Elsevier

Journal of Monetary Economics

Volume 124, November 2021, Pages 48-65
Journal of Monetary Economics

Monetary policy surprises and their transmission through term premia and expected interest rates

https://doi.org/10.1016/j.jmoneco.2021.07.009Get rights and content

Highlights

  • Use a term structure model to decompose high frequency yield movements around FOMC meetings into expectations and term-premia.

  • Once small sample bias is addressed in the model, expectations are found to play an important role in the transmission of policy shocks.

  • The model is used to extract three instruments for policy shocks representing action, expected path of interest rates and its uncertainty.

  • Using local projections, we document the transmission mechanism of these shocks.

Abstract

Monetary policy moves the yield curve. What is the economic interpretation of such moves and what are their macroeconomic consequences? Applying an affine term structure model to high-frequency yield curve movements around FOMC announcements, we shed new light on these questions. Estimation is subject to restrictions addressing estimation bias in previous studies. By imposing additional structure, expectations and term premia are decomposed into three components interpreted as monetary policy action, expected path and its uncertainty. In a local projections model, the shocks identified by the three components provide insights into monetary policy transmission in the context of existing theories.

Introduction

A classic question in macroeconomics concerns the transmission of monetary policy surprises into the economy. The interest in this question stems from the notion that empirical impulse-responses can guide the development of theory (eg, Christiano et al., 1999). This research strategy, however, rests on the assumption that one can identify the relevant impulses (shocks) in the data. The traditional approach to this identification problem relies on monthly or quarterly vector auto-regressions (VAR) combining macroeconomic data with a short-term nominal interest rate, taken as a proxy for a policy instrument. Various identification schemes have been proposed within this approach (see Ramey, 2016, for a review). What they have in common, however, is that the identified shocks at best reflect monetary policy surprises relative to the mathematical expectations of the regression model.1 Furthermore, VAR-based identification is limiting once financial data are included. How does one invert the reduced-form VAR residuals to identify monetary policy shocks when, at monthly or quarterly frequency, financial markets react to monetary policy and policy makers partially base their decisions on information contained in asset prices? At the same time, ignoring financial data is inefficient, as asset prices may reveal expectations and uncertainty about future monetary policy and some sectors, for instance the housing market, are sensitive to asset prices (long-term interest rates).2

High-frequency (HF) data can ameliorate the identification problem (Bagliano, Favero, 1999, Cochrane, Piazzesi, 2002, Gürkaynak, Sack, Swanson, 2005a, Kuttner, 2001, are early contributions). The idea is that, up to a measurement error, the announcement of the outcome of a policy meeting is the only (exogenous) event impacting on asset prices in a tight enough window around the announcement. Asset price movements in that window can thus provide instruments for policy shocks.3 The dynamic effects of the shocks identified by the HF instruments can then be studied in a standard empirical macroeconomic model. Gertler and Karadi (2015) carry out such an exercise and arrive at a stark conclusion: monetary policy transmits into the economy almost exclusively through changes in term premia, with expected future interest rates left almost unaffected.4 This finding presents a challenge to quantitative-theoretical models used for monetary policy analysis. In most models, monetary policy transmits through changes in the conditional mean of the nominal pricing kernel, not its variance, the relevant part for movements in term premia (eg, Atkeson and Kehoe, 2009). Furthermore, in practice, communication aimed at managing expectations of future monetary policy is an integral part of modern central banking (eg, Woodford, 2005).

In this paper, we revisit the relevance of expected future interest rates vs. term premia in the monetary transmission mechanism. However, we go beyond this basic decomposition. By imposing additional structure on estimated expectations and term premia, we decompose the HF yield curve movements in terms of components that can be assigned economic interpretation. These structural components are then used to identify policy shocks in local projections and study their dynamic effects. Our focus is on the nominal yield curve in the period 1996–2007, characterized by conventional monetary policy. In more detail, the analysis proceeds as follows

First, we employ an estimated affine term structure model (ATSM) to decompose the HF movements in yields around Federal Open Market Committee (FOMC) announcements into expected future interest rates and term premia.5 Importantly, the ATSM is estimated subject to restrictions (Joslin et al., 2011), leading to more precise estimates of expected interest rates and term premia than those obtained from VARs, the framework used by Gertler and Karadi (2015).6 The estimates from the restricted ATSM show that expected interest rates are as important as term premia in explaining yield curve movements, including those around FOMC announcements. For instance, at the 10-year maturity, the two components have about the same variance.7 Second, we use principal components (PCs) of the estimated HF changes in expectations and term premia around FOMC announcements as basis to construct orthogonal instruments for monetary policy shocks. A particular rotation is applied to a subset of the PCs to obtain components with an economic interpretation: (i) action, taking the form of a change in the current policy rate; (ii) change in the expected path of future policy rates; and (iii) change in uncertainty about future monetary policy.8 Finally, we use the instruments in a local projections (LP) macro model (Jordà, 2005) to trace out the dynamic effects of the policy shocks, identified by the instruments, on macro variables. Most of the estimated responses can be justified through the lenses of existing theories, although we also document some new patterns. The analysis delivers especially tight findings for the housing market, a sector which, through mortgage finance, is closely related to the term structure.9

We view our analysis as the natural next step in the line of research using HF data to identify monetary policy shocks. The first HF studies used a single asset, fed funds rate futures for the current month, to identify a single monetary policy shock—an action—capturing an unexpected change in the current policy rate (eg, Beechey, 2007, Gürkaynak, Sack, Swanson, 2005a, Kuttner, 2001). Recognizing the complexity of monetary policy announcements, the work of Gürkaynak et al. (2005b) extended the single-shock approach to two shocks: action and statement (see also Campbell et al., 2012). In this case, the shocks are identified from HF changes in a spectrum of fed funds rate futures with maturities up to a year. Under the assumption that term premia for such a short horizon are small, the fed funds rate futures reflect expectations of the policy rate for the coming year. In this approach, the statement does not affect the current rate but captures any changes in expectations for the policy rate one year ahead, not inferred from the action itself.10 We extend this approach to information contained in the entire yield curve (up to 10-year maturity). This is possible due to the ATSM, which allows us to extract expectations separately from term premia, while avoiding the problems, in this task, inherent in a VAR. Two orthogonal instruments (action and expected path) are extracted from the expectations part of the yield curve. Unlike action, the expected path component is restricted not to affect the current short rate. The third orthogonal instrument (uncertainty) is obtained from term premia. This instrument affects neither the current short rate nor its expected future path and can be interpreted as any residual uncertainty surrounding future monetary policy not already inferred from the other two components.11 Term premia and uncertainty in our framework are thus closely related. The three instruments have very different loadings on the HF changes in yields: action has a declining pattern across maturities, expected path has a tent-like pattern with a peak at the 2-year horizon, and uncertainty has an increasing pattern. To provide support for the economic interpretation of the components, we compare the first two components to those obtained by previous studies from fed funds rate futures (Gürkaynak et al., 2005b) and the third component to implied and estimated interest rate volatility.12

The interpretation of the three instruments is derived solely from their HF effects on the yield curve. Further structural content of the shocks they identify is based on the responses of macro and financial variables in the LP model. The effects of the shock identified by action are consistent with a standard monetary policy shock in a New-Keynesian model, including its extensions with the financial accelerator (Bernanke et al., 1999) and time-varying term premia (Rudebusch and Swanson, 2012). The shock identified by the expected path component is associated with a strong response of interest rate expectations and produces responses of other variables that are consistent with both the Fed information effect (Nakamura and Steinsson, 2018) and the Fed response to news channel (Bauer and Swanson, 2020).13 Finally, the responses to the uncertainty component are a little less clear-cut to map into existing theories. We propose a hypothesis, based on the LP, that could be explored in future research. In the data, term premia and various measures of monetary policy uncertainty increase in response to the shock. The effect on output, however, is mixed and we ascribe it to a fall in excess bond premium (Gilchrist and Zakrajšek, 2012), a variable capturing tightness in the corporate credit market. Specifically, an increase in the term premium increases the 30-year mortgage rate. New home sales and demand for mortgages decline, thus possibly allowing more credit to flow to the corporate sector. This effect may be counteracting any negative effect of uncertainty on output.

For all three instruments, our analysis uncovers a particularly tight connection between monetary policy and the housing market. Regardless of the shock, an increase in the 10-year bond yield, no matter whether occurring due to expectations or term premia, is associated with a similar increase in the 30-year mortgage rate and a sharp contraction in the housing market (new home sales and house prices).

HF intra-day data have been increasingly used to study various phenomena. Besides the context most directly related to us, the literature can be divided into two mutually non-exclusive categories: yield curve decomposition (including real and inflation components) and identification of shocks. The first category includes, for instance, Beechey (2007), Beechey and Wright (2009), Bauer (2015), Gertler and Karadi (2015), Hanson and Stein (2015), and Hördahl et al. (2015). Daily data are sometimes also used (Abrahams et al., 2016). Some studies employ ATSMs, while others use regressions. The second category includes, eg, Bernanke and Kuttner (2005), Miranda-Agrippino and Ricco (2015), Nakamura and Steinsson (2018), Cieslak and Schrimpf (2019), Jarocinski and Karadi (2020), and Bauer and Swanson (2020).14 In terms of the housing market, a subset of our findings is consistent with those reported by Hamilton (2008), who follows a different methodology.15

The paper proceeds as follows. Section 2 discusses the HF data, Section 3 introduces the ATSM and the necessary notation, Section 4 provides an overview of the estimation method and the restrictions imposed, Section 5 reviews the estimates, applies the model to the HF data, and carries out the LP analysis. Finally, Section 6 concludes. Robustness checks and technical details related to the estimation are included in an Online Appendix.

Section snippets

High-frequency data

In order to study the HF yield curve reactions, we measure yields at various maturities in a narrow window around FOMC announcements. In doing so, we build on the literature studying monetary policy shocks within the HF approach. As noted in the Introduction, this literature focuses on short maturities, whereas we explore the reaction of the entire yield curve. Our HF data source is Refinitiv Tick History, except the 3-month T-bill rate, which has substantial gaps in the database at the

The ATSM framework

The aim of this section is to provide a brief overview of the ATSM and introduce concepts and notation used in the rest of the paper. An underlying assumption behind an ATSM is the fundamental principle of finance, applied to default-free zero-coupon bonds of different maturities. Specifically,Et[Mt+1Rt+1(j)]=1,where the expectation operator is with respect to information in period t, the scalar Mt+1>0 is a kernel that prices all bonds and Rt+1(j) is a one-period gross return on a bond of any

Estimation of the ATSM

This section provides an overview of the estimation method and the restrictions imposed. All technical details are contained in the Online Appendix.

Results

The results are presented in the following steps: (i) we inspect the impact of the restrictions on the estimated models (Section 5.1), (ii) extract and analyse three main components of monetary policy surprises from the HF data (Sections 5.2–5.4), and (iii) use the components as instruments in a local projections model (Section 5.5).

Conclusions

HF changes in the yield curve around FOMC announcements are used to advance our understanding of monetary policy surprises and their effects on the macroeconomy. To this end, we adopt a three-stage procedure. First, we decompose high-frequency movements in the yield curve around FOMC meetings into expectations and term premia. Unlike existing work on the topic, we carry out this decomposition using term structure models (and we also correct for a small sample bias in the estimates of the two

Declaration of Competing Interest

None.

References (64)

  • M.G. Kendall

    A note on bias in estimation of autocorrelation

    Biometrica

    (1954)
  • K. Mertens et al.

    The dynamic effects of personal and corporate income tax changes in the united states

    American Economic Review

    (2013)
  • V.A. Ramey

    Macroeconomic Shocks and Their Propagation

  • P. Tillmann

    Monetary policy uncertainty and the response of the yiled curve to policy shocks

    J Money Credit Bank

    (2020)
  • A. Atkeson et al.

    On the Need for a New Approach to Analyzing Monetary Policy

    NBER Macroeconomics Annual, Volume 23

    (2009)
  • M.D. Bauer

    Inflation expectations and the news

    International Journal of Central Banking

    (2015)
  • M.D. Bauer

    Restrictions on risk prices in dynamic term structure models

    Journal of Business and Economic Statistics

    (2018)
  • M.D. Bauer et al.

    Correcting estimation bias in dynamic term structure models

    Journal of Business and Economic Statistics

    (2012)
  • M.D. Bauer et al.

    The Fed’s Response to Economic News Explains the “Fed Information Effect”

    Working Paper Series

    (2020)
  • M. Beechey

    A Closer Look at the Sensitivity Puzzle: The Sensitivity of Expected Future Short Rates and Term Premia to Macroeconomic News

    Finance and Economics Discussion Series

    (2007)
  • B. Bernanke et al.

    The Financial Accelerator in a Quantitative Business Cycle Framework

  • B.S. Bernanke et al.

    What explains the stock market’s reaction to federal reserve policy?

    Journal of Finance

    (2005)
  • A.S. Blinder et al.

    Central bank communication and monetary policy: asurvey of theory and evidence

    J Econ Lit

    (2008)
  • N. Bloom

    The impact of uncertainty shocks

    Econometrica

    (2009)
  • J.R. Campbell et al.

    Macroeconomic effects of federal reserve forward guidance

    Brookings Pap Econ Act

    (2012)
  • J.H.E. Christensen et al.

    An arbitragefree generalized nelsonsiegel term structure model

    Econom J

    (2009)
  • L.J. Christiano et al.

    Monetary Policy Shocks: What Have We Learned and to What End?

  • A. Cieslak et al.

    Non-monetary news in central bank communication

    J Int Econ

    (2019)
  • J.H. Cochrane et al.

    Decomposing the Yield Curve

    Mimeo

    (2008)
  • T. Cogley et al.

    Evolving Post-World War II U.S. Inflation Dynamics

    NBER Macroeconomics Annual 2001, Volume 16

    (2002)
  • F.X. Diebold et al.

    Modelling bond yields in finance and macroeconomics

    American Economic Review: Papers and Proceedings

    (2005)
  • G.R. Duffee

    Information in (and not in) the term structure

    Review of Financial Studies

    (2011)
  • Cited by (14)

    • What does anticipated monetary policy do?

      2023, Journal of Monetary Economics
    • The Fed's dual shocks and the housing market

      2022, Economics Letters
      Citation Excerpt :

      These insights are also supported in Fig. 3 with a six-month lag. Recent work by Kaminska et al. (2021) highlights the tight relationship between monetary policy and the housing market. Their detailed yield-curve-based analysis depicts circumstances where monetary contraction improves stock returns and economic activity.

    • Interest rates and foreign spillovers

      2022, European Economic Review
    • Nominal U.S. Treasuries Embed Liquidity Premiums, Too

      2023, Journal of Financial and Quantitative Analysis
    View all citing articles on Scopus

    We thank Ambrogio Cesa-Bianchi, Refet Gürkaynak, Silvia Miranda-Agrippino, Juan Rubio-Ramírez, Eric Swanson, Stan Zin, an anonymous referee, and seminar and conference participants at the Bank of England, the University of Surrey and the 3rd RCEA Macro-Money-Finance conference for valuable comments and suggestions. The views expressed are those of the authors and not necessarily of the Bank of England or its Monetary Policy Committee.

    View full text