The record backs Warsh on 2021 and 2013 but not against forward guidance as such

What Warsh says is wrong

Warsh's critique has two parts: five named faults in forward guidance, and an older, separate critique of buying bonds. At Jackson Hole on 28 August 2026, his hundredth day as Chair, he said he had set out to change the form and function of the Chairman's forward guidance⁠^{Source 5}. The practice was essential in the 2008 crisis, he said, but 'has overstayed its welcome'⁠^{Source 5}. The faults, in his order:

• Ambiguity. In normal times guidance 'risks creating ambiguity in the name of clarity', a line he was already using as an academic in 2016⁠^{Source 5,Source 20}.
• Leading astray. 'Oversharing policy deliberations and overcommitting to future decisions can lead markets, businesses, and households astray.' This is the closest he comes to the reader's 'wrong signals'⁠^{Source 5}.
• Tied hands. Quasi-commitments on rates through the cycle mean, in his words, that 'we inhibit our own freedom to make the right calls'⁠^{Source 5}.
• The ((hall-of-mirrors|hall of mirrors)). If markets rely on the Fed's guidance and the Fed on market prices, both are 'more likely to be blinded to new developments'⁠^{Source 5}.
• Forecasts as guidance. Publishing forecasts to illustrate the Fed's reaction function works 'better in the lab than in the field', and he cites one study for it⁠^{Source 5}.

His one named example is 2021, hedged: guidance that year 'might well have' slowed the response to inflation, a view he backs with a Romer and Romer retrospective not in our archive⁠^{Source 5}. On buying bonds the 2026 speech says only that unconventional policies 'may suit genuine crises but should otherwise be used sparingly, if at all'⁠^{Source 5}.

The fuller version is his November 2010 speech as a governor, days after the second round of quantitative easing was announced⁠^{Source 21}. Expanding the balance sheet 'is not a free option', he said then; it makes the Fed 'more of a price maker than a price taker' in Treasuries and obscures price signals about US indebtedness⁠^{Source 21}. A 2025 lecture, known here only through the Riksbank governor's summary, argued for a permanently smaller balance sheet and a higher bar for future purchases⁠^{Source 22}.

The core he defines is a firm two percent target for PCE inflation, maximum employment, short-term rates as the predominant tool, and 'a quieter Fed, more purposeful in its communications'⁠^{Source 5}. What has changed so far is modest: a task force is reviewing FOMC communications, and the FOMC has raised rates once, in September 2026 after nine months on hold, taking the monthly-average funds rate from 3.63% to 3.75%⁠^{Source 2,Source 6,Source 23}. The dot plot is still published and, in Governor Waller's words, still 'serving that signaling role'⁠^{Source 6}.

The hall of mirrors has a Fed side too. Over 1994–2008 a ten percent fall in stocks between meetings predicted a cut of roughly a third of a point at the next one, the pattern Warsh called 'asset price dependent' in 2016⁠^{Source 24}. In 2020 the Fed's Vice Chair called it an affliction the Fed 'guards against and does not suffer from'⁠^{Source 25}.

2021: a promise that outlived its conditions

The 2021 episode is Warsh's one named example and the strongest case for him. In September 2020 CPI inflation was 1.4% and unemployment 7.8%⁠^{Source 1,Source 26}. The FOMC promised rates near zero until maximum employment and inflation at two percent and on track to exceed it moderately for some time⁠^{Source 8}. It also signalled that hikes waited until the tapering of bond purchases was well under way⁠^{Source 8}.

The conditions were met far sooner than the promise allowed. Core PCE inflation passed two percent in March 2021, at 2.1%, and CPI inflation passed four percent in April 2021, at 4.2%, reaching 7.0% in December as unemployment fell to 3.9%⁠^{Source 1,Source 7,Source 26}. The guidance did not change through 2021, leaving the public to wonder what 'for some time' meant, and the funds rate averaged 0.08% over the year⁠^{Source 2,Source 8}.

The first hike came in March 2022, with CPI inflation at 8.5% and core PCE inflation at 5.6%, 11 months after CPI passed four percent⁠^{Source 1,Source 2,Source 7}. From the September 2020 change of purchase goal to the first hike the Fed's assets grew another USD 1.89 trillion as purchases ran on⁠^{Source 17}. Hikes were tied to the end of tapering, a tie Kiley and Mishkin say 'may have created inertia'⁠^{Source 9}. The chart below shows inflation and the funds rate across the episode, with the guidance, the September 2021 signal and the first hike marked.

How much of the delay was the guidance is contested. Governor Waller says it 'tied the hands of the FOMC in 2021 and unnecessarily delayed rate increases', and agrees that locking in a long policy path is unlikely to end well⁠^{Source 6,Source 8}. The only size in the archive is Kohn's estimate, reported second-hand, of at least several months and perhaps half a year⁠^{Source 9}. Kiley and Mishkin say that assessment 'is not universally shared'⁠^{Source 9}.

The data date the gap, not its cause: the delay is tangled with that year's diagnosis that inflation was transitory. An IMF paper on the cost of falling behind the curve starts from the observation that major central banks 'were slow to react' to the 2021 surge⁠^{Source 27}.

2013: the reaction nobody meant to cause

The cleanest measured case of a market move policymakers did not intend is one Warsh's speech never mentions. On 22 May 2013 Bernanke first signalled the end of asset purchases in testimony to Congress, and the 19 June press conference reinforced it⁠^{Source 3,Source 12}. Fed staff call the move that followed, the taper tantrum, 'unintentionally induced by policymakers'⁠^{Source 12}.

The 10-year Treasury yield rose from 1.94% on 21 May to 2.98% on 5 September, a rise of 104bp, while the 2-year yield rose only 26bp⁠^{Source 3,Source 4}. The term premium rose from −0.02% to 0.60% on the Federal Reserve Board's Kim-Wright estimate, 60% of the yield rise: the spike was mostly compensation for holding bonds rather than a repricing of near-term policy⁠^{Source 3,Source 10}. The BIS reading is that markets had fixated on one aspect of the guidance and a recalibration triggered a disruptive reaction, amplified by the unwinding of leveraged bets on low rates⁠^{Source 28}. The chart below shows the 10-year and 2-year yields and the term premium through 2013, with the taper hint and the September delay marked and the 21 May to 5 September window shaded.

The side effect reached households: the 30-year mortgage rate rose from 3.51% to 4.57%, a rise of 106bp⁠^{Source 11,Source 12}. Fed research comparing outcomes with pre-shock forecasts finds 'no observable negative effect on the aggregate economy', though residential investment clearly fell⁠^{Source 12}. An IMF study uses the shock to test how eighteen large emerging markets reacted, depending on how well their inflation expectations were anchored⁠^{Source 13}. On 18 September the FOMC surprised markets by delaying the taper, and the 10-year yield fell 17bp that day⁠^{Source 3,Source 12}.

One caveat: the no-harm result starts from a yield below two percent, and today's 10-year yield is 2.6 times that level⁠^{Source 3}. The authors warn that a similar rise from a higher starting point could be more contractionary⁠^{Source 12}.

Design decides: the lower-bound years

The era Warsh calls essential tests his first fault: the design of guidance, not guidance as such, decided the side effects. The funds rate averaged 0.13% over the 85 months from December 2008 to December 2015⁠^{Source 2,Source 5}. The Fed moved from qualitative guidance in December 2008 to calendar dates from August 2011, 'at least through mid-2013'⁠^{Source 14,Source 28}. From December 2012 it used a threshold: no hike while unemployment stayed above six and a half percent and projected inflation stayed below two and a half⁠^{Source 28,Source 29}.

Guidance moved yields most when it was new. The 10-year yield fell 16bp on 16 December 2008 and 20bp on 9 August 2011, then moved only 7bp, 2bp and 6bp on the two 2012 extensions and the threshold⁠^{Source 3}. The 16 December 2008 statement also cut the funds rate to its zero range⁠^{Source 2}. The BIS found the same diminishing returns in futures, though the December 2012 day also carried asset-purchase news, and the September 2012 extension was announced alongside the third round of purchases⁠^{Source 3,Source 28}. The table below lists the moves on the guidance, purchase and speech days compared in this article.

The threshold was outrun by the data. Unemployment was 7.9% when it was set and fell below six and a half percent in April 2014, 16 months later; the first hike came 20 months after that, with unemployment at 5.0%⁠^{Source 2,Source 26,Source 29}. Doh, Gruber and Song find that after the threshold the Fed's own words changed: policymakers' speech tone stopped responding to unemployment-forecast revisions as they started to 'lean against the data'⁠^{Source 29}. Their lesson is that the threshold may have been too precise, the opposite of Warsh's objection⁠^{Source 29}.

The cross-country test by Ehrmann and co-authors on seven economies gives the design verdict. Pooled, guidance did not change how 2-year yields react to news⁠^{Source 14}. Calendar guidance with a year and a half or less to run more than doubled the reaction; long-horizon calendar guidance eliminated it⁠^{Source 14}. The BIS asked already in 2014 whether guidance is a crisis tool only, which is Warsh's question, and the survey literature reads guidance as generally effective⁠^{Source 9,Source 28}.

Buying bonds: repair or stimulus

Warsh's critique of buying bonds is older, and the research draws his dividing line for him. The second round of quantitative easing was announced on 3 November 2010, days before a speech in which he set out his reservations⁠^{Source 21}. It added USD 565bn to the Fed's assets by June 2011, while the 10-year yield rose from 2.67% to a peak of 3.75%⁠^{Source 3,Source 17}. His claims were never tested in the archive; the survey's deeper point is that QE began as market stabilisation and shifted to demand stimulus, one tool for two aims⁠^{Source 9}.

Withdrawing intervention had its own side effect. Between their 2015 peak and the August 2019 trough the Fed's assets fell USD 756bn to USD 3.76 trillion, most of it in the 2017–19 runoff⁠^{Source 17}. The authors find that runoff 'increased repo rate distortions, the severity of rate spikes', and it ended in the September 2019 repo spike⁠^{Source 30,Source 31}. That day the effective funds rate printed 5bp above the top of its target range; from the August trough the Fed's assets grew USD 406bn by year-end⁠^{Source 17,Source 30}.

March 2020 was a genuine crisis by any measure. The 10-year yield rose 64bp in seven trading days while the Fed cut to zero, breakeven inflation fell 115bp from mid-February to 0.50%, and the VIX closed at 82.69⁠^{Source 3,Source 18,Source 32}. D'Amico and co-authors find that while the stated goal was market functioning, purchases narrowed price dislocations without lowering the yield curve⁠^{Source 16}. After the goal changed to accommodation on 16 September 2020, the Treasuries bought cut the 10-year-equivalent yield by about seventy-eight basis points⁠^{Source 16}.

Over the whole programme the Fed's assets grew USD 4.71 trillion by the first hike and peaked at USD 8.97 trillion in April 2022, ending only as hikes began; today they are USD 6.76 trillion, 75% of the peak⁠^{Source 17}. The lesson the authors draw is that the communicated purpose decides whether purchases repair or stimulate, and that changing it works as long as the flexibility does not cost purchase predictability⁠^{Source 16}. The chart below shows total assets since 2006; they include lending facilities and mortgage bonds, so they are not the Treasury purchases the research measures.

Where the evidence cuts the other way, and the 2026 test

The question asks only whether side effects occurred; a fair test needs what the tools achieved. In the same cycle, by Waller's account, the market tightened before the Fed did⁠^{Source 8}. After the September 2021 signal the 2-year yield rose from 0.25% to 1.85% on the eve of the first hike, a rise of 160bp, and the 10-year yield rose 83bp⁠^{Source 3,Source 4}. Waller puts the rise at nearly two hundred basis points by mid-February, a larger figure than this series shows⁠^{Source 8}.

The dot plot he criticises is read two ways by the one study he cites. Engstrom finds that private forecasts anchor to stale dots, which predicts forecast errors three and four quarters ahead, and that reliance on the December 2021 dots 'materially slowed' the rise in forward rates in 2022⁠^{Source 15}. But the median dot is the most accurate one-quarter forecast of the funds rate, and the author writes that the results 'do not imply that the SEP is harmful overall'⁠^{Source 15}.

Silence has a price too. Waller says saying nothing about the path 'could surprise markets and create volatility', and in early October 2026 futures priced an eighty-five percent chance of another hike by December⁠^{Source 6}. Since the 2022 liftoff the size of statement surprises, measured by their standard deviation, has almost doubled and that of press-conference surprises more than tripled, which the authors link, untested, to uncertainty about the reaction function⁠^{Source 33}. Data-based guidance can act as an automatic stabiliser, and a euro-area model predicts that projections seen as precise anchor expectations⁠^{Source 9,Source 34}.

The 2026 record is a set of conditions to watch, not effects. The BIS says uncertainty about the reaction function and 'more concise communication' were reflected in higher bond volatility over June to September, and in July especially at the front end⁠^{Source 19}. The Jackson Hole speech was read as hawkish, and the 2-year yield rose 14bp that day⁠^{Source 4,Source 19}. In this article's data, 2-year realised volatility since June averages 5.3bp, 1.3 times the quiet first five months of the year but 0.8 times the 2023–25 average of 6.7bp, and below 2022's 7.7bp⁠^{Source 4}.

Breakeven inflation stayed within 22bp over June to September, which the BIS calls remarkably stable⁠^{Source 18,Source 19}. The 10-year yield hit 5.31% on 5 Oct 2026, the highest in data back to 2006, and the Kim-Wright term premium reached 1.10% on 5 October, the highest since 14 Apr 2010⁠^{Source 3,Source 10}. The BIS puts about forty percent of the summer's rise in the 10-year yield in the term premium; this article does not attribute the October highs to the communication change⁠^{Source 19}. The chart below shows 2-year and 10-year realised volatility since 2022.