In a recent PIIE Policy Brief , David Reifschneider and David Wilcox argue that the high inflation in 2021 was largely the result of special factors and that many of these factors are likely to fade in 2022 and 2023. They argue that long-term expectations of inflation have remained stable. Thus, they conclude, with the proper caveats, that most of the high inflation will go away on its own and the Federal Reserve does not have to hit the brakes too strongly. It just has to get the policy interest rate close to the neutral rate. The landing is likely to be smooth.
They may be right (and from the bottom of my heart, I hope they are, for the sake of the US economy). But I am more worried than they are.
Uncomfortably Like the 1970s
Let me begin with a graph that was motivated by the highly relevant forthcoming book by Alan Blinder on soft and hard landings, which he discussed at Princeton's Markus Academy in February 2022.
Blinder looks at Fed tightening episodes since 1965. He identifies 11 episodes and excludes 2 (the global financial crisis and the COVID-19 crisis) where what happened to activity clearly did not come primarily from monetary tightening. He concludes that there was a soft landing in 6 of the remaining 9 cases, and a harder landing in the other 3. I read his implicit bottom line as optimistic and in line with the Reifschneider-Wilcox piece: We can probably land softly from where we are today.
I am not convinced. I think that, in most of the episodes Blinder has identified, the movements in inflation to which the Fed reacted were too small to be of direct relevance to the current situation, and the only comparable episode to today, if any, is the episode that ended with the Volcker disinflation of the early 1980s.
I find that a scary comparison. The figure below shows why. It plots the core inflation rate (in red) and the real policy interest rate (in blue), constructed as the federal funds rate minus core inflation over the previous 12 months. (The results would be even more striking if I had used headline inflation, but, for the usual reasons, we want a series that removes, at least in part, the transitory components of inflation.)
I draw two main conclusions from the figure.
First, the figure shows how little action there was in either inflation or the real policy rate relative to the 1970s episode and today. The other episodes identified by Blinder are hardly visible. This leads me to think that the only potentially valid comparison (with all the caveats listed below) is between the 1975–83 episode and the current one.
Second, it shows what happened when the Fed got seriously "behind the curve" in 1974–75. The real rate and inflation moved in opposite directions, very much like today. Look at the two green ellipses. In early 1975, core inflation was running at 12 percent and the real policy rate was equal to about −6 percent, a gap of about 17 percent. Today, core inflation is running at 6 percent and the real policy rate is equal to −6 percent, a gap of 12 percent—smaller than in 1975, but still strikingly large (had I used core personal consumption expenditures [PCE] instead of core consumer price index [CPI], the two numbers would be 14 percent and 10 percent, respectively). It then took 8 years, from 1975 to 1983, to reduce inflation to 4 percent, with an increase in the real rate from bottom to peak of close to 1,300 basis points, and a peak increase in the unemployment rate of 600 basis points from the early 1970s.
Today is obviously different in many ways.
As for the Fed, Arthur Burns did not feel he had a mandate to execute a hard landing and William Miller did not have much time to act. Even so, together, they still had increased the real policy rate from −6 percent to about 0 percent when Paul Volcker took over. And it took Volcker another 7 percent real rate increase to finish the job and get inflation down to 4 percent by 1983. Today's Fed is more independent and is likely to react faster.
In the 1970s, persistent inflation led long-run inflation expectations to become increasingly deanchored. They have not yet done so today. The Fed has a record of low inflation for nearly 40 years for which it can justifiably take credit, something the 1970s Fed just did not have. The broader economic context is also different, but perhaps less than some assume. In the 1970s, the situation was complicated by two major energy price shocks. But today we might be in for a similar experience: The COVID-19 recovery already increased oil prices substantially, and the war in Ukraine could lead to a further major increase. Energy, however, plays a smaller role in production and consumption today than it did then, and this is good news.
What else that is relevant is truly different? Is it reasonable to think that a 200-basis-point increase in the policy rate, so only 1/6 of the rate increase from 1975 to 1981, will do the job this time when the gap between core inflation and the policy rate is 2/3 of what it was in 1975? And that unemployment will barely budge? I wish I could believe it.
Historical comparisons are only suggestive. In their analysis, Reifschneider and Wilcox rightly focus on the specific mechanisms underlying inflation and on the specific shocks likely to affect the US economy in the coming years. One can thus potentially disagree with them on two grounds: a different view of the mechanisms determining inflation, and different forecasts of shocks.
The Mechanisms Determining Inflation
How one thinks about the dynamics of inflation going forward depends mainly on four underlying assumptions: the extent of real wage catchup, the degree of anchoring of inflation expectations, the slope of the Phillips curve (how much inflation moves for a given move in unemployment), and the response of prices to wages. Let me take these four dimensions in turn.
Catchup of Wages
The standard specification of the relation between wage inflation, price inflation, and the state of the labor market has the following generic form:
Δw = Δpe – a(u − u*)
where Δw = w − w(−1), where w is the logarithm of the nominal wage and w(−1) is the log of the nominal wage last period, so Δw is nominal wage growth; Δpe = pe − p(−1), where p(−1) is the logarithm of the price level last period and pe is the logarithm of the expected price level this period, so Δpe is expected price inflation this period; u is the unemployment rate, and u* is the natural unemployment rate, so (u − u*) is the unemployment gap. Other factors, such as actual or expected productivity growth, should also enter the equation; to focus on my main point, I ignore them here.
The interpretation is standard and familiar: The lower unemployment, or the higher expected price inflation, the higher nominal wage growth.
This formulation does, however, have a strong implicit and questionable implication, that bygones are bygones. Suppose that, this year, actual inflation turns out to be higher than expected, so real wages end up lower than they would have been had expectations been correct. The specification above implies that next year, this loss will be irrelevant for wage setting: Workers do not try to make up for the lower real wage. Put another way, they accept the lower real wage and try to maintain it, given labor market conditions.
Contrast this with a specification that conforms better with what theory suggests and what we think workers care about, namely keeping the same real wage (adjusted for productivity growth, which I ignore for simplicity here), given labor market conditions:
w = pe − a(u − u*)
Namely, the nominal wage depends on the expected price level and the state of the labor market. Bygones are not bygones. For given labor market conditions, workers want a given (expected) real wage, independent of what may have happened to real wages by accident in the past. If, after an episode when actual inflation turned out to exceed expected inflation, real wages as a result ended too low, they want to catch up.
As we shall see, which of the two specifications is relevant does strongly affect the dynamics of inflation and what we can expect inflation to be in the near future.
Anchoring of Inflation Expectations
This is a familiar debate, and anchoring plays a major role in the Reifschneider-Wilcox discussion. We can think of workers, or more generally wage setters (workers and firms if we think of wages as determined by bargaining), as forming expectations of the price level this year according to:
Δpe = (1 − λ) Δp* + λ Δp(−1)
where Δp* is a constant, perhaps the target inflation rate announced by the Fed, and λ = 0 corresponds to fully anchored expectations. Whatever happened to inflation in the past, wages are set on the basis of a constant expected rate of inflation. λ = 1 corresponds to fully deanchored expectations, so expected inflation this year reflects actual inflation last year. The more expectations are deanchored, the more inflation last year will affect inflation this year.
Flatness of the Phillips Curve
At some general level, we think of the Phillips curve relation, whatever its exact specification, as capturing the pressure of the state of the labor market on wages. This is captured by the coefficient a in the two wage-setting specifications above. This estimated coefficient has decreased over time (Blanchard 2016)—that is, the Phillips curve has "flattened," meaning a smaller move in inflation for a given move in unemployment.
If one assumes that the coefficient a will remain small, then even large negative unemployment gaps (overheating) have a limited effect on wage inflation, given other variables. If one worries that the relation is nonlinear and that, given the tight labor market, the effect of unemployment will be stronger, then one worries that overheating may lead to substantial pressure on wage inflation.
To derive what these alternative assumptions about wage behavior imply for inflation dynamics, we need to close the model with an equation for price setting. I shall assume that prices are set according to:
p = w + x, or equivalently
Δp = Δw + Δx
where x is all the factors that affect the markup of prices over wages, from other input prices to stockouts, among others (again, I ignore productivity).
I shall take that equation as a given, and not question the implicit assumption that (given x) markups are constant. In doing so, I shall avoid another important discussion: The empirical evidence is that, over the past few decades, movements in wages, or more specifically movements in unit labor costs, have not been passed through fully to prices. This has led some to assume that higher nominal wage growth may be partly or largely offset by decreases in markups and thus have a limited effect on price inflation. My interpretation of the evidence, however, is that when there is little action in labor costs, and most of the movements are either noise or transitory, firms may just ignore these short-run movements. But this will not hold if there is a large and sustained increase in wages. Then, firms will react and adjust prices. In the current context, the assumption of constant markups seems more appropriate.
Forecasts of Unemployment and Price Shocks
We now have the elements we need to characterize the inflation mechanism. But inflation dynamics depend not only on the relations described above but also on the forecasts of two variables, the factors affecting prices given wages (x), and the unemployment gap (u − u*).
Reifschneider and Wilcox do a thorough job of summarizing what we know about both, and I do not disagree much with them. Let me briefly discuss each one.
The catch-all variable x is the variable that, to a first approximation, led to much higher inflation in 2021. It should not be thought of primarily as supply shocks but as production and distribution disruptions, reflecting the inability of firms to meet high demand and leading to higher prices, from chips to cars. It is reasonable to expect many of those prices to stabilize, and in many cases to decrease, leading to a partial decrease in x.
While COVID-19-induced factors are likely to fade, the still fuzzy implications of the Russia-Ukraine war suggest that energy prices and some food prices may further increase, and we may face further supply disruptions. Uncertainty is high, but it seems reasonable to assume that between the decrease in COVID-19 factors and the potential increase in commodity prices, x may only slightly decrease relative to where it was in 2021.
Turning to the unemployment gap: There is little question that the labor market is overheating. While the unemployment rate is still a bit higher than it was pre-COVID-19, other more relevant measures—such as the ratio of vacancy to unemployment, which stood at a historically high 1.15 at the end of 2019 and was 1.73 in December 2021, an unprecedented number—indicate a much tighter labor market. Predicting what will happen depends on both the supply side, the reentry of those who have left the labor market (which would decrease the pressure), and the demand side, which depends itself on two main factors: the course of fiscal policy, and how much of their accumulated savings households decide to spend. Uncertainty is again very high, but it seems reasonable to assume that, despite the current announced course of fiscal tightening in the United States, aggregate demand will remain strong and the labor market will remain tight for the foreseeable future, giving substantial power to workers in bargaining.
Putting Things Together
Suppose there is no catchup, expectations remain anchored, and the Phillips curve remains flat. Then, putting the wage and the price equations together gives:
Δp = Δp* − a(u − u*) + Δx
This delivers an optimistic message: Ignoring the unemployment gap, and assuming that Δx is either close to zero or negative, implies that inflation this year will be equal to or possibly even lower than target inflation. And even if unemployment continues to be lower than the natural rate (i.e., there is a negative unemployment gap), the coefficient a is small, and so this barely has an effect on inflation. Thus, we can expect inflation to return to target on its own; there is no need for a large increase in policy rates by the Fed.
But make the opposite set of assumptions—that there will be catchup and that expectations become fully deanchored—and the effect of unemployment on inflation is larger than in the recent past. Then putting the wage and price equations together gives:
Δp = Δp(−1) − a(u − u*) + x
The message is much less optimistic. It says that, even absent shocks and ignoring the unemployment gap, inflation will remain the same as last year. Furthermore, unless x completely reverses itself, the positive value of x will lead to higher inflation. And, if there is indeed a negative unemployment gap and the slope of the Phillips curve is large, this will further fuel inflation. There is then no question that a strong monetary contraction will be needed.
Which of these two views is closer to the truth? Reifschneider and Wilcox argue that the empirical evidence, based on data going back a few decades, mostly supports no catchup, anchored expectations, and a flat Phillips curve. And they are on solid empirical ground with that as a characterization of the past: There is little evidence of catchup in recent decades, measures of long-run expectations remain anchored (short-run inflation expectations have started increasing, though), and the estimated Phillips curve is indeed very flat.
The issue, however, is how much the past few decades, characterized by stable inflation and nothing like COVID-19 or war shocks, are a reliable guide to the future. There are good reasons to doubt it. What I believe is central here is salience: When movements in prices are limited, when nominal wages rarely lag substantially behind prices, people may not focus on catching up and may not take variations in inflation into account. But when inflation is suddenly much higher, both issues become salient, and workers and firms start paying attention and caring. I find the notion that workers will want to be compensated for the loss of real wages last year, and may be able to obtain such wage increases in a very tight labor market, highly plausible, and I read some of the movement in wages as reflecting such catchup.
We know from history that inflation expectations can deanchor fast. It is fascinating to read what happened in the 1970s through the series of Brookings papers written by Robert Gordon (1970, 1977) (we should all reread these papers, to be reminded that these discussions are not new and we are sometimes reinventing the wheel).
In his 1970 paper, Gordon found that λ was about 0.6, rejecting the accelerationist hypothesis (defined as λ = 1.0), a conclusion strongly endorsed by the participants at the Brookings meeting. By 1977, all his estimates were equal to 1. True, it took seven years for the coefficient to reach 1, and this was in an environment of higher inflation and a less credible central bank than today. But with much wider and faster diffusion of information, I can see expectations adjusting more quickly now than then.
In that respect, I am puzzled by one of the Reifschneider-Wilcox findings that short-run inflation expectations do not appear to help predict inflation. This matters because, in contrast to long-run expectations, short-run expectations, as Reifschneider and Wilcox indicate, have partly deanchored. I would have thought that in setting wages for this year, whether unilaterally or through bargaining, the relevant variable in the mind of workers and firms would be short-run inflation. Again, the explanation may be lack of salience. When inflation is stable, workers and firms may just not think about it and go with their stable long-run inflation expectations. In the current environment of high inflation, I would expect them to be much more aware and make wage demands based on short-run inflation expectations.
Finally, I also find it plausible that an exceptionally tight labor market may have a larger effect on wages than suggested by the empirical estimate of a. This is discussed in a parallel blog post by my colleague Joseph Gagnon (2022), based on earlier work with Christopher Collins (2019). His estimate is that, when the unemployment gap is negative, a 1 percentage point decrease in the unemployment rate increases inflation (other things equal) by 0.6 percent, about 8 times the coefficient estimated by Reifschneider and Wilcox. Based on this estimate, an unemployment gap of −2 percent might add 1.2 percent to inflation, not a negligible contribution.
To conclude, Reifschneider and Wilcox have done a great service in stating their assumptions about specification and forecasts. If the world had not changed, I would accept their conclusions. There are good reasons, however, to think that the Phillips curve will, as it has done many times in the past, shift, and that the landing will be harder than Reifschneider and Wilcox conclude. Part of the inflation will indeed go away on its own, but the Fed may have to increase interest rates by more than 200 basis points to get back to its target. In that respect, I think it is wrong to think in terms of a "terminal rate," as financial investors and commentators do. Unless one thinks that all that will be needed is a monotone adjustment of the policy rate to the long-run neutral rate and no more, it makes little sense to talk about a terminal rate. The adjustment is likely to have a bump, with the policy rate staying above the neutral rate until inflation is under control and the rate returns to the long-run neutral rate.
Blanchard, Olivier. 2016. The Phillips Curve. Back to the 60s? American Economic Review, Papers and Proceedings, May.
Gagnon, Joseph E., and Christopher G. Collins. 2019. Low Inflation Bends the Phillips Curve. PIIE Working Paper 19-6, April. Washington: Peterson Institute for International Economics.
Gagnon, Joseph E. 2022. Why Inflation Surged in 2021 and What the Fed Should Do to Control It. RealTime Economic Issues Watch blog, March 11. Washington: Peterson Institute for International Economics.
Gordon, Robert. 1970. The Recent Acceleration of Inflation and Its Lessons for the Future. Brookings Papers on Economic Activity 1, no.1: 8-41.
Gordon, Robert. 1977. Can the Inflation of the 1970s Be Explained? Brookings Papers on Economic Activity 8, no. 1: 253-77.
Reifschneider, David, and David Wilcox. 2022. The case for a cautiously optimistic outlook for US inflation. PIIE Policy Brief 22-3. Washington: Peterson Institute for International Economics.
1. The source of these statistics is the same as for the figure; core PCE inflation data are from US Bureau of Economic Analysis via the Federal Reserve Bank of St. Louis (FRED).
2. Source: US Bureau of Labor Statistics.
The data underlying this analysis are available here.