When will the Fed respond to inflation?
Price volatility, inflation expectations, and the Fed's new reaction function
Matt Klein is arguably the most important economics commentator writing today. During the intellectual revolution that followed the global financial crisis, Klein, then at FT Alphaville, played a crucial role in propagating the emerging expert understanding of macrofinance to informed-but-non-expert observers. In 2018, together with Alexandra Scaggs, he was poached by Barron’s, who offered him his own column. Like many other FT subscribers, the Policy Tensor immediately got a Barron’s subscription to keep up. Now, Klein has gone free-lance. His Substack newsletter is expensive: $18/mo. (The Policy Tensor charges only $5/mo on an honors system, meaning that you should subscribe if you think I am contributing to the larger conversation.) But not reading him is so costly to the best informed that we just grit our teeth and shell out the change. Hope you get rich, Matt Klein!
His latest dispatch is free to access for non-subscribers. In it, he has this fantastic graph decomposing the latest inflation reading. It shows that pandemic-related shocks are the biggest contributor to the headline number and that underlying inflation is still near the Fed’s target.
That’s one reason for the Fed to wait and watch instead of responding to the recent CPI readings by removing accommodation. Klein’s decomposition is helpful to understand what is happening to US inflation at this time. But it is very context-specific. In general, I like Matt’s attention to detail. But the little guy inside my head always wants a more systematic picture.
The Fed’s reaction function has changed over the past few years. Since inflation expectations became anchored in the mid-1990s, the Fed fought tight labor markets on the understanding that inflation obeys the Phillips curve—that low unemployment will cause inflation to accelerate. As Blanchard put it most nakedly, “The notion that a tight labor market will lead to inflation is impossible to contradict.” Well, Blanchard is completely mistaken. But so were the Fed technocrats in general. US inflation is a function of global slack, not domestic slack. This is what led to the Fed’s policy error during, say, 1996-2018. The Fed has since seen the light.
The Fed has gone through an intellectual revolution over the past few years, which is reflected in the rise of Lael Brainard. In my polemic, The Making of the Mother of All Economic Booms, I argued that what is driving the changing reaction function of the Fed is ‘the threat from below’ revealed by 2016. What became manifest is that, fed up with elite incompetence, angry working-class people from ‘flyover country’ are ready to burn the world constructed by the elites to the ground. US elites in general and Fed technocrats in particular have come to understand that the only way to contain ‘the threat from below’ is to restore broad-based growth. And the only way the technocrats know how to restore broad-based growth is to run the economy really hot.
Okun’s ideas from the early-1970s about a high-pressure economy have therefore come back with a vengeance: it is only when labor markets are tight that working class wages start to grow as fast as professional class salaries. This is reflected in the robust pattern known as the Wage curve. In the time-series, and controlling for productivity growth, 1 percent higher unemployment rate predicts a 20 percent lower growth rate of real compensation per hour. In the cross-section of commuting zones (CZs), adjusting for human capital differences between them, 1 percent higher unemployment rate predicts 18 percent lower wages.
So, that’s how you can have your cake and eat it too: tight labor markets don’t imply higher inflation because the Phillips curve is dead, but they do imply higher wage growth because the Wage curve is alive and kicking.
OK, so the Fed has made what is effectively a class-political decision to foster a high pressure economy. Given the hold of theory on economists’ minds, this is being justified with a model of Fed policy called ‘average inflation targeting’, meaning that the Fed is verbally committing to deliver on the 2 percent target over the cycle. Moreover, most of the FOMC is now convinced that the Fed’s response function ought to be symmetric, meaning that the Fed will respond as aggressively to lowflation as highflation.
As Powell put it, ‘the key to the whole thing’—the whole thing being the idea of running the economy really hot to deliver broad-based growth—‘is that inflation expectations are anchored on target’. That’s the Fed’s new synthesis between the lessons of the 1970s stagflation crisis—which was caused by an expectations spiral—and the commitment to a high pressure economy to contain the threat from below. In order to understand the Fed’s new reaction function, therefore, one has to watch inflation expectations like a hawk. The Fed will have to respond by removing accommodation if inflation expectations threaten to get deanchored on the upside. So the key question is not what is happening to inflation per se, but what is happening to inflation expectations.
There’s not much disagreement between the Fed and the Market. Breakeven inflation rates have risen over the past year. But they are very close to target.
The problem with the breakevens—the difference between yields on Treasuries and inflation-protected Treasuries (TIPS)—is that they are confounded by a time-varying risk premium. When investors demand lower compensation for systematic risk, inflation compensation gets compressed; conversely, when investors demand higher compensation for systematic risk, inflation compensation spreads widen.
The Minneapolis Fed reports a different measure of market-based inflation expectations that backs out market-implied probabilities from cap and floor contract prices posted by the dealers. These are more reliable than inflation breakeven spreads because option prices contain more fine-scale information than bond yields. The next figure displays the options-implied probability that CPI would average over 3 percent over the next 5 years. This probability rose as high as 44 percent in May and has since fallen back to 32 percent as of last week. So, investors have become less fearful about high inflation than they were at the peak of the panic in May.
So, market-based inflation expectations are again well within the Fed’s comfort zone. Put another way, there’s little disagreement between the Fed and the dealers (the market-makers) that the present bout of inflation is transitory. This is why, despite the apparently alarming headline inflation numbers, the yield on the benchmark 10-year note has continued falling from around 1.7 percent in April to 1.4 percent in mid-July—the puzzle that Robert Armstrong has been struggling to understand in his FT Unhedged newsletter.
However, the Fed cares less about the inflation expectations of traders in Manhattan than the expectations of price-setters in the economy out there. Consumer surveys are one proxy available. Consumer expectations of inflation over the next three years have risen somewhat from 3 percent before the pandemic began to 3.5 percent now. But they are still within the Fed’s comfort zone.
A more direct measure of the inflation expectations of price-setters in the economy can be obtained by sorting the 174 price series released by the BEA on volatility. Essentially, the lowest volatility, say, quintile, corresponds to the price-setting behavior of firms that change the prices of their products very infrequently, perhaps because it is very costly for them to change prices often. This means that they must choose prices in line with their inflation expectations. The lowest quintile of prices series by volatility therefore contains a strong signal of inflation expectations.
Put another way, the lowest quintile of price series by volatility (Q1) contains predictive information about future inflation. Changes in Q1 predicts 1-year forward CPI (t = 2.2), after controlling for lagged CPI, whereas Q2-Q5 do not—whether one looks at them separately or jointly (t = 2.83).
The lowest volatility quintile with the strongest expectations signal shows a non-trivial jump in recent months, suggesting that infrequent price-setters’ inflation expectations may be starting to move.
Looked at over a longer horizon, the lowest volatility quintile of the CPI series shows a return to pre-recession levels. This signals a regime change. If the pattern suggested by this graph continues to hold, we may be getting back to a normal monetary cycle and leaving the “New Normal” behind. This is good news for the technocrats—they don’t want to be stuck forever at the zero lower-bound.
The central message of this dispatch is that, in order to understand and predict the Fed’s response, we must pay attention to inflation expectations rather than inflation per se. Market-based expectations and survey-based consumer expectations have both risen but they are still within the Fed’s comfort zone. The expectations of the slowest-moving price-setters have also started to move, although so far they too are within the Fed’s comfort zone. The Fed will have to begin removing accommodation and proceed with liftoff if expectations continue rising much further. Moreover, the rapidity of the Fed’s response will depend on how fast these expectations climb. This is as it should be. Investors can and should care about the earnings gains for a rapidly growing economy and price stability: you want the Fed to deliver on its new commitments.
Finally, the grapevine has it that Biden is going to appoint a new Fed chair when Powell’s term expires in Feb 2022. The White House should appoint Lael Brainard. More than anyone else, she’s responsible for the intellectual revolution at the Fed. She should be the one to consummate it. I’ve noted her stellar resume before. She will provide a pair of steady hands during the handover to fiscal policy.
Postscript. Perhaps a more systematic approach to understanding inflation dynamics is latent factor analysis. Basically, we extract the common component of the 173 price series for which continuous data is available since 1995. (More precisely, we compute year-on-year percentage change in the price series and, in an abuse of notation, also call them the price series.) The latent factor is a synthetic series that is, by construction, the most contemporaneously correlated with all the price series. We can think of the individual prices series as linear functions of the latent factor plus an idiosyncratic error term. In effect, the latent factor captures the dynamics of underlying inflation. We recenter and rescale the latent factor to have the same location and scale as the unweighted mean of the price series.
The next figure shows the latent fact and the unweighted mean of the price series. The diachronic pattern of the later factor is similar to the lowest quintile of the price series by volatility, suggesting that it too contains a very strong signal of inflation expectations.
However, the two differ in the detail. This is especially striking if we zoom into the past decade. Whereas the lowest quintile of the price series by volatility show a spike in recent months, recent readings of the latent factor suggests that underlying inflation jumped up at the beginning of the pandemic and has since been trending down.
We also report the last 12 months of readings for the unweighted mean of the price series. In Jan 2021, mean year-on-year CPI was 0.38 percent. By June, it had climbed to 5.61 percent—tracking headline consumer price inflation. Meanwhile, underlying inflation, as measured by the latent factor, was 2.06 percent in January. By June, it had fallen to 1.67 percent per annum. If this analysis is correct, the Fed is quite right to believe that the present spike in headline will dissipate soon.
You can find the python code and the dataset on my GitHub.