David Wilcox
BloombergDecember 11, 2022

From the early days of the inflation outbreak in 2021, analysts and policymakers have feared that if rapid price gains persisted more than a few months, expectations might become unanchored, making the job of bringing actual inflation back down to 2% more difficult.

Bloomberg Economics has run 72 models to explore how expectations might have been expected to behave, given what has happened to actual inflation. The big takeaway: long-term expectations — which are a better predictor of future inflation than short-term — are at the lower end of what historical experience would suggest.

The Fed’s inflation fight is far from over. But with long-term expectations still anchored, Powell & Co have one important thing working in their favor.

  • Historically, short-term inflation expectations have been fairly closely correlated with recent inflation, especially in categories that are most salient for consumers, such as food and energy.
  • True to form, various measures of short-term expectations have moved up substantially since early 2021. The University of Michigan’s short-term measure — the series with the longest history — has increased by a little more than one would have predicted based on the historical degree of sensitivity to actual inflation.
  • These upward moves are not particularly concerning, though. Short-term expectations have not been useful predictors of future actual inflation.
  • Experts focus more on longer-term expectations because they are more useful for predicting future inflation. These measures historically have been much less sensitive to recent actual inflation — and have remained so during the past 18 months.
  • Michigan’s measure of expectations over the next 5-10 years has drifted up a little since the onset of the pandemic, but only to about the levels that prevailed in the first half of the 2010s. Our modeling exercise suggested they have moved up by less than historical patterns would have suggested.
  • Inflation is still a long way from the Fed’s 2% target, but encouraging signs are accumulating. The October CPI report was more favorable than expected, and supply-chain pressures have receded almost back to average. The containment of long-term expectations is another encouraging development.

The Short and Long of Expectations

If any evidence was needed that the inflation problem is not yet solved, measures of short-term inflation expectations provide it. All are up — and some are up a lot. Every observation in the Michigan short-term series this year has been higher than any recorded during the preceding decade.

Short-Term Expectations Have Broken From Their Anchor

Source: University of Michigan, Federal Reserve Bank of New York, Morning Consult and Federal Reserve Bank of Cleveland Research Department, Survey of Professional Forecasters, Bloomberg Economics. Note: The last observation for Michigan is November; for FRB-NY is October; for Morning Consult is the week ending Dec. 3, and for the SPF is 4Q.

Another potentially problematic aspect of short-term expectations is that the various available measures don’t agree. The nearly-3-percentage-point divergence between the series from Morning Consult and the Cleveland Fed versus the one from Michigan suggests no one should pretend to know with precision what households think.

Long-Term Expectations Remain Much More Stable

Source: University of Michigan, Federal Reserve Bank of New York, Survey of Professional Forecasters, Bloomberg Economics.

Michigan’s measure of expectations over the next 5-10 years has drifted up about 0.5 percentage point since the onset of the pandemic, but even so remains no higher than it was a decade ago. The readings from the early 2010s did not prove problematically high; they were followed by inflation that, on average, ran below the FOMC’s 2% target.

The longer-term measure from the Survey of Professional Forecasters is remarkable for its stability through the Covid period. (The SPF asks forecasters for their five- and 10-year-ahead expectations. The series shown in the graph above is inferred by calculating the expectation that must hold for years 6-10 to reconcile the 5- and 10-year series.) The most recent readings for this series are slightly below the key 2% level.

There is no sign in the SPF series of fear on the part of professional forecasters that the Fed might give up on its fight to control inflation or move the inflation target up from its current level.

Short-Term Expectations Higher Than Experience Would Suggest

Historically, recent inflation experience has played an important role in shaping households’ near-term expectations. Changes in the prices of food and energy especially have been closely correlated with near-term expectations — a sensible finding given that households come into contact with such prices frequently, whenever they visit a grocery store or fill up their gas tank.

There is no consensus best model describing the responsiveness of inflation expectations to actual inflation. Recognizing the model uncertainty, Bloomberg Economics estimated 36 models of near-term expectations, differing by explanatory variables and estimation sample period. (See more details in the Methodology section at the end of this analysis.)

In the graph below, each gray line represents the predictions of one model. The orange line shows how Michigan’s measure of 12-month-ahead expectations has evolved. The dashed white lines show the 15th and 85th percentiles of the model predictions since the beginning of 2020. Through most of this period, the Michigan series has been above the 85th percentile of model simulations.

Short-Term Expectations Up More Than History Would Suggest

Source: University of Michigan, Bloomberg Economics. Notes: The estimation sample period for all models ends in December 2019. The gray lines show the in- and out-of-sample simulations of each model. The dashed white lines show the 15th and 85th percentiles of the post-sample simulations.

Long-Term Expectations Have Been Well-Behaved

Inflation modeling experts pay more attention to longer-term expectations because they have been more useful in predicting future movements in actual inflation. Longer-term expectations are less sensitive to recent fluctuations in actual inflation.

As the chart below shows, the longer-term Michigan measure has increased since the beginning of the pandemic, but almost always has been within the interval marked by the 15th and 85th percentiles of model simulations. Lately, the Michigan measure has been closer to the 15th percentile.

The Long-Term Anchor Still Looks Firmly in Place

Source: University of Michigan, Bloomberg Economics. Notes: The estimation sample period for all models ends in December 2019. The gray lines show the in- and out-of-sample simulations of each model. The dashed white lines show the 15th and 85th percentiles of the post-sample simulations.

Why It Matters

The ongoing anchoring of long-term expectations is important. It should make the Fed’s job of bringing actual inflation under control much easier.

Too much of a good thing is also possible. During the latter half of the 2010s, the FOMC wrestled with the fact that inflation persistently fell short of the 2% target. The committee will keep a close eye on the various gauges of longer-term inflation expectations to make sure they don’t fall into the same low-expectations trap again.

Methodology

The results reported in the third and fourth charts above are based on estimates of 36 models for each of short- and long-term inflation expectations, as measured by the University of Michigan’s Survey Research Center. These models have the following specifications:

  • Right-hand-side variables:
    • CPI (overall) 12-month percent change
    • CPI energy 12-month percent change
    • CPI gasoline 12-month percent change
    • CPI gasoline log level of the index
    • CPI core 12-month percent change, CPI energy 12-month percent change, CPI food 12-month percent change
    • CPI energy 12-month percent change, CPI food 12-month percent change
  • Additional combinatorics:
    • The above six right-hand-side variables only
    • The above six right-hand-side variables plus the first lag of the left-hand-side variable (either short- or long-term expectations)
    • The above six right-hand-side variables plus 12th lag of the left-hand-side variable (but not lags 1-11)
  • Sample periods for regression estimation:
    • January 1994–December 2019
    • January 2005–December 2019