The Fed's Five Task Forces: A Roadmap

On 17 June, Federal Reserve Chair Kevin Warsh established five task forces to reconsider how the Fed operates. They cover communications, balance sheet policy, data, productivity and jobs, and inflation frameworks. Together they amount to the most consequential review of U.S. monetary policy at least since the FOMC adopted a formal inflation target in 2012.

The review is overdue. Five years of above-target inflation are eroding credibility the Fed spent decades building. Meanwhile, mounting threats to Fed independence may have made self-examination harder: any institution under threat is less inclined to ask publicly what needs repair.

This post introduces a series addressed to the five task forces. Three are already up: one on communications and two on the balance sheet (here and here). Four more follow today (see here, here, here, and here). In addition, there is a post that summarizes our recommendations. All three of the existing posts first appeared before the announcement of the task forces, so we have added a postscript to each.

The Challenge of Unobservables

The five task forces look like five separate assignments. We view them as very closely connected.

In each area, the Committee steers by a number nobody observes. Estimating the inflation trend means filtering noise out of the raw data. No one measures it directly. The statistical agencies infer real quantities rather than collect them: the Census Bureau gathers revenue, the BLS gathers prices, and the BEA derives the rest — so the Fed never sees the shock driving inflation, only a reconstruction of it. Models and statistical procedures attempt to extract potential output (y*), the natural rate of interest (r*), and the natural rate of unemployment (u*) from history. Rather than analyze how banks behave, the Fed uses aggregate data to estimate the inflection point of reserve demand. And outsiders try to reverse-engineer the FOMC reaction function from insufficient information.

The combination of reliance on history and stylized models makes Fed estimates of key unobservables fragile, so that when the structure of the economy shifts, they fail together. We view 2021 as an important instance of common failure. Observers could not readily decompose the enormous shocks to the economy in real time. Key, widely used measures failed to signal increases in the trend of inflation. The output gap showed slack when the reverse was true. And the FOMC's Summary of Economic Projections (SEP) lagged inflation rather than anticipating it. The balance sheet had its own large shock in September 2019, when the Committee underestimated the level of reserves the banking system needed.

The five task forces face a common challenge of dealing with unobservables. In each case, we recommend increased transparency as one element of the FOMC's reforms.

What We Already Argued

Communications. In June, we agreed with Chair Warsh that more talk is not better talk. In our view, the problem is that too little of what the Fed says clarifies the FOMC's reaction function. The dot plot publishes conclusions with the reasoning removed: no forecasts attached to the dots, no names, no way to see who disagrees or – most important – why. Worse, the Committee's actual behavior proved asymmetric, responding more forcefully to shortfalls in employment than to overshoots in inflation. That is not equal treatment of the dual mandate, and the SEP as currently published gave no way to detect it.

The fix is a credible framework paired with communications that let the public hold the FOMC accountable: publish the matrix linking each participant's rate path to that participant's own forecasts, and identify the participants. Abolishing the dot plot before a more credible regime exists to replace it would just reduce accountability.

Balance sheet policy. In February and again in April, we examined how the Fed decides what its balance sheet should look like. The Committee assembles its estimate of the aggregate reserve level banks need from partial components, none of which describes how banks behave. Lacking knowledge about aggregate reserve demand, efforts to shrink the Fed balance sheet pose a risk of short-term market instability like that of September 2019, when the spread between the secured overnight financing rate (SOFR) and the interest rate on reserves spiked. That spread, not the federal funds rate, is what the Committee should monitor and work to keep within a wide range. Funding stress now appears in the repo market, which the federal funds rate no longer captures. To limit that risk, the Committee will need to implement measures that reduce reserve demand before it considers reducing reserve supply.

One fiscal point deserves stating plainly. Shrinking reserves matched by sales of Treasury securities does not reduce the government's combined interest bill and may well raise it. Whatever the case for a smaller balance sheet, it is not that one.

The Four New Posts

With this introduction, we are also releasing four new posts that address the Fed’s Task Force mandates: (1) Rethinking the Fed’s Inflation Framework; (2) How Should the Fed Measure Inflation?; (3) What Should the Fed Measure?; and (4) Productivity and Employment: AI’s Promise and Pitfalls. A fifth and final post gathers and summarizes the recommendations in the entire series.

Rethinking the Fed's inflation framework. Warsh charged the Inflation Frameworks Task Force with revisiting how the Fed understands and responds to the drivers of inflation. Understanding them turns on whether the Committee can distinguish a demand shock that drives prices and output in the same direction from a supply shock that pushes them in opposite directions in real time. This question is primarily one about data, and it is the subject of the two posts (here and here) that follow. This first post takes up the question of how the Committee should respond.

The Chair has committed to retaining the 2 percent inflation target for now. Within that constraint, we recommend making the FOMC reaction function explicit and implementing "makeup" strategies only in exceptional circumstances, if ever. Flexible average inflation targeting (FAIT) failed for a clear reason: the FOMC never specified the critical parameters, so nobody could tell what the framework promised. In effect, FAIT just boosted discretion. Price-level targeting — an alternative to inflation targeting — assumes something the measurement posts (here and here) show to be doubtful: that the price index we use is clean enough to anchor a path for the aggregate price level measured in decades. That leaves conventional inflation targeting as the best we can do – for now.

How should the Fed measure inflation? The FOMC asks a single number to do two different jobs: read the inflation trend for internal purposes, and serve as the standard against which the public holds the Committee to account. These jobs have incompatible requirements. The first requires the best available estimate and should be revisable as methods improve. The second needs something simple, familiar, and hard to game. The FOMC can and should separate them, but both should be transparent.

We also warn about the statistical measures now in wide use to estimate trend inflation. It turns out that these estimates are least useful when price increases are broadly based, as they were in 2021 and are again under tariffs.

What should the Fed measure? The private sector is already using high-frequency prices and scanner data. The Fed should too, but not for the reason usually given. Measuring prices faster and more precisely does almost nothing to improve the estimate of the inflation trend, because precision is not the binding constraint. The noise that obscures the trend is genuine movement in relative prices, not sampling error.

The value of new data lies in what they reveal that the Fed does not currently observe: the quantities that help distinguish a demand shock from a supply shock, the many attributes that determine the quality of a good or service, and how often firms reset prices. These are all critical to keeping prices stable.

Productivity and employment. The faster artificial intelligence advances, the less reliable the Fed's usual policy guides become. This applies with particular force to the unobservable “stars”: y*, r*, and u*. Even long before AI, estimates of the output gap (y-y*) were consistently wrong in the same direction for years at a stretch: from 2002 to 2014, every real-time reading was too low, showing 1.5 percentage points more slack on average than we now believe existed. A rapid technological transition is precisely the kind of event that leads such estimates to be persistently wrong.

The answer is not to call for better estimates of the stars. The uncertainty will remain too large. Instead, policymakers need to develop strategies that rely less on the stars.

The Common Thread

One recommendation appears in every post in this series, and we reached it independently each time: boost transparency. That means publishing the inputs, not just the output. Show how broad price increases are in order to reveal when the filtered inflation trend is misleading. Provide the forecasts behind each participant's dot. Reveal all the indicators behind the Committee's judgment on reserve adequacy. Disclose the revision record behind estimates of the unobservable guides to policy.