Estimate revisions are changes in analysts' forecasts of a company's earnings. Stocks whose forecasts are being revised up have tended to outperform those being revised down, which makes revisions a widely used signal in quantitative models.
How the signal is built
Revision ratio = (number of upward revisions - number of downward revisions) / total estimates over the past one to three months. Alternatives use the percentage change in the average (consensus) forecast.
An illustrative example
| Company A | Company B | |
|---|---|---|
| Analysts covering | 10 | 10 |
| Raised estimates | 6 | 1 |
| Cut estimates | 2 | 7 |
| Revision ratio | (6 - 2) / 10 = +0.4 | (1 - 7) / 10 = -0.6 |
The numbers are made up. A revisions strategy favours A and avoids B.
Why it may work
- Analysts revise gradually, so a revision tends to be followed by more in the same direction.
- Investors may under-react to the information the revision carries. See post-earnings-announcement drift.
- Research such as Chan, Jegadeesh and Lakonishok (1996) found revisions predicted returns beyond price momentum.
Cautions
- Coverage. Smaller companies may have few or no analysts, which limits the signal. Screens based on analyst numbers also change a portfolio's size tilt.
- Timing. Revisions must be dated as of when they were published. See point-in-time data.
- Short half-life. The signal decays within months and needs fairly frequent updating. See signal decay and turnover.
- Herding. Analysts follow each other, so revisions are correlated.
Frequently asked questions
What are estimate revisions?
They are changes in analysts' forecasts of a company's earnings, usually tracked over the past one to three months.
How is a revision signal calculated?
One common way is the number of upward revisions minus downward revisions, divided by the total number of estimates.
Do estimate revisions predict returns?
Research has found that stocks with upward revisions tend to outperform, though the effect decays quickly and depends on analyst coverage.
What is the difference between revisions and earnings surprise?
A surprise compares reported earnings with what was expected. A revision is a change in the forecast itself.