The Fed and the Secular Decline in Interest Rates

Abstract

This paper documents a striking fact: a narrow window around Fed meetings captures the entire secular decline in U.S. Treasury yields. Yield movements outside this window are transitory and wash out over time. This is surprising because the forces behind the secular decline are thought to be independent of monetary policy. Long-term bond yields decline when the Fed cuts the short rate and when the Fed lowers its long-run forecast of the federal funds rate (the “dot plot”). These results are consistent with the view that Fed announcements provide guidance about the long-run path of interest rates.

The Fed’s ability to affect real rates of return, especially longer-term real rates, is transitory and limited. Except in the short run, real interest rates are determined by a wide range of economic factors, including prospects for economic growth—not by the Fed.

     Ben Bernanke, “Why are interest rates so low?”

One of the most important macroeconomic trends over the last several decades has been the secular decline in interest rates. The decline can be linked to a fall in inflation expectations (e.g., Sargent 1999; Kozicki and Tinsley 2001; Cieslak and Povala 2015; Bauer and Rudebusch 2020) as well as a decline in real interest rates. While the Federal Reserve has the ability to control interest rates in the short run, economists believe that the Fed has limited ability to affect real interest rates in the longer run. Accordingly, other economic forces are seen as the more likely driver of the decline in real interest rates. Indeed, the most prominent explanations for the secular decline are a global savings glut (Bernanke 2005), a lack of capital investment opportunities (Summers 2014), and a slowdown in productivity growth (Gordon 2017). What is common among these forces is that they are slow-moving and supposedly lie outside the control of monetary policy.

This paper documents a surprising fact in light of these theories: A narrow time window around monetary policy meetings of the Federal Reserve captures the entire secular decline in long-term interest rates over the last decades. As panel A of Figure 1 shows, changes in the 10-year U.S. Treasury yield that occurred within this window not only add up to the entire cumulative yield change since 1989, but also capture the low-frequency movements of the 10-year yield strikingly well. To put this finding differently, yield changes outside this window were transitory and offset over time. This is exactly what panel B shows.

The puzzling nature of this fact remains when we look at how other bond yields changed around Federal Open Market Committee (FOMC) meetings. The empirical pattern holds for both the 5-year Treasury yield and the 5-year/ 5-year Treasury forward rate. That is, even the decline in longer-term forward rates—which are presumably less affected by the monetary policy stance—is fully captured by a narrow window around Fed meetings. Additionally, we can decompose nominal yields into real yields and risk-adjusted inflation expectations since 1997, when Treasury inflation-protected securities (TIPS) were introduced. Consistent with the notion that inflation expectations have been relatively stable over the last two decades, the decline in long-term nominal yields comes entirely from a decline in real yields. This reinforces the puzzling nature of the pattern: while most economists believe that the Fed can influence long-term inflation expectations, they are less likely to believe that the Fed can impact long-term real interest rates.

The pattern is puzzling not only in light of existing theories on the secular decline, but also in light of theories on the workings of monetary policy. In standard New Keynesian macro-models (e.g., Clarida, Gali, and Gertler 1999; Galí 2015), changes in the nominal short rate transmit to real rates because of the stickiness of prices. However, in these models, short rate changes have no impact on real rates beyond the horizon over which prices in the economy adjust. In addition, the real economic effects of monetary policy do not seem to be long-lasting (Ramey 2016). Thus, it is surprising that there is a systematic downward drift in real, long-term bond yields around FOMC meetings.

I conduct various analyses to investigate the puzzling pattern further. The window used to construct Figure 1 includes 3 days. For each FOMC meeting, it includes the day prior to the meeting, the day of the meeting, and the day after the meeting. I start by decomposing the pattern into the individual days. I find that days prior to FOMC meetings account for 42% of the total yield decline occurring around FOMC meetings, FOMC meeting days account for 47%, and days following meetings account for 11%. The exact daily decomposition over the last two decades depends on whether bond yields are computed as of 3 p.m. or 5 p.m. ET. For the period after 1994, I conduct a finer decomposition of the pattern on a 5-minute frequency using intraday prices of 10-year on-the-run U.S. Treasury Notes. In line with the daily pattern, I find that yields start to drift downward on days prior to FOMC meetings and continue their decline during the next day until they drop sharply at the FOMC announcements. Yields then continue to decline until they remain roughly flat during the day following the FOMC meeting.

Next, I relate the change in the 10-year U.S. Treasury yield observed around FOMC meetings to the state of the economy and financial markets. I find that all financial and economic variables considered have fairly low explanatory power for the pattern. In addition, none of the variables is significantly related to the change in the 10-year yield around FOMC meetings. For example, there is little evidence that economic news prior to the FOMC meetings predicts the change in long-term yields around FOMC meetings. I also do not find that the 10-year yield decline around FOMC meetings is larger during recessions or when past equity returns were negative. Variables measuring uncertainty in financial markets, such as the VIX and the MOVE index, are the most promising and exhibit a slight (albeit statistically insignificant) relationship with long-term yield changes.

I, then, turn to how the pattern relates to the Fed’s actions and communications. I find that the short rate actions taken by the Fed are important for explaining the pattern. Until 2008, when the short rate reached the zero lower bound (ZLB), the 10-year yield decline is almost entirely concentrated at meetings at which the Fed changed the target for the federal funds rate. As an alternative test, I split the sample of FOMC meetings based on whether the monetary policy was unexpectedly hawkish or dovish using the current month’s federal funds futures contract (Kuttner 2001). I find that the yield decline up until the ZLB became binding occurred solely at meetings at which the new target of the fed funds rate was set below the market’s expectations. Over the entire sample period (including the ZLB periods), meetings with negative “Kuttner shocks” account for 77% of the yield decline observed around FOMC meetings. In other words, yields declined by a total 518 bps at these meetings.

While the Fed’s monetary policy actions explain a large part of the pattern, particularly during the first three decades, they cannot explain the continued downward drift of Treasury yields once the ZLB was reached. Constrained by the ZLB, the Fed revised its communication strategy and began to communicate its expected path of the federal funds rate to the public. Since 2012, the Fed has released the FOMC members’ forecasts for the federal funds rate over the next 3 years as well as over the longer run in the so-called “dot plot.” The forecasts for the federal funds rate over the longer run are particularly interesting since they show a protracted decline of more than 180 bps over the past decade. Anecdotal evidence reveals that the decline in these forecasts is driven by Fed officials’ view that the natural rate of interest has declined.

I test how the release of the Fed’s long-run forecast influences long-term yields, that is, the market’s expectation for the long-run level of interest rates, by regressing the change in long-term real yields on the meeting day (i.e., when the dot plots are released) on the change in the Fed’s long-run forecast relative to the prior meeting. The results are stark. They show that a 100-bps decrease in the Fed’s forecast for the long-run level of interest rates leads to a more than 70-bps decrease in long-term real yields. Thus, the results imply that the release of the dot plots lowered bond yields by around 130 bps over the last decade on days when they were released.

Importantly, the reaction of bond yields likely reflects only the causal impact of the information release, and not the causal impact of monetary policy itself. That is, long-term yields might have declined by the same amount without the Fed communication (but potentially on a different date). Generally, the evidence presented in this paper cannot distinguish between the Fed causing the secular decline in yields (for example, because the Fed might have brought down inflation expectations in the 1980s) and Fed meetings revealing longer-run trends in yields that are caused by economic forces outside the Fed’s control.

What, then, might explain the observed pattern? In the last part of the paper, I discuss potential explanations. One explanation for the observed pattern is the possibility that the Fed provides guidance to the market about the long-run level of interest rates at FOMC meetings—something I call “Long-Run Fed Guidance.” This guidance might be implicitly released through the Fed’s short rate actions or explicitly through the dot plot. In response to Long-Run Fed Guidance, the market might update its beliefs about the long-run level of interest rates. Over the past decades, this could have meant that the market came to know about the secular decline at FOMC meetings. In addition, Long-Run Fed Guidance might also lead to more coordination in financial markets such that informed market participants preferably trade on their own long-run information ahead of FOMC meetings.

Long-Run Fed Guidance might arise because the Fed has better information about long-run trends in interest rates or simply because the market overestimates how much control the Fed has over interest rates in the longer run. Long-Run Fed Guidance could explain (i) why long-term yields declined around FOMC meetings, (ii) why yields declined mostly when the Fed surprised the market with a lower federal funds rate, and (iii) why long-term bond yields responded sensitively to the Fed’s long-run dot plot, that is, the explicit guidance by the Fed about the long-run level of the federal funds rate. This makes the explanation appealing. At the same time, it is an open question why the bond market would rely so much on the Fed’s guidance. After all, the aggregation of information seems to be a key feature of financial markets.

Other channels may have also contributed to the pattern. Investors might have updated their perception of the monetary policy rule around FOMC meetings (Bauer and Swanson 2023; Bianchi, Lettau, and Ludvigson 2021; Bauer, Pflueger, and Sunderam 2024). Because of institutional or behavioral frictions (Hanson, Lucca, and Wright 2021), long-term yields might have responded excessively to these updates (Hanson and Stein 2015). As an alternative, the Fed might have more influence over the natural rate of interest than what is implied by standard macro-models (e.g., McKay and Wieland 2021; Rungcharoenkitkul and Winkler 2021). Additionally, the fact that FOMC meetings reveal important news for long-term bonds could imply that there is a risk premium associated with the FOMC uncertainty (Savor and Wilson 2013; Ai and Bansal 2018). I discuss the benefits and challenges associated with each explanation.

Contribution to the literature

This paper relates to the literature on the secular decline in interest rates. A large literature has examined the decline in trend inflation that occurred after the Great Inflation (e.g., Clarida, Gali, and Gertler 1999; Drechsler, Savov, and Schnabl 2020) as well as the decline in real interest rates (e.g., Bernanke 2005; Summers 2014; Gordon 2017). This paper documents that the secular decline in interest rates was realized in a narrow window around monetary policy meetings.

The paper also speaks to the literature studying the reaction of financial assets to monetary policy. The overarching theme in the literature is that long-lived assets are surprisingly sensitive to monetary policy. Starting with Bernanke and Kuttner (2005), a number of studies have relied on high-frequency identification to examine the effects of monetary policy on financial assets, for example on interest rates (e.g., Hanson and Stein 2015; Nakamura and Steinsson 2018; Binsbergen and Grotteria 2024). Some papers also focus on the effect of the Fed’s communication on asset prices (e.g., Gurkaynak, Sack, and Swanson 2005; Schmeling and Wagner 2024; Gómez-Cram and Grotteria 2022; Swanson 2021). I add to this literature by providing evidence that a particular form of communication, namely the Fed’s forecast for the long-run level of the federal funds rate—revealed through the dot plot—has a strong effect on long-term yields. Other studies have focused on asset price movements outside the high-frequency window. In their influential work, Lucca and Moench (2015) document the “preannouncement drift,” that is, the empirical pattern that equity returns are excessively high in the 24 hours leading up to the FOMC meeting. Relatedly, Cieslak, Morse, and Vissing-Jorgensen (2019) argue that the entire equity premium is earned in the even weeks of a biweekly cycle of the FOMC announcements.¹  Bianchi, Lettau, and Ludvigson (2021) and Bianchi, Ludvigson, and Ma (2022) provide evidence for the effect of monetary policy on asset valuations at lower frequencies. I add to this evidence by showing that there are large and systematic movements of long-term nominal and real bond yields in the 3 days around FOMC meetings.

1. Data

1.1. FOMC meetings

The FOMC is the committee within the Federal Reserve that is responsible for conducting monetary policy.² It consists of 12 voting members—the 7 members of the Board of Governors, the president of the Fed New York, and 4 of the remaining 11 Reserve Bank presidents, who serve 1-year voting terms on a rotating basis. Monetary policy decisions are made by the FOMC based on a majority vote.

Since 1981, the FOMC has typically carried out eight scheduled meetings per year.³ The majority of monetary policy decisions since 1994 have been made during these scheduled meetings, while only relatively few changes in monetary policy were decided during unscheduled meetings (typically in the form of conference calls). By contrast, prior to 1994, these unscheduled meetings accounted for a large fraction of federal funds rate target changes.⁴

Meeting dates

Importantly, I want to obtain the dates when the meeting information was released to the public. For example, for scheduled FOMC meetings before 1994, the market learned about any change in monetary policy typically on the day following the meeting through the open market operations of the New York Fed. In order to correctly identify release dates of meeting news, I use two sources: dates of scheduled and unscheduled FOMC meetings from the website of the Federal Reserve and dates that the market associated with a change in monetary policy, as identified by Kuttner (2001, 2003). My main sample therefore starts in June 1989. After 1994, monetary policy was conducted mostly in the form of scheduled meetings and the Fed always released a statement if there was a change in the federal funds rate. I use the day when the statement was released (this is typically the second day of the FOMC meeting).

In total, my sample contains 283 FOMC meetings, of which 256 are scheduled meetings and 27 are unscheduled meetings. A further description and a detailed list of the FOMC meeting dates can be found in the Internet Appendix.

Dot plot

In recent times, the Fed releases the “Statement of Economic Projections” (SEP) on a quarterly basis—essentially at every other meeting— simultaneously with the monetary policy statement. The SEP contains the forecasts of all (voting and nonvoting) FOMC meeting participants for GDP growth, unemployment, inflation and the federal funds rate over the short run and the longer run. The market and the financial press refer to the forecasts for the federal funds rate simply as the “dot plot” because of a prominent chart that reveals the individual (anonymized) forecasts. The dot plot has been released since the FOMC meeting on January 25, 2012.

I collect all individual forecasts of the federal funds rate of all FOMC meeting participants, that is, all “dots,” from Bloomberg. My sample contains all 38 dot plot observations since January 25, 2012, up to and including the FOMC meeting on June 16, 2021. In my analysis, I concentrate on the forecast for the federal funds rate over the “longer-run” (this is discussed further below). To get a forecast that is representative of the average Fed view, I compute the mean of the individual forecasts for the long-run level of the federal funds rate.⁵

1.2. Bond market data

Daily data

I use the interpolated zero-coupon (continuously compounded) U.S. Treasury yield curve constructed by Gürkaynak, Sack, and Wright (2007) (GSW). For the yield curve interpolation, GSW use a large set of U.S. Treasury Notes and Bonds, but exclude the two most recently issued securities (with maturities of 2, 3, 4, 5, 7, 10, 20, and 30 years). Analogous to the data on nominal yields, I use interpolated zero-coupon yield curves for U.S. Treasury inflation-protected securities constructed by Gürkaynak, Sack, and Wright (2010) to obtain data on real yields and breakeven inflation. Treasury inflation-protected securities were first issued in 1997 and the data provided by Gürkaynak, Sack, and Wright (2010) is available from 1999 onwards. For both data sets, the quotes used to construct the daily yield series are as of 3 p.m. ET.

In addition, I obtain yields of the 10-year on-the-run U.S. Treasury Notes (Ticker: “USGG10”) and the 10-year on-the-run U.S. TIPS Notes (Ticker: “GTII10”) from Bloomberg (BB). I use mid-quotes. The on-the-run security is the most recent issued security of a certain maturity, for example 10 years. This data has the advantage that yields are based on the traded quotes of individual securities instead of being extracted from the interpolated yield curve. On-the-run securities are also more liquid than off-the-run securities and they might, therefore, incorporate information more quickly. The disadvantage of this data is that the duration of the 10-year on-the-run issue has changed as the coupon rates have decreased over time because bonds are typically issued close to par value. Furthermore, the issuance of a new on-the-run bond requires a switch between securities. Bloomberg uses quotes as of 5:15 p.m. ET to construct the daily yield series.

Intraday data

I also use intraday bond pricing data on the on-the-run 10-year U.S. Treasury Notes from GovPX. This data is based on quotes from the interdealer Treasury market. In particular, the data features the quotes that dealers submit to various voice-assisted brokerage systems (Fleming and Remolona 1999). I use mid-quotes.

2. A New Fact: The Decline in Interest Rates around FOMC Meetings

In this section, I show that a 3-day window around FOMC meetings fully captures the decline in interest rates over the last decades. I define this 3-day window such that it includes, for every FOMC meeting, the day prior to the meeting, the day of the meeting, and the day after the meeting. I set the day of each FOMC meeting to the day when monetary policy news were revealed to the public. As elaborated in Section 1, this is typically the day after the FOMC meeting prior to 1994 and the meeting day since 1994. I then distinguish between days that fall into the 3-day window around FOMC meetings and days that do not. Doing so, I construct a hypothetical time series of the 10-year U.S. Treasury yield that takes into account only yield changes realized during the 3-day FOMC window and ignores the yield movement on days outside of the 3-day window, that is, it sets these yield changes to zero.

The results of this analysis are shown in panel A of Figure 1. The dark gray line shows the actual evolution of the 10-year U.S. Treasury yield. The line reveals that the 10-year yield declined by around 7% between June 1989 and June 2021. This is the secular decline in interest rates that has received substantial attention in the literature. By contrast, the red line shows the hypothetical time series of the 10-year U.S. Treasury yield when we consider only yield changes that were realized within the 3-day FOMC window. As the figure shows, the yield changes that occurred within this window not only add up to the entire cumulative yield change since 1989 but also capture the slow-moving component of the 10-year yield strikingly well. The 3-day window includes roughly 10% of trading days since FOMC meetings take place every 6 weeks on average. Thus, 10% of trading days capture the secular decline in interest rates.

We can also illustrate the main result differently. Instead of constructing a hypothetical time series of the 10-year U.S. Treasury yield that considers only the yield movements on the 3 days around FOMC meetings, we can create an alternative hypothetical time series taking into account only the yield movements on days outside of the 3-day FOMC window—essentially, the other 90% of trading days. This time series is shown as the light gray line in panel B of Figure 1. Given the previous analysis, the results are hardly surprising: This hypothetical yield fluctuates around zero over the entire sample period and exhibits no cumulative decrease in the yield over the sample period. Putting it differently, these days do not account for the secular decline in long-term Treasury yields.

We can decompose the (continuously compounded) 10-year U.S. Treasury yield |$ y_{10y} $| into

where |$ y_{5y} $| is the 5-year Treasury yield and |$ y_{5y5y} $| is the 5-year/5-year forward rate, that is, the 5-year yield 5 years from now under current market prices. Similarly, for roughly the last two decades for which we have data on TIPS available, we can split the 10-year U.S. Treasury yield into

where |$ y_{TIPS,10y} $| is the 10-year yield on TIPS and |$ y_{Bkeven,10y} $| is the 10-year breakeven inflation.

I repeat the main analysis using these yield components, that is, I add the yield changes that occurred within the 3-day window around Fed meetings for the 5-year yield, the 5-year/5-year forward rate, the 10-year TIPS yield, and 10-year breakeven inflation in Figure 2. The figure shows that the main empirical fact also holds for the 5-year yield (panel A) and the 5-year/5-year forward rate (panel B). In other words, the 3-day window around FOMC meetings accounts for the secular decline even in longer-term forward rates. That is, even the decline in longer-term forward rates, which are supposedly less affected by the current monetary policy stance, is fully captured by the narrow window around Fed meetings. When looking at the 10-year TIPS yield (panel C) and 10-year breakeven inflation (panel D), we see that the entire decline in nominal interest rates since 1999 was due to a decline in real interest rates, while breakeven inflation was quite stable. Given the prior results, it is therefore unsurprising that we find a large decline in the TIPS yields around FOMC meetings, while breakeven inflation shows little movements around these meetings.

I also vary the starting point of my main analysis in Figure 3. First, I start the analysis in 1980, shortly after Paul Volcker became the new chairman of the Fed. For this period, it is somewhat harder to determine when the market learned about monetary policy news, mainly for two reasons (for the same reasons it is not feasible to extend the sample further back). First, the Fed was deliberately opaque at the time (Lindsey 2003), and the market learned about Fed actions through the Fed’s open market operations. In addition, as there was also a lot of volatility in money markets, it was not always clear to markets whether a change in the fed funds rate represented a Fed action. Second, monetary policy was mainly conducted during unscheduled meetings. I rely on two sources to collect the meeting dates prior to June 1989 (this is further discussed in the Internet Appendix): the website of the Federal Reserve for the scheduled meeting dates and the federal funds target rate series constructed by Thornton (2005) for unscheduled meetings.

Panel A of Figure 3 shows that the main result also holds when starting the analysis in 1980 (although there are some larger deviations for the period between 1980 and 1990). This is interesting, as the time period after 1980 was mainly characterized by a decline in long-term inflation expectations (e.g., Cieslak and Povala 2015), which is quite different from the period since 2000 (or even 1990) during which inflation expectations have been fairly stable. Panel B shows that the main pattern also holds when starting in 1994. This was the year when the Fed started to communicate its decisions via the release of FOMC statements.

Figure 4 shows that the documented pattern is not an artifact of the interpolated zero-coupon yield curve, but also holds when using quotes from on-the-run 10-year U.S. Treasury and TIPS Notes. In other words, the pattern also holds when we look at yield changes of traded, highly liquid securities. Looking directly at market prices rules out the concern that the yield curve interpolation distorts the analysis.

Finally, Figure 5 documents that the main pattern also holds when we expand the window to include more days around FOMC meetings. In particular, the figure considers only yield movements that occurred in a symmetric 7-day window around FOMC meetings. That is, in addition to including the days |$ t-1 $|⁠, t, and |$ t+1 $|⁠, the window includes the days |$ t-3 $|⁠, |$ t-2 $|⁠, |$ t+2 $|⁠, and |$ t+3 $|⁠. This shows that the documented pattern is unlikely to be due to any market microstructure or liquidity events that would likely be mean-reverting.

Could this be due to chance?

One might wonder whether it is possible that the empirical pattern arose by coincidence. To address this question, I conduct the following placebo test. For each month over the sample, I randomly draw days in the amount equal to actual FOMC meetings within this month. These days represent “placebo FOMC meetings.” Repeating this procedure for all months in the sample, gives a simulated path of placebo FOMC meetings. Analogous to the analysis above, we can then compute the cumulative yield change that occurred in the 3-day window around these placebo FOMC meetings. Panel A of Figure 6 shows the outcome of 1,000 simulated paths. The figure illustrates that the likelihood of obtaining a yield decline as large as the yield decline around actual FOMC meetings (shown as the red vertical line) is near zero. Thus, the pattern is highly unlikely to have arisen solely due to coincidence.

However, yield changes within the 3-day window around FOMC meetings not only add up to the total yield decline, but also capture the slow-moving component of the 10-year U.S. Treasury yield strikingly well. To formalize this observation, we can ask how much of the total movement in the 10-year yield over a 2-year horizon is explained by yield movements within the 3-day window around Fed meetings over the same horizon.⁶ Panel B of Figure 6 compares the variation explained by the actual FOMC meetings (shown again as the vertical red line) with the variation explained in the 1,000 simulated FOMC meeting paths. Again, we draw the conclusion that the actual path of FOMC meetings is a statistical outlier: Less than 1% of the simulations reach a higher |$ R^{2} $| than the actual Fed meetings. To conclude, the empirical pattern is highly unlikely to have occurred just by chance.

Next, I compare the average daily yield changes occurring within the 3-day FOMC window with the average yield changes occurring outside the FOMC window. Table 1 shows that the average daily change within the FOMC window ranges from –0.94 bps for the 1-year Treasury yield to –0.49 bps (–0.63 bps) for the 30-year off-the-run (on-the-run) Treasury yield. The average yield change within the FOMC window is significantly different from zero for all maturities and forward rates.⁷ By contrast, the average yield change occurring outside of the FOMC window is close to zero and statistically insignificant. Of course, this is not surprising given the empirical evidence shown above. The table repeats the key message of the paper: yield changes observed within the FOMC window account for the secular decline in yields; yield changes outside the window do not. The last rows document the average daily change in TIPS and breakeven yields over the last two decades. The average daily change within the 3-day FOMC window is between –0.70 bps for the 10-year on-the-run U.S. TIPS yield and –0.57 bps for the 30-year on-the-run U.S. TIPS yield. All changes are statistically different from zero. By contrast, the average change in breakeven yields and inflation swap rates is close to zero.

2.1. Daily decomposition

The main analysis uses a 3-day window around FOMC meetings, that is, it includes day |$ t-1 $|⁠, day t (the meeting day), and day |$ t+1 $|⁠. Table 2 shows the cumulative change in the 10-year U.S. Treasury yield that occurred on these days separately. The table documents that yields already declined on the days leading up to FOMC meetings. The cumulative yield decline per decade on this day is between 70 and 99 bps. In total, the 10-year yield declined by 339 bps on days prior to FOMC meetings. Slightly more important are FOMC meeting days, where yields declined by 374 bps in total. The days following FOMC meetings are less important for the pattern. Yields declined by 88 bps on these days over the last decades.

The table also documents that the importance of the FOMC meeting days depends on the yield measure used to conduct the analysis. As outlined above, the GSW data uses bond prices as of 3 p.m. ET, while the Bloomberg data uses prices as of 5:15 p.m. ET. Thus, the latter include any potential movements that occur in the 3 hours following the FOMC meetings, while the GSW only allow 1 hour to pass after the FOMC statement is released. In more recent times with expanded Fed communication, such as press conferences, it might take time for yields to incorporate all information released by the Fed (Gómez-Cram and Grotteria 2022). The following section sheds further light on this issue by using intraday data.

Intraday evidence

For the meetings after 1994, it is possible to determine the time at which the Fed communicated the meeting outcome to the public. For these meetings, I compute the cumulative decline in interest rates during the 3-day window on a 5-minute basis using intraday prices of the on-the-run 10-year U.S. Treasury Notes. I obtain the exact release times of the monetary policy announcements from Gorodnichenko and Weber (2016) and the website of the Federal Reserve. In total, this period contains 212 FOMC meetings at which the Fed announcement was mostly at or slightly after 2 p.m.⁸ The total yield decline for this subset of meetings is 492 bps (454 bps) in the GSW (Bloomberg) data.

The U.S. Treasury market is an over-the-counter market that operates around the clock. I restrict the analysis to the time interval from 9 a.m. to 5 p.m. ET to focus on times when liquidity is high. Whenever I refer to the “closing” or the “opening” price in the following, this is with regard to this time interval. Using the closing yield 2 days prior to the meeting day as a starting point, I compute the cumulative yield change that occurred within the next 3 trading days for every 5-minute interval. I, then, sum the cumulative yield change across all meetings. To ensure that the analysis based on intraday data aligns with daily yield data, I also compute the cumulative yield change based on daily data for the same subset of meetings taking into account that different data sources use closing prices at different times.

Figure 7 shows that yields gradually decline ahead of FOMC meetings, which is somewhat similar to the “predrift” in the equity market (Lucca and Moench 2015). Leading up to the FOMC meeting at 2 p.m. ET, the cumulative decline across all meetings is 216 bps. This is followed by a sharp drop in yields when the meeting information is released after 2 p.m. Two hours later, yields have decreased to 389 bps, and they continue to decline further. Potentially, it might take time for the market to digest the monetary policy news, as, for example, various experts speak on financial media. This is also consistent with intraday volatility in the 10-year U.S. Treasury Note remaining elevated for a substantial period of time after the meeting announcement. Relatedly, Gómez-Cram and Grotteria (2022) also document a strong positive correlation between the price movements directly following the announcements and the price movements during the Fed chair’s press conferences for several assets. On the day following each FOMC meeting, yields drop at the opening and then stay roughly flat until the end of the day. The intraday data aligns well with the daily data.

2.2. State dependence

Next, I test whether the pattern is state-dependent. Putting it differently, I test whether the movement in the 10-year yield can be predicted (or explained) by various economic and financial variables. To test this, I run the following regression

where |$ \Delta_{t_{i}-2,t_{i}+1}y_{10y} $| measures the 3-day change in the 10-year U.S. Treasury yield around FOMC meeting i and |$ X_{t_{i}-2} $| is a predictor variable measured 2 days ahead of the FOMC meeting, that is, as of day |$ t_{i}-2 $|⁠. I use various financial and economic variables as predictors. First, I consider the first two principal components of the yield curve extracted using the 1-year, 2-year, 5-year, 10-year, and 30-year U.S. Treasury yields. Next, I include variables capturing the news released ahead of the FOMC meeting. I include a dummy variable that indicates whether the stock market return in the past month was negative. Following Bauer and Swanson (2023), I consider the quarterly change in the S&P 500 index, in commodity price indices, and in the slope of the yield curve. Fourth, I consider macro variables that capture economic activity such as the Chicago Fed National Activity Index and the Brave-Butters-Kelley Index (Brave, Butters, and Kelley 2019). I use the values from the month prior to the FOMC meeting in the regression. I also include the surprise in the latest nonfarm payroll announcement occurring prior to the 3-day FOMC window. The surprise is defined as the actual value minus economists’ consensus forecasts.⁹ Next, I include an indicator of whether the Fed meeting took place during a recession. Finally, I include dummies indicating whether the volatilities in the equity market (as measured by the VIX) and in the Treasury market (as measured by the MOVE index) are above average.

Table 3 finds little evidence that the pattern depends on financial or economic conditions. None of the variables are statistically significant at the 10% level. The R-squared for any set of explanatory variables is below 1.1%, indicating that the explanatory power of changes in the 10-year U.S. Treasury yield around FOMC meetings is fairly low. For example, the variables that are important for the “Fed response to news” channel (Bauer and Swanson 2023) have little explanatory power for long-term Treasury yield changes around FOMC meetings. Similarly, it is not the case that the 10-year yield declined mostly at FOMC meetings when past equity returns were negative (although including the variable reduces the statistical significance of the intercept) or when the economy was in a recession. Variables related to financial uncertainty are the most promising, and long-term yields declined by more when uncertainty in financial markets was high (albeit not significantly so).

The Internet Appendix provides further results. First, I show that the results are not explained by other news releases that fell into the 3-day window around FOMC meetings, such as macroannouncements and corporate earnings news. Second, I split up the decline that took place around scheduled versus unscheduled meetings for both the 5-year yield and the 5-year/5-year forward rate. I find that unscheduled meetings are more important for the decline in the 5-year yield, but do not matter for longer-term forward rates such as the 5-year/5-year forward rates.

3. The Role of the Fed’s Action and Communication

3.1. The Fed’s short rate actions

In this section, I relate the observed pattern to the Fed’s monetary policy actions and communications. The conduct of monetary policy has changed substantially over the past decades. While the Fed was opaque in the 1970s and 1980s—often deliberately so, going as far as intentionally hiding the target of the federal funds rate—it started to become more transparent over time. In the 1990s, the Fed put more emphasis on the market understanding the Fed’s intended federal funds rate target. Since 1994, the Fed has released a statement whenever there was a change in the policy rate. Up until the federal funds rate hit the ZLB in 2008, changes in the federal funds rate were the Fed’s main policy tool.

I start by investigating how the pattern is related to the Fed’s short rate actions. To do so, I split the sample of FOMC meetings into two groups: one where the Fed changed the target for the federal funds rate and one where it did not. Panel A of Figure 8 shows that meetings at which the Fed changed the federal funds rate (80 out of 177 meetings until 2008) can account for almost the entire decline in long-term yields around FOMC meetings up until the ZLB became binding in late 2008. Including the ZLB episodes, meetings with target changes (97 out of 283 meetings) saw a decline in the 10-year U.S. Treasury yield of 435 bps (the total decline was 677 bps across all 283 meetings).

Alternatively, we can focus on meetings at which the Fed set the target for the federal funds rate below the market’s expected level (which can be extracted from the current month’s federal funds futures contract (Kuttner 2001; Bernanke and Kuttner 2005)). This set of meetings strongly overlaps with the meetings during which the Fed changed the target rate, but there are some differences, especially during the later part of the sample when the federal funds rate was close to zero. Panel B shows that the meetings with a negative “Kuttner shock” can account for 77% of the decline around FOMC meetings. In total, yields declined by 518 bps at the 103 (out of 283) FOMC meetings with a negative surprise in the federal funds rate target.

Table 4 analyzes this more formally by regressing the 10-year U.S. Treasury yield change in the FOMC window on a dummy that is one if the Fed changed its target for the federal funds rate. Columns (1) and (2) focus on the period until 2008 (when the ZLB was reached), whereas columns (3) to (5) focus on the entire sample period.¹⁰ The regression coefficient in column (2) shows that the yield change is on average –4.36 bps at meetings with a target change and –0.88 bps at meetings with no target change. Columns (4) and (5) include the surprise component of the first federal funds futures to get the unexpected monetary policy actions (Kuttner 2001). The coefficient in column (3) shows that the average yield change at meetings with a negative Kuttner shock is –5.03 bps, while it is –0.93 bps at meetings with a positive or zero Kuttner shock. Column (5) interacts the main explanatory variable with a dummy indicating whether the stock market return over the past month was negative. The results show that yields did not decline by more when past stock returns were negative, suggesting that the “Fed put” is not the main driver behind the pattern.

To conclude, the Fed’s short rate actions seem to explain the majority of the pattern up until the ZLB limited any further short rate action. That is, the 10-year U.S. Treasury yield decline was realized around FOMC meetings at which the Fed surprised the market with a lower short rate.

Once the ZLB was reached, the Fed engaged in more communication with markets and, in particular, provided more forward guidance. At the same time, long-term yields continued to drift downwards. Next, I examine whether the continued decline in long-term yields can be linked to the Fed’s communication. I focus on the so-called “dot plot,” a communication tool that features very prominently in financial news.

3.2. The Fed’s dot plot communication

Since January 25, 2012, the Fed has released its forecast for the federal funds rate in the dot plot. The dot plot is contained in the SEP and is released simultaneously with the FOMC statement on a quarterly basis; see Internet Appendix Figure D.1 for the first dot plot released at the meeting on January 25, 2012. The dot plot contains the (voting and nonvoting) FOMC members’ forecasts for the federal funds rate over the next 3 years as well as their forecasts for the federal funds rate over the “longer run.” These longer-run projections “represent each participant’s assessment of the value to which each variable would be expected to converge, over time, under appropriate monetary policy and in the absence of further shocks to the economy.” Fed officials also report (with a 5-year delay) the horizon over which the convergence is expected to occur. The reported horizon varies between 2.5 and 10 years with the median response being approximately 5–6 years (see the Internet Appendix for more details).

In the following, I focus on these long-run projections since they seem most relevant for long-run changes such as the secular decline in yields. The time series of these projections is shown in Figure 9. Three things are noteworthy. First, the long-run projections for the federal funds rate (dark gray dots) decline steadily over the sample period. The projections do not show any cyclical behavior, which is in stark contrast to the Fed’s forecast for the federal funds rate in 1 year (light gray diamonds). For example, the long-run forecasts do not drop during the COVID-19 outbreak. This underpins the long-run nature of these forecasts. Second, the figure shows that the Fed’s forecasts for long-run inflation (medium gray squares) have been stable at 2%. This means that the forecasts align with the Fed’s stated inflation target of 2%.¹¹ Thus, the decline in the Fed’s forecast for the long-run nominal rate is entirely driven by the Fed’s view about the long-run real rate. This is also exactly how FOMC members discuss these long-run forecasts. For example, one respondent stated in the SEP released on December 17, 2014, “I have lowered my long-run value of the federal funds rate by 25 bps based on my view that the long-run real rate of interest is somewhat lower.” The Internet Appendix contains additional quotes from FOMC members, including their views on which factors have driven down the long-run rate. When asked for the reasons behind lower forecasts, Fed officials mention lower potential GDP growth, higher global savings, and lower model estimates for the natural rate of interest. Third, the total decline in the long-run projections between January 2012 and June 2021 was 183 bps.

I examine how bond yields reacted to the release of the long-run dots. That is, I ask how the Fed’s disclosure of its expectation about the long-run level of the federal funds rate affects the market’s expectations. To do so, I regress the change in the 10-year U.S. TIPS yield on the day of the information release on the change in the Fed’s expectation from the past meeting to the current meeting. Let |$ t_{i} $| be the date of meeting i, then the regression specification is

where |$ \Delta_{i-1,i}\mathbb{E}^{F}\left[\text{Long-term fed funds rate}\right] $| is the change in the Fed’s forecast for the long-term federal funds rate from meeting |$ i-1 $| to meeting i and |$ \Delta_{t_{i}-1,t_{i}+1}y_{TIPS,10y} $| is the change in the 10-year U.S. TIPS yield on the day of meeting i (and the next day). I use the 2-day change in the yield, that is, the change from |$ t_{i}-1 $| to |$ t_{i}+1 $| (using a 3-day window leaves the results unchanged). The unit of observation in the regression is an FOMC meeting during which the dot plot is released. For example, when the unit of observation is the meeting day on June 16, 2021, the dependent variable is the change in the 10-year TIPS yield from the close on June 15, 2021, to the close on June 17, 2021, while the explanatory variable is the change in the FOMC forecast from March 17, 2021 (the last prior FOMC meeting with a dot plot release) to June 16, 2021. I focus on U.S. TIPS (instead of U.S. Treasury) yields since changes in the forecast are due to real interest rates.

To measure the causal effect of the information release on bond yields, the following assumptions have to hold. First, for there to be no reverse causality, the dot plot must not be influenced by the yield movements on the meeting day. In other words, the dot plot must reflect the FOMC members’ forecasts as of the day prior to the meeting. This seems plausible, as it takes time to prepare the SEP. Second, there cannot be an omitted variable that is correlated with the Fed’s forecast and the market’s reaction. This might be the case if the bond market reacts to other information released during the meeting and this information is correlated with the dot plot. However, this seems unlikely if one believes that the dot plot contains the pieces of information that are most relevant for long-term bonds. Nevertheless, I include various controls in the empirical specification: (i) the meeting-to-meeting change in the Fed’s forecast for the federal funds rate level at the end of next year; (ii) the quarterly change in the log of the S&P 500 index, in the log of the Bloomberg Commodity index, and in the slope of the yield curve; and (iii) the Chicago Fed National Activity Index, the Brave-Butters-Kelley Index (for both measures I use the prior month’s value), and the most recent nonfarm payroll surprise. These variables address potential concerns that business cycle news revealed between meetings may be an omitted variable (Bauer and Swanson 2023; Karnaukh and Vokata 2022) if, for example, the Fed reacted more strongly to negative news than the market expected (Cieslak 2018; Schmeling, Schrimpf, and Steffensen 2022).

Figure 10 documents a strong positive relationship between the change in long-term real yields and the change in the Fed’s forecast for the long-term level of the federal funds rate. This holds for the 10-year U.S. TIPS yield (panel A), as well as for the 5-year/5-year U.S. TIPS forward rate (panel B). The relationship is not driven by outliers.

Table 5 shows the regression results. There is a statistically and economically strong relationship between the long-run dots and bond yields. A 100-bps decrease in the Fed’s expectation for the long-run level of the federal funds rate leads to a 77-bps decrease in the 10-year U.S. TIPS yield on the FOMC meeting day and the day after (column 1). The |$ R^{2} $| is also high; the dot plots for the long-term federal funds rate explain 21% of the 2-day movement in the 10-year U.S. TIPS yield. Including the control variables in columns (2), (3), and (4) changes the regression coefficient only slightly. Columns (5) to (8) use the 2-day change in the 5-year/5-year forward rate as the dependent variable. This forward rate reflects the market’s (risk-adjusted) expectation of short rates further out in the future. Columns (5) to (8) document that the Fed’s long-run forecast has a similar effect on the 5-year/5-year U.S. TIPS yield to that on the 10-year U.S. TIPS yield. It is worth reemphasizing that the estimates likely reflect only the causal effect of the Fed’s information release and not the Fed’s causal effect on long-term interest rates. In other words, the results do not directly imply that the Fed is responsible for the decline in long-term yields. This view is consistent with the Fed’s view that its forecasts are only responding to economic forces outside its control.

The results suggest that the market updates its expectation about the future path of real short rates in response to observing the Fed’s dot plot forecasts. They also imply that long-term rates declined by around 134 bps (⁠|$ \approx 0.73\times 183 $| bps) in response to the release of the dot plots over the sample period. This evidence is also consistent with the Fed’s perceived effectiveness of the dot plot. For example, the former Fed vice chair Stanley Fisher stated in the speech “Monetary Policy Expectations and Surprises” in 2017 that “the SEP [which contains the dot plot] in particular has been useful in providing information on policymakers’ assessment of the potential growth rate of the economy and r*, the equilibrium real interest rate, both of which help guide the market’s expectations of the eventual path of policy” (Fischer 2017).

Taking stock, the Fed’s short rate actions and communication of the dot plot can explain most of the pattern documented in the previous section. That is, meetings where the Fed surprised the markets with a lower federal funds rate and where the Fed lowered the long-run dots, that is, its forecast for the federal funds rate over the longer run, account for a large fraction of the observed decline in the 10-year U.S. Treasury yield. In addition, the Fed conducted asset purchases (“quantitative easing”) to influence longer-term interest rates. Potentially, the market might perceive quantitative easing as a signal that interest rates were going to stay lower for longer (Eggertsson and Woodford 2003). In total, I find that the 10-year U.S. Treasury yield declined around FOMC meetings with quantitative easing announcements by slightly more than 100 bps. This number drops to 61 bps when excluding the meetings with a negative Kuttner shock.¹² To conclude, the evidence suggests that the pattern is strongly related to the Fed’s actions and communications.

4. Potential Explanations

Before discussing theories that could explain the observed pattern, it is worthwile to reflect on the forces behind the secular decline in interest rates. The interest rate decline started in the 1980s when inflation was at double-digit figures. Subsequently, long-term inflation expectations declined substantially until the mid- to late 1990s (e.g., Sargent 1999; Kozicki and Tinsley 2001; Cieslak and Povala 2015; Bauer and Rudebusch 2020). The fall in interest rates over the last two to three decades is then seen mostly as a real phenomenon. For example, proxies for long-term real interest rates showed a marked decline of more than 4% since 1990 (e.g., see Internet Appendix Figure E.1).

Why is the Fed not seen as the main culprit for the decline in real interest rates? According to the mainstream view, economic forces outside the Fed’s control have driven down real interest rates requiring the Fed to adjust monetary policy to these forces. In the words of Larry Summers, the Fed is a “follower rather than a leader with respect to real interest rates.”¹³ Thus, common explanations for the decline in real interest rates point to other economic forces such as a slowdown in productivity (Gordon 2017), a lack of capital investment opportunities or so-called “secular stagnation” (Summers 2014), a rise in the savings of emerging economies (Bernanke 2005), a fall in the price of capital due to technological change (Eichengreen 2015), changes in demographics (Gagnon, Johannsen, and López-Salido 2021; Carvalho, Ferrero, and Nechio 2016), an increase in the liquidity and safety premium of Treasuries (Del Negro et al. 2017), or a decrease in sovereign default risk (Miller, Paron, and Wachter 2021). In addition, estimates of the natural rate or the so-called r*, that is, the (unobserved) rate at which monetary policy is neither contractionary nor expansionary, have declined (Laubach and Williams 2003). At the same time, it is challenging to draw definite conclusions about trends in the natural rate of interest since r* estimates are quite imprecise.

To summarize, while the Fed might control inflation in the long run, it is unlikely to affect the economic forces behind the decline in real interest rates. It is, therefore, puzzling that a narrow window around monetary policy meetings accounts for the secular decline in nominal and, particularly, in real interest rates.

The persistent downward drift in long-term real interest around FOMC meetings is also puzzling from a monetary economics perspective. The Fed mainly controls the nominal short rate (the federal funds rate, which is an overnight rate). In standard New Keynesian macro-models (e.g., Clarida, Gali, and Gertler 1999; Galí 2015), changes in the nominal rate transmit to real rates because prices are initially sticky. However, in these models there is no effect on the real rate beyond the horizon over which prices in the economy adjust. In addition, the real economic effects of monetary policy do not seem to be long-lasting (Ramey 2016). Thus, it is surprising that the prices of long-term real bonds exhibit large and systematic movements around FOMC meetings. Viewed from this perspective, the main fact echoes the findings of prior studies documenting that real, long-lived assets respond surprisingly sensitively to monetary policy (Bernanke and Kuttner 2005; Hanson and Stein 2015).

In the remainder of the paper, I discuss mechanisms that could have contributed to the observed pattern. This is a challenging task since classical theories are silent about the empirical evidence. Two things complicate the matter. First, the forces behind the decline in interest rates likely changed over time, and second, the pattern is constructed using a 3-day window.

4.1. Long-Run Fed Guidance

One potential explanation for the observed pattern is the possibility that the Fed provides guidance to the market about the long-run level of interest rates at FOMC meetings. It might do so implicitly by changing the short rate or explicitly by communicating a long-run forecast, as is the case in the dot plot. Following Fed guidance, the market might update its beliefs about the long-run level of nominal interest rates. Over the past decades, this could have meant that the market came to know about the secular decline at FOMC meetings. I refer to this idea as “Long-Run Fed Guidance.”

Long-Run Fed Guidance for the inflation component of nominal interest rates can be rationalized in the following way. If the Fed controls inflation over the long run, then the Fed’s inflation target matters for long-run inflation. Nominal bond yields will then incorporate the Fed’s inflation target and the market will pay close attention to any information released by the Fed about its inflation target. As a recent example, the Fed is quite hesitant to talk about the possibility of raising the inflation target from 2% to 3% despite substantial news coverage of the issue.

Long-Run Fed Guidance for the real rate component of nominal interest rates is harder to rationalize since the Fed has presumably limited influence over the economic forces determining trends in real rates, including the natural rate of interest. However, it is possible that the market perceives the Fed to have better information about the level of the natural rate of interest. The market might perceive the situation as such since the natural rate is by definition a key input into any monetary policy decision. To get a sense of where the natural rate currently is, the Fed constantly monitors the effects of monetary policy. If monetary policy has contractionary effects, then the Fed concludes that the natural rate is below current rates, and vice versa for expansionary monetary policy. Thus, the Fed, by closely monitoring the effects of monetary policy, obtains an implicit estimate for the level of the natural rate. In its constant discussions about whether current monetary policy is expansionary or contractionary, the Fed might provide valuable guidance to markets. This might lead the market to update its beliefs when observing the Fed’s actions and communications.

Several explanations are possible for why the market might change its beliefs in response to FOMC meetings. One possibility is that the Fed has superior information about the long-run level of interest rates. Another is that the market might only perceive this to be the case. Finally, the power of the Fed might exist because the market overestimates how much control the Fed has over interest rates in the long run. After all, a common mantra on Wall Street is “don’t fight the Fed.”

One appealing feature of this explanation is that it is consistent with all the evidence documented in the paper. First, the dot plot evidence described in the previous section is a direct test of Long-Run Fed Guidance. It shows that the market pays close attention to the long-run guidance released by the Fed and that the market updates its beliefs accordingly. Second, Long-Run Fed Guidance can explain why trends in inflation and real interest rates show up around FOMC meetings. Third, short rate decisions implicitly reveal the Fed’s stance on the neutral level of interest rates and, therefore, can explain why long-term rates respond to “monetary policy shocks.” Fourth, Long-Run Fed Guidance might lead to more coordination in financial markets such that informed market participants preferably trade on their own long-run information ahead of FOMC meetings. This could explain why yields start to drift downwards prior to an FOMC meeting.

But the explanation does not come without challenges. First, based on this explanation, one might expect that the pattern depends on the underlying drivers of the secular decline or changes in the Fed’s communication of monetary policy. The robustness of the pattern over the past decades suggests that this is not the case. Second, it is an open question why the bond market would rely so heavily on the Fed’s guidance. While some studies have argued that the Fed possesses an information advantage about the short-run fundamentals of the economy (Romer and Romer 2000; Campbell et al. 2012; Nakamura and Steinsson 2018), a recent study by Bauer and Swanson (2023) challenges this interpretation, known as the “Fed information effect.” Long-run guidance might be different from the Fed information effect since it is not about the short-term trajectory of the economy, but rather unobserved trend variables such as the natural rate of interest or trend inflation. But whether the market truly updates its beliefs in response to the Fed’s guidance is an open question.

4.2. Other channels

Several other channels are possible. Since the drivers behind the secular decline in interest rates have likely changed over time, it is possible that the importance of each channel has also changed over time. The discussion below is not meant to be exhaustive.

Previous work has highlighted the excess sensitivity of long-term yields and forward rates to changes in monetary policy (Hanson and Stein 2015), potentially due to investors reaching for yield or extrapolating the short rate (Hanson, Lucca, and Wright 2021). Similarly, several papers—often focusing on the stock market—stipulate that monetary policy works mostly through the risk premium channel (Drechsler, Savov, and Schnabl 2018; Lagos and Zhang 2020; Jeenas and Lagos 2024; Pflueger and Rinaldi 2022; Kekre and Lenel 2022). Potentially, investors might have changed their perception about the monetary policy rule (Bauer, Pflueger, and Sunderam 2024), about monetary policy regimes (Bianchi, Lettau, and Ludvigson 2021), or about the Fed put (Cieslak and Vissing-Jorgensen 2021) around FOMC meetings, and this might have translated into excess movements of long-term yields.

One advantage of this explanation is that prior studies have already documented evidence for the excess sensitivity of long-term assets (e.g., Giglio and Kelly 2018) and the Fed’s effect on risk premia (e.g., Bernanke and Kuttner 2005; Hanson and Stein 2015; Cieslak, Morse, and Vissing-Jorgensen 2019). However, the explanation leaves some questions unanswered. Why are interest rate changes persistently negative around FOMC meetings, and why do they capture the general decline in interest rates? Why do the Fed’s long-run dots have a substantial impact on long-term yields?

Alternatively, recent studies entertain the possibility that the Fed can influence the natural rate of interest. McKay and Wieland (2021) argue that expansionary monetary policy lowers the future path of the long-run real rate by bringing forward the purchase of durable goods. Relatedly, Rungcharoenkitkul and Winkler (2021) argue that the Fed’s natural rate estimate can become self-fulfilling because the Fed faces a “hall of mirrors“ problem. The explanation is appealing because it naturally leads to the observed pattern. However, it is at odds with almost all theories on the secular decline in interest rates. Thus, it is debatable whether this channel can be quantitatively important over the past decades.

Finally, investors might require compensation for being exposed to news (Savor and Wilson 2013; Ai and Bansal 2018). It is, therefore, possible that some of the yield movements reflect compensation for the exposure to FOMC news. One attractive aspect of this explanation is that if some uncertainty is resolved in the run-ups to the meetings (Laarits 2019; Hu et al. 2022), this could explain why bond returns, and also equity returns (Lucca and Moench 2015), are high prior to FOMC meetings. However, a necessary condition for the existence of a risk premium is that the news released at or before FOMC meetings affects long-term bonds. Thus, this explanation can only coexist with other explanations. While a news-related risk premium could rationalize the high bond returns around FOMC meetings, it does not explain why FOMC meetings captured the trend in interest rates so well over the past decades. Moreover, it is silent as to why long-term yields declined when the Fed cut the short rate or lowered its long-run dot plot forecasts.

I ultimately leave it for future work to examine the outlined channels and their quantitative importance for the pattern. This seems like an interesting avenue for future research since it might shed more light on how monetary policy influences asset prices and the real economy, a fundamental question in asset pricing and monetary economics.

5. Conclusion

Prior studies have documented that real, long-term assets are surprisingly sensitive to monetary policy (Bernanke and Kuttner 2005; Lucca and Moench 2015; Hanson and Stein 2015; Bianchi, Lettau, and Ludvigson 2021). This is puzzling because the Fed only controls the federal funds rate. I add to this evidence by documenting a large decline in long-term nominal and real interest rates around Fed meetings. Moreover, this decline lines up remarkably well with the secular decline in interest rates. One potential explanation for this puzzle is that guidance by the Fed about the long-run level of interest rates leads market participants to update their expectations of the future path of short rates. This might explain why not only long-term bonds, but also other long-duration assets, such as stocks, respond so sensitively to the Fed’s actions and communications.

Code Availability: The replication code is available in the Harvard Dataverse at https://doi.org/10.7910/DVN/OBH3KV.

Acknowledgement

I am deeply grateful to Alexi Savov for his invaluable guidance and advice, and I also want to thank Philipp Schnabl, Anthony Saunders, and Toomas Laarits for their excellent support. I received helpful comments from Viral Acharya, Yakov Amihud, Anna Cieslak, Ignacio Cigliutti, Fernando Cirelli, Itamar Drechsler, Quirin Fleckenstein, Simon Gilchrist, Marco Grotteria, Arpit Gupta, German Gutierrez, Sam Hanson, Franz Hinzen, Carl-Wolfram Horn, Kasper Joergensen, Kose John, Sydney Ludvigson, Odhrain McCarthy, Robert McDonald, Emanuel Moench, Holger Mueller, Andres Sarto, Marcos Sonnervig, Jeremy Stein, Marti Subrahmanyam, Adi Sunderam, Quentin Vandeweyer, Kaushik Vasudevan, Luis Viceira, Olivier Wang, Chao Ying, Motohiro Yogo, Nicholas Zarra, Steven Zheng, and various seminar and conference participants. I gratefully acknowledge NYU Stern’s Center for Global Economy and Business (CGEB) and NYU Stern’s Salomon Center for generously providing funding for data purchases. I thank Emily Lydic for editorial assistance and Margaret Underwood for research assistance. Supplementary material can be found on The Review of Financial Studies website.

Footnotes

Author Notes

Authors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online.

References