“My recommendation for the presentation of economic projections and their use in policy explanations is to align them better with the rhetoric of uncertainty and data dependency and the true state of economic knowledge.” Former Federal Reserve Board Vice Chairman Donald Kohn, January 2019.
When it comes to forecasting, we usually cite famous Yankee catcher and baseball philosopher Yogi Berra, who reputedly said: “It's tough to make predictions, especially about the future.”
For central bankers, this is more than just a minor headache. Given the lags between policy actions and their effects, forecasting is unavoidable. That puts uncertainty about the economic outlook at the heart of the policymakers’ daily job. Indeed, no one knows the future path of the economy or interest rates—not even those making the decisions.
Communicating this inevitable monetary policy uncertainty is difficult, but essential. In the United States, the Federal Open Market Committee (FOMC) uses a variety of means for this purpose. In two earlier posts, we discussed the evolution of FOMC communications and the usefulness of the quarterly survey of economic projections (SEP). Here, we examine a key aspect of FOMC communications that receives insufficient attention: the explicit publication of policymakers’ range of uncertainty about the future path for the policy rate. Buried near the end of the FOMC minutes, published three weeks after the SEP release, this information is consumed only by die-hard devotees.
To be sure, there was a period in the years following the September 2008 peak of the financial crisis when the FOMC could reliably project the policy rate several years ahead: the federal funds rate target would remain close to zero. Having exhausted the capacity for conventional monetary policy easing, but wishing to do more to stimulate the economy, policymakers turned to forward guidance. In an effort to lower long-term interest rates (and long-term risk premia more generally), they committed themselves to keep the policy rate low for a long time. In the terminology of Chicago Fed President Charles Evans, the goal was to provide Odyssean guidance, whereby Committee members metaphorically tied themselves to the mast. The FOMC’s large-scale asset purchases, which eventually boosted the balance sheet by a factor of nearly five, functioned at least in part as a mechanism for reinforcing this commitment.
When the FOMC began publishing the SEP in 2012, it may have had a similar goal: reduce uncertainty about future policy rates. While statements and speeches directly communicated policymakers’ intent, by bringing the growth, unemployment, inflation, and interest rate projections of all 19 FOMC participants together in one place, the SEP made clear the degree to which everyone shared the commitment to keeping interest rates “low for long.”
But the function of the SEP shifted years ago. Large-scale-asset purchases concluded in 2014, and the Fed’s December 2015 rate hike marked a clear end to the zero-rate commitment. In the past few years, the two key policy questions for the FOMC have been: What should be the normal level of interest rates? And, how quickly should we get there? The SEP usefully reflects the variety and evolution of FOMC participants’ responses to these two questions. Following President Evans again, the contrast is between Odyssean commitment and Delphic guidance—the latter is Greek for Yogi Berra forecasting, it is explicitly not about commitment. The more the policy rate path depends on the evolution of the economy, the greater is the unavoidable uncertainty.
Today, not only is that uncertainty substantial, but it is just as important to communicate as the interest rate commitment was half a dozen years ago. This brings us to an important, but largely overlooked element of the FOMC’s communication framework: historical projection error ranges that are now included in the FOMC minutes three weeks after the initial SEP release. Based on the work of Federal Reserve Board economists David Reifschneider and Peter Tulip (and recently highlighted by our friend Mickey Levy), Table 2 in the minutes reports estimates of error ranges (measured as the root-mean-squared prediction error) for projections of real GDP growth, the unemployment rate, inflation, and the short-term interest rate!
To see how informative these error ranges are about the uncertainty associated with the FOMC projections, consider the latest version from the March 19-20, 2019 meeting minutes. There we see that, for the unemployment rate, the median projection for 2021 is 3.9 percent, with an error range of plus or minus 1.7 percentage points. This tells us that, given historical experience, there is a 70 percent chance that in 2021, the unemployment rate will be between 2.2 and 5.7 percent. For inflation, the median is 2.0 percent, with an error range of plus or minus 1.1 percentage points, so the confidence interval goes from 0.9 to 3.1 percent. (For GDP growth, the median projection is 1.8 percent, with a standard error of 1.9 percent – that is, the 70 percent confidence interval is from -0.1 to +3.7 percent.)
Uncertainty regarding the future level of unemployment and inflation (and real growth) translates directly into uncertainty about the path of future policy. Here, again, the FOMC is remarkably clear about the degree of imprecision. In March 2019, the error range for the 2021 projection of the short-term interest rate is plus or minus 2.5 percentage points. Given the median projection of 2.6 percent, this means that the Committee believes there is a 70 percent chance that, in 2 years, the target interest rate will be between 0.1 and 5.1 percent! (The 50 percent confidence interval for the policy rate over this same two-year horizon is plus or minus 1.6 percentage points.)
Since 2017, the FOMC also has published a figure following each SEP that helps visualize the uncertainty about the interest rate path. The following fan chart is taken directly from the March 19-20, 2019 minutes. It makes clear that, while the median suggests little change in the target path over the next 2-plus years (in red), there is considerable uncertainty that increases with the forecast horizon.
Uncertainty in the March 2019 projections of the federal funds rate (with 70% confidence interval), 2019 to 2021
Source: Figure 5, Minutes of the Federal Open Market Committee, March 19-20, 2019.
To underscore the value of what already is being published (however deep in the minutes), we collected information from all 30 published SEPs and combined them with the historical uncertainty measures in the Reifschneider and Tulip paper to generate a history of the uncertainty in the FOMC’s two-year ahead projections. The result is in the chart below. The black line is the two-year ahead median, while the gray area is the 70 percent confidence interval. Note that “two years ahead” is only an approximation, since the projection is always for the end of the calendar year that is two years ahead. We show the projections as of each publication date. For example, the March, June, September, and December 2015 projections for the end of 2017 are plotted as four consecutive points in 2015. (The fact that “two-years ahead” is closer to December than it is to March explains much of the jagged pattern in the confidence interval: uncertainty declines as the forecast horizon shortens.)
Uncertainty in the two-year-ahead projections of the federal funds rate (quarterly with 70% confidence interval), 2012-March 2019
Source: Minutes of the Federal Open Market Committee, 2012 to 2019; Table 5 in Reifschneider and Tulip; and authors’ calculations.
In our view, this information about the uncertainty in the projections can serve as an extremely valuable tool. As we look at the version of the SEP published with the FOMC’s meeting minutes, it is striking how similar it is (fan charts and all) to a typical “inflation report” of an inflation-targeting central bank. The Bank of England’s (BoE) Inflation Report remains the classic example.
Central banks that produce inflation reports typically use them both to communicate and to guide their deliberations. On the former, the reports have both a backward- and forward-looking function. Retrospectively, they provide an evaluation of how policymakers have performed, allowing parliamentarians, financial market participants and the public at large, to hold independent central bankers accountable for their actions. Prospectively, the reports provide projections to help people understand how policymakers will likely respond to changing conditions. Publication of the information in the reports also allows participants on a monetary policy committee to focus public attention on the changing issues that foster uncertainty.
By creating accountability and transparency, inflation reports have a powerful influence on internal committee dynamics. The obligation to publish both an expected value and a range for projections of the state of the economy and policy (something like the first figure above) focuses internal discussions. In particular, the need to reach a consensus significantly influences the nature of committee discussions.
While the SEP uncertainty table and fan charts strongly resemble what other central banks publish, the FOMC does not engage in the consensus building associated with the production of a BoE-like inflation report. The uncertainty measures that we describe above are based on historical forecast errors, rather than an agreement among the FOMC members. Furthermore, unlike the BoE’s nine-member Monetary Policy Committee, who all sit in the same building, it seems unrealistic to think that the 19-strong FOMC, spread among 13 locations from Boston to Atlanta to San Francisco, are prepared to forge a consensus four times per year to publish a BoE-style report.
It is nevertheless critical, as Don Kohn suggests, that Federal Reserve policymakers find an agreed-upon mechanism for clearly communicating uncertainty. Mr. Kohn has proposed that the Fed stop publishing the median projections that currently draw considerable attention from financial market observers. Others suggest eliminating the dot plot altogether (see, for example, here). We do not share these views, because we find the dot plots and their medians useful for constructing a rough FOMC policy reaction function (see our earlier post).
Instead, we favor highlighting the range of uncertainty around the median projections by publishing the equivalent of Table 2 and the fan charts prominently and quickly. One simple approach would be to release these materials as part of the first version of the SEP (that is published together with the meeting statements), instead of waiting three weeks until publication of the minutes. And, rather than feature the table with the median projections, start with a chart like the one we reproduce above. The FOMC should not bury this important and valuable product at the end of a 30-page document that appears weeks after the SEP’s initial release.
Highlighting the inevitable uncertainty by publishing the fan charts and historical forecast error table together with the initial SEP would help shift the public discussion. Rather than constantly facing questions about the median path of policy rates―and giving explicit answers that can add to volatility in in financial markets―a focus on uncertainty associated with the outlook would help to align FOMC members’ public comments with the risks that they perceive. If FOMC participants—Governors and Reserve Bank Presidents—also were to concentrate their public comments on explaining the sources of uncertainty, this would help counter any excessive public attention on the median projections. It also would help shed light on whether FOMC member disagreements arise from differences in interpretations of the current state and likely evolution of the economy, or from differing views about likely future policy actions and the responses to those actions.
To conclude, in a world where policymakers are rightly not committed to a specific interest rate path, the FOMC can and should exploit existing tools to improve communications regarding the uncertainty of the future policy path. The Committee already has revealed there is only an even chance the policy rate will be between 1.0 and 4.2 percent by the end of 2021. That range probably far exceeds what most observers believe about FOMC policy uncertainty. Consequently, with only a modest change in practice—one that places such uncertainty measures front and center in their published material—the Fed can materially improve its communications. Making the evolving scale and sources of uncertainty a focus of the Chairman’s post-meeting press conferences, and of FOMC members’ public remarks, would then follow naturally.