A client recently asked how we formulated our estimate of 30 basis points for the steady-state term premium. In response, we were motivated to write this commentary, with guidance from Jonathan Wright, a senior adviser at LH Meyer.
Let’s begin the discussion with how our estimate for the steady-state term premium evolved. ▪ When we first provided estimates of the steady-state term premium, in late 2002, we judgmentally set the term premium at 100 basis points, intentionally quite a bit below what we judged the consensus was at that time.
▪ Following the financial crisis, we lowered the estimate to 75 basis points.
▪ More recently, in early 2016, we lowered the estimate to 30 basis points as a result of consultations with our senior adviser Jonathan Wright, who cited a credible range of 0 to 50 basis points.
We have an empirical basis for our estimate of the steady-state term premium.
▪ We start with the Kim-Wright measure of the term premium on the ten-year Treasury yield.1 Their model produces a measure of the term premium at a given point in time, not its steady-state value. ▪ As can be seen in Chart 1, there is a lot of variation in this measure over short timeframes. But some lower frequency trends can also be observed over the period from the 1960s until today. ▪ We use a ten-year moving average of the Kim-Wright term premium measure as a proxy for its steady-state level. The ten-year average peaks in the mid-1980s and holds about steady through the rest of that decade.
▪ By the early 2000s, this ten-year moving average declined toward 100 basis points. ▪ Following the recession, the ten-year moving average had declined further, and 75 basis points seemed like a good estimate. (The ten-year moving average at the time was closer to 50 basis points.) ▪ In early 2016, the ten-year average fell to zero.
The average of the Kim-Wright term premium measure over (for example) ten years is just one of several considerations that informs our judgment. Here are other inputs:
▪ Starting from the estimate of the term premium today, based on the Kim-Wright measure, we ask which factors push that away from the steady-state, and then make a judgmental correction for these factors. For example, the term premium today is around zero or slightly negative and there is perhaps some scope for it to rise as the balance sheet normalizes (though much of that is already priced in). Other than that, we don’t see strong forces pushing the term premium up from here.
1 Don H. Kim and Jonathan H. Wright, “An Arbitrage-Free Three-Factor Term Structure Model and the Recent Behavior of Long-Term Yields and Distant-Horizon Forward Rates,” Federal Reserve Board, 2005. Their model produces estimates for 1990 to today. We use a proxy to approximate the evolution of the term premium before 1990.
▪ We take the historical average of the slope of the term structure but make a judgmental adjustment for the fact that deflation risk and the effective lower bound are likely to remain bigger concerns going forward than they were for most of the past 50 years.
▪ We look at the estimated slope of the yield curve five to ten years hence, as reported in long-horizon surveys.
▪ In the end, there is some judgment involved in combining the information from these various approaches. Still, one thing all these measures agree upon is that the steady-state term premium should be appreciably lower now than it was in the past.
Of course, we want a story to go along with the movements in our proxy for the steady-state term premium, specifically why it was so high in the mid-1980s through mid-1990s and has declined gradually thereafter. ▪ The term premium is usually thought of as the premium that investors demand to take duration risk. That
duration risk is usually viewed primarily as inflation risk. Consistent with this reason, the term premium was very high during the latter 1970s through much of the 1980s, coincident with the Great Inflation and its immediate aftermath. Our proxy for the steady-state term premium reached as high as 400 basis points during this period.
▪ Thereafter, as inflation fell and stabilized near 2%, our proxy for the steady-state term premium steadily declined to below 100 basis points.
▪ We see the decline to near zero following the Great Recession as carrying this story further. There is no more concern about secular stagnation and deflation—related in part to being in slower growth, low r star regime—than was the case previously. This, in turn, has resulted in another shift in the size of the term premium. In recent years, investors have been less concerned about inflation risk and more concerned about the risk of deflation and the accompanying implication for an extended period of a low term premium.
But, as noted above, there are other factors that affect the term premium. For example, both the Fed’s asset purchases and now the prospect of balance sheet normalization have had and will continue to have effects on the term premium. And, given how long these Fed actions have affected the term premium, the ten-year moving average has been influenced by these persistent, but ultimately temporary, considerations.
▪ In order to remove these balance sheet influences, we make an adjustment to our ten-year moving average proxy for the term premium using the results of some Fed research.
▪ A study by Engen, Laubach, and Reifschneider (2015) estimates that the peak cumulative effect of the balance sheet programs, in late 2014 when QE3 was completed, was to lower the term premium by 120 basis points; since then, the cumulative effect has been diminishing. They estimated that the effect is about negative 50 basis points today, and still shrinking [link].2
▪ In Figure 2, we adjust the quarterly series for the Kim-Wright measure of the term premium to remove the effect of QE by simply adding to that measure the effect on the term premium from QE estimated by Engen et al.3
▪ The black line in Figure 2 is the adjusted term premium and the black dashed line is the ten-year moving average of the adjusted term premium (an adjusted proxy for the steady-state term premium). ▪ The result is that the decline in the ten-year average of the adjusted term premium flattened out after the beginning of QE1 and is about 40 basis points today.4 Our current estimate of the steady-state term premium, 30 basis points, also takes into account a wider range of factors, including the increased likelihood that the Fed balance sheet (relative to GDP) will be permanently larger than in the pre-crisis period.
However, the steady-state term premium today depends on expectations about the balance sheet in the future and its corresponding effects on the term premium in the longer run.
▪ The estimates from the Engen et al. study assumed the balance sheet will shrink to the same percentage of GDP as in the pre-crisis period. That is unlikely to be the case, given that the FOMC appears to want to
2 Engen, Eric M., Thomas Laubach, and David Reifschneider (2015). “The Macroeconomic Effects of the Federal Reserve’s Unconventional Monetary Policies,” Finance and Economics Discussion Series 2015-005. Washington: Board of Governors of the Federal Reserve System, http://dx.doi.org/10.17016/FEDS.2015.005.
3 This adjustment assumes that the effect of QE on the term premium will ultimately go away, or become very small when the balance sheet is normalized. However, it seems quite likely that if the real neutral rate remains as low as we now expect, that asset purchases will become part of conventional monetary policy during most recessions and that its effect on the ten-year moving average of Kim Wright measure may never fully go away.
4 There are other factors that might be affecting the term premium today that might not continue into the future in which case we should also adjust the ten-year moving average to remove their effect. An example would be the global saving glut, though some form of that may still be influencing U.S. rates today.
remain in the current operating regime that features a soft floor on policy rates, which will require a larger balance sheet than was the norm before the crisis [click here for our commentary]. ▪ While much of the effect of balance sheet normalization on the term premium has already taken place, the normalized balance sheet, which we expect to be $3.5 trillion in 2023, will be larger than in the pre-crisis period in terms of its size relative to GDP. If the larger balance sheet is to be a permanent feature of the economic landscape, its effect on the term premium should be included in the steady-state measure.
The path of the ten-year moving average for the Kim-Wright measure will continue to be an important input into our judgment about the steady-state level of the term premium. We believe this measure can be particularly useful in identifying trends and changes in the trend, after also taking into account a wider range of factors influencing movements in the term premium.