clock menu more-arrow no yes mobile
Christina Animashaun/Vox

Filed under:

How I (and US policymakers) got inflation wrong

Bad takes, revisited.

Dylan Matthews is a senior correspondent and head writer for Vox's Future Perfect section and has worked at Vox since 2014. He is particularly interested in global health and pandemic prevention, anti-poverty efforts, economic policy and theory, and conflicts about the right way to do philanthropy.

Let’s get this out of the way first: I was wrong about inflation.

When prices began ticking up a little faster than normal early last year, I wasn’t overly concerned. I’d been covering economic policy since 2008, and in that whole time (in fact, in my whole lifetime!), the US had never had a problem with excess inflation.

In fact, our main inflation problem was that we had too little of it. The Federal Reserve, which is tasked with managing the money supply to keep inflation steady and unemployment low, set a low inflation target of 2 percent a year, and kept falling short. If there is, as most economists believe, some trade-off between inflation and unemployment, that means the Fed’s policies on inflation were unduly hawkish and kept many people out of work during the long recovery from the Great Recession.

So I wrote a long piece last summer arguing that high inflation was unlikely in the 2020s. When fears arose last fall that rising inflation expectations — that is, people thinking inflation was going to be higher in the future — could in turn lead to more inflation now, I wrote a newsletter citing research that cast doubt on that theory. This January 1, I predicted that average US inflation for the year would be below 3 percent.

Well, in February, core inflation as measured by the personal consumption expenditures (PCE) index grew by 5.4 percent, and seems to still be heating up. That’s my preferred metric, but data released on April 12 showed that the more widely publicized consumer price index grew by 8.5 percent compared to the year before, the highest rate in four decades. The US would be very lucky to keep inflation below 5 percent for the year at this point; my prediction of 3 percent looks, three months later, ridiculous.

At the time I wrote my July 2021 piece, “Don’t worry about inflation,” a prescient copy editor noted that this headline might look bad if I was wrong and inflation got increasingly worse. I responded that I stood by it, and if I was wrong, I would write a groveling follow-up piece.

So here we are.

In my defense, I wasn’t alone. A lot of people much better credentialed than me argued that inflation was going to cool down substantially in 2022. When I made my prediction, I cited the Federal Reserve’s policymakers, whose median forecast as of December 15, 2021, was for 2.7 percent inflation in 2022, per the core PCE measure (their favorite metric as well as mine).

But in March, the Fed updated its projection from 2.7 percent to 4.1 percent. Private forecasters surveyed in November said they expected 2.3 percent inflation; by February, it was up to 3.1 percent.

Put another way, the people you’d expect to have the surest handle on where inflation is going have admitted they got it wrong, too — and by a lot. “The models really led us astray,” Karen Dynan, a professor of economics at Harvard and former assistant secretary of the treasury for economic policy, told me, offering the below chart as further evidence.

Forecasters failed to predict surging inflation
Predictions from this time in 2021 about the year’s inflation were far off the mark.
PIIE/Karen Dynan

At the same time, the error wasn’t inevitable. Economists like Larry Summers, Olivier Blanchard, and the team at the Committee for a Responsible Federal Budget were all warning about high inflation in early 2021. They got it right, or at least more right than I did.

So, why did so many forecasters, and I, get this so wrong?

I can’t tell you what’s in Fed chair Jerome Powell’s heart, despite my past entreaties to the Fed media department. But I can say in regards to my case that I unfairly dismissed the most boring, Econ 101 explanation for why inflation happens: that there was too much money sloshing around for the amount of stuff the economy was able to produce — meaning the price of that stuff went up.

How we got to that point is a complicated story, involving a massive pandemic, trillions in relief money, and a jittery, uneven reopening. So let’s dive in.

A very brief history of inflation theories

To understand what I got wrong, it helps to understand the theory of the economy I was reacting against. And to understand that theory, it helps to understand the theory that theory was reacting against. So let’s start all the way back in 1958.

The world’s most famous inflation theory — really, less a theory than a chart — is the Phillips curve, illustrated below by the Federal Reserve Bank of St. Louis.

An illustration of the Phillips curve St. Louis Fed/Kristie Engemann

In 1958, A.W. Phillips, a New Zealand-born economist, published a paper titled “The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861–1957.” In the paper, he plotted exactly that: unemployment on the x-axis, and the growth of wages on the y-axis.

There was, he concluded, a relationship: When unemployment was lower, wages rose faster; when unemployment was higher, wage growth was weaker. Because wages are a major cost for businesses, those businesses sometimes have to pay for higher wages by raising their prices. So the relationship was not just between unemployment and wages, but between unemployment and prices generally.

This makes some intuitive sense. The best explanation of the Phillips curve I’ve heard comes from Barry Pepper’s Wall Street trader in Spike Lee’s classic film 25th Hour: “More jobs means fewer people looking for work, means it’s harder to find good people to fill those jobs, means you got to raise wages to get them, means inflation goes up. You got it?”

The idea that there was a straightforward trade-off between unemployment and inflation proved incredibly popular in the years after Phillips’s paper, including with policymakers in the United States.

As Brad DeLong argued in his excellent history of the Great Inflation of the 1970s, US policymakers of the previous decade thought they could just move leftward on the Phillips curve, to a point with higher inflation and lower unemployment, and increase jobs without much pain. They didn’t want to give up Great Society spending programs or the military buildup demanded by the Vietnam War, and were willing to accept that all that spending drove unemployment down and inflation up.

They were wrong. In the 1970s, the US experienced something bizarre: stagflation, or a combination of slow growth/elevated unemployment and high inflation. The Phillips curve, in other words, broke down: There was not a simple trade-off between inflation and unemployment; as it turned out, both could rise simultaneously.

The end of the Phillips curve — and the rise of NAIRU

So the Phillips curve was gradually supplanted among policymakers by a concept called NAIRU: the non-accelerating inflation rate of unemployment. Fed policy from the late 1980s onward can largely be explained by this idea.

This theory proposes that there is a relationship between inflation and unemployment, but that relationship isn’t linear. If unemployment is above the NAIRU rate as estimated by the Fed — say, 8 percent when NAIRU is 4.4 percent — then the labor market doesn’t have much to do with inflation. But if the unemployment level falls below the NAIRU, then businesses will be compelled to raise prices. And, importantly, they start raising prices faster and faster, meaning the inflation rate would accelerate, not just rise modestly as unemployment falls.

This happens because of expectations: After the initial price increases, businesses and consumers start to expect inflation and plan for it. Workers might demand more money because they know $1,000 today will be worth a lot more than $1,000 a year or even a month from now. Businesses will raise prices for the same reasons — and because they need to pay their workers more.

By the mid- to late 2010s, NAIRU held considerable sway at the Fed. It was, I argued in 2019, bad for workers. By creating a fear of “too low” unemployment, NAIRU led to slower economic recoveries with more people forced out of work.

Thankfully, its reign was coming to an end. By late 2015, unemployment was falling below 5 percent, bottoming out at 3.5 percent in January/February 2020, and at no point did inflation seem to spiral up the way it should if we’d gone below NAIRU. All this made NAIRU theory look more than a little silly.

The Fed eventually came around to that perspective. In 2020, the Fed under Jerome Powell removed references to NAIRU from its statement of strategy, which is meant to explain the institution’s policy approach to the public and to markets. Before, the Fed committed to acting when employment was above what the Fed considered its maximum level; under Powell, it switched to only caring about “shortfalls of employment from its maximum level,” not overly high employment. In other words, the Fed obsession with keeping inflation low — even at the expense of jobs — was over.

Why did the intellectual tide turn against NAIRU? Basically, because it was hard to see support for it in the data. For one thing, no one knew what the actual NAIRU level was that is, how low unemployment could go without causing inflation. It could have been 5.5 percent unemployment; it could have been 3 percent; it could have been 0.5 percent!

What’s worse, the level would certainly change over time, as the economy itself changed and had different labor needs. Before the 2008 financial crisis, the Fed’s estimate of NAIRU was 5 percent. By January 2011, it had raised the estimate to 6.6 percent. By 2015, the Fed estimate had fallen back down to 5.3 percent.

In other words, NAIRU didn’t seem all that helpful — and Powell’s Fed acted accordingly.

What our measure of unemployment missed

Fast-forward to July 2021. At the time I wrote my piece, inflation was at 3.6 percent.

Nearly a year later, it shows no sign of relenting. Inflation is now a full-blown national worry for the first time in decades. Has NAIRU been vindicated?

Not really. I don’t think any of the reasoning about NAIRU above, which profoundly affected my view of inflation in the pandemic, was wrong per se. I still don’t think NAIRU, taken at face value, is an adequate view of the labor market and inflation, and I don’t think that our 3.6 percent unemployment rate right now is too low, as NAIRU would have it.

But if NAIRU can’t explain what’s happening with inflation now, what can? I think there’s a simpler theory to consider — one that foregrounds excessive spending during the pandemic and how it might make better sense of our current realities.

NAIRU uses unemployment as a sort of proxy. It’s a proxy for how tight labor markets are: how hard it is for businesses to find new workers, how much they have to pay them to join or stay, etc. It’s also a proxy for spending in the economy. Jobs are the main way people get money, so when more people have jobs, more people have money to spend.

Both of these proxies broke down during the pandemic: The unemployment rate wasn’t a good proxy for the overall labor market, and stimulus and unemployment insurance benefit payments made it a terrible proxy for spending.

Let’s unpack this a bit. We already knew from the Great Recession that the unemployment rate is an imperfect measure of the labor market, because it doesn’t include people who not only lost their jobs but also stopped looking for work. For that reason, many economists and journalists, myself included, have come to rely on the employment-to-population ratio among workers ages 25 to 54, or “prime-age EPOP,” which took these missing workers into account.

But in the aftermath of Covid, the various measures of labor market tightness started to diverge, as former Obama chief economist Jason Furman has documented:

If you looked at prime-age EPOP, formerly the sophisticate’s preferred metric, the labor market didn’t look very tight at all. It still had a long way to go before you’d expect wages to start rising substantially, possibly triggering inflation and a wage-price spiral.

But if you looked at another measure, the share of people quitting their jobs, the market looked extremely tight. And as Furman and Harvard research associate Wilson Powell have found in other work, the quits rate does a similarly good job as prime-age EPOP at explaining people’s wages.

So the unemployment or prime EPOP rate might not necessarily capture the labor market.

At the same time, those metrics — which in the past correlated very well with spending — really ceased being a good measure of consumer spending during the pandemic.

The stimulus put a lot of money in our pockets

The US responded to the recession with an unprecedented surge in government spending programs, sending out $3,200 per person in stimulus checks, up to $600 a week in bonus unemployment insurance, boosted food stamp benefits, monthly child tax credit payments in 2021, and more. As of the end of 2021, JPMorgan Chase found that household checking account balances among low-income families were still about 65 percent higher than in 2019; balances among families in the top quarter of the distribution were 30 to 35 percent higher, too.

While the balances remained elevated, JPMorgan Chase researchers also found that families were steadily spending them down over time. That meant accelerated consumer spending, but spending that wasn’t necessarily connected to whether people are getting jobs.

In particular, as many have noted, it meant accelerated spending in goods. During the pandemic, people have been understandably less interested in going to restaurants, taking yoga classes, having people come to their homes to fix appliances, etc. So they bought stuff instead. A lot of stuff. I certainly did; I still don’t fully understand how I wound up with three mechanical keyboards, but it happened.

Past stimulus checks during non-pandemic episodes have been disproportionately spent on durable goods, rather than services, suggesting that the stimulus checks might have accelerated this phenomenon just as the virus did. And because prices of goods tend to be less “sticky” than prices of services (meaning they tend to rise and fall more easily), this especially contributed to inflation.

This surge in spending led to big, well-publicized shortages in certain areas, most famously cars, as demand for durable goods outstripped the economy’s ability to produce them (sick workers limiting production was a factor, too, if a smaller one). That provoked localized price spikes on a few goods. And because oil producers slowed production in expectation of a big post-Covid recession, they too struggled to keep up with demand, so gas prices rose — which Putin’s invasion of Ukraine only worsened.

For a while, many commentators thought you could wave off inflation fears by saying it was just limited in a few sectors. But at this point, an “inflation in a few places” theory doesn’t really fly.

Some goods, like oil and cars, have specific narratives like a chip shortage or low drilling that could explain inflation. But as Bloomberg’s John Authers has detailed, inflation is still rising even if you exclude those goods. The Dallas Fed’s “trimmed mean” inflation measure, which purposely removes “outliers” where prices are rising extremely fast or extremely slow from the data, started to shoot up recently, too. Check out that soaring line on the righthand side of the chart below.

If you remove outlier categories, inflation’s still shooting up Federal Reserve Bank of Dallas

Inflation now is pretty broad-based.

And the explanation seems fairly straightforward: Due to a combination of rapidly growing wages through all of 2021, plus trillions in government fiscal support, there has just been too much money around combined with insufficient goods and services to spend it on.

That’s led to not just inflation but accelerating inflation, as wage increases contribute to price increases and higher expectations of future inflation contribute to higher immediate inflation. That’s why you’ve started to see inflation in categories beyond just gas and cars. It’s a situation similar to what NAIRU would predict, except I would argue it’s not really about low unemployment.

This story — that total spending is what matters — has a long pedigree in economics. After the Great Recession, a number of economists (like Bentley professor and influential blogger Scott Sumner, former Obama chief economist Christina Romer, and monetary policy expert Michael Woodford) proposed that the Fed switch from targeting inflation to targeting nominal gross domestic product, or NGDP, which is basically the sum of all spending in the economy, not adjusted for inflation. That sounds technical, but it basically means that the Fed should aim for total spending to grow at a steady rate.

When spending grows too slowly, you get 2008-2016, a period when inflation was reasonable but economic growth and job growth were too slow. When spending grows too quickly, as NGDP-targeting supporter and economist David Beckworth concludes it is right now compared to the historical trend, you get an inflationary situation like our current predicament.

During the Great Recession, the NGDP clique were widely seen as monetary doves, because they wanted the Fed to be aggressively driving down unemployment and not worrying about inflation. Now folks like Sumner come across as hawks, because he wants the Fed to tighten and keep inflation under control. But it’s all just the same position: Spending should grow steadily, neither too fast nor too slow.

Am I just being wildly wrong … again?

Plenty of people disagree with the theory outlined above.

One obvious objection is that consumer spending overall isn’t much higher than the pre-pandemic trend. It just happens to be more concentrated in goods than services.

This is true, but I think it ignores the fact that the economy isn’t back to its full level of pre-pandemic health. People are still getting sick and missing work. Some people are still afraid to go to stores or use in-person services. So pre-pandemic spending levels might be more spending than the economy, as currently constituted, can absorb.

“I don’t think you need any special exotic theories to say that, after you completely ripped an economy apart, it couldn’t grow past its pre-pandemic potential,” Furman told me.

Skanda Amarnath, executive director of the group Employ America, which pushes for the Fed to prioritize reducing unemployment, agreed that spending is a part of the story, including spending due to federal stimulus.

“You shouldn’t be a denialist about fiscal policy accelerating cyclical recoveries [recoveries from recessions], and cyclical recoveries have a role in driving inflation,” he told me. But he pushed back on the idea that the labor market is excessively tight, leading to wage gains, which in turn lead to more inflation. It’s hard to make sense of what’s happened to durable goods in that story, he noted, as most durable goods are made overseas and aren’t too sensitive to prices in the US.

This is fair, and I think the idea of higher wages driving prices higher is the weakest part of the story outlined above. As economist Preston Mui has noted, employers in the US have a lot of power to set wages, which weakens the argument that workers can simply demand higher wages because prices are rising. But I think a story about inflation expectations can still make sense here. Firms expect more inflation in the future, and know their competitors do, too, so they can afford to raise prices now without losing out on business.

This expectations story has come under profound challenge too. David Reifschneider and David Wilcox, two longtime Fed economists, wrote a smart piece for the Peterson Institute for International Economics arguing higher inflation expectations are unlikely to be driving inflation this year. What has mattered most in models of inflation, they write, is long-run inflation: how much higher prices will be in several years, not a few months. Long-run expectations aren’t actually that high right now, indicating that people expect prices to come down, and inflation won’t accelerate the way my theory suggests it currently is.

But Reifschneider and Wilcox’s evidence comes largely from a long period when the US didn’t have a serious inflation problem; it had pretty stable, consistent 1 to 3 percent inflation year after year. That’s very easy to plan for, and is a context where long-run inflation expectations are probably the most important thing.

But when inflation is rising quickly, like right now, short-run inflation expectations might start to matter more. “If you look back to the 1970s and ’80s when inflation was high and exciting and noticeable,” short-run expectations mattered more, Dynan says. “With inflation [today] getting close to those levels, it turns out people are noticing it and it’s playing a more important role.”

What now?

On March 16, the Federal Reserve raised interest rates for the first time since 2018, the beginning of its attempt to tackle inflation. The rate hike was greeted with near-unanimous support from economists, even progressive ones, who recognize the inflation problem has gotten quite serious. But there are some dissenters whose views deserve to be taken seriously.

J.W. Mason, a professor at John Jay College and fellow at the Roosevelt Institute, outlined the case against a hike on his blog. The basic rationale is simple: Rate hikes hurt the economy. The entire reason they work against inflation is that they hurt the economy. They control inflation by causing “less spending in the economy, lower incomes, and less employment.” And Mason simply did not think the current inflation problem is serious enough to run that risk.

I empathize profoundly with that view. Rate hikes are a bit like chemotherapy: They try to prevent future pain and harm by inflicting substantial pain right now. They kill jobs and lower wages, and the damage is especially severe for the lowest-paid workers. I think I would agree with Mason and oppose all rate hikes if I put very little odds on the possibility of inflation spiraling further into the double digits.

I still put a low probability on that possibility — but it’s not zero. 2021 taught me that inflation could get a great deal worse than I expected, and that I shouldn’t discount the possibility that inaction by the Fed would lead to accelerating, not just higher, inflation. That might have been what happened in the 1970s, and it could happen now.

More to the point, I worry that inaction might simply delay the inevitable, and that the rate hikes that would be necessary to control inflation in a year would be larger, graver, and more devastating to people than the modest hikes the Fed has planned this year.

After years of dallying by the Fed in the 1970s, Fed chair Paul Volcker successfully ended inflation — but only at enormous cost. He directly caused two deep recessions in the early ’80s that drove the unemployment rate to its highest level since the Great Depression. The process worked, Reagan adviser Michael Mussa later said, because the Fed proved it was willing “to spill blood, lots of blood, other people’s blood” to get inflation under control. (There’s a reason the policy was dubbed the “Volcker shock.”)

A basic principle of governance should be limiting the amount of blood spilled. And at the moment, I think we can do that by hiking mildly now, rather than dramatically in a little while.

But I’m open to being proved wrong here, too. After all, I’ve been wrong before.

Supreme Court

The courts were never going to save America from Donald Trump

Supreme Court

The Supreme Court just crushed any hope that Trump could be removed from the ballot


The conspiracy theories about Kate Middleton’s disappearance, explained

View all stories in Politics

Sign up for the newsletter Today, Explained

Understand the world with a daily explainer plus the most compelling stories of the day.