Forecasting Investor Behavior: An Empirical Analysis on Sell-Side Analyst Recommendations and Differences of Opinion

Abstract

Miller’s (1977) theory of differences of opinion states that stocks with high differences of opinion (controversial stocks) tend to be overvalued when there are binding short constraints. Generally, existing literature has supported Miller’s (1977) theory using different proxies for differences of opinion. Because studies on consensus recommendations of sell-side analysts have suggested that analyst recommendations correlate with investor behavior, my goal is to provide further evidence of this conclusion by testing Miller’s (1977) theory exclusively with analyst recommendation data. Using empirical data from 1994 to 2015, I investigate whether Miller’s (1977) theory will hold using the standard deviation of recommendations as a proxy for differences of opinion as well as utilizing consensus recommendations to identify short-constrained stocks. Stocks become short-constrained when short sale supply is low and short sale demand is high, which occurs when investors are pessimistic. I find that the most unfavorably recommended stocks have the most restrictive short constraints, indicating that analyst recommendations and investors share similar opinions on stocks. In line with existing literature, my results also support Miller’s (1977) theory by finding a significantly negative relationship between the standard deviation of recommendations and stock returns only within this subset of unfavorably recommended, short-constrained stocks, such that stocks with a low standard deviation of recommendations generate annual abnormal returns of 4.5 percent more than stocks with a high standard deviation of recommendations. Because both the mean and standard deviation of recommendations have a strong relationship with returns, a strategy of buying stocks with the lowest standard deviation and lowest mean (most favorable) of recommendations and shorting stocks with the highest standard deviation and the highest mean (least favorable) of recommendations yields annual abnormal returns of 9.8 percent. Overall, this paper provides further evidence that analyst recommendations correlate with investor behavior.