The conventional wisdom regarding the evolution of short-term interest rates and long-term bond spreads is being challenged by some high profile economists in a recently published NBER working paper, with potentially significant implications for both monetary policy and term structure dynamics. Using a newly constructed dataset of short-maturity interest rates that spans over 200 years, the authors reassess very long run trends in the neutral rate of interest, commonly referred to as r*, and decomposition of the term premia. Among their conclusions (edited for brevity) are that:
- “Short-maturity rates have secularly fallen faster than long-maturity rates -- resulting in clear evidence of secularly rising term spreads. This is contrary to consensus literature that focused on falling term spreads over the recent decades in context of the ‘great moderation’ and falling inflation volatility.”
- “In order to jointly match long-run patterns in inflation volatility and term spreads, important qualifications in the existing literature are required: either term premia remain a close approximation of sovereign default risk and that risk has in fact risen; or else, factors other than default risk are long-run determinants of term premia time variation – plausibly liquidity factors.”
Among my own conclusions (assuming my understanding of the paper is correct, and no guarantees that’s the case) are that:
- Their analysis indicates the long run mean for r*, the neutral real rate of interest, is about 1.25% (see Figure A.1 below, excerpted from the paper, page 41 of 56). Presuming a scenario where inflation is near the Fed’s 2.00% target, that would imply the neutral Fed funds rate is about 3.25%.
- Full disclosure: I’ve long been a bit wary of both the usefulness of the term premium concept and the robustness of its calculation given the wide variation in model estimates, once famously (I flatter myself) calling it the flux capacitor of term structure theory. Common defenses of the term premium are roughly that “the levels of the model estimates may vary widely, but they’re all showing the same trend and that’s what you need to focus on.” Not sure how this latest challenge of the conventional methods will fit into the whole theory, but it’s not putting a dent in my skepticism as of yet. Hawking radiation this is not.
For complete discussion, see the new working paper from the National Bureau of Economic Research (NBER), Rethinking Short-Term Real Interest Rates and Term Spreads Using Very Long-Run Data, by Rogoff, Rossi and Schmelzing, published October 29, 2024.
Refresher Definitions and Concepts
r* or r-star – the natural or neutral short-term rate of interest that neither stimulates nor restricts economic growth when the economy is at full employment. It follows that r* should be the target of the primary monetary policy rate, which in the US is the overnight fed funds rate, when the economy is in equilibrium.
- The Fed should set the funds rate above r* to cool the economy and lower inflation, and set it below r* to stimulate the economy and combat deflation.
- The neutral rate is not directly observable and economists use models to estimate it. It is considered to be an underlying characteristic of the economy and may fluctuate over time.
Term spread – the difference in yield between two points on a yield curve, conventionally calculated as the longer-term minus the shorter-term yield.
- For example, if the overnight (O/N) fed funds rate is 2.00% and the 10-year Treasury (10yT) rate is 5.00%, the term spread between the O/N and 10yT rate is 3.00% or 300 bp (spread = 5.00% - 2.00%).
Term structure – the term spreads across the curve calculated against a fixed point, typically using e.g. the O/N fed funds or the 3-month T-bill rate as the short-term rate.
Term premia – Standard finance theory posits that long term yields can be separated into two components: expectations of short-term rates over the same time period plus a term premium. A simplistic calculation decomposing the yield on the 10-year Treasury can be written as:
Yield on 10yT = Avg expected T-bill yields over next 10 years + Term premium
Unfortunately, neither of the components of the nominal yield can be directly observed, so each must be estimated.
- The interest rate expectations component can be estimated via surveys or forecasted through the use of term structure models. This component is often further separated into two parts:
- Average expected future short-term real interest rates, plus
- Average expected inflation until the bond matures.
- The term (or risk) premium component is the additional compensation that risk-averse investors require for holding longer-maturity bonds instead of rolling short-term securities. This can be estimated using joint macroeconomic and term structure models. Several popular models also split the term premium into two parts:
- The real risk premium which is the compensation investors require to bear risk associated with variable future short-term interest rates, and
- The inflation risk premium which reflects the uncertainty of future inflation.
Evolution of term premia estimation
A variety of models have cropped up over the decades to estimate the term premium. The methodology of the models and the estimates tend to vary widely.
For an excellent overview of the term premia and several models used to estimate it, see the BIS article, Term premia: models and some stylized facts from 2018. An example of four different model estimates of the 10yT term premium, and the decomposition of the 10T nominal yield into 2 components using one of the models, is excerpted below.
The variation in the expectations component among the 4 models is also shown (left hand graph below), along with a complete decomposition of the 10-year yield using the Hordahl and Tristani joint macroeconomic and term structure model (right hand graph, below).
Updated estimates through 2024 of the term premium for the 10yT using the four most popular models currently continue to vary widely, from slightly negative to over 150 bp. See Will the True Treasury Term Premium Please Stand Up, for further details.
