Dot-ology: What can we learn from the dot plot?
“[I]f properly understood, the dot plot can be a constructive element of comprehensive policy communication,” Federal Reserve Board Chairman Jerome Powell, March 8, 2019.
The primary objective of monetary policy is to keep inflation and unemployment low and stable. To be effective, central bankers must address shocks to inflation and unemployment, while ensuring that what they say and do is not a source of volatility. One way to make a commitment to stability credible is for policymakers to broadcast their likely responses to shocks—their reaction function. Such transparency escalates the cost of reneging, helping to anchor expectations about the future that influence current behavior (see our primer on time consistency). And, because they can anticipate how policy will respond to changes in economic and financial conditions, it improves everyone’s economic and financial decisions.
With such a stability-oriented policy strategy, the policy path will depend on what happens in a changing world. Only under specific circumstances―such as when the short-term interest rate is at or near its effective lower bound―will policymakers be inclined to commit to a specific future policy path.
Recently, we wrote about the remarkable evolution of Federal Reserve monetary policy communication over the past quarter century. Today, the Federal Open Market Committee (FOMC) publishes statements, minutes, and quarterly forecasts for growth, inflation, unemployment, and interest rates. In this post we take up a narrow question: What can we learn from the information published in the FOMC’s quarterly Summary of Economic Projections (SEP)?
Our answer is: quite a bit. The data allow us to estimate not only an FOMC reaction function, but also a short-run projection of the equilibrium real rate of interest (r*)―one that is consistent with projected economic conditions over a two- to three-year horizon—in addition to the long-run r* that is implicit in each SEP. While there is almost surely room to improve on the SEP, we conclude, as a friend and expert Fed watcher once suggested, “Don’t ditch the dots.”
Initiated to bolster the FOMC’s commitment to keep interest rates close to zero for an extended period, the SEP includes medians of economic growth, inflation and unemployment projections for the next two or three years, as well as a plot of the policy rate projections for all of the FOMC participants. “Dot plots” like the one released following the latest FOMC meeting reproduced below attract intense attention both in financial markets and in the media.
FOMC participants’ assessments of appropriate monetary policy: Midpoint of target range or target level for the federal funds rate (end of period), March 20, 2019
A bit like pathologists analyzing a biopsy, dotologists study these plots in an effort to divine the intentions of policymakers. When will the next interest rate move come? Will the move be an increase or decrease? How many changes are coming over the next year? What about the next two years? The questions go on and on (and on)!
The lack of definitive answers has fueled criticism of the dot plot and the SEP. We are not among the critics. As Yogi Berra reminded us, it’s tough to make predictions—especially about the future. The dots are simply a collection of individual baseline forecasts. They do not convey the considerable uncertainty that surely accompanies every one. The FOMC minutes, published three weeks after each SEP, convey some of that uncertainty. For example, for the December 2018 version (the latest available), the median policy rate projection for the end of 2021 was 3.1 percent. But, reflecting the estimated 70 percent confidence interval of plus or minus 2.4 percent, the range is from 0.7 to 5.5 percent. (For a normal distribution, this is roughly plus and minus one standard deviation.)
Given this level of uncertainty, it should be clear that the dots and the SEP do not represent any sort of guarantee. Instead, when the policymakers wish to make a commitment, they usually do so through the post-meeting statement. This is exactly what happened in 2012, when the FOMC began publishing the SEP. (See the discussion of such forward guidance in the 2016 report of the U.S. Monetary Policy Forum.)
To the extent that the dot plot is merely a collection of forecasts, the format in which it appears would seem to limit its usefulness further. The SEP provides no means to connect the inflation, unemployment and interest rate forecasts of individual respondents. The reported medians (and ranges) do not necessarily reflect any particular FOMC participant’s view or reaction function. Moreover, the mix of individuals changes from year to year, as members enter and exit the FOMC. So, one might be skeptical about using the dots to construct a coherent story about policymakers’ likely reaction to changing circumstances.
On closer evaluation, however, the medians do contain useful information. To see this, we (temporarily) transformed ourselves into dotologists. From the 32 SEP publications since January 2012, we collected data on the median values for the following: the policy interest rate, inflation (as measured by FOMC’s preferred price index, the core PCE), and unemployment. Each SEP has forecasts for three or four years, giving us 105 data points. Treating these observations as if they come from a single policymaker, we estimate a simple Taylor rule where the policy interest rate (i) is set equal to the equilibrium real rate of interest (r*), plus current inflation (p), plus a coefficient (a) times the inflation gap (p-p*) and another coefficient (b) times the unemployment gap (u-u*):
i = r* + p + a (p-p*) - b (u-u*)
Two things emerge from bringing the FOMC’s median projections to this simple relationship. First, the SEP-implied short-run reactions to changes in inflation (1+a) and unemployment (b) are roughly 2¼ and ½, respectively. That is, for each percentage point projected inflation is above or below the target of 2 percent, the FOMC will move its policy rate a bit more than two percentage points. Policymakers appear to be far less sensitive to the unemployment gap, moving interest rates by only about half a percentage point for each percentage point that projected unemployment moves relative to their estimate of the equilibrium rate (u*). (While the estimated ratio of a to b is surprisingly high, this regression fits reasonably well, accounting for more than 70 percent of the variation in interest rate forecasts.)
Interpreting the dot plot in this way casts light on the frequent—and potentially misleading effort—to divide policymakers into hawks and doves (see, for example, here). It is common to label those FOMC members who expect to hike rates hawks (and vice versa). But, if the goal is to anticipate future decisions, the reason why someone wants to change policy matters. Is it because, compared to the FOMC overall, a person puts a relatively higher weight on deviations of inflation from target (1+a) than on deviations of unemployment from the norm (b)? Or, is it because they have different forecasts of inflation and unemployment? From meeting to meeting, it could surely be the latter. But, absent a persistent forecasting bias, the former seems more likely to endure over time.
Returning to the dots, they can provide additional interesting insights into the thinking of FOMC participants. Specifically, we can estimate the Committee’s implied level of r* (the constant term in the above equation) over short intervals—say, two to three years—and (after adjusting for the 2 percent inflation target) compare them to the longer run policy rate projections reported in the SEP.
The following chart shows the results of this exercise. The red line is the estimate of the short-run r* computed from the Taylor rule (we plot only the first value for each year). The shaded area depicts a 95-percent confidence interval around that measure. The black line is the median value of the long-run r* (the median projected federal funds rate in the longer run minus the 2-percent inflation objective). The dotted black lines show the range of participants’ estimates of the r* in the longer run. Note that the short-run r* starts well below zero in 2012, rises gradually and approaches one-half of one percent in 2018. Put differently, the FOMC’s current forecasts for interest rates, inflation and unemployment are consistent with a short-run r* of about 0.5 percent.
SEP implied short-run and long-run equilibrium real interest rate (r*), 2012-18
Over the same period, the FOMC’s estimate of the long-run r* declines consistently. Starting above 2 percent (with a range from 1.25 to 2.5 percent) in 2012, the March 2019 SEP estimate of the long-run real interest rate today is between 0.4 and 1.5 percent, with a median of 0.75 percent. (The median is almost exactly equal to the current estimate from the Laubach and Williams model published by the Federal Reserve Bank of New York.)
The convergence between the short-run and long-run measures of r* helps us interpret Fed policy. At the start, when the short-run r* was sub-zero and far below the long-run measure, the Fed supplemented conventional interest rate policy with an array of unconventional measures, including forward guidance and large-scale asset purchases (LSAPs). The Fed halted LSAPs at the end of 2014, when the short-run r* rose above zero. At the end of 2015, the Fed began to raise interest rates for the first time since 2006. The convergence between the two measures of r* in 2017-18 coincides with the normalization of monetary policy.
Our brief foray into dotology leaves us in agreement with Chairman Powell: even as currently published, the dots are quite informative. They help us to understand the FOMC’s behavior as inflation and unemployment change, and they highlight the changing perception of what is neutral. Furthermore, even over the turbulent period of the past seven years, the pattern is relatively stable. The implied levels of the short-run (and long-run) equilibrium real rate of interest evolve gradually, not abruptly, as new data prompts FOMC members to update their views. Against this background, we see little to substantiate claims of a stunning shift in FOMC strategy over the past few months. Given forecasts of inflation and unemployment, combined with estimates of equilibrium rates, the current policy rate is consistent with what a simple model based on SEP history suggests.
Our conclusion is that, while there are likely to be ways to improve the SEP (see, for example, Levy), publication of the dots does more good than harm. They are a useful tool, providing information that is difficult to convey in other ways.
There is no need to throw out the baby with the bathwater.