A momentum screen based on buying stocks with rising analyst earnings estimate revisions in light of empirical findings that stocks with their estimates revised often outperform the market over at least the next 12 months.
Earnings estimates are created by equity analysts in order to project the growth and profitability of a company on a quarterly and/or yearly basis. In 1969, Burton Fabricand was apparently the first to write about the strong link between estimate revisions and subsequent stock prices, showing that portfolios of stocks selected on the basis of large estimate revisions significantly beat the market over a three month holding period. Following this, in 1979 Leonard Zacks (founder of Zacks Investment Research) published an article entitled "EPS Forecasts - Accuracy is not enough." He found that "there was no correlation between forecast EPS growth and stock growth… consistent with efficient market concepts”. In contrast, “portfolios of companies whose consensus forecasts underestimated actual actual earnings growth outperformed the market on average”. He concluded that, to achieve excess returns, stock selection must be based on anticipating changes in the consensus expectation, rather than changes in actual earnings. More recently, researchers McKnight and Todd examined a portfolio of European stocks and found that the 20% with the highest net upward revisions outperformed the lowest 20% by over 16% a year. This is a fascinating study and it's well worth a read. This effect was persistent/robust and not concentrated to small stocks, stocks with low analyst coverage, or stocks with low book-to-market ratios.
How well does it work?
In addition to the results found by McKnight and Todd, the AAII 'Estimated Revenue up 5%' screen is the best-performing screen in its extensive database, up a massive 30% compounded since inception, compared with 2.3% for the S&P 500. Zacks' track record has also been impressive - buying a portfolio of Zacks #1 Rank stocks over the 17 years from 1988 apparently generated an average annual return of 33% versus the S&P 500 return of 12%.
Why does it work?
Researchers Barberis, Shleifer, and Vishny suggested that the kind of momentum is rooted in investors’ conservatism bias and the fact that investors do not update their beliefs adequately based on the strength and weight of new information. Interestingly, McKnight’s work found that the positive returns realized on a buy portfolio were large and persistent, whereas the sell portfolio generated a near zero return, i.e. the bad news was quickly priced in whereas the good news diffused slowly. This may be because conflicts of interest generated by investment banking relationships may encourage analysts to report overly optimistic earnings, and investors, aware of these biases, respond by being cynical and adopting a “wait-and-see” approach when it comes to good news.
Interestingly, returns for a trading strategy based on forecast revisions appear to exceed one based on recommendations or even changes in recommendations. One possible explanation is freshness. Barber et al found that nearly 50% of all recommendations are left unchanged when they are revisited, approximately 300 days after they were first made. Earnings forecasts, in contrast, are generally more responsive to short-term news events (being revised on a quarterly or even monthly basis). Interestingly, GLG Partners’ researchsuggested that this frequency of update might actually mean that a positive earnings estimate revision was a weaker signal than a buy or strong buy recommendation, but the evidence from McKnight's work appears to suggest otherwise.
An indicative Revised Earnings Estimates screen might include:
- Consensus EPS estimate for the current year greater than one month ago (or up by some threshold level, e.g. 5-10%), with at least one upward revision and no downgrade
- Consensus EPS estimate for next year above one month ago, with at least one upward revision and no downgrade AND/OR
- Revenue estimate for the current year greater than one month ago, with at least one upward revision and no downgrade
- Revenue estimate for next year above one month ago, with at least one upward revision and no downgrade.
Another factor worth considering is the degree of agreement in the so-called consensus estimate and the degree of historic earnings surprise – one research note, "Analyst Forecasts and Stock Returns” found that:
- Stocks with lower variance (dispersion) between analyst estimates perform better than ones with high variance (dispersion).
- Stocks with less error in forecasts (low standard deviation with earnings surprises) also had higher stock performance.
How can I run this Screen?
- The Zacks Rank: Harnessing the Power of Earnings Estimate Revisions
- Analyst Forecasts and the Cross Section of European Stock Returns
- EPS Forecasts - Accuracy is not enough
- Earnings Estimates - Good Stock Screen?
- AAII: Est Rev Up 5% Screen
- Whisper Numbers and Earnings Estimate Disparities
- Analysts’ Forecasts As Earnings Expectations
- The Rewards to Meeting or Beating Earnings Expectations
- How do analysts forecast earnings and what do they do with these forecasts?
- A Review of Research Related to Financial Analysts' Forecasts and Stock Recommendations