A modernized Taylor Rule?
admin | October 19, 2018
This document is intended for institutional investors and is not subject to all of the independence and disclosure standards applicable to debt research reports prepared for retail investors.
St. Louis Fed President James Bullard has transformed in recent years from a centrist or moderately hawkish policymaker to one of the most extreme doves on the Committee. He has been objecting to rate hikes for quite some time, even as a solid consensus has emerged on the FOMC to keep moving rates back toward the Committee’s view of neutral. Bullard offered a conceptual framework for his dovish views this week by suggesting a series of three tweaks to the traditional Taylor Rule. Bullard’s modified, or in his words, “modernized” Taylor Rule, which prescribes that the federal funds rate need not rise any further, provides a window into the thought process of a dove and, in my own hawkish view, a sense of the vulnerabilities of that approach.
Taylor Rule basics
Most readers will undoubtedly be familiar with the Taylor Rule concept, but it may help to clarify the ensuing discussion to walk through the basics. The Taylor Rule is a formula that provides a recommendation for the federal funds rate based on how the economy is performing. In effect, the Taylor Rule starts with a “neutral” or equilibrium real rate of interest, what many within the Federal Reserve system call r*. We can then add inflation, or more precisely, the Fed’s inflation target of 2% to generate an equilibrium nominal policy rate. Then, the Taylor Rule says that the actual policy rate should deviate from that equilibrium based on whether the unemployment rate is running above or below estimates of equilibrium and whether inflation is running above or below the Fed’s target, i.e. an unemployment gap and an inflation gap.
There are two key unobservable variables that we have to estimate to perform a Taylor Rule calculation: r* and the equilibrium unemployment rate. Dr. Taylor originally used 2% for r*, but most economists and policymakers now believe that the equilibrium real rate of interest has fallen, though there is some debate about how much. Estimates of the equilibrium unemployment rate have also fallen in recent years.
With those values in hand, one can either estimate the coefficients on the two gap terms or assume their values. The original Taylor Rule calculated them based on historical experience, but the gap coefficients at this point are mostly assumed (in the earliest versions of the rule, the coefficients are 1.5 for the inflation gap term and 1 for the unemployment gap term.
St. Louis Fed President Bullard gave a presentation this week in which he offered three tweaks to the traditional Taylor Rule construction.
- Lower r*. Most economists agree that the 2% estimate of r* that prevailed 25 years ago, when the Taylor Rule was first rolled out, no longer holds. The FOMC dot projections suggest that r* is now around 1% (the median estimate for the “longer-run” funds rate is 3%, minus 2% for inflation). That is what I currently use in my own Taylor Rule calculations. I am open to the possibility that r* may be (or move) higher than that, depending on how the economy performs going forward.
In particular, in most economic models, r* is tied fairly closely to the trend rate of growth in the economy. There are various ways to express that relationship. My default has always been that over a very long period (close to 50 years leading up to the financial crisis), the fed funds rate averaged about 50 bp below nominal GDP growth. Most Fed officials appear to believe that there are factors post-crisis – such as the demands for liquid instruments dictated by financial regulations and the persistent bid for high-quality short fixed income U.S. assets by foreign investors/governments – that have pushed r* down relative to equilibrium growth. At this time, the FOMC pegs r* at 1% and equilibrium real GDP growth at 1.8%, a gap of 80 bp. Looking ahead, I believe that the investment boom that has ensued in 2017 and so far this year will continue and will ultimately help to raise the underlying trend of productivity growth and in turn the potential pace of economic growth. If that occurs, then I would expect estimates of r* to increase commensurately.
Bullard takes a different, shall we say, novel approach to estimating r*. He argues that the risk-free rate of return in the economy can be equated to the one-year Treasury bill yield. Subtracting the 12-month inflation rate can then give us a real interest rate. He then satisfies the “equilibrium” concept by taking a longer-run moving average of this real one-year bill rate. Based on that measure, he concludes that r* is roughly zero, and thus that an equilibrium nominal interest rate (r* plus inflation) is 2%.
Perhaps I am missing something, but Bullard’s choice of metric here seems obviously misguided to me. The one-year bill yield is indeed closely tied to expectations of monetary policy, but it is correlated to expectations of actual monetary policy, not assessments of what constitutes a neutral or equilibrium policy stance. Using a moving average may help to diminish the difference between the two, but when the Fed has run a dramatically easy monetary policy for over a decade, as it has since the crisis, no moving average will be enough to bring the two together. So, Bullard’s calculation tells us where monetary policy is (or has been), not where it should be in equilibrium.
To take the point to its absurd extreme, imagine a world where r* is fixed but unobservable. The Fed has no control over r*, but it can heavily influence the one-year Treasury bill rate with its monetary policy. Thus, under Bullard’s framework, if the FOMC wanted to push estimates of r* down, it could just keep easing to push the actual one-year rate to a low level. This would in reality prove nothing, as r* is fixed, but monetary policy would end up far too easy, yet Bullard’s construction would suggest that the equilibrium rate was dropping.
To be sure, r* is unobservable, so estimating it will always be problematic. The simplest way is trial and error. When policy is easy and moving back towards neutral, as is currently the case, the Fed has in the past continued to raise rates until the economy slows down (at which point, we will know that the Fed has overshot neutral). Or, to put it another way, the Fed has to hike until it breaks something.
- Bullard’s second tweak to the Taylor Rule regards the “disappearing Phillips Curve.” Bullard takes a widely-held view that the relationship between the real economy and inflation (in this case, between the unemployment gap and inflation) has become far weaker than in earlier decades. Thus, he suggests that the coefficient in the Taylor Rule relating the unemployment gap to the prescribed policy rate should be slashed from 1 to 0.1.
I am somewhat sympathetic to the idea that inflation is less sensitive to the economy’s deviations from some unobserved equilibrium. Indeed, this is a sign of the Fed’s success. To the extent that the Fed has convinced the public that it will always get inflation back to the 2% target, it would take an egregious policy mistake to actually get a significant deviation of inflation from 2%. I suspect that this is a big part of the evolution of the slope of the Phillips Curve.
However, it does not necessarily follow that Fed policy should then also be insensitive to the unemployment rate gap. As Governor Quarles pointed out on the very same day, inflation in the U.S. has apparently become another example of an economic theorem known as Goodhart’s Law: when an indicator becomes a target of policy, that indicator loses its value as a gauge of the state of the economy. Instead it becomes a referendum on the credibility of the central bank’s commitment to do what it says it will do.
In other words, the fact that inflation is quite steady near the Fed’s target does not necessarily mean that the Fed should drastically cut its response to the unemployment gap. As Chairman Powell has suggested, in recent economic cycles, the downturn has come following/as a result of financial market trouble more than an excessive acceleration in consumer price inflation. It may well be the case that a significant unemployment gap is an accurate predictor of an overheating economy, even if the overheating ultimately shows itself as something other than consumer price inflation, for example excessive asset valuations. In this case, Bullard’s proposed tweak would be muffling an important signal.
Regardless, Bullard’s suggestion for the unemployment gap coefficient would mean that the Taylor Rule would not deviate very much from equilibrium based on the unemployment gap term. A 1 percentage point unemployment gap would only move Bullard’s Taylor Rule result by 10 bp. Even a 5 percentage point gap, as we saw (roughly) after the financial crisis, would only imply 50 bp of easing relative to equilibrium.
- Bullard’s final Taylor Rule tweak is to replace inflation with inflation expectations. He argues that we have sufficient history from TIPS breakevens that we can reliably use them as an improvement over the spot inflation rate in determining the inflation gap. In particular, Bullard proposes using the five-year TIPS breakeven rate minus 30 bp, to account for the historical difference in rates of increase between the CPI and PCE deflator.
While I agree with Bullard in theory that inflation expectations may be a better measure to use than actual spot inflation in a Taylor Rule, there are several questionable elements to this proposed change. First, I am not convinced that the movements in TIPS breakevens are a clean reflection of changes in inflation expectations. It has been widely researched and reported on that a substantial proportion of the movement in TIPS breakevens reflects changes in risk and liquidity premia. In fact, though Bullard is a big believer in TIPS breakevens, my sense is that the FOMC as a whole pays significantly less attention to them today than it did, say, a decade ago. One particularly relevant argument in recent years is that the Fed has distorted the TIPS market (and Treasury yields generally) via its massive QE purchases. It may be no coincidence that TIPS breakevens fell to unprecedentedly low levels in the 2014-2016 period, when the Fed’s holdings of Treasury securities and thus the impact of QE was at its peak.
In addition, even if we stipulate that a TIPS breakeven is the right variable to use, it is not clear to me that the 5-year breakeven rate is the proper measure. This gauge, the shortest “on-the-run” breakeven rate, tends to be disproportionately influenced by near-term considerations, such as swings in oil prices. This makes sense, as TIPS investors do not have much insight into inflation three or four years out, so they project out from what they know, often ascribing more persistence to the latest swing in energy prices than history would dictate. To the extent that Fed officials routinely discuss that they tend to look beyond short-term fluctuations in inflation, it would stand to reason that we not include a breakeven measure heavily influenced by those high-frequency swings. Indeed, most Fed officials who discuss TIPS breakevens have focused on the 5-year, 5-year forward rate, which measures inflation expectations over the second half of the next 10 years, a gauge that would wipe away any influence from near-term energy price swings and focus on longer-run underlying expectations.
In any case, as with Bullard’s tweak to the unemployment gap coefficient, one natural consequence of this change would be to dampen movements in the Taylor Rule prescription for the fed funds rate. Particularly for the 5-year, 5-year forward breakeven rate, but even for Bullard’s preferred 5-year breakeven rate, in the absence of QE distortions or massive oil price swings, movements are generally pretty small. Both of these TIPS breakeven rates have meandered within a 50 bp range since the beginning of last year.
In any case, as the Goodhart’s Law discussion above implies, if the Fed’s inflation target is considered credible, then inflation (and inflation expectations) will not move much, right up until the Fed loses its inflation-fighting credibility. So, this term in the Taylor Rule essentially drops out anyway. Bullard’s suggestion regarding the Phillips Curve effectively removes the unemployment gap term from the equation as well (it will contribute no more than 10 to 20 bp in all but the direst scenarios). Thus, Bullard’s “modernized” Taylor Rule would amount to the fed funds rate never deviating by much more than a quarter point or two from the equilibrium level, except in cases where inflation moves up or down sharply due to short-term fluctuations (which the Fed typically prefers to ignore and could easily do by focusing, as it usually does, on core inflation rather than headline). That sounds like a nice uneventful world, but it is hard to believe that such a policy path would work out very well. Even if Bullard’s modifications suit his current purpose of justifying a halt to rate hikes, I cannot imagine that he or other Fed officials would be comfortable following his modernized Taylor Rule during a recession, which would require doing little to nothing while the economy crumbled.