Anatomy of a rise: Monetary policy and the post-Covid surge in long-term interest rates

The sharp rise in long-term interest rates since 2020 is difficult to explain from slow-moving fundamentals. This column shows that narrow windows around nonfarm payroll releases and speeches by prominent Federal Reserve policymakers capture 80-90% of the observed increase in long-term US yields, despite covering only 24% of trading days. These events primarily shift expectations about the policy-rate path rather than the natural rate of interest. This suggests that natural rate of interest may anchor long-term rates less firmly than standard theory assumes, allowing shifts in monetary policy perceptions to generate persistent, and potentially self-validating, movements in yields.

From a conventional economic perspective, the swings observed in longer-term interest rates – and their sharp rise in recent years – is quite puzzling. Standard theory posits that, when longer-term inflation expectations are stable (as they have been), long-term rates are primarily determined by ‘real’ forces such as the supply and demand for safe assets, demographics, and the rate of technological progress. Since these factors are generally slow-moving, longer-term rates should also move gradually. 

However, this is not what we have observed in recent years. Since August 2020, when the cost of longer-term US government borrowing was near its lower point, the 10-year yield on US Treasuries has risen by over 400 basis points (see Figure 1, which also shows the 5-year/5-year forward rate, and the average expected short rate over the next 10 years).

Figure 1 US long-term interest rates have risen since 2020

To explain such a sudden rise within the framework of standard models, one would have to argue for radical changes to the underlying drivers. This could, in principle, be the case: the Covid pandemic brought fiscal challenges during the early part of this period, while recent progress in AI might have boosted (expected) growth in the later part.² This raises the question: 

Can recent long-term rate increases be explained from fiscal- or AI-related news?

In a recent paper, Christensen and Rudebusch (2026, henceforth “CR”) forensically examine whether worsening fiscal prospects or AI breakthroughs could explain the rise in the so-called natural rate of interest (r*), which is widely considered the main driver of long-term rates. They perform an event-window analysis to assess whether long-term rates rose in response to important news about these factors. If worsening fiscal prospects, or AI breakthroughs, caused investors to revise upwards their estimate of ‘neutral’, long-term rates should rise when that information arrives. 

To test this hypothesis, CR compile a set of key dates associated with fiscal-policy developments and major generative-AI news and analyse the cumulative movement in long-term rates on those days. Surprisingly, their findings reveal no substantial evidence that these events contributed to the observed rise in long-term rates. Specifically, they report that windows around fiscal policy events account for less than 20% of the increase, whereas AI-related event windows actually show a negative contribution. 

Next, CR turn to monetary policy. Conventional economic theory suggests that monetary policy should not have any influence over long-term real rates, as its effects are typically neutral over horizons longer than the period during which prices remain sticky.  However, prior evidence challenges this notion. Hillenbrand (2025) shows that the entire observed decline in long-term US Treasury yields in the decades preceding Covid occurred within a narrow three-day window surrounding Fed interest rate decision meetings (‘FOMC windows’). This finding suggests that monetary policy might have played a role in driving down long-term rates during that period. 

To explore whether a similar relationship exists also in the post-Covid era, CR conduct a Hillenbrand-style analysis for the period since 2021. Their finding reveal that the pattern has changed: changes in long-term rates during FOMC windows do not account for any of the observed increase in long-term interest rates. Lustig (2026) arrives at a similar conclusion, further reinforcing the idea that monetary policy is unlikely to be responsible for the recent upward trend in long-term rates. This leaves the drivers of the post-Covid surge in long-term rates as an unresolved puzzle.  

The role of news unrelated to r*

Standard economic theory suggests that persistent movements in long-term rates should mainly reflect changes in r*.  Accordingly, news that captures the attention of financial markets, but that is unlikely to provide meaningful information about r*, should not play a significant role in driving trends in long-term rates.  

To test this hypothesis, which is implicit in the work by CR, we examine different news events that are closely monitored by market participants due to their relevance to the near-term stance of monetary policy, even though they are not generally considered informative about r*. In particular, we focus on narrow, three-day windows surrounding: 

1. nonfarm payroll (NFP) releases (note that we drop the 10 March 2023 release as this coincided with the failure of Silicon Valley Bank); and
2. speeches from prominent monetary policymakers at the Federal Reserve Board.

Over time, speeches have become an increasingly popular tool for central bankers to communicate their policy reaction function, the systematic mapping from macroeconomic conditions (e.g. inflation and output) to policy decisions. Similarly, NFP releases are widely seen as providing a strong signal about the near-term trajectory of interest rates (Bauer et al. 2025). Importantly, neither NFP releases nor Fed speeches are commonly believed to have a direct impact on r*.

Our analysis reveals that the combined set of NFP and speech windows accounts for a disproportionately large fraction of the observed rise in long-term rates. While these three-day windows cover only 23.9% of trading days over our 1 August 2020-3 September 2026 sample period, they account for: 

• 90.5% of the observed rise in the 10-year nominal Treasury yield;
• 91.3% of the observed rise in the nominal 5-year/5-year forward rate; and
• 81.0% of the observed rise in the average expected short rate over the next 10 years.

Figure 2 visualises these striking patterns by plotting the cumulative changes in the various rates since 1 August 2020.

Figure 2 The majority of post-2020 rises in long-term interest rates occurred on a limited number of days, marked by Fed Board speeches and NFP data releases

Note: Event windows, cumulating to the blue lines, only consider changes from the close of day at (T-1) to the close of day at (T+1), where T dates the event (a speech or NFP release) itself. Last data point: September 3rd, 2026.

Why may long-term real rates be only weakly tied to r*?

These observations beg the question: why are longer-term real rates so sensitive to news that is unlikely to be informative about r*? This question also applies to Hillenbrand’s (2025) prior observations related to FOMC events, assuming one believes that the Fed does not have superior information about r*. 

We develop one potential explanation to such patterns is developed in Beaudry et al. (2025). We show why and when real rates can potentially deviate from r* for prolonged periods of time, and why the long end of the yield curve might not be firmly pinned down by real forces. The reason is that long-term rates determine the growth rate of retirement savings. From this narrow life-cycle perspective, higher long-term rates boost activity and inflation, as they allow households to meet their retirement goals with less savings today, leaving more resources for current consumption. Embedding this ‘interest income effect’ in an otherwise standard New Keynesian model, and considering a set of reasonable calibrations, we find that the permanent component of monetary policy (normally equated to the central bank’s estimate of r*) is weakly anchored. In particular, if the central bank sets policy based on a view of r* that is either too high or too low, the economy need not deliver strong error-correcting signals (such as inflation running away from target) that would quickly reveal the mistake. This makes perception errors with respect to r* rather slow to correct, leaving long-term real rates susceptible to self-fulfilling dynamics. 

This perspective makes the theoretical construct of r* much less binding for monetary policy than conventionally thought, as it provides a very weak anchor to long-term rates.  Long-term real rates may be best viewed as quasi-indeterminate from a purely fundamental standpoint – in the sense of not being exclusively driven by the real forces thought to affect r*. This leaves open the possibility that monetary policy can emerge as an unwitting contributor to trends in long-term real rates. For example, after a large inflationary shock, a central bank might raise rates and erroneously infer that r* has risen and adopt a ‘higher-for-longer’ stance. In our model, the boost coming from the interest income effect roughly offsets the standard contractionary channels that would slow the economy when rates are persistently held above neutral (r*), so – on balance – activity need not fall sharply, nor inflation undershoot the target. This dynamic validates the initial ‘higher-for-longer’ belief, further reinforcing the view that r* has risen – and pulling interest rates up along the yield curve. 

In light of the puzzle raised by CR, we would argue that non-r*-related news has long been a key driver of interest rate trends – both pre- and post-Covid. Which specific non-r* news moves long-term rates can change over time, depending on what investors coordinate on. For example, this can vary with how the Fed communicates (primarily through speeches or mainly via policy rate decisions?), or with perceptions of what the Fed is responsive to. 

In such an environment the volatility observed in long-term real rates can easily exceed that of their underlying ‘secular’ drivers – a possibility that both policymakers and investors are wise to explicitly account for in their decision-making. 

Source : VOXeu