Navigating in Cloudy Skies
“Interest rates are still low by historical standards, and they remain just below the broad range of estimates of the level that would be neutral for the economy‑‑that is, neither speeding up nor slowing down growth.” Federal Reserve Board Chairman Jerome Powell, November 28, 2018.
“We’re a long way from neutral at this point, probably.” Video Interview of Chairman Powell by Judy Woodruff, PBS (at the 8:05 mark), October 3, 2018.
Stargazers hate clouds. Even modest levels of humidity and wind make it hard to “see” the wonders of the night sky. Very few places on our planet have consistently clear, dark skies.
Central bankers face a similar, albeit earthly, challenge. Even the simplest economic models require estimation of unobservable factors; something that generates considerable uncertainty. As Vice Chairman Clarida recently explained, the Fed depends on new data not only to assess the current state of the U.S. economy, but also to pin down the factors that drive a wide range of models that guide policymakers’ decisions.
In this post, we highlight how the Federal Open Market Committee’s (FOMC’s) views of two of those “starry” guides—the natural rates of interest (r*) and unemployment (u*)—have evolved in recent years. Like sailors under a cloudy sky, central bankers may need to shift course when the clouds part, revealing that they incorrectly estimated these economic stars. The uncertainty resulting from unavoidable imprecision not only affects policy setting, but also complicates policymakers’ communication, which is one of the keys to making policy effective.
The concepts of the natural rates of interest (r*) and unemployment (u*) have been around for many decades―we can trace the first back to Knut Wicksell at the end of the 19th century and the second to Milton Friedman and Edmund Phelps in the late 1960s. The former is the real (inflation-adjusted) rate of interest that prevails when the economy is growing steadily with normal use of resources (see Williams). The latter is the frictional unemployment rate that prevails when production is normal and inflation is stable (see Barnichon and Matthes).
Taking those two stars, and adding a third one ―the inflation target, p*―we have a natural guide for monetary policy. When unemployment differs from u* or inflation deviates from p* (or both), then policymakers aiming at price stability and maximum sustainable employment will set interest rates away from r*. But, implementing such a strategy typically requires having some sense of the level of the stars. Getting the numbers right turns out to be an extremely complex task.
To see the practical difficulties, start with r*. For central bankers, the short-term policy rate is what’s relevant. In the United States, this is still the federal funds rate―the overnight unsecured interbank lending rate. When U.S. policymakers speak of a neutral rate of interest, they mean the nominal federal funds rate that adds the Fed’s 2-percent inflation target (p*) to their estimate of r*. Since 2012, the FOMC has included in its quarterly Summary of Economic Projections (SEP) each participant’s estimate of the policy rate in the “longer run”—when inflation equals the FOMC target. We can think of this longer-run rate as an estimate of the neutral policy rate. The figure below shows the evolution of the median of the longer-run rate estimates (red diamond), along with the high (beige square) and low (blue square) end of the range.
FOMC estimate of the policy interest rate in the longer run, January 2012-September 2018
Source: Various issues of the FOMC’s Summary of Economic Projections (SEP).
What is most striking about this chart is the decline in the FOMC’s projections of the longer-run, or neutral, rate across all participants. Indeed, the highest most recent estimate (September 2018) is below the lowest estimate back in January 2012, when the Committee began publishing the data. Furthermore, the median estimate of the neutral rate of interest has fallen by more than 1 percentage point over this relatively short interval. Since the inflation objective remains unchanged at 2 percent, this means that the consensus estimate of r* is now 1 percent. It is interesting to note that the median longer-run interest rate in the SEP is currently a bit higher, but has fallen more, than the model-based estimate of Holston, Laubach and Williams (HLW). The latter was 0.95 percent at the beginning of 2012 and is 0.56 percent today.
Looking more closely at the above chart, notice that the dispersion in the estimates (the distance from the blue to the beige squares) hardly changes. The current 1-percentage-point range is almost exactly the seven-year average of 1.1 percent. If anything, as a proxy for uncertainty, this range seems too narrow. For example, HLW calculate that the standard error of their forward-looking (one-sided) estimate of r* is 1.5 percent. Taking this model-based estimate at face value, there is a 70 percent chance that r* is between +2.1 percent and −1.0 percent. These wide bounds are consistent with a neutral nominal interest rate that is anywhere between 4.1 percent and 1.0 percent! By this loose standard, we already are close to neutral. Clearly, this imprecision complicates both decision-making, leaving plenty of scope for debate and compromise among FOMC members, and communication, especially when the FOMC members’ individual policy targets differ.
Adding to the problems of estimating the natural rate of interest (r*) is the fact that policymakers also need an estimate of the natural rate of unemployment (u*). In standard monetary policy models, u* plays a role virtually identical to that of r*. Consider, for example, the following modified Taylor rule based on the unemployment gap (between the current unemployment rate and u*):
Policy Rate = r* + Current Inflation +
0.5 × (Current Inflation – Inflation Target) – (Current Unemployment Rate – u*)
In this “balanced approach”—which appears to have been favored at the Federal Reserve under former Chair Yellen (see here)—any change in u* feeds one-for-one into the Fed’s rate target, precisely as r* does.
How have FOMC participants’ perceptions of u* evolved? The answer is in the following chart, which is analogous to the previous one. Here, we plot estimates of the unemployment rate in the “longer run” taken from the SEP. Much like the FOMC’s estimates of the interest rate in the longer run, the range of estimates for the unemployment rate has declined markedly over the past 6 years. The median estimate (red diamond) fell by 1.1 percentage points since 2012, and is now very close to the Congressional Budget Office’s contemporaneous (vintage) estimate of u* (blue dotted line). And, once again, uncertainty as measured by the high-low gap remains a substantial 0.6 percentage points. (As is the case with r*, a model-based estimate of u* produces a standard errors of about 0.8 percentage points, suggesting a wider confidence range.)
FOMC estimates of the unemployment rate in the longer run and CBO vintage estimates of the natural rate of unemployment, January 2012-September 2018
Sources: Various issues of the FOMC’s Summary of Economic Projections and (for the CBO data) ALFRED.
Just how much do these changing (and uncertain) estimates of r* and u* matter for policy setting? The next chart shows current policy target rates calculated using the Federal Reserve Bank of Atlanta’s Taylor Rule Utility. The largest differences—reflecting variation in the estimates of r*—are between the left half and right half of the chart. But, even using the lower estimate of r* from HLW, the range of rule-based target rates stretches from 2.73% (using core inflation and a low weight on the unemployment gap) to 4.32% (using headline inflation and a higher weight on the unemployment gap). The lower estimates in each case—the red bars—use a measure of the output gap (the percentage difference between current real GDP and the CBO estimate of potential GDP) rather than the unemployment gap. Importantly, even the lowest target from this exercise is about 50 basis points higher than the current federal funds rate.
Current policy rates implied by various specifications of the Taylor Rule, December 2018
Notes: HLW is the latest Holston Laubach Williams one-sided estimate of r*. The black bars represent a modified Taylor rule based on the unemployment gap (using the CBO’s latest estimate of u*), while the red bars are based on the output gap between current real GDP and the CBO estimate of potential GDP. Headline and core are measures of inflation based on the price index of personal consumption expenditures. A value of b=0.5 corresponds to the original Taylor rule, while b=1 reflects the “balanced approach” that doubles the weight on resource utilization. Source: FRB Atlanta, Taylor rule utility.
The conclusion from all of this is that the dispersion in estimates of r* and u* leads to sizable uncertainty about the appropriate policy rate. What should policymakers do? The classic response (initially articulated by Brainard in 1967) is to proceed with caution: go slowly, take small steps and watch closely to see what happens. While recent research on robust policies describes circumstances under which this “policy attenuation” argument need not apply, in practice, the FOMC’s gradualist approach looks to be a wise one. Importantly, it provides time to learn more about r* and u*.
This brings us back to the statements by Chairman Powell and Vice Chairman Clarida. As policy approaches a more neutral level, the challenge of knowing exactly what happens next grows. Lacking a clear view of the stars—even if they are close by—the FOMC has less ability (or reason) to provide guidance about the likely future course that they will steer. If private observers—including those in financial markets—understand the models policymakers use to guide their decisions, then the incoming data will lead them to anticipate the FOMC’s response to news.
At the same time, uncertainty about key policy drivers makes communication unavoidably complex. As Chairman Powell indicated in late November (see opening quote), the current federal funds rate (about 2.2 percent) is only a bit below the bottom of the broad range of the FOMC’s neutral-rate estimates (2.5 percent). At the same time, as the Chairman suggested in October, the policy rate is still significantly below neutral—at least based on the median FOMC estimate of 3 percent. So, why did the U.S. equity market jump when the Chairman spoke in November? The likely reason is the difficulty any policymaker has in conveying all of the inevitable uncertainties without risk of misinterpretation. (Some market analysts share this view.)
A less benign interpretation is that market participants are suspicious about the Fed’s commitment to price stability in the face of persistent criticism from President Trump about rising policy rates. In our view, such skepticism is unwarranted, and is inconsistent with low Treasury bond yields. The FOMC has no incentive to deviate from its largely apolitical inflation-targeting framework. When the Fed ultimately chooses to change course―whether it be to speed up, slow down, or even reverse direction―we have little doubt that the shift would come not as a reaction to short-run political pressure, but as a response to revisions of FOMC participants’ projections for growth and inflation.
More bluntly, the truth is in our stars and the incoming data, not on Twitter.