Transitory: The Word the Fed Wishes It Could Take Back

The Word

Every political era is compressed into a single word. For 2021 that word is transientIt was the Federal Reserve's official description of the inflation that appeared that spring meaning temporary self-correcting not something for which interest rates are raised. Inflation the rate at which consumer prices rise had been below the Federal Reserve's 2 percent target for nearly a decade. In January 2021 the consumer price index (CPI) the government's main indicator of what a basket of everyday goods and services costs stood at1.4 percent. In December it was 7 percent the highest reading since 1982. The Federal Reserve spent all of that year with its policy rate at zero still buying $120 billion in bonds a month to pump money into the economy. This article is about the costliest forecast error of the decade: how it happened and why it looked more reasonable in real time than hindsight admits

What the Fed Saw in April

The first alarming figure came in May 2021 when the April CPI report was 4.2 percent versus expectations close to 3.6. Go back and read that report today and the transitory story really holds up on a first pass. Used car prices which rose 21 percent in three months alone accounted for a third of the monthly increase. The global chip shortage had crippled new car production so buyersThey flocked to the used market. Airline tickets and hotels were recovering from a pandemic bottom that most airlines never expected to see again. and base effects the statistical peculiarity of comparing this year's prices with last year's unusually low ones inflated year after year by construction. Each hot component had a visible unique and specific story attached to the pandemic. Bottlenecks are cleared. Base effects disappear. Inflation returns on its own to 2 percent. That was the thesis and it was not stupid

Why the Thesis Missed the Demand Side

Thesis lost the other half of its own ledger. Roughly $5 trillion of fiscal support covered in our CARES Act entry had hit household and business balance sheets since 2020. In 2021 a third round of stimulus checks arrived just as vaccines allowed the economy to reopen. Consumers weren't simply catching up on delayed purchases. They were spending savings that were unprecedented in any previous recovery and that spending hit chainssupply chains that were still operating below capacity. Too much money chasing too few goods is the oldest recipe for inflation in economics. It was simply hidden within twelve months of plausible and unique stories

Breadth Was the Tell

Here's the mechanical problem. Each month a different category could explain the figure: cars then rent then food then energy. Any hot price is a bottleneck story. The acceleration of the median price across the entire basket is something else entirely and in fall 2021 the median was shifting. Rent deserves its own phrase here. It's about a third of the CPI basket moving slowly because leases reset gradually and once rents started rising there was none lefttemporary story that covered it

A Worked Example: What Breadth Actually Looks Like

Let me nail down the idea of breadth with round illustrative numbers not the actual 2021 CPI weights. Suppose a price index has 10 categories each equally weighted at 10 percent of the basket and you want to know whether inflation is broad or narrow. In the first month only 2 of the 10 categories are above 3 percent annualized. Call it a 20 percent spread reading 2divided by 10. That's a bottleneck story: two categories easily named easily excusable

Now suppose that three months later 7 of the 10 categories are above 3 percent. Diffusion is now 70 percent 7 divided by 10. Even if the headline number only went from 3.0 percent to 4.5 percent a move of one and a half points sounds modest on its own. The jump in diffusion from 20 to 70 percent indicates something the headline is hiding: Acceleration stopped being two noisy categories and became propertyof the entire basket. If we weight it the same math holds. At most 20 percent of the basket's weight drove the first reading. At least 70 percent drives the second. That's the arithmetic case for looking at breadth rather than a single number: the same headline can be a bottleneck or genuine inflation and the difference only appears once the categories are counted

Apply that lens to the actual data from 2021 and the pattern will track what really happened. In the spring the story really focused on two or three categories including used cars. By the fall this article already notes that rent food and energy had added up.The CPI report didn't need to print the word breadth for that change to be visible. It just had to count how many line items it forgave each month and notice that the count kept increasing

Case Study: Ford, GM, and the Chip Shortage That Ate the Used Car Market

The clearest real-world example of how a bottleneck can disrupt an entire CPI printout is in the auto industry. Beginning in late 2020 a global shortage of semiconductors the chips that power everything from infotainment displays to transmission control units forced automakers to halt assembly lines. Ford and General Motors repeatedly reduced production through 2021 in some cases building partially finished pickup trucks and parking them waiting for chips thatThey hadn't arrived yet. Toyota which had stockpiled chips after supply disruptions related to the 2011 tsunami held out longer than most before having to cut production as well

The supply of new cars was reduced to the demand that the pandemic had not reduced at all so buyers who could not get a new vehicle bid on a used one. That is the mechanism behind the 21 percent increase in three months in used car prices already cited in this article. It is also in miniature exactly why the so-called transition seemed reasonable in real time.up to date with orders. What no one took into account was how long chip delivery times would stretch some semiconductor orders reportedly saw wait times exceed a year in late 2021 and how much a single supply shock could contaminate a price index broad enough to look like general inflation from a distance. The chip shortage was real and temporary in the sense that it eventually subsided. It still helped produce a CPI figure that held up for two more years.transitory and the consequent turned out not to be opposites

The Retirement of the Word

On November 30 2021 in testimony before the Senate Banking Committee Chairman Jerome Powell formally surrendered. He said it was a good time to retire the word transitional. The pivot that followed occurred at full speed. Bond buying slowed within months and in March 2022 the Federal Reserve began the fastest rate-hike cycle in four decades the subject of the next entry in this series. The cost of delay is the counterfactual on which allThe world is still arguing. If the tightening had started six months earlier inflation would likely peak lower and the 2022 repricing would extend more smoothly. Instead the Fed compressed roughly two years of tightening into one and the violence of that rally is what bankrupted the bond market destroyed the traditional 60/40 portfolio and ultimately helped bring down Silicon Valley Bank

The Defense the Fed Deserves

It's easy to dive into the transitory now that we know how the movie ends. Before you do that it's worth trying to make the decision because the honest version of this story is more useful than the easy version

Get started with the framework. In August 2020 the Federal Reserve adopted a new approach called flexible average inflation target built specifically because the previous decade had taught it the opposite lesson from what 2021 seemed to demand. Throughout the 2010s the Fed had repeatedly tightened or worried aloud about doing so ahead of an inflation that never showed up and in doing so arguably stifled labor market gains for workers often lower-wage workers and minority workers who tend to see only real wage growth once the recovery intensifies for a while.time. The new framework explicitly said: do not take preventative measures. Let inflation exceed the target for a while to make up for years of being below it and wait for real sustained evidence before applying restrictive measures. Applied to 2021 that framework did exactly what it was created to do. It just wasn't built for a pandemic

Second the data itself was truly unreadable in a way that is now easy to forget. Seasonal adjustment models are based on decades of normal years and 2020 had no normal pattern around which to adjust. Supply chains were tangled in ways without modern precedent: a demand shock and a supply shock landed at the same time around the world in the middle of a public health emergency. No one inside or outside the Federal Reserve had a clear historical playbook for that combination.specific

Third the Federal Reserve was not the only one that made a mistake. Market measures of expected inflation discounted through the Treasury Inflation Protected Securities (TIPS) market remained relatively contained for most of 2021. Most private bank economists had similar forecasts. The strongest dissenting voice former Treasury Secretary Larry Summers warned as early as February 2021 that the magnitude of the pandemic stimulus was at risk.of overheating the economy and it turned out he was right. But Summers had also warned about the risk of inflation at other times in the previous decade when it didn't materialize so from inside the Fed in early 2021 weighing an economist's warning with a market that wasn't pricing in a serious inflation risk and a staff forecast that matched the Fed's own record wasn't an obviously dumb decision. It seems obvious only from a December vantage point

How I Actually Read Fed Language Now

My reading after reviewing this entire episode is that the real lesson is not "the Fed is bad at forecasting." Everyone is bad at eighteen-month forecasting including me. The lesson is about which parts of a forecast to trust

The way I would actually use this now is by separating a central bank's diagnosis from its forecast. Diagnosing what's driving inflation right now tends to be spot on because it's mostly reading data that already exists. The Federal Reserve correctly identified that used cars airfares and base effects drove the April 2021 number. That part of the transitory story was accurate all along. The forecast what will happen to that inflation over the next year is another animalbecause it requires that the diagnosis be complete and that no new causes appear. That is the part that failed. The fiscal stimulus and pandemic-era savings were never fully included in the original diagnosis and when they were six months of increase had already been lost

So when I read a central bank statement now I ask a specific question: Is this statement referring to what has already happened or what will continue to happen? I trust the former much more than the latter. I also look at the breadth of any impression for the reason discussed above and pay attention to whether the strongest disagreement comes from someone with a track record right or wrong or from someone who says the same thing every cycle. Just because Summers is right in 2021 doesn't mean Summers is always right.reason. It means that your specific argument that a stimulus of this size relative to the size of the output gap was historically unusual deserved more weight than the "another economist is worried about inflation again" typically does

I also want to be honest about the limits of my own vision. I didn't live through 2021 as an investor or as a policymaker. I'm piecing this together from data and after-the-fact testimony which is exactly the advantage hindsight gives you and exactly why I'm trying not to take pride in a call where a lot of very smart very well-informed people got it wrong in real time

The Bottom Line

The transitional hypothesis was defensible in April 2021 shakier in the summer and dead by Thanksgiving and the Fed maintained it about six months after the moment the evidence changed. The price was the highest inflation in forty years and the brutal tightening of the 2022 recovery strategy. The framework that produced the failure was a reasonable response to the mistakes of the previous decade the data was indeed unusually difficult to read and the Fed was not the only oneYou were wrong. None of that changes the outcome and it shouldn't. For anyone learning macro 2021 is the canonical case study in how anchoring frame-lag and category-by-category excuses can blind the best-staffed economics institution to the oldest standard in its own textbook: Breadth is what it says not a single number