Why Disagreement Mutes News While Policy Uncertainty Amplifies It
This chapter explains the mechanism behind the two opposing state effects in Treasury markets: why wider disagreement about a macro release usually makes its yield impact smaller, while greater uncertainty about the future policy path can make the same release more powerful. The key is to separate two information problems—how noisy the signal about the economy is, and how uncertain the rule connecting the economy to policy is. That distinction lets a reader explain the signs without treating either result as a generic statement about market volatility.
The measurement chapter established that the BIS study estimates a change in the news-response slope, not a standalone effect of “uncertainty.” The model now gives that slope an economic meaning. Investors are learning about the state of the economy and about the policy regime that will translate that state into future short rates. Forecast disagreement interferes with one part of that learning. Uncertainty about policy can increase the value of another part.
How a noisy consensus weakens the release signal
Before a macroeconomic announcement, forecasters submit private expectations. Those forecasts are not identical because people may have different information, different models, or different judgments about recent data. The public consensus combines those signals into a number that the market can use as a baseline. The cross-sectional spread of forecasts is the observed disagreement.
The important point is that disagreement is not merely a description of forecasters' personalities. In the BIS model, it changes the quality of the consensus as information about the underlying economic state. A tight cluster of forecasts suggests that the pre-release signal is relatively precise, even if the cluster turns out to be wrong. A wide cluster suggests that the consensus contains more noise. It is harder to know how much of the gap between the eventual release and the consensus reflects a genuine change in the state and how much reflects an unreliable starting point.
A simple analogy is a measurement taken before a test. If several instruments produce nearly the same reading, the average is a useful starting estimate. If they disagree widely, the average may still be the best summary available, but it tells us less about the true value. When the official measurement arrives, it can correct the prior estimate; yet the observer must first disentangle a new signal from uncertainty already present in the instruments.
The model represents this with a filtering step. Agents begin with a prior belief about the latent macroeconomic state. They then observe the consensus, whose error becomes larger when disagreement is higher. The filtering gain—the weight placed on the consensus relative to the prior—falls as the consensus becomes noisier. Agents lean less on the pre-release number and carry more residual uncertainty into the announcement.
That last point creates an apparent tension. If disagreement leaves more uncertainty about the state, should the official release not become more informative? It can, in one channel. A release may resolve more of the remaining uncertainty when the prior is diffuse. But the event-study surprise is measured against the consensus, not against the agent's fully filtered estimate of the state. A noisy consensus can therefore make the measured surprise a poor proxy for the structural information that actually changes beliefs.
This is the first route to attenuation. Suppose two days produce the same measured gap between a release and the consensus. On the day when the consensus is reliable, that gap is more likely to represent news about the underlying state. On the day when forecasts were scattered, part of the gap may simply reflect that the baseline was noisy. The market can respond less to the same observed surprise because it places less confidence in what the surprise reveals.
The distinction matters for Treasury yields because the state matters only through the expected path of policy and rates. A release that is a weak guide to the economic state is also a weak guide to future short rates, unless it carries some separate information about the policy regime. Lower signal quality therefore reduces the expected revision in the yield curve. This is why the empirical finding that disagreement dampens yield sensitivity is naturally interpreted as an information-channel effect rather than as a statement that disagreement directly pushes bond prices in one direction.
The model does not require investors to agree about every detail. It only requires that observed disagreement be related to the noise in the public consensus. The empirical paper cannot observe the true consensus error directly, so this is a modeling link, not a direct measurement of each investor's information set. That qualification matters: a wide survey distribution could also reflect a difficult-to-predict release or a changing macro regime, and those conditions could affect yields through other paths.
Why an uncertain policy path makes relevant news more valuable
Now consider a different uncertainty. Investors may understand the release reasonably well while remaining unsure how forcefully the central bank will respond. A strong data print could be met with a small adjustment, a large adjustment, or a response that depends on which policy regime is in force. In this setting, the release carries two kinds of information. It says something about the economic state, and it may help reveal the policy mapping that turns that state into the short rate.
The BIS model represents the policy mapping with a latent regime. In one regime, the short rate responds more strongly to the macroeconomic state; in another, it responds less strongly. The regime also affects the typical level of the announcement, so an unusually high or low release can update the probability that one regime is more likely. A macro surprise can therefore change not only the expected state but also the market's estimate of how policy will react to that state.
This is slope learning. If investors are already certain that the policy slope is high or low, a release has little regime information to add. If they are uncertain between the alternatives, a diagnostic release can shift the perceived policy slope. That shift matters more for longer-horizon expected short rates because a change in the policy response can persist through the anticipated path. Under the model's pricing approximation, the yield response contains a state-learning channel and a slope-learning channel.
State learning is the more familiar channel. The release changes the market's estimate of the current macroeconomic condition. If the state is expected to persist, the estimate changes the expected sequence of short rates and therefore the Treasury curve. The size of the state-learning response depends on how much uncertainty remains after the pre-release information has been processed and how precisely the announcement measures the state.
Slope learning is different. The release changes the perceived relationship between the state and policy. A given economic outcome can imply a very different future rate path when investors are unsure whether the central bank will react strongly or weakly. In that situation, short-rate uncertainty does not simply make investors more fearful. It makes policy-relevant information more valuable. The same surprise can generate a larger revision because it helps answer a question that was previously unresolved.
This logic also explains why “uncertainty amplifies yields” is an inaccurate shorthand. The model predicts a larger response to the surprise, not a predetermined upward move in yields. If the surprise points toward a tighter policy path, the response may be a larger increase. If it points toward easing, the response may be a larger decline. For a countercyclical release such as initial jobless claims, the sign of the underlying news matters even when the information value of the news is amplified.
The two learning channels can operate at the same time. A release can update beliefs about the state through its measured economic content and update beliefs about the policy regime through its diagnostic content. The resulting Treasury response is the maturity-specific change in expected future short rates, with nominal components such as inflation beliefs and risk compensation potentially affecting the observed yield as well. The model isolates the dominant expectations channel in the narrow event window; it does not claim that term premia never move.
The diagram shows why the opposing effects are not contradictory. Disagreement weakens the link between the observed surprise and the information investors can safely use. Policy uncertainty strengthens the value of information that identifies the future rate-setting rule. Both channels meet at the same outcome: a revision to expected future policy rates and, consequently, to Treasury futures prices and yields.
Why measured disagreement can have a threshold
The distinction between the measured and structural surprise makes the disagreement result more subtle. The measured surprise is the actual release minus the public consensus. The structural surprise is the part of the release that is genuinely unexpected relative to the agent's filtered pre-release belief about the state. When the consensus is noisy, the two need not coincide. The gap between them is a measurement problem: the observed surprise contains movement in the baseline that the agent has already partially discounted.
More disagreement can therefore produce two opposing effects. First, it can increase residual state uncertainty after the consensus, making the official release more useful for learning about the state. This is an information effect. Second, it can make the measured surprise a noisier proxy for the structural surprise, mechanically shrinking the observed event-study coefficient toward zero. This is an attenuation effect.
The BIS model shows that the combined effect is not necessarily monotonic. The product of the attenuation factor and the announcement's filtering gain is hump-shaped in consensus-error variance. When consensus error is relatively small, the additional information released by resolving state uncertainty can dominate, and more disagreement can increase the observed response. When consensus error is already large, the filtering gain is close to its useful limit while the measured surprise continues to become noisier. Attenuation then dominates and the response falls.
The threshold is determined by the relationship among announcement noise, prior state uncertainty, and consensus error. It is not a universal number that can be read from a market dashboard. In the paper's calibration, typical macroeconomic releases lie in the attenuation-dominated region because announcement noise is small relative to consensus error or because the prior state uncertainty is already sufficiently limited. Moving from the fifth to the ninety-fifth percentile of disagreement in the simulation lowers the observed response by roughly 46 percent, while the attenuation factor itself falls by roughly 28 percent. Those are calibrated model results, not a universal empirical elasticity.
This boundary is valuable because it prevents a simplistic interpretation of the empirical negative interaction. The data do not show that disagreement can never make a release more informative. They show a response pattern consistent with the market operating, for the studied releases and sample, where the loss of reliability in the measured surprise outweighs the extra state uncertainty left for the announcement to resolve. Another market, release, or regime could occupy a different part of the curve.
What the mechanism leaves open
The information-channel explanation competes with a trading-channel story. If investors hold more heterogeneous beliefs, they may be more willing to take opposite sides of a trade after a surprise, reducing its immediate price impact. That story can explain a dampening effect. The BIS authors prefer the learning interpretation because it also explains the amplification associated with short-rate uncertainty: greater uncertainty about the policy path might reduce optimal directional positions rather than increase them. Their model produces both effects within one Bayesian framework and predicts that disagreement should dampen reactions more for releases with high diagnostic content.
That comparison is evidence in favor of a coherent mechanism, not proof that trading behavior is irrelevant. The study does not observe each investor's beliefs or positions. It also differs from rational-inattention models, where information-processing capacity and attention allocation jointly determine signal precision and disagreement. The BIS framework treats disagreement as arising from heterogeneous forecasting models and signal precision as exogenous to uncertainty, which makes the two conditioning variables easier to separate empirically but also imposes a structure on the explanation.
The short-rate-uncertainty measure has a similar boundary. Total uncertainty about future short rates includes uncertainty about the policy slope, future economic states, and shocks. The clean theoretical amplification channel is specifically slope uncertainty. The reduced-form interaction with the composite measure should therefore not be read as a pure estimate of uncertainty about the central bank's reaction function.
The model is also deliberately stylized. It uses a two-day learning problem, a binary latent policy regime, and a linearized response around typical surprises. It treats the term-premium component as second-order in the 20-minute window. The paper's simulated method-of-moments exercise rejects the joint specification and leaves possible maturity-specific term premia, market microstructure, and different horizon decay rates for future work. These limits do not erase the mechanism; they mark the distance between a useful explanation and a complete model of the Treasury market.
The central result survives those qualifications as a disciplined way to think. Forecast disagreement concerns whether the release is a reliable signal about the state. Policy uncertainty concerns how much investors can learn about the future rate-setting rule. One can reduce the response to the same measured surprise; the other can increase the value of a policy-relevant surprise. The next test is whether a single release can change roles across regimes—precisely what the contrasting post-COVID behavior of CPI and nonfarm payrolls allows us to examine.