When forecasters disagree, the usual response is to average them and move on. The consensus figure gets quoted, the spread gets discarded, and a number that was assembled from wildly different views is presented as though it were a single view.
That discards the more informative part. The width of the range describes how much genuine uncertainty exists about an outcome, and there’s a substantial research literature on what that width does and doesn’t tell you.
Some of what it says is counterintuitive, and some of the most confident claims about it have weakened considerably over time.
Why the Range Matters More Than the Midpoint
Published market outlook for 2026 pieces vary widely, and the variation is the signal worth extracting.
A consensus figure loses several things that the underlying distribution contains:
- The width, which indicates how contested the outcome is
- The shape, since a cluster with two outliers differs from an even spread
- The direction of the outliers, which shows where the tail risks are seen
- Whether the range has widened or narrowed, which carries more information than its level
- How many forecasters contributed, since a wide range among three people means something different from a wide range among thirty
None of that survives the averaging.
What Dispersion Actually Measures
The first difficulty is that dispersion is not a clean measure of one thing.
Academic work modelling how forecasts are produced identifies several contributors: the genuine uncertainty in the underlying outcome, the differing private information each forecaster holds, and the strategic incentives forecasters face. One analysis notes that when analysts’ incentives to bias are modelled explicitly, the average dispersion and forecast error computed from the model are almost identical to those documented in the data.
That’s an uncomfortable result. It suggests a meaningful share of observed disagreement reflects the incentives forecasters operate under rather than differing views about the world.
The practical implication is that dispersion mixes at least three ingredients: real uncertainty, genuine disagreement, and systematic bias. They point in different directions, and separating them from the headline number isn’t possible.
Whether It Predicts Anything
The research on dispersion as a return predictor has an interesting arc.
For years, higher dispersion was documented as predicting lower subsequent returns, an effect attributed variously to short-sale constraints and to overpricing. More recent work has complicated that. Using US data from 1981 to 2014, one study found that the return predictive power of aggregate dispersion only exists prior to 2005, with investor sentiment able to explain the dispersion effect only in that earlier period, and neither sentiment nor institutional ownership nor put options explaining the significant weakening afterwards.
The same research draws a distinction worth holding onto: the level of dispersion and the change in dispersion behave differently, with levels reflecting uncertainty and changes reflecting shifts in information asymmetry.
An effect that worked for two decades and then largely stopped is a familiar pattern in financial research, and it’s a reason to treat any dispersion-based signal cautiously.
Three Reasons Forecasts Diverge
Distinguishing the causes changes how a wide range should be read:
- Genuine uncertainty, where the outcome depends on variables nobody can currently observe
- Different information, where forecasters have access to different inputs
- Different assumptions, where the same inputs are processed through different models
- Incentive effects, where a forecaster’s position or role shapes the number they publish
The third is the most tractable for a reader. Two outlooks reaching opposite conclusions usually agree on most of the picture and diverge on one variable, and finding that variable is more useful than the range itself.
How to Use Dispersion Practically
Given what the research supports and what it doesn’t:
- Treat a wide range as a statement about uncertainty, not as a trading signal
- Watch changes in the range rather than its absolute level
- Look for the disputed variable, since that’s what to monitor as data arrives
- Be sceptical of narrow ranges, which can reflect shared assumptions rather than shared evidence
- Note who is disagreeing, because a spread among specialists differs from one among generalists
The fourth point deserves emphasis. Clustered forecasts feel reassuring and can indicate that everyone is anchored to the same starting point rather than that the outcome is genuinely predictable.
The Limits
Dispersion tells you how much people disagree. It doesn’t tell you who is right, and the evidence that it predicts returns has weakened enough that building a strategy on it would be unwise.
What it does reliably provide is a map of where the uncertainty sits. An investor who knows which variables the professionals are arguing about knows what to watch, and that’s a more durable use of the information than any attempt to trade the spread itself.
