In the latest press conference Powell received a question about the neutral rate of interest. His answer was vague and the Chair did not provide neither an estimate nor a view on whether it has risen or not. In this note, we present two recent papers on R* (“Global Natural Rates in the Long Run: Postwar Macro Trends and the Market-Implied r* in 10 Advanced Economies ” by Josh Davis et al. (2023), and “Global R*” by Cesa-Bianchi et al. (2023)) and offer a critique. In part II, we review another set of papers and we conclude. Overall, the real-time estimate of R* is subject to high uncertainty, and it remains very difficult to conclude whether R* has risen or not.
The press conference
Question from Megan Cassella: “Hi Chair Powell, Megan Cassella with Barron’s. Thanks for taking our questions. I wanted to see if you could talk about the neutral rate. You mentioned today that you’re still debating whether rates are sufficiently restrictive and you’ve recently said that evidence is suggesting policy is not too tight right now. So, I was curious if you could elaborate on that at all, and whether that means the neutral rate in your view has risen”.
Chair Powell answer: “Yeah, so first thing to say is that it’s very important, it’s a very important variable in the way we think about monetary policy, but you can’t identify it with any precision in real-time. And we know that. So, you have to just take that, you have to take your estimate of it with a grain of salt. What we know now is within a range of estimates of the neutral rate policy, is restrictive, and it’s therefore putting downward pressure on economic activity, hiring, and inflation. So, we do talk about this, there’s not any debate or attempt to sort of agree as a group on what, whether R* has moved or not, some people think it has, some people don’t think it has. Ultimately, it’s unknowable, and so really again, what we’re focused on is looking at the data and giving ourselves a little more time now to look carefully at the data by being careful in our moves, does it feel like monetary policy is restrictive enough to bring inflation down to 2 percent over time? That’s the question we’re asking ourselves. I think years from now economists will be revising their estimates of R* as it existed on November 1, 2023. You can’t, we can’t really wait for that in making policy. We have to look, we have to have those models and look at them and think about them and ultimately, we’ve got to look at the effects that policy is having, accounting for the lags, which makes is difficult.”
(Source: page 22 of press conference transcripts, available here).
Paper #1: Global Natural Rates in the Long Run: Postwar Macro Trends and the Market-Implied r* in 10 Advanced Economies
A new NBER paper “Global Natural Rates in the Long Run: Postwar Macro Trends and the Market-Implied r* in 10 Advanced Economies ” by Josh Davis (PIMCO), Cristian Fuenzalida (Graham Capital Management), Leon Huetsch (UPenn), Benjamin Mills (PIMCO), and Alan Taylor (UC Davis) estimates the drivers of yields in a unified macro-finance framework, which allows for both macroeconomic factors and risk premia. Trends in the natural rate and the inflation rate account for most variation in yields, while bond risk premia explain little.
What the paper does
The authors propose a hybrid model that bridges the finance and the macro approach to determining the drivers of the yield curve. The finance literature commonly uses term-structure models with factors describing level, slope, and curvature of the yield curve, but neglect fundamental macroeconomic factors explaining yields. The macro literature pays great attention to macroeconomic factors but often neglects financial market information such as risk premia. Davis et al. (2023) show that the two approaches give contrasting results for the natural rate and bond risk premia, implying different narratives for the historical trends in bond yields (an evidence labelled as “the natural rate puzzle”). In this paper, the authors develop a simple model that encompasses both types of information: Yields are determined by slow-moving macroeconomic trends in the natural rate and inflation, but also by a cyclical factor that captures financial market information (and other high-frequency movements) in a reduced form. The model uses data on bond yields, output growth, and inflation trends to estimate the natural rate and the cyclical factor.
Main results
A decomposition of the model fit shows that trends in inflation and natural rates together account for the vast majority in bond yield movements of the sample. This main result is shown in Figure 1, which plots the R^2 of in-sample regressions of bond yields on the potential drivers for each economy in the sample. Model fit is very high across specifications and is largely explained by macro trends for every economy (that is, the red bars explain most of the fit of the models when trends pi* and r* are put as explanatory variables). In further results, the authors show that future excess returns are predicted by the cyclical factor but not by macroeconomic trends. The R^2 in these predictive regressions is generally much lower.
Figure 1. Main result of Davis et al. (2023). Specification 0 includes no trend, specification 1 adds the inflation trend, and specification 2 adds both the inflation trend and the natural rate trend.
Another important result of the paper is that the estimated natural rate shows a stronger decline over the past 50 years than estimated in previous work. Natural rates may have remained stable until the 1990s, before starting a rapid decline toward or below zero. This pattern is strongly correlated across countries, and the dispersion of natural rate estimates across countries has decreased over time. The authors study two potential drivers of the decline in the natural rate: population aging and weaker economic growth. They find both drivers to explain a large share in the variation of natural rates. Finally, the authors do not find evidence of an increase in R* following the pandemic (sample ends in June 2023) – see Figure 2 below.
Figure 2. Estimates of R* for the US, Davis et al. (“our estimates” in the figure) vs other models
A critique of the paper
Overall, it is appealing that the paper attempts to study macro and finance drivers of bond yields in a joint framework, hence proposing to bridge the macro and finance literatures that are answering the same question with different methods. We have two main concerns.
The analysis in the paper is subject to endogeneity concerns. The authors focus on decomposing the drivers of bond yields and of natural rates through a statistical approach and interpret regressors that explain most of the variation in the dependent variable as important drivers. However, their model cannot fully disentangle the causal mechanism underlying fluctuations in macroeconomic or financial factors. For example, trends in the natural rate and inflation may be driven by a set of unobserved factors (which could be financial) or may feature a causal relationship between each other. Similarly, the paper considers two possible drivers of the natural rate and finds these to explain natural rates well. While this confirms previous findings on the importance of demographic changes and weak productivity for trends in the natural rate, it does not control for a host of other explanations which may affect the natural rate.
The cyclical factor is defined in a reduced form that is convenient but hard to interpret. While the authors focus their interpretation of the factor on financial market information, it can in theory be driven by other macroeconomic factors capturing cyclical information. To the extent that it is not clear if the cyclical factor represents financial factors instead of non-financial information about the business cycle, it is questionable if the paper bridges the macro and finance approaches to explaining bond yield trends.
Policy implications
This paper’s findings reinforce concerns about secular stagnation restricting monetary policy through the effective lower bound in times of weak economic growth. The model estimates of the natural rate are lower than in most previous work, with current natural rates being estimated to be near zero or negative across developed economies. The authors suggest that weak economic growth and aging populations are the main drivers of low natural rates. Given current forecasts of ongoing growth challenges and continued aging in the population, the natural rate may remain close to zero, although this statement is subject to estimation uncertainty (plus/minus 75bp confidence intervals).
The findings of the paper suggest a “new” reading of the decline in bond yields across advanced economies: bond risk premia have been remarkably stable over the previous decade with cyclical variations around a small but positive average. Yields have decreased due to downward macroeconomic trends, mainly driven by weak productivity and aging populations. To the extent that productivity growth will remain anaemic and countries will face demographic issues, the downward pressures on R* is expected to continue.
Paper #2: Global R*
A new paper (“Global R*”) by Ambrogio Cesa-Bianchi, Richard Harrison, and Rana Sajedi from the Bank of England suggests that the secular decline in global real interest rates since the 1970s is mostly driven by a productivity slowdown and increased longevity. The authors estimate a structural model featuring different drivers of movements in the real interest rate. The focus of the paper is on long-run trends and does not consider factors associated with business cycle fluctuations.
What the paper does
This paper develops a structural model to estimate the trend in the global equilibrium real interest rate. The authors consider five potential drivers of the interest rate trend: Productivity growth, population growth, longevity, government debt, and the relative price of capital. The model is estimated using average trends of these five drivers across 31 countries and generates a simulated path for the trend in the real interest rate.
Main results
The main result of the paper is shown in Figure 3. The figure reports the estimated trend of the global real interest rate (solid black line) along with a decomposition into the different structural drivers that the model considers. The global equilibrium real rate has trended up from 1950 until the mid-1970s to a value of around 2.7%, before starting a decline until the end of the sample in 2020, when it reaches is -0.7%.
The increase in longevity and the slowdown in productivity growth have contributed most strongly to the downward trend in real interest rates. Slower productivity growth decreases expected returns on investment, resulting in weaker demand for capital. Higher longevity lengthens the retirement period and increases the level of wealth households wish to hold to fund their retirement. This increases the supply of capital. Over the sample period, population growth has consistently contributed positively to the trend in interest rates, but the effects are quantitatively weak compared to the two main drivers. Government debt and the relative price of capital are not found to have any significant effect on the trend in real rates.
Figure 3. Main result of Cesa-Bianchi, Harrison and Sajedi (2023).
Note: x-axis in Figure 3 shows time in years (i.e. “11-15” stands for 2011-2015)
A critique of the paper
The model estimates are subject to large estimation uncertainty. This can be seen in two ways. First, the authors show that the structural model estimates are in line with empirical estimates of the trend in interest rates from a Bayesian vector autoregression. Both indicate a decline in the trend component of interest rates over the previous decades and suggest negative levels of the trend in 2020. Empirical estimates such as the VAR shed light on the estimation uncertainty surrounding the point estimate: the 95% confidence interval of the VAR model indicates the level of the trend component falls between +1% and -1%. Second, to give a sense of the uncertainty surrounding their model estimate, the authors re-estimate their structural model for a set of alternative, reasonable parameter choices. All model simulations indicate the same pattern of interest rate trends peaking in the 1970s and declining thereafter. However, the range of trend estimates in 2020 spans around 3 percentage points and is skewed to the downside. The decline in the interest rate trend since 1970 could therefore be larger than the main result suggests.
The model likely overestimates the importance of increased longevity for the decline in interest rates. This is because the model does not account for pension systems, which constitute a large share of retirement wealth across many developed countries. Income streams from pensions are usually guaranteed while income from private wealth is subject to asset price fluctuations, such that the model overestimates the desired wealth of households at retirement if agents are risk averse.
Policy implications
The results suggest that the future trend of the global interest rate may remain at a low level unless there is a substantial reversal in the trends of longevity or productivity growth. While short-term fluctuations around this trend may be large and cannot be determined using the framework laid out in this paper, the results suggest that monitoring trends in longevity and productivity growth can give important guidance on the appropriate framework for monetary policy. Once the disruptions to supply chains and the effects of the stimuli following the Covid recession have fully faded out, long-term trends in these two important drivers may revive concerns from the previous decade about low R* (this might also explain some of the hesitancy of the Bank of Japan to address rising inflation through contractionary monetary policy).