Tolerable Inflation, Intolerable Uncertainty

arXiv:2609.31512 · econ.GN, q-fin.EC · Submitted 2026-09-25 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.

Ruth: I'm Ruth.

Warren: And I'm Warren, here with Iris.

Ruth: We're looking at the paper: "Tolerable Inflation, Intolerable Uncertainty".

Warren: Here's the gist of the paper. The paper develops a framework analyzing optimal monetary policy rules by examining how different inflation forecasts interact, establishing that equilibrium concepts are discretionary (Markov-perfect).

Ruth: The central theme revolves around the identification failure regarding an optimal monetary policy rule, which is characterized by three walls separating observable data from the optimum: Wall I (the curve), Wall II (the rule), and Wall III (the shock).

Warren: One major contribution is the identification of an "observational-equivalence bill." This bill arises because optimal tolerance and unwarranted drift are observationally equivalent in the inflation history, meaning a central bank cannot empirically demonstrate that a warranted deviation is justified ex post as effectively as in real time.

Ruth: This equivalence carries a quantifiable cost, which is proportional to the point-years inflation spends above target, paid in second moments rather than first moments.

Title and authors: Ruth: Let's talk about the title of this paper. "Tolerable Inflation, Intolerable Uncertainty." What does that really mean for the policy discussion?

Warren: It suggests we have a tolerance threshold for inflation, but the uncertainty around where that target actually lies is what’s truly problematic or intolerable >

Iris: It frames the issue as a tension between managing actual inflation and dealing with the unknown underlying forces driving it, which is what they focus on in this work >

Ruth: Exactly. They're moving beyond just chasing a single inflation number and looking at how much flexibility we need to build into our policy rule given how uncertain those numbers really are >

Warren: The authors are showing that the uncertainty itself imposes a cost, which is measured by those second moments we talked about earlier, making it a real price for having an announced target >

Iris: It’s interesting because it ties this identification failure directly to the agents holding finite patience; they generate predictive variance that shapes this relationship between tolerance and uncertainty >

Ruth: So, when you think about the authors' motivation here, they seem less interested in finding a perfect rule and more interested in quantifying the cost of being imprecise about that rule >

Warren: It’s a quantitative way of saying that ambiguity isn't free; it has to be priced into the system somehow >

The paper's summary: Ruth: So, summarizing the core findings, they are establishing that equilibrium concepts in this setting are discretionary, meaning they can only be found by looking at inflation forecasts interacting with each other >

Warren: And the central theme is that we have an identification failure regarding what the optimal monetary policy rule should even be >

Iris: They map this failure using three walls: the curve of observable data, the line representing the policy rule, and then a shock that pushes things around >

Ruth: And they highlight this "observational-equivalence bill," which is essentially saying that optimal tolerance and unwarranted drift are indistinguishable when looking at historical inflation data alone >

Warren: This equivalence means a central bank can't prove ex post if it’s warranted to deviate from the target as effectively as it could have in real time, and that cost is measured by how much inflation overshoots the target in point-years >

Iris: And then they introduce Normalized Uncertainty, which cleans up the raw uncertainty measure by dividing it by a fitted envelope to get that dispersion we discussed earlier >

Ruth: This correction reveals a specific shape for forecast variance: flat below the announced target and rising linearly above it, with a sharp kink precisely at the announced number >

Warren: It’s important to see how this affects practical application; it means uncertainty doesn't just get bigger randomly; it gets structured by the target itself >

The paper's improvements: Iris: The authors suggest two main ways to improve this framework. First, they propose an "observational-equivalence bill" that can be incorporated into AI systems to distinguish between optimal tolerance and unwarranted drift >

Ruth: That’s the first improvement—using the same design, which they say hasn't been shown before, but linking it specifically to second moments only for better identification >

Warren: That makes sense from a modeling standpoint; if we can isolate that specific cost structure, we might be able to price that ambiguity more accurately than just looking at first moments >

Iris: The second improvement is implementing the Normalized Uncertainty correction by dividing raw uncertainty by the fitted envelope, which they call "distancenormalized dispersion" >

Ruth: That sounds like it removes the predictable footprint that tolerance and public doubt leave on measured variance, making the uncertainty measure cleaner for policymakers >

Warren: So, in practice, this means we can use a measure that is more focused on the actual deviation from the target rather than just raw noise or historical averages >

Conclusion: Ruth: To wrap up "Tolerable Inflation, Intolerable Uncertainty," the main implication is that while we can argue for an optimal policy rule, its realized stance remains undecidable because of these structural identification failures >

Warren: The verdict on whether a bank’s stance is too tolerant or too tight cannot be settled ex post or in real time because the configurations separating the dial from the rule depend on counterfactual menus that equilibrium histories never play out >

Iris: And they also point out a limitation: key inputs like optimal gap and accommodation share are only bracketed by published readings, meaning every ingredient of the optimum sits at least one layer behind what we can actually estimate >

Ruth: So observable data register episodes and their second moments, but the precise verdict on policy remains elusive because of these structural identification failures in "Tolerable Inflation, Intolerable Uncertainty" >

Warren: It’s a sobering thought for anyone trying to use this framework to make concrete policy decisions right now >

Iris: I just want to say that while the model is rigorous about what it measures, the paper leaves it clear that we are dealing with an impossibility in terms of ex post verification of the rule >

Eric Vansteenberghe

Banque de France · University of Paris 1 Panthéon-Sorbonne

econ.GN, q-fin.EC

Submitted: 2026-09-25

Updated: 2026-11-13

Comments: 85 pages, 10 figures, 18 tables. Upcoming Banque de France working paper. JEL: C18, D81, E31, E52, E58. The measures are constructed in arXiv:2411.05938. Replication code: https://github.com/skimeur/normalized-uncertainty

Code: https://github.com/skimeur/normalized-uncertainty

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 73/100

The gist: The paper develops a framework analyzing optimal monetary policy rules by examining how different inflation forecasts interact…

Key concepts

Observational-Equivalence Bill
This arises because optimal tolerance and unwarranted drift in inflation history are observationally equivalent. A central bank cannot empirically prove a warranted deviation ex post as effectively as in real time. This equivalence carries a quantifiable cost measured by the point-years inflation spends above target, paid in second moments.
Normalized Uncertainty (NU)
NU modifies measured uncertainty by dividing the raw second moment by the fitted envelope. This removes the predictable component of uncertainty related to distance from target. The resulting law dictates that forecast variance is flat below the announced target and rises linearly above it, with a kink precisely at the announced number.
Three Walls
These walls separate observable data from the optimal policy rule: Wall I (the curve), Wall II (the rule itself), and Wall III (the shock). The model shows that while an optimal rule can be argued, its realized stance remains undecidable because configurations separating the dial from the rule depend on counterfactual menus equilibrium histories never play out.

Terminology

Summary

The paper develops a framework analyzing optimal monetary policy rules by examining how different inflation forecasts interact, establishing that equilibrium concepts are discretionary (Markov-perfect). The central theme revolves around the identification failure regarding an optimal monetary policy rule, which is characterized by three walls separating observable data from the optimum: Wall I (the curve), Wall II (the rule), and Wall III (the shock). p. 1

Where the paper says it

It then adds walls of its own: the supply share st is latent in real time and carries most of the optimum’s irreducible variance; parameter uncertainty manufactures tolerance that no preference generated; the configurations that would separate the dial from the rule differ only in counterfactual menus that equilibrium histories never play; and the state that fixes the optimum resolves only after the moment in which the bank must act. The consequence is a statement in the oldest register of the identification literature: a tolerated gap that optimal policy prescribes and a gap that unwarranted drift would produce are observationally equivalent in the sense of Sargent (1976)—two readings of one history that no accumulation of equilibrium data separates.

One major contribution is the identification of an observational-equivalence bill. This bill arises because optimal tolerance and unwarranted drift are observationally equivalent in the inflation history, meaning a central bank cannot empirically demonstrate that a warranted deviation is justified ex post as effectively as in real time. This equivalence carries a quantifiable cost, which is proportional to the point-years inflation spends above target, paid in second moments rather than first moments. p. 2

Where the paper says it

We call the resulting cost the observational-equivalence bill. It is proportional to the point-years inflation spends above target, it is paid in the second moments rather than the first, and it cannot be engineered away—because what charges it is precisely the impossibility of demonstrating that the tolerance is optimal.

A second contribution introduces Normalized Uncertainty (NU), which modifies measured uncertainty by dividing the raw second moment by the fitted envelope, thereby removing the predictable component of uncertainty related to distance from target. p. 2

Where the paper says it

Residual belief imprecision—the part the anchor does not explain—is what remains once that structural component is removed, and, so measured, it is both smaller and differently behaved than the raw second moment implies. That removal is the paper’s second contribution, and it inherits a property from the first: because the bill is by construction a function of the distance from target, whatever removes the distance-predictable component removes the bill along with it.

This law priced by NU dictates that forecast variance obeys a one-sided relationship: it is flat below the announced target and rises linearly above it, with a kink precisely at the announced number. p. 5
Where the paper says it

Where genuine imprecision matters, a coefficient appears that the raw measure hid: Barro’s null becomes a growth effect significant at five percent, and loan pricing, indifferent to raw uncertainty once the policy stance is held, prices the purged measure at nearly nine-tenths of a percentage point per standard deviation. 2 The model and the purge The evidence documents a one-sided law: the subjective uncertainty of an inflation forecast is flat while the expected gap sits below the announced target and rises with the gap above it, with the kink at the announced number (Section 3.

The bill remains because it is charged by the equivalence itself—the cost of having said the number and then being unable to prove why inflation is not sitting at it. p. 3
Where the paper says it

2 prices both sides. The bill therefore cannot be engineered away by better communication, but neither is it a fixed feature of having a target: it is a feature of having said a number, and it accrues in proportion to the time spent above it.

The model further establishes that while an optimal policy rule can be argued, its realized stance remains undecidable from equilibrium data due to these three walls. p. 1

Where the paper says it

The consequence is a statement in the oldest register of the identification literature: a tolerated gap that optimal policy prescribes and a gap that unwarranted drift would produce are observationally equivalent in the sense of Sargent (1976)—two readings of one history that no accumulation of equilibrium data separates. 1 The verdict on a central bank’s stance can be argued; it cannot be settled.

The verdict on whether a bank's stance is episode-optimal, too tolerant, or too tight cannot be settled ex post or in real time because the configurations separating the dial from the rule depend on counterfactual menus that equilibrium histories never play out. Furthermore, key inputs—such as the optimal gap and accommodation share—are bracketed only by published readings, meaning every ingredient of the optimum sits at least one layer behind what is estimable. p. 1
Where the paper says it

The optimal accommodation is a well-defined object—the supply share of the shock times a structural lookthrough intensity—and it is unidentifiable, not for want of data but by construction. It inherits both of the great identification failures of empirical macroeconomics: the intensity λ ∗ = 1/(1+ϵpκ) combines a conventionally calibrated demand elasticity ϵp with the Phillips slope κ, an object seventy years of academic research have not pinned; and the dial is revealed only through φπ(1−λ), one layer behind a policy-rule coefficient whose own identification is still in dispute.

The paper concludes that while observable data register episodes and their second moments, the precise verdict on policy remains elusive due to these structural identification failures. p. 1
Where the paper says it

The consequence is a statement in the oldest register of the identification literature: a tolerated gap that optimal policy prescribes and a gap that unwarranted drift would produce are observationally equivalent in the sense of Sargent (1976)—two readings of one history that no accumulation of equilibrium data separates. 1 The verdict on a central bank’s stance can be argued; it cannot be settled.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved AI systems can incorporate an observational equivalence check to distinguish between optimal tolerance and unwarranted drift in policy settings, as this is identified by the same design that, to our knowledge, has not been shown before and is linked to second moments only.

  2. The AI can implement a Normalized Uncertainty correction by dividing measured uncertainty by the fitted envelope of the law. This correction removes the predictable footprint that the bank’s tolerance and the public’s doubt leave on measured variance, leading to a measure that is distancenormalized dispersion.

  3. The system can perform policy recommendations based on identified bounds, such as choosing a tolerated share lambda e which is the optimum λ∗e = ιs drawn over the published intensity band, while understanding that the verdict on any past stance remains undecidable.

Abstract

How much of an inflation deviation a central bank should tolerate is a well-defined optimum in the New Keynesian model, but it cannot be identified: optimal accommodation of supply shocks and an unwarranted drift of the effective inflation target are observationally equivalent. This impossibility has a cost. A public that cannot verify the bank's intention prices each expected point of overshoot, so uncertainty about inflation is flat below the announced target and rises linearly above it: the observational-equivalence effect. Professional forecasts, swap and option markets and realized inflation all show this pattern, with the kink at the announced number. A second and larger effect produces the same shape: inflation is more volatile when it is high, and in realized inflation that level effect is mostly energy. Both effects load on the distance from the target, so the data admit the observational-equivalence effect without isolating it, and its cost cannot be estimated with the data at hand. We therefore cannot say how optimal monetary policy should change on its account, but three policy conclusions survive: in the textbook model, not observing the slope of the Phillips curve and the composition of shocks never calls for tolerating less; the announced target remains worth having, because it compresses beliefs at the target; and the central bank can monitor the point-years inflation spends above target, on which the effect is charged, and read inflation uncertainty net of the distance from the target. Normalized Uncertainty provides that reading. In French loan-level data, raw inflation uncertainty is unrelated to corporate loan rates after controls, whereas a one-standard-deviation increase in Normalized Uncertainty is associated with rates 89 basis points higher. The correction also reveals a negative association with growth across countries and alters euro-area time-series results.

Related papers