Research · Essay
Consensus Is a Position
A belief held by everyone stops being a belief. It becomes inventory.
By NDI Research ✦ ✦3 minutes’ reading
In Brief
- In markets, shared expectations do not remain opinions. They are expressed as positions, and positions have exits.
- The more widely an assumption is held, the more its failure costs, and the less information its continued success provides.
- We do not reject consensus. We reconstruct it, list the assumptions it depends on, and ask which of them are fragile.
Consensus is usually described as a forecast: the median of what informed people expect. It is treated as a neutral starting point, a default to be accepted unless there is strong reason not to. We think this description is incomplete in a way that matters.
In most domains, a widely held belief costs nothing to hold. In markets, every belief that matters is eventually expressed as exposure. Consensus that rates will fall becomes duration. Consensus that volatility will stay low becomes short optionality, leverage and risk-parity weights. A shared expectation is therefore not only a view of the future. It is a distribution of positions in the present.
§ IWhat must be true
Our first step with any consensus expectation is to reconstruct it. We ask: what must be true for this expectation to be correct? The answer is rarely a single proposition. It is a chain of conditions, each with its own probability, most of which are never stated explicitly.
| Consensus: volatility remains subdued through the quarter | Our estimate | Fragility |
|---|---|---|
| Policy path is broadly known and priced | 0.78 | Low |
| No large forced deleveraging in systematic strategies | 0.71 | Moderate |
| Dealer gamma remains positive at current levels | 0.64 | High |
| Earnings dispersion stays within its recent range | 0.82 | Low |
| Funding markets absorb quarter-end without stress | 0.69 | Moderate |
Each condition, taken alone, looks likely. Their joint probability does not. This is the most common shape of a fragile consensus: a stack of individually reasonable assumptions whose conjunction is far less probable than the confidence with which the conclusion is held.1
§ IIReflexive cost
The size of a consensus position changes the payoff of being wrong. When few participants hold a view, its failure is absorbed quietly. When most hold it, failure requires many of them to exit at once, through the same instruments, against limited liquidity. The move that follows is not proportional to the information that triggered it. It is proportional to the inventory that has to be cleared.
This produces an asymmetry that point forecasts cannot see. A crowded consensus that continues to be right generates small, steady returns and very little new information. A crowded consensus that fails generates a discontinuity. The expected value of holding the consensus view can be negative even when the consensus is more likely than not to be correct.
“The question is never only whether the consensus is right. It is what happens to everyone holding it if it is not.”
§ IIINeither follower nor contrarian
None of this is an argument for opposing consensus. Most of the time, consensus is a reasonable aggregation of real information, and a system that reflexively disagrees with it will be wrong more often than a system that ignores markets entirely. Contrarianism is just another default.
What we do instead is give consensus no privileged status. It enters our models as one input among several — informative, but subject to the same scrutiny as any other forecast. Where our reconstruction agrees with it, we agree. Where our reconstruction finds a fragile assumption carrying more weight than it can bear, we record the difference as a probability gap and examine it further. That process is described in The Consensus Gap.