Do changes in consensus analyst earnings forecasts that disagree with contemporaneous stock returns signal exploitable mispricings? In their November 2014 paper entitled “To Follow or Not to Follow – An Analysis of the Profitability of Portfolio Strategies Based on Analyst Consensus EPS Forecasts”, Rainer Baule and Hannes Wilke investigate the power of a variable that relates consensus earnings forecast momentum to stock price momentum to predict stock returns. Specifically, the variable is the ratio of (1+change in consensus earnings forecast) to (1+stock return) over the last six months. Their consensus earnings forecast metric is a rolling average of consensus estimates for the current and next years weighted according to proximity of the current-year forecast to the end of the firm’s fiscal year (for example, three months before the end of the fiscal year, the rolling 12-month metric is 3/12 of the forecast for the current year plus 9/12 of the forecast for next year). They measure predictive power via a portfolio that is each month long (short) the fifth of stocks with the highest (lowest) last-month variable values. They evaluate both raw excess portfolio performance (relative to the risk-free rate) and four-factor portfolio alpha (adjusting for market, size, book-to-market and momentum factors). They limit the stock universe to the widely covered and very liquid components of the S&P 100 Index. Using monthly analyst consensus earnings forecasts and total returns for S&P 100 stocks during February 1978 through December 2013 (a total of 278 stocks listed for at least one month), they find that:
- A hedge portfolio that is long (short) the fifth of stocks with highest (lowest) last-month earnings forecast-stock return momentum ratio generates an average gross monthly excess return of 0.45%.
- The average gross monthly excess return for the long (short) side of the portfolio is 0.89% (0.44%).
- Average gross monthly excess returns decrease systematically across variable quintiles, indicating reliability of predictive power.
- The hedge portfolio generates an average gross monthly four-factor alpha of 0.70% (about 8.8% annualized).
- The average gross monthly alpha for the long (short) side of the portfolio is 0.43% (0.27%).
- Average gross monthly alphas decrease systematically across variable quintiles, again indicating reliability of predictive power.
- The hedge portfolio clearly outperforms one based on consensus earnings forecast momentum alone.
- Results are generally similar but weaker for a more complex alternative variable that facilitates handling of negative and very small consensus earnings forecasts.
In summary, evidence suggests that investors may be able to exploit the forecasting ability of expert analysts by focusing on level of divergence between consensus earnings forecast momentum and corresponding stock price momentum.
Cautions regarding findings include:
- Performance results are gross, not net. As noted in the paper, “profits generated by the described strategies might be substantially lowered by transactions costs.” Transaction costs in this case consists of monthly portfolio reformation frictions and costs of shorting half the portfolio.
- Any costs of timely consensus earnings forecast tracking and monthly data processing would also lower performance.
- The strategy may incorporate data snooping bias in selection of the momentum measurement interval and the consensus earnings forecast weighting scheme.
