Estimate Dispersion: Why the Same Consensus Average Can Hide Different Risks
Look beyond the consensus average with highs, lows, and standard deviation. Identify bullish and bearish assumptions without mistaking analyst disagreement for a probability range.
Estimate dispersion shows how far apart institutions' forecasts are. Even with the same average, a wider spread may signal greater disagreement about key assumptions. The spread itself should not be interpreted as a probability interval for actual earnings.
What highs, lows, and standard deviation reveal
The difference between the highest and lowest estimates makes the distance between the most optimistic and pessimistic forecasts easy to see. Standard deviation summarizes how far forecasts lie from the mean, but its value differs depending on whether a sample or population calculation is used. FactSet's official documentation describes statistics including the mean, median, high, low, and standard deviation. Rather than selecting just one statistic, read it alongside the contributor count and check whether a single extreme value widened the overall range.
Do not directly compare spreads on different scales
A company with larger earnings may naturally have a higher standard deviation than a smaller company. The spread relative to the mean can be a supplementary measure, but that ratio is also unstable when the mean is near zero or negative. For a company with both profit and loss forecasts, it is more useful to examine the proportion of institutions expecting a move between profit and loss and their assumptions than a single average. Also distinguish dispersion caused by stale estimates from disagreement about information available at the same point in time.
Translate disagreement into business scenarios
Identifying whether forecasts differ because of shipment volumes, margins, or exchange rates tells you what to check at the release. Institutional forecasts are not independent random samples and share information, so standard deviation alone cannot readily produce a confidence interval. An event missed by every institution can put actual results outside the range. Do not treat optimistic and pessimistic estimates as firm forecast boundaries; record the conditions under which each assumption would fail.
Recording whether actual results fell within the forecast range is a useful review exercise. But a few successful instances do not establish a probability of success for the next release, because the business environment and contributor group may change. It is more productive to track which assumption proved most wrong and whether the next consensus incorporates that information.
Worked example
Hypothetical example: Group A forecasts KRW 9 billion, KRW 10 billion, and KRW 11 billion, while group B forecasts KRW 5 billion, KRW 10 billion, and KRW 15 billion. Both averages are KRW 10 billion. Using the population method, their standard deviations are about KRW 816 million and KRW 4.082 billion, respectively. Group B's wider range indicates greater disagreement, not a guarantee that actual profit will fall between KRW 5 billion and KRW 15 billion. If B's highest forecast assumes a new facility starts operating early, that schedule is the first item to verify.
What to check
- Have you checked the standard deviation method and units?
- Have you examined outliers and stale estimates separately?
- Is a mean near zero distorting relative ratios?
- Have you avoided mistaking the range for a probability interval?
Sources and primary materials
This is educational material about research methods. For an actual company decision, verify the filings and primary sources for the relevant period.