The Federal Open Market Committee votes. The tally is recorded. The statement is released. And across decades of the most consequential monetary policy decisions in modern economic history, one feature recurs with striking regularity: the vote is unanimous.
The FOMC is composed of twelve voting members — seven governors appointed by the President and confirmed by the Senate, and five of the twelve regional bank presidents on annual rotation. They are, by any measure, among the most analytically sophisticated policymakers in the world. They have access to the largest dedicated economic research staff on earth. They deliberate in private, with transcripts sealed for five years. And when the deliberation ends, they vote — overwhelmingly, almost reflexively — as one.
This investigation asks a simple question: does the formal unanimity of FOMC votes reflect genuine analytical consensus, or does it mask unresolved disagreement that correlates with the institution's most consequential failures?
The method is direct. We begin not from the votes but from the failures — the episodes the Federal Reserve itself has acknowledged, in retrospective assessments and academic reckonings, as its worst mistakes. We then trace each failure backward to the meeting where the critical decision was taken. We examine the transcript. We compare what was said in the room with what appeared in the public statement. We identify who warned, who was overridden, and what institutional mechanism converted private doubt into a public yes vote. Then we test the counter-case — the rare episode where unanimity broke — to ask whether visible dissent produced a different outcome. This is a case-study investigation, not a statistical comparison — six episodes examined in detail, not a base-rate analysis of all FOMC votes. The question is whether those episodes reveal a structural pattern or a collection of coincidences.
There is a term for what this investigation is looking for. Social psychology calls it collective irrationality — the documented phenomenon in which intelligent individuals, each acting rationally on their own terms, produce collective outcomes that none of them would have chosen alone. The mechanism is not stupidity. It is the interaction effects of a system that changes what people say, what they notice, what they fear, and what they are rewarded for. The FOMC transcripts are an opportunity to test whether this phenomenon operates inside the most analytically sophisticated policy committee in the world, and if so, through what specific mechanisms.
Five failures. One exception. Five decades. Four chairmen.
We went into the archives.
The Deference
The first failure is the oldest and the most elementary. Between 1965 and 1972, the Federal Reserve kept monetary policy too loose for too long, seeding what would become the Great Inflation — the defining macroeconomic catastrophe of the postwar era. The question for this investigation is not why the Great Inflation happened. That has been studied extensively. The question is what the Committee knew at the moment the critical decisions were taken, and why the votes were unanimous despite that knowledge.
The documentary record begins in October 1964 — more than a year before the inflation became visible in the price data. Federal Reserve staff produced an internal assessment identifying a specific, quantitative threshold for danger.
"In another view, an over-all rate of 88 per cent may be accompanied by enough evidence of extensions of delivery dates and upward price pressures to seriously threaten price stability."
The warning was precise: 88 percent capacity utilization, combined with extended delivery schedules, would threaten price stability. This was not a forecast of distant trouble. It was a specific, measurable condition that, if breached, would indicate the economy had moved past the point where inflation could be easily contained. By late 1965, the economy had reached that threshold. Staff reports noted upward pressures on prices were strong and that capacity utilization had reached levels not seen since the 1955–57 expansion.
Individual members of the Committee received the staff's analysis and responded with alarm. George H. Ellis, President of the Federal Reserve Bank of Boston, warned in September 1965 that the speed of credit expansion had become the critical variable — the factor that would determine whether price pressures could be contained or would escape into self-reinforcing inflationary expectations.
"In that atmosphere the extent to which demand pressures were stimulated by rapid — he almost said unsustainable — bank credit expansion could play a vital role in determining whether price pressures were contained or escaped to feed inflationary expectations."
Ellis hesitated mid-sentence. He almost said "unsustainable." The transcript records the self-correction — the moment where a sitting Federal Reserve Bank president pulled back from the word that described what he actually saw. That hesitation is a small artifact. But it is the unanimity mechanism made visible in real time: the private assessment was unsustainable; the institutional register permitted only rapid.
The Committee had the staff's threshold. It had Ellis's warning. It had its own internal data showing the economy at or beyond the 88 percent limit. The vote, meeting after meeting through late 1965 and into 1966, was unanimous. The reason is in the transcript.
"The Administration was strongly opposed to a change in policy."
Chairman Martin did not argue that the staff was wrong. He did not claim that Ellis's concern was analytically misplaced. He reported a political constraint and indicated that he did not believe it appropriate for him to lend his support to those who favored a change in the face of that opposition. The analytical question — whether the economy was overheating — was answered internally. The policy question — whether to act on that answer — was subordinated to the Administration's preferences.
By January 1966, the Committee's private deliberation had reached a point that the transcript records with uncomfortable clarity. George W. Mitchell, a Board governor, proposed what amounted to an institutional shrug.
"The Manager had implied that monetary policy could do little to tranquilize such expectations, and he also saw little that could be done, so perhaps the Committee had to lay the problem aside for the moment."
"Lay the problem aside for the moment." The Committee had identified the problem. The staff had quantified the threshold. A regional bank president had nearly called the credit expansion unsustainable. The Chairman had acknowledged that the right policy would be to tighten. And the institutional response was to set the problem down and proceed as if it did not exist — not because anyone disputed the analysis, but because acting on it would require confronting the President of the United States.
The consequences accumulated. By 1967, with a $23 billion Treasury borrowing requirement driven by the Vietnam War, the System expanded bank reserves and the money stock at very rapid rates to accommodate the federal deficit. Darryl R. Francis, President of the Federal Reserve Bank of St. Louis, identified the accommodation as the primary driver of the inflation that was now accelerating — the same Francis who would become the central dissenting voice of the next episode. And by October 1972, John J. Balles, President of the Federal Reserve Bank of San Francisco, framed the Committee's failure not as a unique event but as an institutional pattern.
"The problem that we are faced with at present — namely, a huge Federal deficit in a period of strong economic expansion, is in fact new wine in an old bottle... I recall a discussion I had with some 'new economists' who believed that it was too early to start fighting inflation."
"New economists who believed that it was too early to start fighting inflation." Balles was naming the intellectual fashion that gave unanimity its respectable cover. The Committee did not vote unanimously to ignore inflation. It voted unanimously to defer to a prevailing doctrine — held by credentialed economists with proximity to the Administration — that said it was too early. There was always an analytical reason to wait. The unanimity was real, but it was a consensus about timing, not about substance. And the timing was always wrong in the same direction.
The vote was unanimous. The inflation that followed would take fifteen years and a Volcker shock to extinguish.
This is the first mechanism: political unanimity. The Committee agreed with the dissenters' analysis. It possessed the data, heard the warning, and acknowledged the risk. The unanimity was not intellectual consensus. It was institutional deference — a collective decision to subordinate the Committee's own analytical conclusion to an external political constraint, dressed in the language of economic prudence. And every individual's behavior was rational on its own terms. A Chairman who defied the Administration would spend political capital he needed for other battles. A Governor who pressed the point after the Chairman signaled deference would create friction without changing the outcome. A regional bank president who forced the issue would be overridden by the Board majority. Each member's calculation was locally sound. The collective result was fifteen years of inflation. The vote recorded agreement. The transcript recorded surrender.
The Redefinition
If the Martin-era failure was a case of knowing the answer and choosing not to act, the Burns era added a more sophisticated mechanism: changing the question so that not acting became the correct answer.
Arthur Burns assumed the chairmanship in 1970 and inherited the inflation that Martin's deference had seeded. The standard monetary response — maintain restraint until price pressures subsided — was the path the staff's own models supported. Governor J.L. Robertson had warned as recently as October 1969 that letting up too early would "risk a renewed surge of inflationary expectations" — a warning that predated Burns's arrival and came from inside the Board itself. By December 1969, staff analysis had gone further, admitting privately that the Phillips Curve — the framework connecting unemployment to inflation that underpinned the Committee's models — had "shifted" in ways that produced less favorable outcomes than predicted. The analytical foundation for the Committee's primary forecasting tool was already cracking before Burns reframed the problem.
Burns rejected the standard response. Not on the grounds that the data was wrong, but on the grounds that the type of inflation the country faced could not be addressed by monetary policy at all.
"Monetary and fiscal tools are inadequate for dealing with sources of price inflation such as are plaguing us now — that is, pressures on costs arising from excessive wage increases."
Burns's diagnosis was that inflation had become "cost-push" — driven by structural wage pressures rather than excess demand. If inflation was cost-push, then monetary tightening would reduce employment without reducing prices. The logical prescription was not restraint but an incomes policy: wage-price controls administered by the government, with the Federal Reserve providing accommodative monetary conditions to support employment during the transition.
This was not a fringe position. It drew on a real intellectual tradition. But inside the Committee, it performed a specific institutional function: it reclassified the dissenter's primary variable — the money supply — as irrelevant to the diagnosis. If inflation was structural rather than monetary, then Darryl Francis's insistence on controlling monetary aggregates was not merely wrong but categorically misplaced. And the Chairman gave the reclassification operational force. When the money supply overshot its targets in August 1970, Burns stated explicitly that he "would not be disturbed by an overshoot, particularly in light of the earlier undershoot." The Chairman was not merely tolerating excess money growth. He was authorizing it by name, in the transcript, as acceptable policy.
Francis, President of the Federal Reserve Bank of St. Louis — who had already identified Treasury accommodation as the driver of inflation during the Martin era — saw the reclassification for what it was. In June 1971, he presented the Committee with a comparison that no member rebutted on the merits.
"At today's meeting, the Committee was given three alternative annual rates of increase in M1 for the third quarter: 8, 9, and 10 per cent. By comparison, money had risen at an average rate of 5 per cent since 1964, a period of accelerating inflation."
The arithmetic was devastating. The Committee was debating whether to grow the money supply at eight, nine, or ten percent — when five percent growth since 1964 had already produced accelerating inflation. Francis did not need a model. He needed only the Committee's own recent history. And he drew the parallel explicitly, warning that if the Committee delayed a return to reduced monetary expansion, the result would not be a gradual return to price stability but a repetition of the failure it had just experienced.
"If such desires resulted in the Committee's delaying a return to a reduced rate of monetary expansion, the 1969–71 period, just like the 1966–67 episode, would be merely a costly pause in a trend of accelerating inflation."
"Merely a costly pause in a trend of accelerating inflation." Francis was not predicting the future. He was performing pattern recognition — identifying, from the Committee's own immediate past, a cycle where premature easing after insufficient tightening produced not recovery but the next inflationary wave. The 1966–67 episode he cited was five years old. The same members who had lived through it were sitting in the room. He was telling them they were doing it again. And the Committee, having lived through the previous failure, voted unanimously to repeat it.
The public heard none of this. Analysis of the period's communications reveals that roughly eighty percent of the Committee's internal debate over monetary aggregate targets was omitted from public statements. Where staff and members privately deliberated over M1 growth rates of eight to ten percent, the public received Burns's narrative about "thoughtful citizens" becoming convinced that an incomes policy was temporarily needed. The specific numerical stakes of the monetary debate — the very numbers that would have allowed outside observers to evaluate the Committee's choices — were filtered out entirely.
The vote was unanimous. Burns's cost-push diagnosis prevailed. The wage-price freeze of August 1971 — what Burns called "shock therapy, to rid the nation to the extent possible of inflationary expectations" — was implemented without monetary restraint. Inflation accelerated. Francis's prediction materialized within two years. The Decision Quality Matrix — the structured framework applied to each episode — classified the entire 1970–71 period with the harshest assessment in this investigation: three Predictable Failures and zero instances of Skill. The middle path, the cost-push diagnosis, and the wage-price freeze — all bad process, all bad outcome. The only non-failure was an unemployment forecast miss attributable to the GM strike.
This is the second mechanism: doctrinal unanimity. Burns did not suppress the dissent by political pressure. He did not need to. He redefined the analytical framework so that monetary tightening was categorically inappropriate for the type of inflation the country faced. Within the new framework, Francis's concern about money supply growth was not overridden — it was rendered irrelevant. The vote was unanimous because the Chairman had changed the question. The dissenter was answering a question the Committee no longer recognized as the right one to ask. And again, each member's behavior was individually rational. Burns's cost-push diagnosis was intellectually defensible — structural wage pressures were real, and the Phillips Curve was indeed shifting. The members who accepted the new framework were following the Chairman's lead on a question where the Chairman had genuine expertise. The collective result was that a sitting Committee member could demonstrate, with the Committee's own numbers, that they were repeating a failure from five years earlier — and the Committee voted unanimously to proceed.
Two episodes, two decades, two chairmen. Martin knew the answer and deferred. Burns changed the question. Both produced unanimity. Both produced predictable failures. And in both cases, the dissenting tradition — Francis at the St. Louis Fed, who had identified Treasury accommodation under Martin and excessive money growth under Burns — saw the pattern, named the pattern, and was overridden by the pattern.
The Insurance
The Greenspan era adds a third mechanism — one that is analytically distinct from the first two and, in an important respect, more forgivable.
In 2003, the Federal Reserve held the federal funds rate at one percent. The stated rationale was insurance against deflation — a tail risk that staff models estimated at thirty-five percent probability by April 2003. Japan's lost decade was the operative precedent. The Committee's risk-management framework treated the costs of deflation as catastrophic and asymmetric: if deflation took hold and the policy rate was already at the zero lower bound, the tools available to reverse it were untested and uncertain. The regret associated with easing unnecessarily was judged to be small. The regret associated with easing too little was judged to be enormous.
This was a defensible analytical position. The Decision Quality Matrix classified the 2003 rate decision not as a Predictable Failure but as Bad Luck: a good process that produced a bad outcome because the insurance against one crisis helped fuel a different crisis. The distinction matters. Martin's deference and Burns's redefinition were process failures. Greenspan's insurance was a process that correctly addressed a real risk and inadvertently created a new one. The staff's Core PCE projection of 1.0 percent for 2004 proved to be a significant pessimistic error — the actual outcome was approximately 2.2 percent. The deflation insurance paid out on a risk that did not materialize.
But the unanimity mechanism was still present. It operated not through political pressure or doctrinal reclassification but through the asymmetry of the loss function itself. In September 2002, a year before the rate reached one percent, Governor Ben Bernanke identified the failure mode with precision.
"One concern I do have about easing now is that it might exacerbate the imbalances in the economy by further heating up sectors that are already strong, such as residential construction and autos."
Bernanke saw it. Residential construction, already strong, would be further heated by additional easing. He named the sector. He identified the transmission channel. He described the outcome — imbalances — that would define the next crisis. Then he voted to ease. The deflation risk, in his assessment, outweighed the housing risk. The loss function was asymmetric, and the asymmetry was rational for each individual member: no governor wanted to be the one who allowed a deflationary spiral because of concern about house prices in Phoenix.
Governor Edward Gramlich provided the sharpest characterization of the institutional posture. In September 2002, he told the Committee that its approach to the recovery had produced a specific kind of complacency.
"One implication of this mixed picture is that the Committee may have gotten lulled to sleep. We have been looking for the smoking gun of a double-dip recession. We haven't seen it, and we have contented ourselves with a moderately performing economy for a longer and longer time."
"Lulled to sleep." Gramlich was naming the institutional condition that unanimity produces — a state in which the absence of crisis is mistaken for the presence of stability. A month earlier, he had noted his discomfort that housing refinancing had become the primary support for the economic recovery — a "weak reed" for a recovery to depend on. President J. Alfred Broaddus Jr. of the Richmond Fed warned in March 2003 that if the funds rate moved below one percent, "the public's concern about exactly how we're going to deal with a deflationary situation is going to intensify quickly" — identifying the signaling problem of rates at the zero lower bound. By September 2004, staff analysis had become more explicit, warning that rapidly rising home values, fueled by low real interest rates, represented a potential threat to financial stability and a misallocation of resources.
The Committee's response was recorded two months later.
"There was little overall concern about a bubble in house prices."
Between Bernanke's September 2002 warning about heating up residential construction and Stern's November 2004 dismissal of a housing bubble, the federal funds rate was cut to one percent and held there for a year. The Committee had alternatives. In August 2004, staff modeled a fifty-basis-point move to bring real rates to zero, but the Committee chose twenty-five basis points to meet market expectations of a "measured pace." Earlier, a Taylor rule-based path had been modeled but not adopted because the Committee preferred to maintain the baseline to "guard against a significant further decline in inflation." The roads not taken were not unknown. They were considered and rejected — each time unanimously, each time in favor of the path that kept the insurance in place.
This is the third mechanism: probabilistic unanimity. No one suppressed the housing concern. No one redefined it out of existence. The Committee heard it, weighed it, and rationally deprioritized it against a tail risk they judged to be more dangerous. Each member's individual risk calculus was defensible. The collective outcome was a one-percent policy rate that the Fed's own staff would later identify as the primary fuel for a "sizzling" housing market. Bernanke warned about the imbalance as a Governor in 2002. Five years later, as Chairman, he would face the consequences of the policy he had voted for — the person who warned becoming the person who inherited, a transformation that only unanimity makes possible.
The Quarantine
If the Greenspan-era failure was a case of rational deprioritization — seeing two risks and choosing the wrong one — the Bernanke-era failure that followed was something different. The Committee did not deprioritize the housing risk. It acknowledged it, discussed it at length, heard detailed warnings from multiple members, and then linguistically contained it so that the consensus could proceed unchanged.
The internal evidence of mounting risk was granular, specific, and early. In June 2005, staff analysis informed the Committee that house prices appeared overvalued by as much as twenty percent based on the historical price-rent ratio. Governor Mark Olson identified the structural vulnerability that mattered — not any single risk exposure, but the compounding effect of multiple exposures layered on top of each other.
"While each of the risk exposures appears to be both manageable and, to an extent, managed as isolated risks, they pose heightened risk in some areas due to the layering of the exposures. It's the layering that really causes the risk."
Governor Susan Schmidt Bies reinforced the warning in the same meeting, stating she was "not as sanguine as many of the staff" about the momentum building in housing. By August, Bies had sharpened her concern to the specific instruments that would become the crisis's accelerant.
"ARM products that have been pushed into the subprime market are much more problematic."
Bies — a Governor with supervisory experience, not a macroeconomist working from models — named the product class, named the market segment, and flagged the affordability risk. Olson identified the structural mechanism (layering). Bies identified the specific instrument (subprime ARMs). Between them, the two Governors had described the failure mode of the financial crisis that would arrive two years later. Their warnings were specific, early, and delivered in the room where the votes were taken.
The archive shows what happened to them. On March 20, 2007, William Dudley presented internal data showing that sixty-day delinquencies for the 2006 subprime ARM vintage were significantly higher than previous years and that spread widening had migrated up the capital structure. One day later, the Federal Reserve's public statement contained a sentence that the internal data directly contradicted.
"There was no sign of spillovers from the subprime market to the overall mortgage market."
The gap between the March 20 internal briefing and the March 21 public statement is the unanimity mechanism operating at its most visible. The data showed delinquencies jumping and credit stress migrating. The public heard that there were no spillovers. The filtering was not accidental. It was the institutional process by which private analytical complexity was compressed into public reassurance.
Chairman Bernanke maintained the containment narrative through mid-2007, stating publicly in May that the effect of subprime troubles on the broader housing market would be "limited." In June 2007, Dennis Lockhart, President of the Federal Reserve Bank of Atlanta, told the Committee that its own baseline forecast was insufficiently pessimistic — and named the specific alternative simulation that showed it.
"The Greenbook outlook reflects the baseline expectation of a diminishing drag on real growth from residential investment... However, as suggested in the Greenbook's first alternative simulation, we may be too sanguine."
"We may be too sanguine." Lockhart was pointing at the Greenbook's own alternative scenario — a staff-produced simulation showing worse outcomes — and telling the Committee that the baseline it was using to justify inaction was the optimistic case, not the central case. The Committee proceeded on the baseline. In June 2007, staff projected real GDP would rise 2.5 percent over 2008. The actual outcome was a decline of 0.8 percent — a forecast error of 3.3 percentage points, among the largest in the institution's history. The benign models were not challenged because challenging them would have required challenging the consensus the models supported.
By August, the gap between the Committee's public posture and the reality reported by its own members had become a subject of internal commentary. Eric Rosengren, President of the Federal Reserve Bank of Boston, told the Committee what its own communications looked like from the outside.
"It is notable that the rather benign outlook of the forecasters is in marked contrast to the angst I hear when talking to asset and hedge fund managers in Boston."
And then the observation that serves as the epigraph for this entire investigation.
"Some of the Boston hedge fund managers have observed that one dependable correlation has been that the announcement of no problem seems to be highly correlated with the actual problem's occurring with a lag of one to two weeks."
A sitting Federal Reserve Bank president, reporting to the Committee that the institution's own public communications had become a contrary indicator. Market participants had learned that when the Fed announced that a problem did not exist, the problem was about to arrive. The correlation was not theoretical. It was observable, and the participants who bore the risk of the Committee's decisions had identified it empirically.
This is the fourth mechanism: containment unanimity. The Committee did not defer to political pressure. It did not redefine the analytical framework. It did not rationally deprioritize. It acknowledged the risk, heard specific warnings from members with supervisory expertise, received internal data showing the risk materializing in real time — and linguistically quarantined it. The word "contained" performed the institutional work that political deference performed for Martin and doctrinal redefinition performed for Burns. It allowed the Committee to register the risk in the transcript while proceeding as if the risk did not require a change in policy. And once again, no individual's behavior was irrational. No member wanted to be the one who triggered a market panic by publicly naming the crisis before it arrived. The Chairman had a mandate to project stability. The staff had a baseline to defend. The Governors had supervisory concerns but no supervisory authority over the entities creating the risk. Each member's restraint was locally rational. The collective result was a 3.3 percentage point GDP forecast error and a financial crisis the institution's own members had described two years in advance. The vote was unanimous or near-unanimous through the critical period. The dissenters' warnings remained in the sealed transcripts. The public heard that there were no spillovers.
The information filtering across the four episodes can now be compared. What the staff told the Committee, and what the public heard, diverged by consistent and measurable margins.
| Episode | What Staff Identified | What the Public Heard | Estimated Dilution |
|---|---|---|---|
| Martin, 1964–66 | 88% capacity utilization threatens price stability | "Upward price movements may be only a flurry" | ~70% |
| Burns, 1970–71 | M1 growth targets of 8–10% debated internally | "Thoughtful citizens demand incomes policy" | ~80% |
| Greenspan, 2003–04 | 35% probability of deflation by late 2004 | A "remote" possibility | ~60% |
| Bernanke, 2005–07 | 20% housing overvaluation; delinquencies jumping | "No sign of spillovers" | ~85% |
In every case, the internal assessment was specific and quantified. In every case, the public communication was qualitative and reassuring. The dilution percentages are estimates derived from the proportion of internal analytical content omitted or reframed in public statements — they are not precise, but the direction is consistent. The filtering was not a conspiracy. It was the ordinary institutional process by which a committee that has voted unanimously communicates a confidence it does not privately share.
Four episodes have now established that the pattern holds across four chairmen, four decades, and four distinct mechanisms. But all four share a common directionality: the Committee was too loose or too inactive. The failures ran in one direction. If unanimity culture is truly structural — if it distorts the decision process itself rather than biasing it toward any particular outcome — then the pattern should also hold when the Committee acts in the opposite direction. The archive contains such an episode.
The Signal
On February 4, 1994, the Federal Reserve raised the federal funds rate by twenty-five basis points — the smallest possible increment — after holding at three percent for seventeen months. What followed was the worst bond market rout in a generation. Long-term rates surged. Mortgage-backed securities collapsed. Orange County, California declared bankruptcy. The Mexican peso crisis accelerated. Leveraged hedge funds were forced into liquidation. Global bond losses exceeded a trillion dollars.
The Committee had voted unanimously to tighten. The decision was preemptive — inflation had not yet materialized, but the economy was gaining momentum and the Committee judged that waiting would force larger, more disruptive moves later. This was, on its face, the opposite of the previous four episodes. The Committee acted rather than deferred. It tightened rather than accommodated. And the result was still a crisis.
The transcript reveals why. The Committee agreed on the direction and magnitude of the move. What it failed to model was the interaction between its announcement and a financial system that had spent seventeen months at three percent building leveraged positions predicated on continued low rates. Governor Susan Phillips had warned two months before the move that delay carried a specific risk beyond inflation.
"If we're too late — sort of a perennial Fed problem — we're going to be chasing [market] rates. And even more of a problem is that we'd lose the confidence of the bond markets, which is crucial for growth."
Phillips identified the sensitivity of the bond market to the Fed's timing. But the Committee's framework for managing that sensitivity was inadequate. Chairman Greenspan made the unusual decision to announce the rate increase immediately and publicly — a departure from the Fed's prior practice of allowing the market to infer policy changes from open market operations. The intention was transparency. The effect was the opposite. Staff analysis later noted that the Fed's firming action may have "contributed to disquiet... by suggesting the Fed knew something the market didn't" — that the public announcement, intended as clarity, was received as a signal of hidden danger.
Short-term interest rates increased considerably more than the twenty-five basis point hike — in some cases by sixty basis points — as leveraged positions unwound simultaneously. The intensity of the yield increase surprised most market observers and, critically, the Committee itself. William McDonough, Vice Chairman and President of the New York Fed, acknowledged the miscalculation at the following meeting.
"So I think we got a good deal more market effect and a good deal more tightening out of 25 basis points than we had in mind when we met on the 4th of February."
The smallest possible move produced a reaction several multiples larger than intended because the Committee had not modeled the leverage embedded in the system it was operating on.
The crisis deepened through the year as the Committee continued tightening — twenty-five, twenty-five, twenty-five, fifty, fifty, seventy-five basis points. By November, Vice Chairman Alan Blinder described the escalating pattern and its predictable effect on market expectations.
"I think this will be like feeding red meat to the bond market lions. They will chew it up and they will ask for more. A Federal Reserve that did 25, 25, 25, 50, 50, 75 does not look to an outside observer like it is about to stop."
Blinder saw what the Committee's sequential unanimity looked like from the outside: a pattern of acceleration that signaled more to come regardless of the economic data. Each unanimous vote to increase the increment reinforced the market's expectation that the next vote would increase it further. The Committee was not just setting interest rates. It was generating a signal through the pattern of its own decisions — and the signal's content was the opposite of its intention.
By March 1995, the Committee had the retrospective assessment. Governor Lawrence Lindsey stated it plainly.
"We actually had a pretty good bubble develop in the bond market, and we paid a fairly significant price for it."
The Committee had created a bubble with seventeen months of three percent rates. It burst the bubble with a twenty-five basis point announcement. It then overcorrected with escalating hikes that told the market the Fed was not about to stop. At each step, the vote was unanimous. At each step, the unanimity prevented the Committee from questioning its own momentum — from pausing to ask whether the pattern of its decisions was itself generating the instability it was trying to prevent.
This is the fifth mechanism: operational unanimity. The Committee agreed on the direction and the magnitude. No member argued for inaction. The disagreement was not about whether to tighten but about the second-order effects of how the tightening was communicated and sequenced — effects that no member modeled because the unanimity around the primary decision suppressed deliberation about the transmission. Phillips warned about the bond market's sensitivity. Blinder identified the signaling problem of escalating increments. Both were overridden not by argument but by momentum — the momentum of a committee that, having voted unanimously for twenty-five basis points, found it natural to vote unanimously for fifty, and then for seventy-five.
The 1994 episode prevents the argument from collapsing into a simple claim that the Fed should always have tightened more. The Fed tightened in 1994. Unanimously. And it blew up the bond market. The problem is not that unanimity biases the Committee toward looseness or toward inaction. The problem is that unanimity suppresses the uncertainty that would force the Committee to think harder about the consequences of its own decisions — in either direction. But this raises a final question: if the problem is unanimity itself, what happens when unanimity breaks?
The Exception
On August 9, 2011, the Federal Open Market Committee voted to commit to maintaining the federal funds rate at exceptionally low levels "at least through mid-2013." The vote was seven to three. Presidents Richard Fisher of Dallas, Narayana Kocherlakota of Minneapolis, and Charles Plosser of Philadelphia dissented — the most formal dissents at a single meeting in decades. This is the counter-case: the episode where unanimity broke, and where the archive can test whether visible disagreement produced a different institutional outcome than the private doubts that went unrecorded in five decades of unanimous votes.
The dissenters' objections were specific and structural. Fisher argued that tying forward guidance to a calendar date was both analytically wrong and institutionally dangerous.
"The '2013' just looks too politically convenient, and I don't want to fall back into people being suspicious about the way we conduct our business. So I just want to plead on that front. I think it's a mistake."
James Bullard of St. Louis — not a formal dissenter but a vocal critic — reinforced the point with a warning drawn from the Committee's own history.
"I do not think we would be wise to tie policy action to the calendar. We've already been burned by this twice, and if we do it today, we'll be headed down for a third time."
Plosser made the deeper analytical case. The economic weakness the Committee was responding to might not be the kind of problem monetary accommodation could solve. If the economy had entered a persistent regime of slower growth — a structural shift rather than a cyclical shortfall — then committing to a fixed date for near-zero rates was not just ineffective but misleading, a promise the Committee would have to either break or extend.
"I think it's too soon to tell the extent to which the disappointing economic data are telling us that the current weakness is temporary and will soon abate, or alternatively, that the economy has entered a more persistent regime of slower growth."
The Chairman's response to the dissent is the most revealing document in the archive for this investigation's purposes. Bernanke did not engage the analytical substance of the objections — the argument that calendar dates were structurally inferior to economic conditionality, or that the economy's weakness might be beyond monetary policy's reach. He counted votes.
"If we were to go back to alternative B as written with the forward-leaning language of the last paragraph, without the 'financial conditions', how many dissents would I have?"
The institutional reflex, even when unanimity had already broken, was to minimize the visible scope of the break rather than to ask whether the dissenters were right. The Chairman was managing the optics of disagreement, not the substance of it. The question was not "is calendar-based guidance analytically sound?" but "how many votes will I lose?"
Bernanke prevailed. The "mid-2013" date was adopted. The statement recorded three dissents. And then — over the following sixteen months — the dissenters' critique worked its way through the institution and replaced the framework they had opposed.
The process was slow. Staff privately warned that the effectiveness of calendar-based guidance hinged on a commitment the Committee could not credibly make — that it would adhere to a strategy requiring future choices that would be suboptimal at the time they were taken. By September 2011 — six weeks after the triple dissent — Bernanke himself admitted in deliberation that "the one thing everybody agrees on is that the mid-2013 language needs further elaboration." The date was extended to late 2014 in January 2012, then to mid-2015 in September 2012. Each extension confirmed the dissenters' structural point: a calendar commitment could not survive contact with changing economic conditions without being serially revised, and each revision eroded the credibility the commitment was designed to create.
On December 12, 2012, the Committee abandoned calendar-based forward guidance entirely. In its place, the FOMC adopted quantitative economic thresholds — what became known as the Evans Rule: the Committee would not raise rates as long as unemployment remained above 6.5 percent and inflation projections stayed below 2.5 percent. This was the framework the dissenters had argued for. State-contingent guidance tied to economic outcomes rather than dates. The transition took sixteen months. The dissenters' three votes — the visible, public, costly act of formal disagreement — had forced a fundamental change in the institution's communication architecture.
The Decision Quality Matrix classifies the original August 2011 adoption of calendar-based guidance as Good Luck: bad process, good outcome. The process was bad because it overrode substantive analytical objections to adopt a framework the staff itself doubted. The outcome was good — narrowly — because the date commitment happened to stabilize markets during a period of extreme volatility. The December 2012 transition to thresholds is classified as Skill: good process, good outcome. The Committee incorporated the dissenters' critique, used staff research to build a better framework, and produced a durable result. The dissent converted a Good Luck outcome into a Skill outcome. It improved the quality of the institution's decision-making process — the one thing the five unanimous episodes never achieved.
This is the answer to the counter-case question. When unanimity broke, it mattered. Not immediately — Bernanke overrode the objections in August 2011 and the calendar date was adopted. But the visibility of the dissent created institutional pressure that private doubt in unanimous votes never creates. The three names in the public statement — Fisher, Kocherlakota, Plosser — could not be sealed in a transcript for five years. They were on the record, in real time, and the analytical argument they attached to their votes entered the institution's deliberative process as a standing critique that had to be answered rather than absorbed. Plosser later described the slow-motion concession as "Chinese water torture" — the Committee adjusting the date forward, meeting by meeting, rather than confronting the structural flaw. But the structural flaw was confronted, eventually, because the dissent had made it impossible to ignore.
The comparison to the five unanimous episodes is direct. Ellis warned about unsustainable credit expansion in 1965 and the warning stayed in the sealed transcript. Francis warned about excessive money growth in 1971 and the warning stayed in the sealed transcript. Bernanke himself warned about housing imbalances in 2002 and voted to ease. Bies and Olson warned about risk layering in 2005 and the warning was filtered into "no sign of spillovers." In every case, the private analytical doubt was absorbed by the institution without altering the policy trajectory. In August 2011, the public analytical dissent forced a sixteen-month reassessment that replaced the framework. The mechanism that makes the difference is visibility. The institutional architecture that suppresses it is unanimity.
The Architecture
The prologue of this investigation named a phenomenon: collective irrationality — the documented condition in which rational individuals produce irrational collective outcomes. The six episodes have now supplied the evidence. In every failure, the individual behavior was defensible. Martin's deference was rational for a Chairman who needed the Administration's cooperation. Burns's cost-push diagnosis was intellectually defensible given real structural changes in wage bargaining. Bernanke's 2002 vote to ease was rational given the asymmetric loss function he faced. Every member who stayed silent in the face of the "contained" narrative in 2007 was rational — no individual wanted to trigger a panic. Every member who voted for the next increment in the 1994 tightening cycle was rational — the direction had been set and the data supported continued action. And the three dissenters in August 2011 were rational in a different way — they calculated that the institutional cost of visible disagreement was lower than the analytical cost of silence. In every case, rational behavior at the individual level. In every case, irrational outcomes at the collective level. This is the signature of a badly designed system.
The five failure mechanisms trace the specific pathways through which the system converts individual rationality into collective failure. Political unanimity under Martin: the Committee agreed with the dissenter's analysis and deferred to the Administration. Doctrinal unanimity under Burns: the Chairman redefined the framework so the dissenter's concern became categorically irrelevant. Probabilistic unanimity under Greenspan: the Committee acknowledged the risk and rationally deprioritized it against a tail risk it judged more dangerous. Containment unanimity under Bernanke: the Committee heard the warnings, received the data, and linguistically quarantined the risk so the consensus could proceed. Operational unanimity in 1994: the Committee agreed on the action and was collectively blind to how its own signaling would interact with the system it operated on. And the exception — August 2011, where three members broke the unanimity and, over sixteen months, the institution's framework changed.
The question this investigation set out to test — whether FOMC unanimity masks unresolved analytical disagreement that correlates with the institution's worst failures — is answered by the transcripts. In every failure episode, specific members identified the exact risk that materialized. In every failure episode, the vote was unanimous or near-unanimous. In every failure episode, the mechanism by which private doubt was converted into public confidence was different. But the conversion happened every time. And in the one episode where the conversion did not happen — where three members made their dissent visible, public, and costly — the institution's decision-making process improved.
There is a science behind this pattern. It is not new, and it is not specific to central banking. The field of social psychology has studied the conditions under which intelligent individuals, acting collectively, produce outcomes that none of them would have chosen alone. The foundational work is Irving Janis's study of groupthink, developed from the Bay of Pigs invasion and Pearl Harbor — episodes where cohesive groups of experienced decision-makers suppressed private doubts to preserve the appearance of consensus. Janis identified eight symptoms. The FOMC transcripts exhibit all of them.
The illusion of invulnerability appears in the 2005 staff assessment that the financial system would "bend but not break." Collective rationalization appears in Burns's cost-push diagnosis — a shared analytical framework that made the dissenter's evidence irrelevant by definition. Self-censorship appears in Ellis's mid-sentence hesitation in 1965 and in Bernanke's 2002 vote to ease despite his own warning about housing. The illusion of unanimity is the central finding: ten-to-zero votes recorded atop transcripts showing sharp analytical division. Direct pressure on dissenters appears in Martin's report that the Administration was strongly opposed. And the belief in inherent morality appears in the language of "insurance" and "precautionary measures" — framing inaction or accommodation as virtuous prudence rather than a policy choice with costs.
But Janis's framework, while accurate, does not fully explain the variation across episodes. The five FOMC mechanisms map to five distinct processes in the broader literature on collective irrationality. Burns's doctrinal redefinition resembles what Bikhchandani, Hirshleifer, and Welch described as an information cascade — a process in which early movers establish a direction and subsequent participants discard their private information in favor of following the visible consensus. The Chairman moved first with the cost-push diagnosis; the Committee followed despite Francis's private evidence to the contrary. The Greenspan era's probabilistic unanimity maps to what Prentice and Miller called pluralistic ignorance — a condition in which most members privately doubt the group's direction but each assumes the others support it, so no one objects. Bernanke and Gramlich privately saw the housing risk. Each assumed the consensus had weighed it correctly. Neither escalated.
The containment unanimity of the Bernanke era — where everyone saw the risk but no one owned the response — is the institutional expression of what Darley and Latané identified as diffusion of responsibility. Bies warned. Olson warned. Lockhart warned. Rosengren warned. The transcript shows the warnings. The vote shows that no one took responsibility for converting the warning into action. And the 1994 operational unanimity, where individually rational decisions produced a collectively irrational outcome, is a coordination failure in the sense Thomas Schelling described — a system in which each participant's expectations about others' behavior determine their own, creating equilibria that no individual intended.
Two additional mechanisms from the literature appear across multiple episodes without mapping neatly to any single one. Incentive misalignment — where going along is safer for career, reputation, or institutional standing than dissenting, even when the individual knows the decision is wrong — operates in every episode. The cost of dissent is measured not in the strength of the argument but in the friction it creates for the Chair; the benefit of dissent — vindication — arrives, if it arrives at all, years after the transcript is unsealed. The calculus is structurally biased toward silence. And group polarization — the tendency of groups to become more extreme after discussion, amplifying the dominant tendency rather than averaging views — appears most clearly in the 1994 tightening cycle, where each unanimous increment made the next, larger increment feel natural, and the Committee's collective position moved further from the center of individual views with each meeting.
The common thread across all seven mechanisms — the five from the failure episodes and the two that operate across them — is the principle named in this investigation's prologue: collective irrationality does not arise from individual stupidity. It arises from the interaction effects between rational agents inside a system that transforms what they say, notice, fear, and are rewarded for. The FOMC transcripts are the documentary proof. In every episode, the individuals are analytically competent and the collective outcome is catastrophic. The gap between individual competence and collective failure is not a paradox. It is the predictable output of the architecture.
Two quotes from opposite ends of this investigation reveal the same structural insight. In August 2007, Rosengren told the Committee that "the announcement of no problem seems to be highly correlated with the actual problem's occurring." In November 1994, Blinder told the Committee that "a Federal Reserve that did 25, 25, 25, 50, 50, 75 does not look to an outside observer like it is about to stop." Both are members telling the institution that its own signaling behavior — the public face of its unanimous decisions — was creating the crisis it was trying to prevent. One described the cost of unanimous inaction. The other described the cost of unanimous action. Both were overridden by the consensus. The institution's architecture does not distinguish between directions. It suppresses the uncertainty signal in both.
The mechanisms are embedded in the institution's design. The Chairman speaks last but signals first — setting the analytical frame that subsequent speakers must either accept or explicitly challenge. The transcript is sealed for five years, which means dissenting in the room carries no public cost and no public benefit; the dissenter's vindication arrives, if it arrives at all, after the damage is done. The norm of continuity — the institutional resistance to reversing direction between meetings — creates the momentum that Blinder identified in 1994, where each unanimous vote makes the next unanimous vote more natural regardless of what the data shows. And the culture of consensus itself — the expectation that the Committee will present a unified front — means that a formal dissent is treated not as analytical information but as institutional disruption. The cost of dissenting is measured not in the strength of one's argument but in the friction it creates for the Chair. Three dissents in August 2011 was described as the most in decades. That frequency — once in decades — is itself the measure of how high the institutional cost is. The five unanimous failures are not anomalies. They are the institution operating as designed.
The dissenter lineage across episodes illuminates the structural dimension. Ellis at the Boston Fed in 1965. Francis at the St. Louis Fed in 1967 and again in 1971. Gramlich and Bernanke on the Board in 2002. Bies and Olson on the Board in 2005. Lockhart at the Atlanta Fed and Rosengren at the Boston Fed in 2007. Phillips and Blinder on the Board in 1993–94. Fisher at Dallas, Plosser at Philadelphia, Bullard at St. Louis in 2011. The pattern is consistent: the members who saw the risk that materialized were either regional bank presidents with direct market contact or Board governors with supervisory backgrounds. They saw the ground truth — the actual condition of credit markets, the actual behavior of borrowers, the actual reaction of traders. The consensus was maintained by the members who saw the models — the abstractions that compressed ground truth into parameters and, in compressing it, lost the signal the dissenters were transmitting. The information that would have improved the decision existed inside the room. The institutional architecture ensured it did not reach the vote — or, when it did reach the vote, as in August 2011, it took sixteen months and three serial extensions of a failed framework before the institution conceded what the dissenters had said on day one. (This lineage of institutional dissent — the structural role it plays and the cost it carries — is the subject The Dissenter examined from a different angle.)
The Vote
The transcripts answer the question this investigation posed. The FOMC exhibits collective irrationality — the systematic production of outcomes that no individual member would have chosen — through mechanisms that vary in form but not in effect. Political deference under Martin. Doctrinal redefinition under Burns. Probabilistic deprioritization under Greenspan. Linguistic containment under Bernanke. Operational blindness in the 1994 tightening cycle. Each mechanism is distinct. Each is individually rational for every participant. Each produces the same collective result: a unanimous vote atop a transcript showing that specific members identified the exact failure that would follow.
The exception proves the architecture. In August 2011, three members paid the institutional cost of visible dissent — and the framework they opposed was replaced within sixteen months. The corrective mechanism exists. It is not that the FOMC lacks the capacity for self-correction, or that its members are unable to identify the risks their own policies create. The five failure episodes and the one exception demonstrate the same underlying truth from opposite directions: the information that would improve the Committee's decisions is present in the room. What determines whether that information reaches the policy outcome is not its quality but its visibility. Private doubt, recorded in sealed transcripts, produces no institutional pressure. Public dissent, recorded in the statement, forces a reckoning — slowly, reluctantly, through what one participant called Chinese water torture, but it forces one. Visible dissent breaks the conditions that produce collective irrationality: it makes private information public, forces the group to confront rather than absorb disagreement, and raises the cost of going along above the cost of speaking up.
The Federal Reserve's most consequential failures are not failures of intelligence. They are the predictable output of a system in which every individual's behavior is locally rational — deferring to the Administration, following the Chairman's analytical lead, deprioritizing the less catastrophic risk, avoiding the word that might trigger a panic, voting with the momentum of the previous meeting — and the collective outcome is catastrophic every time. The transcripts show twelve of the most capable economic minds on earth, with the best data and the best staff, producing the worst decisions in the institution's history. Not because they lacked the information. Because the system they operated within — its incentives, its norms, its information flows, its asymmetric costs of dissent — converted their individual competence into collective failure as reliably as it converted their private doubts into public confidence. The transcripts are the proof. The votes are the evidence of what the transcripts cost. And the rarity of dissent — once in decades — is the measure of how well the architecture works.
Search the Archive Yourself
The FOMC Insight Engine contains 352,000+ searchable passages from Federal Reserve transcripts, Tealbooks, minutes, and speeches spanning 1936–2025. Every claim in this article can be verified.
Explore the Archive →