Bias in Analyst Recommendations The Curious Case of Bankrupt Companies
The Issue of Bias in Analyst Recommendations
Yet to date, no analytical investigation of the extent of the bias in analyst recommendations has been reported. This article assesses the degree of bias in analyst recommendations by evaluating the quality of the recommendations for a sample of bankrupt companies in the period prior to their bankruptcy filings. It is well known that, prior to the government investigation of analyst recommendations, the percentage of "sell" recommendations was extremely small. In particular, as of mid-2000, less than 1 percent of the 28,000 stock ratings provided by First Call/ Thomson Financial were "sell" recommendations. In contrast, approximately 74 percent of stock ratings were "buy" or "strong buy."
To the extent that analysts provide valuable advice to the investment community, one would have expected the percentage of "sell" recommendations for firms nearing a bankruptcy filing to be significantly greater than the percentage of "sell" recommendations for the general population of equity securities. Surprisingly, this study finds that the percentage of "sell" recommendations in the period prior to bankruptcy filings is virtually identical to the percentage of "sell" recommendations for the general population of stocks. In the few situations where analysts reported a "sell" recommendation, that recommendation was preceded by a significant drop in the stock's price. It is also interesting to note that in several cases, the bankruptcy was preceded by a "buy" or "accumulate" recommendation. There were also numerous cases where the verbal descriptions provided with the recommendations differed in tone and substance from the actual recommendations themselves. A "strong buy," for example, may have been accompanied by a description of the firm's lack of liquidity, as well as the fact that it is currently in technical default and only suitable for speculative accounts.
Analyst Recommendations: The Case of LTV Corp.
Consider the case of LTV Corp.: Stanford C. Bernstein gave LTV an "outperform" rating only 60 days prior to its bankruptcy, while indicating that it had only a 50 percent chance of surviving downside risk over the next year or two. They indicated that "business is lousy and the company is bleeding badly."2
Merrill Lynch, in one of its own analyst reports for LTV prior to bankruptcy, gave LTV a "buy" recommendation while their analysts simultaneously cut the earnings estimates for the company. Salomon Smith Barney placed a "buy" recommendation on LTV while also indicating that it had a disappointing and confusing most-recent quarter. Similarly, eight days before LTV's bankruptcy, Morgan Stanley gave the firm a "neutral" rating, yet it indicated that the fundamentals were the worst in 30 years.
The Study: Methodology and Data
To perform the analysis, we reviewed both analyst recommendations and information provided to investors in the available analyst reports for the 10 largest firms (by revenue) filing for bankruptcy each year from 1998 through 2000. These 30 firms were identified using the Bankruptcy Yearbook & Almanac. We reviewed analyst reports from the period two years prior to bankruptcy through the date of the bankruptcy filing. The analyst reports were obtained from the Investext Plus database maintained by Thomson Financial. Of the 30 companies identified, five were eliminated due to the unavailability of data, resulting in a sample of 25 firms.
Several analyses were performed. First, we analyzed the final recommendations prior to bankruptcy for each of the 25 firms in the sample during the year prior to their bankruptcy. Each of these final analyst reports was reviewed to determine the analyst's qualitative assessment of the firm's leverage, liquidity, operations and profitability. These categorizations were then compared and contrasted to the analyst's recommendations. The study then analyzed the relationship between stock price movements and changes in analyst recommendations during the two-year period prior to bankruptcy.
Results—The Year Prior to Bankruptcy
During the final year prior to bankruptcy, the sample included 99 recommendations issued by 44 different brokerage firms covering the 25 bankrupt firms in the sample. Only 2 percent of the 99 recommendations were identified as "sell" recommendations, with 27 percent rated as "attractive" through "strong buy." When "hold" and "underperform" ratings are added to the "sell" ratings, these still constitute only 16 percent of the total recommendations. It is surprising to note that for final recommendations in the year prior to bankruptcy, there were more than 10 times as many "buys" and "strong buys" as there were "sell" recommendations.
Given that our sample consists entirely of major bankrupt firms, it is particularly egregious that the proportion of "sell" recommendations is similar to that of the broad spectrum of primarily healthy companies. The proportion of "sell" recommendations for the general population of firms during a similar time period ranges from 1 to 4.7 percent. The fact that only 2 percent of the ratings are "sell" recommendations for a sample of soon-to-be-bankrupt companies clearly indicates the unreliability of the ratings.
The 99 analyst reports in the sample during the year prior to bankruptcy were each categorized by leverage, liquidity, operations and profitability. The ratings were based on the information included in the body of the analyst reports. With four categories per analyst report, there were a total of 396 ratings. Each rating indicated whether the researchers believed that the analysts expressed a "high degree of concern," "caution," "no caution" or were "positive" about the particular category (e.g., leverage). Given that only 2 percent of the recommendations were "sell," it is surprising that 39 percent of the ratings (156) expressed either "caution" (97) or a "high degree of concern" (59). Conversely, it is also troubling that 224 of 398 ratings (56 percent) exhibited "no caution" and that 16 of the ratings were "positive" in the year prior to bankruptcy.
Interestingly, 21 of the 99 final analyst reports in the sample mentioned bankruptcy in the body of the report. Yet again, this must be contrasted to the small percentage of brokerage firms issuing a "sell" recommendation. Of the brokerage firms covering more than a single bankrupt firm in the sample, the brokerage firms with the highest percentage of accuracy are CIBC Oppenheimer (three of four correct) and Raymond James (three of five correct). The poorest performers are Credit Suisse First Boston (one out of eight correct), Lehman Brothers (zero of three), Salomon Smith Barney (zero of three), Morgan Stanley Dean Witter (one of five), B.T. Alex Brown (zero of four) and Warburg Dillon Read (zero of three).
For those 21 brokerage firms predicting bankruptcy for a sample firm, the earliest was 13.7 months prior to bankruptcy; the most recent was just 10 days before the filing. The median time was 4.0 months and mean time was 4.9 months prior to bankruptcy. Unfortunately for the investor, by the time the median accurate bankruptcy prediction was made by the analysts (4.0 months prior to the filing), the typical stock had already lost approximately 70 percent of the value it had two years prior to bankruptcy.
Close to 50 percent of the stocks in the sample received at least one downgrade in the two years prior to bankruptcy. Interestingly, at the first downgrade for a stock by a brokerage firm, the median cumulative return from two years prior to bankruptcy was -47 percent. For stocks with a second downgrade by the same brokerage firm, the median cumulative return at the time of the second downgrade (i.e., following the first downgrade) was -82 percent. In other words, by the time of the first downgrade, nearly half of the equity value had already been lost, while four-fifths of the value was lost by the time of the second downgrade. Moreover, in the two days surrounding these two respective downgrades, the rate of return was -3.6 and -12.7 percent, both statistically significant declines. There were also a number of upgrades in analysts' recommendations in the two years prior to bankruptcy. The first upgrade for a stock by a brokerage firm in the two-year period prior to bankruptcy, although statistically significant, was less pronounced (+1.3 percent). The second upgrade (i.e., following an upgrade by the same brokerage firm) was not statistically significant. The median cumulative returns from two years prior to bankruptcy at the time of the first and second upgrades were -22.1 and -25.9 percent. In other words, even in the case of upgrades, the equity values had declined by the time of the upgrades. Unlike the large difference in equity values between the first and second downgrades, there was relatively little movement between the first and second upgrades. However, the stocks proceeded to decline in value following the second upgrade.
This analysis assesses the nature and extent of the bias in brokerage reports covering the largest bankruptcies from 1998 through 2000. It is troubling that only 2 percent of the analyst reports in the sample recommended that investors sell in the final analyst report prior to bankruptcy. Surprisingly, there is an approximate 10-fold increase (22 percent) in the number of brokerage reports accurately predicting the possibility of bankruptcy. Moreover, approximately 40 percent of the analysts suggested either a high degree of concern or caution in their recommendations. And 50 percent of the stocks in the bankrupt sample had at least one downgrade.
The data suggests that significant additional information is provided in the contents of the analyst reports than is revealed merely by the analyst recommendations. In particular, the information provided in the analyst reports is often more negative than the actual recommendation associated with the stock.
1 The authors are professors at Boston University's School of Management and managing directors of The Michel/Shaked Group, a firm providing expert witness services nationwide. The authors would like to thank Adriana Babisova, Chris Grimm and Reshmi Hegde for their assistance. Return to article