In May 2018, James Bullard of the St. Louis Fed made an observation that should have ended careers. Speaking to his colleagues on the Federal Open Market Committee, he described a pattern that had become institutional muscle memory.
"In both of those cases, there were special factors that came in and moved core PCE inflation lower. And then in both cases, we spent an entire year saying, 'Well, core inflation is lower, but it's not really lower because there are special factors, and therefore we're going to wait until the special factors roll out.'" — James Bullard, FOMC Meeting, May 2018
Then Bullard delivered the counterfactual that transforms this from methodological critique into institutional indictment:
"But if you look at the Dallas Fed trimmed mean in both of those cases, it did not drop, and you would not have had to tell those stories." — James Bullard, FOMC Meeting, May 2018
You would not have had to tell those stories. The alternative measure existed. It was tracked internally. It would have eliminated the need for narrative justification. The Committee chose the measure that required the stories.
The Superior Measure They Chose Not to Use
Four months after Bullard's observation, Federal Reserve staff produced an internal memo that stated the matter plainly:
"If you want a sense of whether trend headline inflation is at, above, or below the FOMC's 2-percent longer-run target, you will likely want to pay more attention to trimmed-mean inflation releases than to ex-food-and-energy inflation releases." — Federal Reserve Staff Memo, September 2018
The Dallas Fed's Trimmed Mean PCE—a measure that removes extreme price movements from the distribution each month rather than mechanically excluding food and energy—had been identified as the superior tool for detecting underlying inflation. Staff analysis showed it had smaller forecast errors. Researcher Evan Koenig called it "a superior communications tool" because it "more successfully filters out headline inflation's transitory variation, leaving only cyclical and trend components."
The Committee knew this. Robert Kaplan of the Dallas Fed proposed formally adopting the trimmed mean as the primary gauge. The proposal was rejected.
The stated reason was that the trimmed mean's historical average ran approximately 25 basis points above total inflation. This would "complicate" communication around the 2% target.
Consider what this means. The measure staff identified as superior for detecting trend inflation systematically ran higher than the measure the Committee chose to emphasize publicly. By omitting the fact that a 2% trimmed mean reading historically corresponded to a 1.75% headline reading, the Committee used core PCE to project confidence in the inflation outlook that the actual trend did not yet justify.
Cherry-Picking and the Narrative Trap
Jeffrey Lacker of the Richmond Fed saw the institutional risk clearly. In September 2011, he warned his colleagues:
"It has the look of cherry-picking the statistic you're looking at. We don't want to get caught shifting which one we are focusing on or appealing to from meeting to meeting." — Jeffrey Lacker, FOMC Meeting, September 2011
Lacker was describing precisely what the Committee would do for the next decade. The record shows asymmetric deployment of inflation measures based on the policy direction the Committee wished to justify—and this asymmetry operated on both sides of the hawk-dove divide.
When Richard Fisher of Dallas wanted to argue for liftoff from zero rates in 2015, he invoked the trimmed mean:
"I prefer the Dallas Fed's Trimmed Mean PCE core inflation measure... Should this low, and still falling, rate of price inflation retard the date of the liftoff from the zero-interest-rate policy we have been operating for more than six years? I think not." — Richard Fisher, Public Statement, February 2015
The trimmed mean was running at 1.6%—closer to target than core PCE's 0.75% to 1.3%. Fisher called it his "GPS" for policy.
But when Chairman Ben Bernanke wanted to dismiss the 2011 commodity price spike and justify continued accommodation, he invoked the same measure for the opposite purpose:
"But for the time being, we're talking about the 'underlying trend,' which basically means indicators like trimmed means and so on that have not really moved much." — Ben Bernanke, FOMC Meeting, March 2011
The same measure. Opposite policy conclusions. Deployed based on which reading supported the preferred stance.
Fisher used trimmed mean to argue inflation was closer to target than core PCE showed—therefore tighten. Bernanke used trimmed mean to argue the underlying trend was stable despite headline volatility—therefore accommodate. Vice Chair Janet Yellen supported this framework, defending core measures as "better forecasters of overall inflation in the medium term than overall inflation itself."
This was not a factional dispute between hawks and doves. It was the institution's relationship to its own data. Leadership utilized trimmed mean not as a statistical check but as a rhetorical device to maintain policy flexibility. By defining the "underlying trend" through a measure that by design excludes volatility, Bernanke and Yellen could dismiss inflationary pressures in 2011 and deflationary pressures in 2019 as "noise."
When the Committee wished to justify continued accommodation on other grounds, core PCE's vulnerability to "special factors" provided alternative flexibility. Cell phone price wars, medical care measurement changes, import price swings—each became a reason to wait another year before normalizing policy.
Fisher had identified this asymmetry as early as 2007:
"Central banks have good reasons to subtract volatile swings in prices, but routinely excluding food and energy—and only food and energy—is not, in my opinion, the best we can do." — Richard Fisher, Public Statement, October 2007
He warned that fixed exclusions risk "throwing out the signal" when those categories represent genuine trends rather than transitory noise.
The trimmed mean, by design, would have prevented this opportunism. It removes whatever is most volatile each month, regardless of category. It doesn't require explanation because it doesn't require exclusion decisions. It simply takes the distribution of price changes and removes the tails, leaving the underlying trend exposed.
This is precisely why adopting it would have been institutionally costly. A measure that automatically filtered noise left no room for discretionary interpretation. It would have forced confrontation with the data rather than negotiation around it.
The Habit of "Transitory"
Thomas Hoenig of Kansas City named the institutional reflex in 2005, sixteen years before it would define the Fed's most consequential failure:
"In this sense, there's a continuation of the view that these things are temporary and that they will back off at some point." — Thomas Hoenig, FOMC Meeting, June 2005
A continuation of the view. Not a fresh analytical judgment in each instance, but an institutional habit—a default setting that characterizes inconvenient data as temporary until proven otherwise. The burden of proof falls on inflation, not on accommodation.
This habit was not accidental. It emerged from a constraint the Committee understood but could not publicly acknowledge.
The Constraint They Could Not Name
To understand why the measurement choice persisted, one must understand what the Federal Reserve was actually optimizing for during the post-crisis decade.
In 1941, John H. Williams, Vice President of the New York Fed, stated the matter with a clarity that would become impermissible in later decades:
"As long as the present situation existed the System could not do much to restrict the supply of credit and, therefore, finds itself in a secondary or tertiary role... although the System could not exercise general monetary controls at this time, it could be of assistance in Treasury financing." — John H. Williams, FOMC Meeting, June 1941
Secondary or tertiary. Not independent. Not mandate-driven. Subordinate to the Treasury's financing needs.
Emanuel Goldenweiser, Director of Research and Statistics, explained why:
"It would not be a feasible policy for the reason that it would increase the cost of Government borrowing without being effective in preventing price rises... It would also raise serious problems about the decline in the capital value of outstanding securities." — Emanuel Goldenweiser, FOMC Meeting, September 1941
The wartime context differs from modern monetary operations, but the revealed principle persists: when debt-management constraints bind, monetary autonomy becomes conditional.
This was 1941. The constraint has never disappeared. It has only become unspeakable.
By 2013, Jerome Powell felt confident dismissing it: "I find myself in disagreement with the paper's assessment that the current fiscal policy challenges might interfere in the near-term with the conduct of monetary policy."
By 2019, Powell had reversed himself: "I want to come back to the issue of the Treasury and how that matters, because there it does seem to me their choices matter a lot in terms of what's left for us to do."
What happened between these statements was the September 2019 repo crisis. Reserves fell to $1.34 trillion—still above the Fed's projected "ample" threshold—but the market broke anyway. Staff analysis was blunt:
"Long-term trends in Treasury issuance have made short-term funding markets more vulnerable to shocks." — Federal Reserve Staff Memo, October 2019
"Historically high levels of Treasury issuance" had made the Fed's balance sheet normalization impossible to complete.
Simon Potter, the SOMA manager, had explained the operational reality two years earlier:
"You wouldn't expect to come out of an FOMC meeting and suddenly find out if you're a Treasury debt manager that you have to issue, say, $60 billion or $70 billion more that month." — Simon Potter, FOMC Meeting, March 2017
The Fed's policy was coordinated with Treasury debt management. Not independent of it.
Lorie Logan had warned in 2017 that combined Treasury financing needs and SOMA redemptions would exceed $1 trillion annually in 2018 and 2019. She was right. The crisis arrived exactly as the mechanism predicted.
The Trap
Narayana Kocherlakota articulated the institutional trap in 2013 with precision that his colleagues apparently failed to absorb:
"We often hear the suggestion that we can mitigate incipient financial instability by tightening monetary policy, but there's quite a tricky term issue here, right? If it's right on us, we are not supposed to tighten. We're supposed to ease." — Narayana Kocherlakota, FOMC Meeting, September 2013
Read that again. If financial instability materializes, the Fed must ease—regardless of inflation. This means the Fed cannot credibly threaten to tighten when tightening itself might trigger the instability that would force easing. The constraint is self-reinforcing.
Two years later, Kocherlakota made the implication explicit:
"Indeed, it appears that the main financial-instability consideration for monetary policy at the moment is that a near-term increase in the target range for the federal funds rate could lead to financial instability." — Narayana Kocherlakota, FOMC Meeting, April 2015
The consideration against tightening was that tightening might cause instability. The consideration against accommodation was inflation. When these conflict, accommodation wins—because instability is immediate and inflation is slow.
This is the institutional logic that made the measurement choice necessary.
The Institutional Selection
The pieces now connect.
An institution that cannot shrink its balance sheet without disrupting Treasury financing. That cannot raise rates without risking the market dysfunction that would force immediate reversal. That must maintain the appearance of price stability commitment while operating under constraints it cannot publicly acknowledge.
Such an institution will rationally prefer an inflation measure that provides flexibility.
Core PCE served this function. Its susceptibility to "special factors" created space for dovish interpretation when accommodation was politically or financially necessary. Its systematic divergence below the trimmed mean meant that measured inflation would appear lower than underlying inflation, reducing pressure to tighten.
The trimmed mean would have eliminated this flexibility. A measure that automatically filtered noise and consistently ran 25 basis points higher would have forced earlier policy confrontation. It would have made visible the gap between stated commitment and actual constraint.
Lacker named this risk: cherry-picking. Bullard named the mechanism: storytelling. The internal record shows the pattern across a decade of policy decisions.
The measurement choice was not methodological preference. It was institutional adaptation to fiscal dominance.
The Filtering
The most consequential information lost between internal analysis and public communication was the acknowledgment that these constraints existed at all.
Staff understood the statistical properties of each measure. They knew the trimmed mean's 25 basis point bias above headline. They knew core PCE's vulnerability to sector-specific volatility. They knew that measurement windows (twelve-month versus twenty-four-month, annualized versus year-over-year) could produce divergent readings from identical underlying data. This was filtered from public communications in favor of whichever framing supported the policy conclusion.
Staff understood the Treasury coordination. What Potter called "you wouldn't expect... if you're a Treasury debt manager" became, in public, a "gradual and predictable" normalization with "appropriate safeguards."
Staff understood the fiscal dominance. What Goldenweiser called "not a feasible policy" because it would "increase the cost of Government borrowing" became, eighty years later, Lael Brainard using "an annualized 24-month measure that looks through the steep declines and subsequent rebound" to show core PCE at 2.3%—smoothing away the immediate inflation spike during a period when the Fed was purchasing two-thirds of Treasury issuance.
The measurement window itself became a tool. Not just trimmed mean versus core PCE, but also twelve-month versus twenty-four-month, annualized versus year-over-year. Each choice created interpretive space. Each interpretation served the institutional necessity of the moment. The multiplicity of defensible measures was itself the source of discretion.
Emmett Rice asked the unanswerable question in 1983:
"Is there a feasible monetary policy that is consistent with no progress in reducing the deficit?" — Emmett J. Rice, FOMC Meeting, November 1983
The question was never answered because answering it honestly would have required admitting that the Fed's independence is contingent, not absolute—that when fiscal needs become acute, the institution adapts its analytical frameworks to avoid forcing confrontation.
What the Record Establishes
The documentary evidence from 90 years of FOMC materials now establishes the following:
The Federal Reserve possessed a superior inflation measure and knew it was superior. Internal staff analysis explicitly identified the trimmed mean as having better forecasting properties and more effectively capturing trend inflation.
The Committee deployed measures opportunistically based on policy preference, not analytical consistency. Hawks like Fisher cited trimmed mean to justify tightening; doves like Bernanke cited trimmed mean to justify accommodation. The same measure served opposite conclusions depending on which reading supported the desired stance.
Participants explicitly identified this as "cherry-picking" and "storytelling." Lacker warned about the appearance of opportunism. Bullard stated the counterfactual directly: with the trimmed mean as primary anchor, "you would not have had to tell those stories."
The Fed's operational independence was constrained by Treasury financing needs. This was stated explicitly in 1941, demonstrated empirically in 2019, and acknowledged by Powell's reversal between 2013 and 2019.
The "transitory" characterization was identified as institutional habit, not analytical judgment. Hoenig named "a continuation of the view" in 2005—the same pattern that would produce the 2021 failure.
The institutional trap was articulated by participants. Kocherlakota explained that the Fed cannot tighten when tightening threatens stability, because instability forces easing regardless of inflation.
The measurement regime—not any single measure—was functional for an institution operating under these constraints. Multiple measures with different properties (core PCE, trimmed mean, various time windows) provided a menu of rhetorical options. Leadership could select whichever reading supported the institutionally necessary conclusion.
The Limits of the Accusation
Precision requires acknowledging what this evidence does not support.
There is no documentation of explicit conspiracy to suppress the trimmed mean for fiscal reasons. The connection between measurement choice and fiscal constraint operates through institutional selection pressure, not coordinated intent.
The Waller caveat stands. In 2021, Christopher Waller warned that during periods of staggered price increases across sectors, the trimmed mean could understate true inflation—showing 2% when the actual measure is 3%. During the 2021 surge, trimmed mean also lagged. No measure would have fully anticipated the regime shift.
No individual is accused of bad faith. The accusation is institutional. The system selected for accommodation bias because the fiscal-monetary equilibrium demanded it. The measurement choice was the mechanism through which this selection operated.
The question is not whether anyone conspired. The question is whether an institution charged with price stability can fulfill that mandate when its operational freedom is constrained by the government's financing needs—and whether the analytical frameworks it adopts reflect genuine methodological judgment or institutional convenience.
Bullard's Year
Return to where we began. A year of waiting, each time, for special factors to "roll out."
In monetary policy, a year is not a neutral delay. A year at the zero lower bound with quantitative easing expanding is a year of additional accommodation. A year of telling markets that inflation isn't "really" what the data shows is a year of shaping expectations around continued support.
The trimmed mean would not have permitted that year. It would have shown—and did show, internally—that underlying inflation was running closer to target than core PCE suggested. It would have forced the Committee to either act or explicitly acknowledge the constraint preventing action.
The institution chose neither. It chose the measure that allowed the year to pass, and then the next year, and then the next.
By the time Hoenig's "continuation of the view" collided with the supply shocks of 2021, the institutional habit was so ingrained that the Committee spent another year—this time with inflation running at forty-year highs—insisting that the pattern was "transitory."
The special factors had become the trend. The stories had become the analysis. And the Committee discovered what Bullard had tried to tell them: you cannot wait forever for factors to roll out when the factors are built into the framework you use to justify waiting.
The FOMC Insight Engine provides semantic search across 90 years of Federal Reserve documents. Every claim in this article can be verified. For research inquiries, contact konstantine.milevsky@gmail.com.