A recommendation level screen would focus on buying the most recommended stocks in the market based on the broker consensus. This is to be contrasted with a screen that focuses on those stocks most recently upgraded - a ‘recommendation revision’ screen. Overall, it appears that the absolute level of published stock recommendations has little (or even detrimental) value to investors, but monitoring the changes in investment recommendations can prove more productive.
Broker recommendations as to whether clients should buy, sell or hold stocks are widely reported in the media. Despite this, the value of these recommendations is hotly debated. The issue is that broker recommendations have historically exhibited a strong upward bias. Researchers Jegadeesh, Kim, Krische and Lee found that the average analyst rating over the 1985 to 1999 period was close to a buy recommendation and sell or strong sell recommendation made up less than 5% of all recommendations. This is because most sell-side analysts work for brokerage houses which often have strong investment banking franchises, creating a potential conflict of interest when firms act as investment bankers to the companies their analysts cover. Following the debacle of the late 1990s, legislators have taken action to address this conflict. The US Global Analyst Research Settlement of April 2003, the EU’s Markets in Financials Instruments Directive of 2004 (MiFID) both targeted the fairness of research. Although this appears to have had some impact, recommendations still have remained relatively optimistic - research by Jegadeesh and Kim in 2006 indicated that analyst recommendations retain a significant element of favourable bias.
Definition of a Consensus Recommendation Level Screen
A somewhat simplistic screen based on pure recommendation levels might, say, buy the most favorably recommended quintile of stocks according to the consensus, and sell the least favorably recommended quintile of stocks.
Does that work?
The evidence is mixed but most research suggests that it doesn't, especially after transaction costs and assuming a more medium term investment horizon. Early work by Womack did find short term excess returns (3-5% over a three day period) as well as longer term post recommendation drift. However, Barber et al found that, while a trading strategy was profiltable before transaction cost, returns after taking into account costs were insignificant. Furthermore, a subsequent paper taking into account the disastrous years of 2000 and 2001 found that the stocks least favored by analysts earned an average annualized market-adjusted return of 13.4% whereas the stocks most highly recommended underperformed the market by 7.1%!
In a 2003 paper, “Analysing the Analysts”, Jegadeesh, Kim, Krische and Lee concluded that i) brokers tend to focus on glamour stocks (i.e. positive momentum, high growth, high volume, relatively expensive), ii) the level of the consensus recommendation adds value because these stocks already have favorable characteristics (i.e. positive momentum stocks) and iii) any excess returns are primarily attributable to changes in recommendations, not the actual recommendation level itself.
As discussed above, analysts tend to be positively biased due to well-known conflict of interest. Although reliable recommendations attract trading commissions and fees, buy recommendations are more likely to generate trading business than sell recommendations, given short-selling constraints. As the tech bubble showed, analysts face additional pressures to distort their recommendations if their brokerage is affiliated with an underwriter of the recommended firm. Management sometimes call up analysts to complain about ratings that are “too low”, and “freezes out” analysts who do not give positive recommendations. Analysts are also subject to the behavioral bias known as “herding” in the sense that they release seemingly correlated recommendations: during bull markets they release more buy recommendations than sell, and vice versa in bear markets. Furthermore, some of the recommendations that enter into the consensus can be fairly stale. Analysts often leave their recommendations unchanged for long periods of time and recommendations generally become less informative over time.
For that reason, a more preferable screening approach is to focus on changes in recommendation levels, which we will be discussing in more detail.
It might be possible to apply a filter to try to correct for the level of bias in the consensus recommendation. This was attempted by Balboa, Gmez-Sala, and Lpez-Espinosa who applied a correction method based on the ratio between buy and sell recommendations and the theoretical balanced situation in which buy and sell recommendations are equal in a given market. Their results showed a significantly higher return than for raw recommendations, although the assumption of a theoretically balanced situation seem open to some debate.
The other possible contrarian use for broker consensus recommendations amongst value investors is as a negative filter to screen out popular / discovered stocks, i.e. to exclude all buys and strong buys as being companies that are too well known by the market.
How can I run this Screen?
- Analyzing the Analysts:When do Recommendations Add Value?
- JP Morgan: The Search for Alpha
- Can Investors Profit From The Prophets? Security Analyst Recommendations And Stock Returns
- Reassessing the Returns To Analysts' Stock Recommendations
- Do brokerage analysts recommendations have investment value
- Excessive optimism and analyst recommendations
- The Value of Adjusting Bias in Recommendations: International Evidence
- Taking Biases out of Earnings Revisions
- After the Global Research Settlement -- Are Analysts Still Biased?
- Do Analyst Recommendations Yield Profitable Trading Strategies?