.5 percent, evidence exists that outliers can skew the mean CAR.
Next, the authors conduct a deeper analysis of the characteristics of influential and noninfluential recommendation changes and discover that approximately 12 percent are influential and that 25 percent of analysts have never provided an influential recommendation change. To control for possible duplicative effects of the analyst variables considered, the authors use a probit test to measure the incremental benefit associated with each variable. The greatest marginal impact is for the following variables: previously influential analyst, change in recommendation away from consensus, analyst Star power, and concurrent earnings and recommendation change forecast. There is also evidence that influential analyst recommendations increase the trading activity for a stock, which may or may not coincide with a more volatile CAR. The increased turnover reflects shareholders adjusting their positions to incorporate the new information.
The authors test the robustness of their results by altering various tenets of their methodology and find that their conclusions remain robust to the variations.