Federal Reserve Chair Kevin Warsh has appointed five task forces to reconsider how the Fed operates: inflation frameworks, data, productivity and jobs, communications, and balance sheet policy. Together they amount to the most consequential review of U.S. monetary policy at least since the Federal Open Market Committee (FOMC) adopted a formal inflation target in 2012. The review is overdue. Five years of above-target inflation are eroding the credibility the Fed spent decades building.

The five assignments look separate, but a common problem runs through all of them. In each area, the FOMC has to steer by something nobody actually observes. No one measures the inflation trend directly; analysts must filter it out of noisy data. The statistical agencies typically infer real economic quantities rather than collect them. This practice blurs the co-movement of prices and output that identifies the drivers of inflation. Economists estimate potential output (y*), the neutral interest rate (r*), and the natural rate of unemployment (u*), collectively the “stars,” from models fit to history, but no one knows the true values. And even outside the Fed, observers must reverse-engineer how the FOMC responds to incoming information – its reaction function.

Reliance on history and stylized models makes all of these estimates fragile, and they fail when the structure of the economy shifts. In 2021, the conventional measures were slow to signal the rise in trend inflation, and the FOMC’s own quarterly forecasts lagged the data rather than anticipating it.

What follows draws on a series of commentaries we published recently at moneyandbanking.com. Each section below closes with our recommendation to the relevant task force; we finish with the thread that connects all five.

Inflation frameworks

Warsh has committed to the 2% inflation target, for now, but wants to revisit the broader framework. We see no case for changing the target itself. Raising it, as some economists proposed after 2008, would buy policy space at the cost of the credibility the Fed is still trying to rebuild, and would look like accommodating the very inflation the public has been suffering. Alternatives such as price-level or nominal-GDP targeting ask the public to trust a decades-long promise, something the FOMC has never demonstrated it can keep.

The bigger challenge lies elsewhere. Without an explicit reaction function, a framework states intentions and nothing more, and intentions collapse under stress. The FOMC’s 2020 strategy revision replaced a symmetric target with “flexible average inflation targeting” (FAIT), a pledge to let inflation run above 2% for a while to offset earlier shortfalls. But the FOMC never provided critical details, so the pledge constrained nothing. And, because it offset only shortfalls and not overshoots, this asymmetric approach likely delayed the response to 2021’s inflation surge. The FOMC removed both flaws in 2025. That was the right call, but the new statement is no more explicit than its predecessor about how policy responds to deviations from target.

Bottom line: Until there is real evidence that a makeup strategy works, the Inflation Framework Task Force should recommend keeping the 2% target as is, and call on the FOMC to publish, in advance, how it will respond to deviations from it, and to keep this disclosure up to date as its reaction function evolves. (Read our framework post.)

Data

Warsh is right that price data are imperfect, but the FOMC already draws on real-time private and alternative data alongside the official statistics, through the Beige Book surveys, business contacts, and market-based measures. The shortcomings are more specific than “the data are bad.”

First, the FOMC needs two different numbers, not one: a headline figure that is simple, familiar, and rarely revised, so the public can hold the FOMC accountable; and a separate, filtered estimate of the underlying trend, to guide policy. These needn’t conflict, so long as the FOMC keeps them separate rather than asking one number to do both jobs.

Second, the trend estimate embeds assumptions that can go stale. The key problem is bias. The process that creates the trend estimate can push it systematically above or below the truth. And, this bias can change precisely when policymakers most need a reliable trend estimate – namely, when a broad shock hits. The Dallas Fed’s “trimmed mean,” for example, discards more from the top of each month’s price-change distribution than the bottom, a calibration built for decades when goods prices routinely fell relative to services. The distribution of prices has shifted, so as of 2021, the trimmed mean began to understate the trend by perhaps a third of a percentage point. The Dallas Fed itself now warns that tariffs may be pushing its measure down artificially.

More frequent price sampling would help much less than it sounds. Sampling error and transitory relative-price movements are unrelated. New data shrink only the sampling component, but by our estimates this accounts for only a few percent of the variation around trend, so eliminating it would barely move the estimate.

What is missing is quantities. Distinguishing a demand shock from a supply shock requires observing price and quantity from the same transaction, and no U.S. statistical agency does this systematically: the Census Bureau collects nominal dollar measures such as revenue, the Bureau of Labor Statistics collects prices, and for most PCE components the Bureau of Economic Analysis divides one by the other to infer the rest. Building a data source that links prices and quantities at the item level is a job for the statistical agencies working together.

Bottom line: The Data Task Force should recommend that the FOMC keep the headline and trend measures explicitly separate and publish its filtering methods. It should also call on the FOMC to advocate in Congress for the funding that the statistical agencies would need to link prices and quantities at the transaction level. For now, the FOMC can continue to employ private scanner data to distinguish demand and supply shocks. (Read our inflation measurement post and our high-frequency data post.)

Productivity and jobs

AI could sharply lift productivity growth, displace large numbers of workers, both, or neither. In a survey of economists across three AI-capability scenarios, disagreement about what a specific scenario would do to the economy accounts for roughly 50 times more of the variance in growth forecasts than disagreement about the technology’s path itself. (One of us is a member of the expert panel surveyed in that work.)

The point is that no one can forecast this, and the Productivity and Jobs Task Force should encourage the FOMC to say so plainly.

The task force also should call on the FOMC to develop policy guides that will be more robust in the face of rapid structural change. The models that policymakers use to infer y*, r*, and u* are unlikely to be reliable, for example, in a period of rapid AI progress. The track record is already weak. Over three decades, real-time output-gap estimates have missed by 1.2 percentage points on average, gotten the sign wrong in nearly one of every five quarters, and stayed wrong in the same direction for years at a time. That pattern helped produce the Great Inflation of the 1970s.

Some policy rules are less sensitive to uncertainty about the stars, but none eliminates it entirely. So this is a problem to manage rather than solve. In the case of AI, that means adjusting policy cautiously while the path is unclear, preserving the option of decisive action once the picture clarifies.

Bottom line: The Productivity and Jobs Task Force should recommend that the FOMC publish a track record for its “star” estimates: average error, typical spread, and past episodes where the sign was wrong. The public could then see how much weight those numbers deserve. At the same time, the task force should encourage the FOMC to undertake a long-run research project aimed at developing more robust policy guides and rules. Finally, the task force should call on the Fed to adopt a “risk management” stance. Such a stance may favor ensuring policy flexibility until it is possible to reliably anticipate the impact of AI on the economy. (Read our productivity post.)

Communications

Warsh’s diagnosis is that the Fed says too much. We think the problem is that it says too little about the thing that matters, namely how the FOMC will actually respond to new information, its reaction function. Communication that doesn’t convey this conditional policy response is superfluous.

The clearest room for improvement is in the Summary of Economic Projections (SEP), the FOMC’s quarterly forecast survey, best known for its “dot plot” of individual rate-path projections. The SEP publishes a distribution of interest rate projections and a distribution of economic forecasts but never links them, so a reader cannot tell whether a participant projecting a high rate expects strong growth, worries about inflation, or simply weighs the same outlook differently. The FOMC should publish that link, putting each participant’s own forecast next to that same participant’s own rate path, ideally with the participant named. That would make the reasoning visible and let the public judge whose analysis has held up over time. Adding a handful of alternative policy scenarios would extend that visibility beyond the central forecast. These scenarios are likely to matter more heading into an AI transition that no one can reliably forecast.

It is worth emphasizing that a reaction function by itself is neither a policy commitment nor a looser form of forward guidance. It says only that if x occurs, expect the FOMC to respond y. In that sense, the forecasts and scenarios are useful primarily to focus the range of circumstances over which the SEP illuminates the reaction function.

Bottom line: The Communications Task Force should recommend enriching the SEP rather than scaling it back: link forecasts to rate paths, name the participants, and add scenarios. A quieter FOMC would be less accountable, not more accurate. (Read our communications post.)

The balance sheet

Warsh wants a smaller Fed balance sheet, and as a way to keep the central bank’s footprint separate from fiscal policy, that’s a reasonable goal. In our view, though, the binding constraint on balance sheet size is the payment system. The Fed limits banks’ use of daylight overdrafts (short-term loans the Fed extends within the business day so a bank can send a payment before matching funds arrive). Without that cushion, banks instead seek to hold sufficient reserves up front to cover the largest gap they expect, during a given day, between the payments they send over Fedwire and the ones they receive. That structural need is what keeps both reserve demand and the balance sheet from shrinking much further without other changes.

Trimming reserve demand can relax that constraint. To reduce the balance sheet safely, the FOMC needs to introduce reforms that lower demand before attempting to lower supply. Cutting supply first risks repeating the money-market disruptions of September 2019, when the Fed misjudged how many reserves banks actually needed and shrank the balance sheet too far.

Two reforms would help most. “Netting” would shrink banks’ settlement obligations substantially. It would let banks offset the payments they’re sending against the payments they’re receiving before drawing down reserves, rather than settling each payment separately. But it would probably take years of study before implementing a new design of the payments system itself. Reducing banks’ need to pre-fund any obligations can go faster, as moves to ease access to the Fed’s discount window and its backstop lending tool (the standing repo facility) are already underway.

Outsiders currently have little way to judge the FOMC’s progress here. Fixing that requires publishing, in real time, the indicators the FOMC actually watches for reserve adequacy: the spread between short-term market rates and the rate the Fed pays on reserves, use of the standing repo facility, and the dispersion of reserves across banks. Openly accounting for when those indicators failed to warn in time (as in September 2019) would build confidence in the effort.

Bottom line: The Balance Sheet Task Force should recommend that the FOMC introduce reforms that reduce reserve demand before cutting reserve supply further. It should also encourage greater transparency about the real-time indicators that the FOMC monitors to judge reserve demand-supply balance. (Read our balance sheet post.)

The thread that connects them

One idea runs through all five recommendations: publish the inputs, not just the output. The skew behind the filtered inflation number. Each participant’s forecast behind their rate path. The revision history behind every “star” estimate. The indicators supporting a reserve-adequacy judgment, and the record of when they failed to warn in time.

None of this asks the FOMC to forecast better. We doubt that’s possible. It asks the five task forces to promote more systematic and transparent FOMC policy, while shortening the interval between the world changing and both the FOMC and the public noticing.