Hostname: page-component-586b7cd67f-t8hqh Total loading time: 0 Render date: 2024-11-24T01:24:29.410Z Has data issue: false hasContentIssue false

Vintage article: the effect of monetary policy shocks in the UK: an external instruments approach

Published online by Cambridge University Press:  09 January 2023

Christoph Görtz
Affiliation:
University of Birmingham
Wei Li
Affiliation:
Erasmus University Rotterdam and Danum Advisors
John Tsoukalas
Affiliation:
University of Glasgow
Francesco Zanetti*
Affiliation:
University of Oxford
*
*Corresponding author. Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

This paper uses vector autoregression model analysis to identify monetary policy shocks on UK data using surprise changes in the policy rate as external instruments and imposing block exogeneity restrictions on domestic variables to estimate parameters from the viewpoint of the domestic economy. The results show large and persistent effects of monetary policy shocks on the domestic economy and point to the critical role of exchange rates and term premia. The analysis resolves important empirical puzzles of traditional recursive identification methods.

Type
Articles
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2023. Published by Cambridge University Press

Preamble

While we started working on this project in the early 2010, we wanted to investigate the power of high-frequency identification in overcoming the empirical puzzles related to the impact of monetary policy in small open economies. With this aim, we combined the identification scheme of monetary policy shocks based on external instruments and high-frequency identification with block exogeneity restrictions to account for the effect of foreign forces on the transmission mechanism of monetary policy in small open economies like the UK. We found that the high-frequency identification is powerful to overcome two prominent empirical puzzles related to the increase in prices and a weak and often statistically insignificant appreciation of the exchange rate in response to a contractionary monetary policy shock, known as the price and exchange rate puzzle, respectively. The analysis also supports the relevance of long-term yields for changes in economic activity.

Since our initial investigation, several studies have improved and refined the estimation of the effect of monetary policy shocks in small open economies using a high-frequency identification approach, and the availability of intraday and even shorter frequency data for several small open economies has further advanced knowledge on the subject. Instead of discussing the details of the vast and substantial new work in this literature, we refer the reader to the second chapter in the Handbook of Macroeconomics by Ramey (Reference Ramey and Ramey>2016) who provides a comprehensive overview on the recent literature on the identification of monetary policy shocks, and the more focused, recent studies on high-frequency identification by Altavilla et al. (Reference Altavilla, Brugnolini, Gurkaynak, Motto and Ragusa>2019) and Miranda-Agrippino and Ricco (Reference Miranda-Agrippino and Ricco>2020).

1. Introduction

This paper assesses the transmission of monetary policy shocks in the UK.Footnote 1 The analysis uses a vector autoregression model (VAR) that comprises domestic and foreign variables and identifies monetary policy shocks using external instruments in the form of surprise changes in market rates that occur within a 2-day window around monetary policy announcements. The estimation imposes block exogeneity restrictions to estimate parameters from the viewpoint of the small open economy. This identification strategy detects large and persistent effects of monetary policy shocks and points to a critical role of exchange rates and term premia for the transmission of monetary policy shocks. The analysis resolves empirical puzzles related to the response of inflation and the exchange rate to monetary policy shocks in recursive identification methods.

Our identification strategy is powerful in addressing two problematic assumptions of recursive identification schemes of monetary policy shocks in small open economies. First, recursive identification methods impose timing restrictions on the effect of monetary policy shocks, restricting the domestic interest rate to have a lagged effect on macroeconomic variables. This restriction is unrealistic for VAR models with financial variables and exchange rates since these variables react immediately to monetary policy shocks.Footnote 2 Our identification scheme relaxes this restriction and allows all the variables in the VAR model to respond simultaneously to exogenous changes in monetary policy. The identifying assumption is that surprise changes in market rates within a narrow time window are mainly due to monetary policy shocks and therefore are orthogonal to movements in nonmonetary shocks. Second, traditional estimation of VAR models for small open economies does not assume independence of foreign variables from movements in domestic variables, which is a central assumption in small open economy models.Footnote 3 Our identification strategy instead imposes block exogeneity restrictions that isolate the foreign economy from movements in the domestic variables and therefore estimates the VAR model from the viewpoint of the small open economy.

Our study reveals several important results. First, monetary policy shocks have large and persistent effects on output and yields of long maturities. A monetary policy shock that raises the domestic policy rate by 25 basis points reduces output (proxied by industrial production) by approximately $0.25\%$ within 1 year. We show that the same VAR model estimated with recursive identification methods leads to a smaller decrease in output of $0.17\%$ . Our findings are similar to those in Cloyne and Hurtgen (Reference Cloyne and Hurtgen>2016), who identify the effect of a monetary policy shock for the UK economy using the narrative approach in Romer and Romer (Reference Romer and Romer>2004) on a new real-time forecast data. With our identification scheme, the 10-year government bond yield increases by $32$ basis points in response to the monetary policy shock whereas it falls by $11$ basis points under recursive identification methods. The rise in the 10-year government bond yield in our identification scheme conforms with the theory of the term structure of interest rates that links movements in long yields to a weighted average of present and expected short yields. This finding reveals that rates with long maturities bear information about expected changes in future economic activity that entail an important propagation channel for monetary policy shocks in small open economies. Our analysis therefore provides support for the relevance of long-term yields for movements in economic activity.Footnote 4

Second, we find that a 25-basis point increase in the domestic interest rate significantly decreases the price level by $0.12\%$ within 2 years. This result sharply contrasts with those from recursive identification methods that instead detect a rise in inflation in response to an unexpected tightening in monetary policy, an empirical phenomenon labeled “price puzzle,” originally established by Sims (Reference Sims>1992) on UK data.Footnote 5 We show that the counterfactual increase in inflation detected by recursive identification methods is tightly linked with the behavior of financial markets, as encapsulated by the counterfactual decrease in the 10-year government bonds. The decline in long-yield bonds is consistent with a strong economy and therefore rising inflation. Thus, our analysis shows that empirically plausible movements in the term premia are critical for a realistic response of inflation to changes in monetary policy.

Third, our identification scheme produces large and empirically plausible movements in the exchange rate. The exchange rate appreciates by $1.6\%$ on impact in response to a 25-basis point increase in the domestic interest rate, which then quickly depreciates. The interest rate differential induced by the contractionary domestic monetary policy is offset by expected future depreciation in the domestic currency, a consistent finding with the exchange rate overshooting hypothesis described by Dornbusch (Reference Dornbusch>1976). We show that this finding is in sharp contrast with the results from an identical VAR model estimated with a recursive identification method, which finds that a contractionary domestic monetary policy induces an extremely weak appreciation in the domestic currency that becomes statistically insignificant from the second period after the shock.Footnote 6

Our study is related to the realm of research that focuses on the monetary transmission mechanism in the UK. Ellis et al. (Reference Ellis, Mumtaz and Zabczyk>2014) and Mountford (Reference Mountford>2005) use sign restrictions on structural VAR and Factor-Augmented Vector Autoregressive models to study the transmission mechanism of an array of macroeconomic shocks in the UK and find a limited role for monetary policy shocks. Cloyne and Hurtgen (Reference Cloyne and Hurtgen>2016) use the narrative approach in Romer and Romer (Reference Romer and Romer>2004) to identify monetary policy shocks on UK data and find similar results to ours. These studies focus on real activity and inflation, and instead they abstract from open economy issues and the inclusion of financial variables, which are central to our investigation. Sims (Reference Sims>1992) is the seminal study to show anomalies related to movements in inflation and exchange rates using a recursive identification scheme estimated on UK data. Cushman and Zha (Reference Cushman and Zha>1997) show that the prize puzzle disappears using block exogeneity restrictions in a VAR model identified with Cholesky restriction on Canadian data. Similar to these studies, our work addresses important empirical puzzles, but we also focus on a broader set of variables and take advantage of the information in financial and exchange rate markets using a less restrictive identification scheme. Our analysis is also closely related to recent studies by Cesa-Bianchi et al. (Reference Cesa-Bianchi, Thwaites and Vicondoa>2016) and Miranda-Agrippino and Ricco (Reference Miranda-Agrippino and Ricco>2017, Reference Miranda-Agrippino and Ricco>2020) that identify the effect of monetary policy surprises on the economy using high-frequency data for the UK. We show that an analysis based on a 2-day window around monetary policy announcements produces robust and consistent results compared to monetary policy instruments derived from data with intraday window.

The analysis also relates to the studies that identify monetary policy shocks using external instruments. Our methodology is similar to Mertens and Ravn (Reference Mertens and Ravn>2013, Reference Mertens and Ravn>2014), who use external instruments to investigate the dynamic effects of changes in taxes and tax multipliers. Stock and Watson (Reference Stock and Watson>2012) use external instruments to estimate the effect of six structural shocks, including monetary policy shocks, to investigate alternative explanations of the Great Recession. Gertler and Karadi (Reference Gertler and Karadi>2015) use a similar methodology to estimate the effect of monetary policy shocks and the role of forward guidance. Similar to these studies, our investigation identifies monetary policy shocks using external instruments, but our analysis focuses on the monetary transmission mechanism in a small open economy for a wider set of variables, including financial variables and exchange rates.

The remainder of the paper is organized as follows. Section 2 outlines the VAR model and the identification scheme. Section 3 presents the selection criteria for external instruments and discusses the findings. Section 4 performs robustness analyses. Section 5 concludes.

2. The identification scheme

In this section, we describe the VAR model and the identification scheme based on external instruments.

Consider the structural VAR model:

(1) \begin{equation} Ax_{t}=\sum \limits _{j=1}^{p}C(j)x_{t-j}+\varepsilon _{t}, \end{equation}

where $x_{t}$ is the vector of variables, $A$ , and $C(j)$ with $j\geq 1$ are conformable coefficient matrices and $\varepsilon _{t\text{ }}$ is a vector of white noise structural shocks. Multiplying each side of equation (1) by $A^{-1}$ yields the reduced-form VAR representation

(2) \begin{equation} x_{t}=\sum \limits _{j=1}^{p}B(j)x_{t-j}+v_t,\text{ } \end{equation}

where $B(j)=A^{-1}C(j)$ , $v_t= S\varepsilon _{t}$ , $S=A^{-1}$ , and $p$ is the number of lags. The variance-covariance matrix of the reduced-form representation (2) is $E[v_{t},v_{t}^{\prime }]=E[SS^{\prime }]=\Sigma$ . Partitioning equation (2) between domestic ( $y_{t}$ ) and foreign ( $y^{*}_{t}$ ) variables yields

(3) \begin{equation} \left [ \begin{array}{c} y_{t} \\[5pt] y^{*}_{t}\end{array}\right ] =\sum \limits _{j=1}^{p}\left [ \begin{array}{c@{\quad}c} B_{11}(j) & B_{12}(j) \\[5pt] B_{21}(j) & B_{22}(j)\end{array}\right ] \left [ \begin{array}{c} y_{t-j} \\[5pt] y^{*}_{t-j}\end{array}\right ] +\left [ \begin{array}{c} u_{t} \\[5pt] u^{*}_{t}\end{array}\right ], \end{equation}

where $B_{11}(j)$ , $B_{12}(j)$ , $B_{21}(j)$ , and $B_{22}(j)$ are conformable coefficient matrices associated with domestic and foreign variables, and $u_{t}$ and $u^{*}_{t}$ are the reduced-form shocks for domestic and foreign variables, respectively. The reduced-form shocks are a function of the structural shocks such that

(4) \begin{equation} \left [ \begin{array}{c} u_{t} \\[5pt] u^{*}_{t}\end{array}\right ] = \left [ \begin{array}{c@{\quad}c} S_{11} & S_{12} \\[5pt] S_{21} & S_{22}\end{array}\right ] \left [ \begin{array}{c} \epsilon _{t} \\[5pt] \epsilon ^{*}_{t}\end{array}\right ], \end{equation}

where $\epsilon _{t}$ and $\epsilon ^{*}_{t}$ are vectors of the structural shocks to domestic and foreign variables, respectively.

To estimate the VAR model, we impose block exogeneity restrictions that prevent changes in domestic variables to affect foreign variables while allowing foreign variables to be a source of economic fluctuations for the domestic economy. These restrictions amount to imposing $B_{21}(j)=0$ for $j=1,2,\ldots,p$ in the reduced-form VAR model in equation (3). They enable us to estimate the coefficients in the VAR model from the viewpoint of the small open economy. Our use of block exogeneity restrictions builds on work by Cushman and Zha (Reference Cushman and Zha>1997) that uses similar restrictions to identify monetary policy shocks with standard Cholesky decomposition methods, showing that they improve the identification of monetary policy shocks in open economies. We use least squares estimation of the reduced form (3) to obtain estimates of the coefficients in each matrix $B_{11}(j)$ , $B_{12}(j)$ , and $B_{22}(j)$ that are necessary to simulate the model and derive the reduced-form residuals for the domestic economy ( $u_{t}$ ) and the foreign economy ( $u^{*}_{t}$ ).

To identify the effect of domestic monetary policy shocks, $\epsilon _{r,t}$ , we partially identify the matrix $S$ . Let $s$ denote the column of $S$ associated with the impact of the structural domestic policy shock $\epsilon _{r,t}$ . Partition $s$ as

(5) \begin{equation} s = \left [ \begin{array}{c} s_{1} \\[5pt] s_{2} \end{array}\right ], \end{equation}

where $s_1$ is the on-impact response of domestic variables to one unit of domestic monetary shock, $\epsilon _{r,t}$ and $s_2$ is the on-impact response of foreign variables to $\epsilon _{r,t}$ . Note that the block exogeneity restrictions that prevent changes in domestic variables to affect foreign variables impy that $s_2=0$ .

We estimate $s_1$ using an instrumental variables approach that follows the methodology developed by Stock and Watson (Reference Stock and Watson>2012) and Mertens and Ravn (Reference Mertens and Ravn>2013). A valid instrumental variable $m_{t}$ is correlated with the primitive monetary shock $\epsilon _{r,t}$ (i.e. $\mathbb{E}(m_{t}\epsilon _{r,t})\neq 0$ ) and is orthogonal to any other structural shock in the VAR model (i.e. $\mathbb{E}(m_{t}\epsilon _{k,t})=0$ , for $k\neq r$ ). The contemporary response $s_1$ is estimated by regressing all reduced-form nonmonetary policy shocks in the domestic block on the reduced-form monetary policy shock $u_{r,t}$ using $m_{t}$ as an instrument. Since $m_{t}$ is orthogonal to nonmonetary shocks, the instrumental estimation yields a consistent estimate for $s_1$ .Footnote 7 We normalize $s_1$ to represent a monetary shock of $25$ basis points.

3. Estimation

This section describes the data and discusses the selection of external instruments. It presents and compares the results from our identification method with those from a standard recursive identification scheme.

3.1 Data

We use monthly frequency data for the period January 1994 $-$ December 2007 on a variety of economic and financial variables for the UK (domestic economy) and the US (foreign economy).Footnote 8 To focus on a stable time period, we chose the sample to coincide with the introduction of inflation targeting and a permanent flexible exchange rate in the UK, which excludes the period of the financial crisis.Footnote 9

The vector of domestic variables is defined as $y_{t}=(g_{t},p_{t},\pi _{t},r_{t},{\unicode[Times]{x00A3}}_{t},f_{t},i_{t})^{\prime }$ . It comprises readings for real activity as measured by industrial production ( $g_{t}$ ) and the manufacturing Purchasing Managers Index ( $p_{t}$ ), inflation ( $\pi _{t}$ ) as measured by changes in the Consumer Price Index index, the domestic policy rate ( $r_{t}$ ), and three financial variables that include the exchange rate index for the pound sterling ( ${\unicode[Times]{x00A3}}_{t}$ ), returns to the Financial Times Stock Exchange (FTSE) Index ( $f_{t}$ ) and nominal yields to 10-year government bonds ( $i_{t}$ ).Footnote 10 The vector of foreign variables ( $y^{*}_{t}$ ) comprises the US policy rate ( $r^{*}_{t}$ ).Footnote 11 Appendix A provides description and sources of the data.Footnote 12

3.2 Selection of external instruments

We construct external instruments for monetary policy shocks by measuring surprise changes in market interest rates in a short-time frame around the Monetary Policy Committee (MPC) policy announcements. The interest rates we consider are as follows: the 1-month libor rate (libor-1m), the 3-month libor rate (libor-3m), the 6-month libor rate (libor-6m) and the 12-month libor rate (libor-12m). These particular interest rates are chosen since they broadly capture the market’s expectations on the policy rate. Important for our analysis, we assume surprise changes in the rates in a short-time frame around MPC announcements are orthogonal to nonmonetary shocks included in the VAR model. As discussed in Kuttner (Reference Kuttner>2001), the inflow of news in a short-time frame that contains monetary policy events is dominated by monetary policy announcements, and policy announcements do not disclose private information of the central bank and do not affect risk premia of the short-term interest rates we consider. Faust et al. (Reference Faust, Swanson and Wright>2004) document that surprise changes in the federal funds futures rate due to policy announcement do not contain the Fed’s private information on macroeconomic indicators. Similarly, Piazzesi and Swanson (Reference Piazzesi and Swanson>2008) establish that surprise changes in the federal funds futures rate in a short-time frame around Fed announcements do not encompass risk premia. Thus, changes in the short-term interest rate before and after monetary policy announcements can be treated as changes in the level of monetary policy.Footnote 13 Piazzesi and Cochrane (Reference Piazzesi and Cochrane>2002) measure monetary shocks as changes in the 1-month Eurodollar rate a day before and a day after Fed policy announcements. They show that monetary policy surprises constructed in this way are powerful estimates of monetary shocks.Footnote 14 A number of studies embed the assumption that surprise changes in market interest rates within a short-time frame are good measures for monetary shocks.Footnote 15

We measure surprise changes in the interest rates in a 2-day window within monetary policy MPC policy rate announcements. Since libor rates are set and reported at 11:00 a.m., 1 hour earlier than MPC announcements at 12:00 p.m., we do not expect libor rates to respond to monetary policy on the same day. However, a careful inspection of the data shows that libor rates occasionally change on the day of policy announcements. For example, the 6-month libor rate jumped to 5.44% from 5.38% on 12 January 1994, when the Bank of England decided to keep Bank Rate unchanged, and they reverted to 5.38% 1 day later. In another instance, on 8 February 1994, the Bank of England cut its Bank Rate by 25 basis points. The rate cut was recorded in the 6-month libor rate on the day of announcement but not 1 day later. Dating errors lead to abnormal changes on announcement dates. To deal with the issue of potential dating errors, we follow the approach in Piazzesi and Cochrane (Reference Piazzesi and Cochrane>2002) and define unexpected changes in libor rates due to policy announcements as the move from 1 day before to 1 day after policy announcements.Footnote 16 As a robustness check, we discuss results from instruments measured in a 1-day window in section 4.

Table 1. Regressing VAR residuals from the monetary equation on external instruments for monetary shocks

Notes: This table reports results for regressing residuals from the monetary policy equation in (2) on various external instruments as shown in the columns. Both VAR residuals from the monetary equation and instruments are expressed in percentage points. The sample is monthly, from January 1994 to December 2007.

To address the problem of weak instruments, we statistically test the degree of correlation of the instruments with reduced-form residuals from the monetary policy equation in the VAR model ( $u_{r,t}$ ) using $F$ -statistics from the first-stage regression. Stock et al. (Reference Stock, Wright and Yogo>2002) suggest that a value of the $F$ -statistics lower than $10$ is evidence of weak instruments. Table 1 shows values of $F$ -statistics from regressing VAR residuals of the monetary policy equation ( $u_{r,t}$ ) on external instruments ( $m_{t}$ ). The $F$ -statistics show that all of the four instruments have values higher than 10 and therefore avoid the weak instrument problem. While all four instruments are statistically valid, we need to ensure that rate surprises do not result from timing shocks, which reflect a surprise in the timing of rate changes rather than a surprise in the expected level of future interest rates. As discussed in Bernanke and Kuttner (Reference Bernanke and Kuttner>2005), this distinction is critical for estimating the impact of monetary policy shocks since shocks that alter the level of interest rates in the near future have stronger impact on the economy than shocks that reflect unanticipated timing in the monetary policy shock.

To investigate whether timing shocks substantially affect libor rate surprises, Figure 1 plots surprise changes in libor rates of different maturities. The left panel plots surprise changes in the 6-month libor rate against surprise changes in the 1-month rate. The right panel plots surprise changes in the 3-month libor rate against surprise changes in the 1-month rate. The 45-degree line corresponds to a one-for-one response of the 6-month libor rate to the 1-month libor rate. Observations lying below the 45-degree line in the northeast quadrant and above the southwest quadrant are those associated with a less than one-for-one effect on the surprise changes in the 6-month libor rate; those lying along a steeper line had a greater than one-for-one effect. Regressing surprise changes in the 6-month libor rate on surprise changes in the 1-month libor rate yields an estimated slope of 0.63, which is significantly different from $1$ at the $1\%$ level. Similarly, a regression of surprise changes in the 3-month libor rate on surprise changes in the 1-month libor rate yields an estimated slope of 0.69, which is also significantly different from $1$ at the $1\%$ level. This analysis suggests that the generally stronger responses of the 1-month libor rate to MPC decisions may capture timing shocks, which have smaller impacts on 3- and 6-month libor rates. Thus, we choose the 6-month rate as our preferred instrument since it avoids the weak instrument problem and is less affected by timing shocks than surprises in the 1- and 3-month libor rate. Surprise changes in the 12-month rate are less affected by timing shocks than surprise changes in the 6-month libor rate. But the instrument is marginally weak for our estimation as shown in Table 1. In other words, we have selected our preferred instrument, 6-month libor rate surprises, to minimize the impact of timing monetary shocks as much as possible. Nevertheless results using the 12-month libor rates are similar to the baseline results. In Section 4, we perform robustness analysis to further investigate the selection of external instruments using principal component analysis.

Figure 1. Surprise changes in libor rates. Notes: This graph shows common and idiosyncratic components of surprise changes in the 1-, 3- and 6-month libor rates due to MPC announcement. The left panel plots surprise changes in the 6-month libor rate against surprise changes in the 1-month rate. The right panel plots surprise changes in the 3-month libor rate against surprise changes in the 1-month rate. The common component among the rates is shown by the strong correlations, while idiosyncratic components are manifested by loosely scattered points along the 45-degree line.

3.3 Findings

Figure 2 shows impulse responses to a 25-basis point surprise in domestic monetary policy (thick line) with $68\%$ confidence intervals (thin lines).Footnote 17 To appraise the differences between our identification method and standard recursive methods, Figure 3 shows impulse responses to a 25-basis point surprise in monetary policy using a standard Cholesky identification (thick-dashed line) with $68\%$ confidence bands (thin-dashed lines) alongside our identification (solid line).Footnote 18

The figure shows that a 25-basis point increase in the domestic interest rate generates a significant decline in industrial production of approximately $0.25\%$ within 1 year after the shock. Figure 3 shows that the same VAR model estimated with the recursive identification method generates a decline in output of approximately $0.17\%$ within 1 year after the shock.Footnote 19 Thus, the response of output is almost twice as strong in our identification method. Our finding is consistent with the results in Cloyne and Hurtgen (Reference Cloyne and Hurtgen>2016), who estimate the effect of monetary policy shocks using the narrative approach in Romer and Romer (Reference Romer and Romer>2004) and find that industrial production declines approximately 0.5% (i.e. 2.3% in response to a 1% monetary policy shock).

Figure 2. The baseline result. Notes: The VAR has $6$ endogenous U.K. variables: log(IP), CPI inflation, the policy rate, log(£ERI), monthly returns to the FTSE and yields to 10-year government bonds; and one exogenous variable: the Fed funds target rate. Two lags of endogenous and exogenous variables are included in the VAR model. Identification is achieved with instruments for monetary policy shocks constructed as surprise changes in the 6-month libor rate in a two-day window around MPC announcements. Cumulative responses from inflation and FTSE returns are reported. The $68\%$ confidence interval is generated using a wild bootstrap procedure with 1000 replications. The sample is monthly, from January 1994 to December 2007.

Figure 3. Compare instrumental identification with recursive identification. Notes: Dashed lines show median impulse responses and the $68\%$ confidence intervals from recursive identification ordering financial variables after the monetary variable. Impulse responses from our instrumental identification are shown in solid lines.

Figure 2 shows that a 25-basis point increase in the domestic interest rate reduces the price level by $0.12\%$ within 2 years after the shock. The response is significantly different from zero and it is surrounded by a large degree of uncertainty, as captured by the large confidence interval. Mountford (Reference Mountford>2005) and Ellis et al. (Reference Ellis, Mumtaz and Zabczyk>2014) estimate weaker negative responses of the price level of $0.04\%$ and $0.05\%$ , respectively. The magnitude of our results is consistent with Romer and Romer (Reference Romer and Romer>2004), who show that a 25-basis point, positive shock reduces CPI (PPI) by $0.9\%$ ( $1.5\%$ ) in 48 months. Cloyne and Hurtgen (Reference Cloyne and Hurtgen>2016) find similar results to ours on UK data. Figure 3 shows the response of the price level identified with recursive identification methods. The price level increases in response to the shock. This counterfactual response is a pervasive issue in the identification of monetary policy shocks using recursive identification schemes, as originally detected by Sims (Reference Sims>1992) and labeled “price puzzle.” Our identification scheme based on external instruments resolves this issue and generates a realistic fall in inflation in response to tightening in domestic monetary policy.Footnote 20

Figure 2 shows that an increase of 25 basis points in the domestic interest rate generates an impact appreciation of $1.6\%$ in the exchange rate index (£ERI). The response shows that appreciation in the domestic interest rate relative to the foreign interest rate is followed by a persistent depreciation of the domestic currency. This result is consistent with the exchange rate overshooting hypothesis described by Dornbusch (Reference Dornbusch>1976) that predicts an appreciation in the nominal exchange rate in response to an increase in the interest rate followed by depreciation in the nominal exchange rate.Footnote 21 The response of the exchange rate from our identification scheme is larger than estimates from alternative identification methods. For instance, Faust et al. (Reference Faust, Rogers, Swanson and Wright>2003) use a VAR model identified by a response-matching approach and find that the peak appreciation of US dollars against UK pounds is less than $1\%$ in reaction to a $25$ -basis point shock.Footnote 22 Bjørnland (Reference Bjørnland>2009) identifies a VAR model using a long-run neutrality restriction of monetary policy on the exchange rate and finds that nominal exchange rates appreciate on impact within the range $0.4-1.0\%$ in four countries: Australia, Canada, New Zealand, and Sweden. Figure 3 compares the response of the exchange rate index against those estimated with recursive identification methods. The exchange rate increases by approximately $0.4\%$ in the first period after the shock, and it then becomes unresponsive to the monetary policy shock. This is a common counterfactual finding in VAR models identified with recursive schemes. Caglayan et al. (Reference Caglayan, Kocaaslan and Mouratidis>2013) find a similar puzzling response of exchange rate movements to the monetary policy shock in a VAR model identified with sign restrictions on UK data.

Figure 2 shows that an increase of 25 basis points in the domestic interest rate generates a fall of approximately $1.8\%$ on impact in the FTSE index and the index declines by approximately $2.6\%$ within 9 months after the shock. The impact response is close to the range of values $0.6-1.1\%$ in Bernanke and Kuttner (Reference Bernanke and Kuttner>2005) from an event study on the effect of an increase of 25 basis points in the nominal interest rate on US equities.Footnote 23 Figure 3 shows that the response of the stock market returns estimated with recursive identification methods is similar on impact, but it becomes insensitive to the monetary policy tightening 5 months after the shock.

Figure 2 shows that an increase of 25 basis points in the domestic interest rate increases the yields to 10-year government bonds by $30$ basis points. The impact reaction of the 10-year yields is stronger than the reaction of the short-term policy rate, showing that monetary policy has powerful and persistent effects on rates with long maturities, consistent with the findings in Hanson and Stein (Reference Hanson and Stein>2015). However, the strong increase in the 10-year yields is unlinked with expectations of higher inflation (inflation falls in response to the shock), but it is generated by a rise in the term premium, which is consistent with the monetary policy channel described in models with financial frictions.Footnote 24 Figure 3 shows that the response of the yields to 10-year government bonds estimated with recursive identification methods is negative and therefore inconsistent with the theory of the term structure of the interest rate that links the increase in the short-term policy rate with a rise in rates of long-term maturities.Footnote 25

4. Robustness analysis

To establish whether our baseline results are robust, we undertake a number of robustness checks on the choice of the external instruments and the specification of the VAR model.

Figure 4. Identification using the first principal components of monetary shocks. Notes: This graph shows median impulse responses and their $68\%$ confidence intervals (dashed lines) for a model identified with the first principal component of surprise changes in the 1-, 3-, 6- and 12-month libor rates. Impulse responses from our baseline instrumental identification are shown in solid lines.

We begin by investigating to what extent our instrument selection affects the results. This robustness check is particularly important since results from the $F$ -statistics in Table 1 – based on the threshold of $10$ suggested by Stock et al. (Reference Stock, Wright and Yogo>2002) – show that all variables are potentially valid instruments. As shown in Figure 1, while the instruments are strongly correlated, there are substantial idiosyncratic components in the instruments as manifested by loosely scattered points along the 45-degree lines.Footnote 26 In some cases, libor rates of different maturities move against each other as a response to MPC announcements, as shown by points in the northwest and southeast quadrants. In addition, as discussed in Section 2, one particular example is the significantly less than one-for-one effect between 1- and 3/6-month libor rate surprises. To investigate the issue of idiosyncratic changes in the valid instruments, we follow the approach in Gurkaynak et al. (Reference Gurkaynak, Sack and Swanson>2005a) and use principal component analysis to extract a common “level” factor of monetary shocks derived from market rates of different maturities.Footnote 27 We then use the principal component of all the statistically valid instruments as the external instrument to identify monetary policy shocks. Gurkaynak et al. (Reference Gurkaynak, Sack and Swanson>2005a) point out that the first principal component of surprising changes in market interest rates provides information on the level shock of monetary policy, which shifts market interest rates in the same direction. Since the first principal component is a level factor that moves the 1-to-12 month libor rates in the same way, it is free of idiosyncratic changes of the libor rates, such as timing shocks that mainly affect short-term libor rates and “future policy path shocks,” mainly captured by libor rates with longer maturities.

We find that the first principal component is a strong instrument for our identification, with $F$ -statistics from the first stage fit equal to $29.6$ . It accounts for $86\%$ of variation in the instruments.Footnote 28 Figure 4 shows the impulse responses to a 25-basis point surprise in monetary policy using the first principal component to identify monetary policy shocks (thick-dashed line) with $68\%$ confidence intervals (thin-dashed lines). Responses from our preferred instrument (6-month labor rate) are reported (black line). The figure shows that the impulse responses from the principal component analysis are remarkably close to our baseline results. We interpret this finding as evidence that our preferred instrument accurately captures most of the information on monetary policy changes conveyed by statistically powerful instruments.Footnote 29

To ensure results are robust to the specification of our instruments, we use a narrower time window to estimate monetary policy surprises and consider changes in libor rates in a one-day window within MPC announcements. Since libor rates are set and reported before 11:00 a.m., 1 hour earlier than MPC announcements at 12:00 p.m., we measure monetary shocks as changes in libor rates in the day after announcements. Figure 5 presents the impulse responses identified with 1-day monetary shocks using the 6-month libor rate (dashed line) alongside the benchmark results. Results are generally similar to the baseline identification, although the responses of inflation and FTSE index are somewhat smaller. As explained in Section 3.2, the different responses may be due to measurement errors on the 1-day window around the policy announcements, which is reflected by the wider uncertainty around the estimated responses in the VAR model.

Figure 5. Measuring monetary shocks in a 1-day window. Notes: This graph shows impulse responses identified using monetary shocks measured as changes in the 6-month libor rate in a 1-day window around MPC announcements (dashed lines). Impulse responses from our baseline instrumental identification are shown in solid lines.

Finally, to ensure our identified monetary policy shocks accurately represent the exogenous component of monetary policy and are not accounting for anticipation effects, we follow the methodology in Miranda-Agrippino and Ricco (Reference Miranda-Agrippino and Ricco>2017, Reference Miranda-Agrippino and Ricco>2020) and regress our identified monetary policy shock on forecast variables from Cloyne and Hurtgen (Reference Cloyne and Hurtgen>2016) that capture information available prior to the MPC meetings. The identification of shocks based on the residual of this regression, which can be interpreted as a cleaned proxy for exogenous monetary policy shocks, produces similar impulse response functions to those of our benchmark specification.Footnote 30

5. Conclusion

This paper assessed the transmission of monetary policy shocks in the UK. Our identification scheme uses surprise changes in the policy rate within a narrow time window as external instruments and imposes block exogeneity restrictions on domestic variables to estimate parameters from the viewpoint of the domestic economy. The estimation detects large and persistent effects of monetary policy shocks on real activity and long-term yields, pointing to a critical role of the exchange rate and term premia. The analysis addresses important empirical puzzles related to movements in inflation and exchange rates in response to exogenous changes in monetary policy detected by recursive identification schemes.

The analysis offers some interesting directions for future research. The results show that unexpected changes in monetary policy generate sharp movements in the term premia that are consistent with the theory of the term structure of the interest rate. It would certainly be interesting to develop small open economy models with financial frictions that introduce yields of different maturities to identify important theoretical channels that link movements in term premia to fluctuations in economic activity. This framework will also shed light on what extent and through what mechanisms the inclusion of frictional financial markets enhances the response of exchange rates to monetary policy surprises. These extensions are open to future research.

A. Appendix: Data sources

The table below describes the data definition and sources.

Footnotes

We are grateful to participants at several seminars for extremely useful comments and suggestions. Some of this work was completed while Wei Li was an intern at the Bank of England and Francesco Zanetti was visiting the Bank of Finland and the De Nederlandsche Bank. They would like to thank these central banks and their staff for their hospitality and support.

1 The UK is an archetypal example of a small open economy since exports and imports equal approximately 30% of GDP, and the size of the overall economy is small compared to the world economy. We believe that our results are therefore informative on the transmission mechanism in small open economies.

2 See Rigobon and Sack (Reference Rigobon and Sack>2004), Faust et al. (Reference Faust, Swanson and Wright>2004), and references therein for a discussion of the issue.

3 See Gali and Monacelli (Reference Gali and Monacelli>2005) for a discussion on the exogeneity assumption of the foreign variables in the context of monetary policy in small open economy models.

4 For a discussion on the topic, see Estrella and Hardouvelis (Reference Estrella and Hardouvelis>1991), Estrella (Reference Estrella>2005) and the recent survey by Wheelock and Wohar (Reference Wheelock and Wohar>2009).

5 See Castelnuovo and Surico (Reference Castelnuovo and Surico>2010) for a recent discussion on the price puzzle. Our results show that identification with external instruments resolves counterfactual dynamics in small open economies.

6 Eichenbaum and Evans (Reference Eichenbaum and Evans>1995) first detected the counterfactual dynamics of exchange rate movements. See Engel (Reference Engel>2014) for a review on the issues and and Benigno et al. (Reference Benigno, Benigno and Nistico’>2011) for further discussion.

7 The instrumental regressions estimate contemporary responses of UK variables to a unit of monetary shock, yielding $s_1/\sigma _{r}$ , where $\sigma _{r}$ is the standard deviation of $\epsilon _{r,t}$ .

8 Our identification method requires changes in realized interest rates to achieve the identification of monetary policy shocks. Since the interest rate remained unchanged in the UK in post-2007 period and monetary policy was conducted using alternative instruments, as outlined in Kapetanios et al. (Reference Kapetanios, Mumtaz, Stevens and Theodoridis>2012), our method fails to identify monetary policy shocks in periods of unconventional monetary policy. Miranda-Agrippino and Ricco (Reference Miranda-Agrippino and Ricco>2017, Reference Miranda-Agrippino and Ricco>2020) overcome the issue by assuming that the transmission mechanism of monetary policy is invariant across times of normal and unconventional monetary policy. Studies by Rogers et al. (Reference Rogers, Scotti and Wright>2016), Liu et al. (Reference Liu, Mumtaz, Theodoridis and Zanetti>2019) and Swanson (Reference Swanson>2017) also consider alternative identification strategies of monetary policy shocks during the post-2007 period of unconventional monetary policy. This topic remains an interesting subject for future research.

9 Zanetti (Reference Zanetti>2014) and references therein provide a discussion on the key policy changes and time series properties in the UK economy. Ikeda et al. (Reference Ikeda, Li, Mavroeidis and Zanetti>2020) studied the change in the effectiveness monetary policy during the financial crisis and the effective lower bound on the short-term interest rate.

10 Industrial production, PMI, and the nominal exchange rate index enter are in logs. The other variables are levels of CPI inflation, the policy rate, monthly returns to the FTSE, yields to 10-year government bonds, and the Fed funds target rate. A time trend is included in the model. Results remain similar if we use the real exchange rate.

11 Since the Euro Area accounts for a large share of UK trade, we have estimated the model including 1-month Euribor to the foreign block and results remain substantially unchanged. The benchmark results are generally robust to the inclusion of additional measures of real activity and financial variables in the foreign bloc. However, the response of inflation becomes insignificant when we include US and Euro Area industrial production or inflation. An appendix that details the findings is available on request to the authors.

12 We include the manufacturing Purchasing Managers Index (PMI) as a measure of real activity since it is more timely available to financial markets (it is usually released at the end of the month or in early days of the next month) than industrial production and is closely monitored by financial markets. Koenig (Reference Koenig>2002) shows that the PMI index is closely linked to the direction of monetary policy changes, and therefore, it is a powerful measure for the identification of monetary policy shocks. Similarly, D’Agostino and Schnatz (Reference D’Agostino and Schnatz>2012) document that the PMI index retains strong predictive and forecasting power of economic activity.

13 In our analysis, we make the same assumption. However, further research is needed to establish whether the same result holds in UK data.

14 Piazzesi and Cochrane (Reference Piazzesi and Cochrane>2002) find that monetary shocks constructed as surprise changes in market interest rates are robust to the omitted-variable problem, the time-varying parameter problem and the orthogonalization problem that usually affect policy-rule-based or VAR-based monetary shocks.

15 Other studies follow this approach. Bernanke and Kuttner (Reference Bernanke and Kuttner>2005) study the impact of monetary policy on equity returns using future rate surprises to measure monetary shocks. Faust et al. (Reference Faust, Rogers, Swanson and Wright>2003) use future rate surprises to identify the effects of monetary shocks on exchange rates. Gurkaynak et al. (Reference Gurkaynak, Sack and Swanson>2005b) study responses of long-term interest rates to monetary shocks. Barakchian and Crowe (Reference Barakchian and Crowe>2013) is the first study that estimates the impact of monetary shocks on output using a measure of the Fed’s private information to identify clean monetary shocks measured from surprise changes in market rates. Gorodnichenko and Weber (Reference Gorodnichenko and Weber>2016) study responses of return volatilities to monetary shocks of firms with different levels of price stickiness.

16 In principle, it would be interesting to use intraday changes in market interest rates around the hour of policy announcements. However, intraday data are unavailable for the time period we consider.

17 We estimate the model in equation (3) using the ordinary least squares and imposing the block exogeneity restriction, $B_{21}(j)=0$ . We include two lags in the estimation (i.e. $p=2$ ). The inclusion of additional lags requires the estimation of a large number of parameters that proves difficult with the relatively short number of observations. We compute confidence intervals using a wild bootstrap procedure with $1000$ replications. For the wild bootstrapping, we randomly draw a vector $\iota _{t}$ of the length of the sample from $\{-1,1\}$ , then multiply $\hat{u}_{t}$ and $m_{t}$ element-by-element by $\iota _{t}$ to generate bootstrapped residuals $\hat{u}_{t}^{bs}$ and the corresponding external shocks $m_{t}^{bs}$ . Bootstrapped samples are generated using the estimated coefficient matrices $\hat{A}$ and $\hat{u}_{t}^{bs}$ . Then, we estimate a VAR model for the bootstrapped sample, and finally implement the identification using bootstrapped instruments $m_{t}^{bs}$ .

18 The ordering of the variables in the VAR model is ( $g_t, \pi _t, r_t, {\unicode[Times]{x00A3}}_t, f_t, i_t, p_t, r^{*}_{t}$ ). The results are robust to different orderings of the variables. A companion appendix that details the robustness of the findings is available on request to the authors. See Carlstrom et al. (Reference Carlstrom, Fuerst and Paustian>2009) and Castelnuovo (Reference Castelnuovo>2012) for a discussion on the issues related to the Cholesky decomposition in closed economies.

19 A weak response of output to the monetary policy shock is also detected by Mountford (Reference Mountford>2005), using sign restrictions on a structural VAR model, and Ellis et al. (Reference Ellis, Mumtaz and Zabczyk>2014), using a factor-augmented VAR model on UK data. They find that GDP falls by approximately $0.2\%$ and $0.6\%$ , respectively, in response to a 25-basis point surprise in monetary policy.

20 Numerous studies focus on possible solutions to the price puzzle. See Castelnuovo and Surico (Reference Castelnuovo and Surico>2010), Engel (Reference Engel>2014) and references therein for a discussion.

21 This finding is consistent with the results in Benigno and Thoenissen (Reference Benigno and Thoenissen>2003) from a theoretical model calibrated on UK data.

22 Note that this study uses interval identification. To make a comparison, we compare the median of their estimates to our median estimates.

23 Bredin et al. (Reference Bredin, Hyde, Nitzsche and O’reilly>2007) perform an event study on UK data and show that a 25-basis point monetary shock reduces the stock market returns by $0.2\%$ , quantitatively smaller than Bernanke and Kuttner (Reference Bernanke and Kuttner>2005).

24 See the survey by Brunnermeier et al. (Reference Brunnermeier, Eisenbach and Sannikov>2012) for a recent review on the topic.

25 Several studies performed on closed economies data are consistent with our identified response in yields with long maturities. Cook and Hahn (Reference Cook and Hahn>1989) and Kuttner (Reference Kuttner>2001) show that long-term nominal yields respond to monetary shocks in the same direction. Hanson and Stein (Reference Hanson and Stein>2015) show that distant real forward rates respond strongly to monetary shocks. Gurkaynak et al.(Reference Gurkaynak, Sack and Swanson>2005a) show distant nominal forward rates respond oppositely to monetary shocks.

26 Correlation between 1- and 6-month libor surprises is 0.74, and that between 1- and 3-month libor surprises is 0.84.

27 To describe the principal component analysis, denote the set of $n$ valid instruments by $Z$ (with $n$ columns). We calculate the first principal component in three steps. First, we demean $Z$ by subtracting its row means and denote the demeaned matrix by $\bar{Z}$ . Second, we calculate the eigenvectors of the covariance matrix of $\bar{Z}$ denote the eigenvectors by $V$ , and sort them in descending order of their associated eigenvalues. Third, we derive the first principal component ( $F_1$ ) by computing $F_1= \bar{Z}V_1$ , where $V_1$ is the first column of sorted $V$ .

28 The fraction of variations explained by the $j$ th principal component is calculated as the ratio of the $j$ th eigenvalue (ranked in diminishing order) and the sum of total eigenvalues. The second principal component explains $11\%$ of variations in the four variables.

29 An appendix that compares principal component analysis on the surprise changes in 1- and 3-month libor rates is available on request to the authors. The analysis shows that the shorter the maturity of the instrument, the weaker the estimated impact of monetary shocks on macroeconomic and financial variables. To the limit, impulse responses from the 1-month libor rate are closer to the responses from recursive Cholesky identification methods. We interpret this finding as showing that the information conveyed in the external instrument is critical for the identification of monetary policy shocks and rates with very short maturities embed a wide range of information that is not informative on the changes in the level of monetary policy.

30 An appendix that details the findings is available on request to the authors.

References

Altavilla, C., Brugnolini, L., Gurkaynak, R. S., Motto, R. and Ragusa, G. (2019) Measuring Euro area monetary policy. Journal of Monetary Economics 108, 162179. “Central Bank Communications: From Mystery to Transparency” May 23-24, 2019 Annual Research Conference ofthe National Bank of Ukraine Organized in cooperation with Narodowy Bank Polski10.1016/j.jmoneco.2019.08.016CrossRefGoogle Scholar
Barakchian, S. M. and Crowe, C. (2013) Monetary policy matters: Evidence from new shocks data. Journal of Monetary Economics 60(8), 950966.10.1016/j.jmoneco.2013.09.006CrossRefGoogle Scholar
Benigno, G. and Thoenissen, C. (2003) Equilibrium exchange rates and supply-side performance. Economic Journal 113(486), C103C124.10.1111/1468-0297.00113CrossRefGoogle Scholar
Benigno, G., Benigno, P. and Nistico’, S. (2011) Risk, Monetary Policy and the Exchange Rate, In NBER Macroeconomics Annual 2011, vol. 26, pp. 247309. National Bureau of Economic Research, Inc., NBER Chapters.Google Scholar
Bernanke, B. S. and Kuttner, K. N. (2005) What explains the stock market’s reaction to federal reserve policy? Journal of Finance 60(3), 12211257.CrossRefGoogle Scholar
Bjørnland, H. C. (2009) Monetary policy and exchange rate overshooting: Dornbusch was right after all. Journal of International Economics 79(1), 6477.CrossRefGoogle Scholar
Bredin, D., Hyde, S., Nitzsche, D. and O’reilly, G. (2007) UK stock returns and the impact of domestic monetary policy shocks. Journal of Business Finance & Accounting 34(5-6), 872888.10.1111/j.1468-5957.2006.02001.xCrossRefGoogle Scholar
Brunnermeier, M. K., Eisenbach, T. M. and Sannikov, Y. (2012). Macroeconomics with financial frictions: A survey, NBER Working Papers 18102, National Bureau of Economic Research, Inc.CrossRefGoogle Scholar
Caglayan, M., Kocaaslan, O. K. and Mouratidis, K. (2013) The Role of Financial Depth on the Asymmetric Impact of Monetary Policy, Working Papers 2013007, The University of Sheffield, Department of Economics.Google Scholar
Carlstrom, C. T., Fuerst, T. S. and Paustian, M. (2009) Monetary policy shocks, Choleski identification, and DNK models. Journal of Monetary Economics 56(7), 10141021.10.1016/j.jmoneco.2009.09.009CrossRefGoogle Scholar
Castelnuovo, E. (2012) Testing the structural interpretation of the price puzzle with a cost-channel model. Oxford Bulletin of Economics and Statistics 74(3), 425452.CrossRefGoogle Scholar
Castelnuovo, E. and Surico, P. (2010) Monetary policy, inflation expectations and the price puzzle. Economic Journal 120(549), 12621283.CrossRefGoogle Scholar
Cesa-Bianchi, A., Thwaites, G. and Vicondoa, A. (2016) Monetary policy transmission in an open economy: New data and evidence from the United Kingdom, Bank of England working papers 615, Bank of England.Google Scholar
Cloyne, J. and Hurtgen, P. (2016) The macroeconomic effects of monetary policy: A new measure for the United Kingdom. American Economic Journal: Macroeconomics 8(4), 75102.Google Scholar
Cook, T. and Hahn, T. (1989) The effect of changes in the federal funds rate target on market interest rates in the 1970s. Journal of Monetary Economics 24(3), 331351.10.1016/0304-3932(89)90025-1CrossRefGoogle Scholar
Cushman, D. O. and Zha, T. (1997) Identifying monetary policy in a small open economy under flexible exchange rates. Journal of Monetary Economics 39(3), 433448.CrossRefGoogle Scholar
D’Agostino, A. and Schnatz, B. (2012) Survey-based nowcasting of US growth: A real-time forecast comparison over more than 40 years, Working Paper Series 1455, European Central Bank.Google Scholar
Dornbusch, R. (1976) Expectations and exchange rate dynamics. Journal of Political Economy 84(6), 11611176.10.1086/260506CrossRefGoogle Scholar
Eichenbaum, M. and Evans, C. L. (1995) Some empirical evidence on the effects of shocks to monetary policy on exchange rates. The Quarterly Journal of Economics 110(4), 9751009.10.2307/2946646CrossRefGoogle Scholar
Ellis, C., Mumtaz, H. and Zabczyk, P. (2014) What lies beneath? A time varying FAVAR model for the UK transmission mechanism. Economic Journal 124(576), 668699.CrossRefGoogle Scholar
Engel, C. (2014) Exchange rates and interest parity. Handbook of International Economics 4, 453522.10.1016/B978-0-444-54314-1.00008-2CrossRefGoogle Scholar
Estrella, A. (2005) Why does the yield curve predict output and inflation? Economic Journal 115(505), 722744.CrossRefGoogle Scholar
Estrella, A. and Hardouvelis, G. A. (1991) The term structure as a predictor of real economic activity. Journal of Finance 46(2), 555576.CrossRefGoogle Scholar
Faust, J., Swanson, E. T. and Wright, J. H. (2004) Identifying VARS based on high frequency futures data. Journal of Monetary Economics 51(6), 11071131.CrossRefGoogle Scholar
Faust, J., Rogers, J. H., Swanson, E. and Wright, J. H. (2003) Identifying the effects of monetary policy shocks on exchange rates using high frequency data. Journal of the European Economic Association 1(5), 10311057.CrossRefGoogle Scholar
Gali, J. and Monacelli, T. (2005) Monetary policy and exchange rate volatility in a small open economy. Review of Economic Studies 72(3), 707734.CrossRefGoogle Scholar
Gertler, M. and Karadi, P. (2015) Monetary policy surprises, credit costs, and economic activity. American Economic Journal: Macroeconomics 7(1), 4476.Google Scholar
Gorodnichenko, Y. and Weber, M. (2016) Are sticky prices costly? Evidence from the stock market. American Economic Review 106(1), 165199.CrossRefGoogle Scholar
Gurkaynak, R. S., Sack, B., and Swanson, E. (2005a) Do actions speak louder than words? The response of asset prices to monetary policy actions and statements. International Journal of Central Banking 1.Google Scholar
Gurkaynak, R. S., Sack, B., and Swanson, E. (2005b) The sensitivity of Long-Term interest rates to economic news: Evidence and implications for macroeconomic models. American Economic Review 95(1), 425436.CrossRefGoogle Scholar
Hanson, S. G. and Stein, J. C. (2015) Monetary policy and long-term real rates. Journal of Financial Economics 115(3), 429448.10.1016/j.jfineco.2014.11.001CrossRefGoogle Scholar
Ikeda, D., Li, S., Mavroeidis, S. and Zanetti, F. (2020) Testing the Effectiveness of Unconventional Monetary Policy in Japan and the United States, IMES Discussion Paper Series 20-E-10, Institute for Monetary and Economic Studies, Bank of Japan.Google Scholar
Kapetanios, G., Mumtaz, H., Stevens, I. and Theodoridis, K. (2012) Assessing the economy wide effects of quantitative easing. Economic Journal 122(564), 316347.10.1111/j.1468-0297.2012.02555.xCrossRefGoogle Scholar
Koenig, E. F. (2002) Using the purchasing managers’ index to assess the economy’s strength and the likely direction of monetary policy. Economic and Financial Policy Review.Google Scholar
Kuttner, K. N. (2001) Monetary policy surprises and interest rates: Evidence from the Fed funds futures market. Journal of Monetary Economics 47(3), 523544.10.1016/S0304-3932(01)00055-1CrossRefGoogle Scholar
Liu, P., Mumtaz, H., Theodoridis, K. and Zanetti, F. (2019) Changing macroeconomic dynamics at the zero lower bound. Journal of Business & Economic Statistics 37(3), 391404.10.1080/07350015.2017.1350186CrossRefGoogle Scholar
Mertens, K. and Ravn, M. O. (2013) The dynamic effects of personal and corporate income tax changes in the United States. American Economic Review 103(4), 12121247.CrossRefGoogle Scholar
Mertens, K. and Ravn, M. O, (2014) A reconciliation of SVAR and narrative estimates of tax multipliers. Journal of Monetary Economics 68, S1S19.CrossRefGoogle Scholar
Miranda-Agrippino, S. and Ricco, G. (2017) Unsurprising shocks: information, premia, and the monetary transmission, Bank of England, Mimeo.Google Scholar
Miranda-Agrippino, S. and Ricco, G., (2020) The transmission of monetary policy shocks. American Economic Journal: Macroeconomics, forthcoming.Google Scholar
Mountford, A. (2005) Leaning into the wind: A structural VAR investigation of UK monetary policy. Oxford Bulletin of Economics and Statistics 67(5), 597621.10.1111/j.1468-0084.2005.00133.xCrossRefGoogle Scholar
Piazzesi, M. and Swanson, E. T. (2008) Futures prices as risk-adjusted forecasts of monetary policy. Journal of Monetary Economics 55(4), 677691.CrossRefGoogle Scholar
Piazzesi, M. and Cochrane, J. (2002) The fed and interest rates - A high-frequency identification. American Economic Review 92(2), 9095.Google Scholar
Ramey, V. (2016) Macroeconomic Shocks and Their Propagation. In: Ramey, V. J. B. Taylor and H. Uhlig (eds.), Handbook of Macroeconomics, Handbook of Macroeconomics, vol. 2, pp. 71162, Elsevier.Google Scholar
Rigobon, R. and Sack, B. (2004) The impact of monetary policy on asset prices. Journal of Monetary Economics 51(8), 15531575.CrossRefGoogle Scholar
Rogers, J. H., Scotti, C. and Wright, J. H. (2016) Unconventional Monetary Policy and International Risk Premia, International Finance Discussion Papers 1172, Board of Governors of the Federal Reserve System (U.S.). Google Scholar
Romer, C. D. and Romer, D. H. (2004) A new measure of monetary shocks: Derivation and implications. American Economic Review 94(4), 10551084.CrossRefGoogle Scholar
Sims, C. A. (1992) Interpreting the macroeconomic time series facts: The effects of monetary policy. European Economic Review 36(5), 9751000.CrossRefGoogle Scholar
Stock, J. H., Wright, J. H. and Yogo, M. (2002) A survey of weak instruments and weak identification in generalized method of moments. Journal of Business & Economic Statistics 20(4), 518529.10.1198/073500102288618658CrossRefGoogle Scholar
Stock, J. H. and Watson, M. W. (2012). Disentangling the channels of the 2007-2009 recession, NBER Working Papers 18094. National Bureau of Economic Research, Inc.Google Scholar
Swanson, E. T. (2017). Measuring the effects of federal reserve forward guidance and asset purchases on financial markets, NBER Working Papers 23311. National Bureau of Economic Research, Inc.Google Scholar
Wheelock, D. C. and Wohar, M. E. (2009) Can the term spread predict output growth and recessions? A survey of the literature. Federal Reserve Bank of St. Louis Review, 419440.Google Scholar
Zanetti, F. (2014) Labour market and monetary policy reforms in the UK: A structural interpretation of the implications. In The UK Economy in the Long Expansion and its Aftermath, pp. 81110, Cambridge University Press.Google Scholar
Figure 0

Table 1. Regressing VAR residuals from the monetary equation on external instruments for monetary shocks

Figure 1

Figure 1. Surprise changes in libor rates. Notes: This graph shows common and idiosyncratic components of surprise changes in the 1-, 3- and 6-month libor rates due to MPC announcement. The left panel plots surprise changes in the 6-month libor rate against surprise changes in the 1-month rate. The right panel plots surprise changes in the 3-month libor rate against surprise changes in the 1-month rate. The common component among the rates is shown by the strong correlations, while idiosyncratic components are manifested by loosely scattered points along the 45-degree line.

Figure 2

Figure 2. The baseline result. Notes: The VAR has $6$ endogenous U.K. variables: log(IP), CPI inflation, the policy rate, log(£ERI), monthly returns to the FTSE and yields to 10-year government bonds; and one exogenous variable: the Fed funds target rate. Two lags of endogenous and exogenous variables are included in the VAR model. Identification is achieved with instruments for monetary policy shocks constructed as surprise changes in the 6-month libor rate in a two-day window around MPC announcements. Cumulative responses from inflation and FTSE returns are reported. The $68\%$ confidence interval is generated using a wild bootstrap procedure with 1000 replications. The sample is monthly, from January 1994 to December 2007.

Figure 3

Figure 3. Compare instrumental identification with recursive identification. Notes: Dashed lines show median impulse responses and the $68\%$ confidence intervals from recursive identification ordering financial variables after the monetary variable. Impulse responses from our instrumental identification are shown in solid lines.

Figure 4

Figure 4. Identification using the first principal components of monetary shocks. Notes: This graph shows median impulse responses and their $68\%$ confidence intervals (dashed lines) for a model identified with the first principal component of surprise changes in the 1-, 3-, 6- and 12-month libor rates. Impulse responses from our baseline instrumental identification are shown in solid lines.

Figure 5

Figure 5. Measuring monetary shocks in a 1-day window. Notes: This graph shows impulse responses identified using monetary shocks measured as changes in the 6-month libor rate in a 1-day window around MPC announcements (dashed lines). Impulse responses from our baseline instrumental identification are shown in solid lines.