Does Monetary Policy Spill Over to Other Economies? | RDP 2019-03: Explaining Monetary Spillovers: The Matrix Reloaded

RDP 2019-03: Explaining Monetary Spillovers: The Matrix Reloaded 5. Does Monetary Policy Spill Over to Other Economies?

Jonathan Kearns, Andreas Schrimpf and Fan Dora Xia

April 2019

Download the Paper 1,776KB

We start with Equation (1) to test whether monetary policy shocks originating from the seven advanced economies spill over to the recipient economies under consideration. To measure the interest rate response, we consider rates of different maturities: 1-month and 6-month interest rates, and 2-year and 10-year government bond yields.^[20] We define that a spillover from an originator central bank j to a recipient economy i is significant if the p-value from the F-test of joint significance of ${\hat{β}}_{i j}$ for the three monetary policy shocks coeffcients is less than 10 per cent.

Figures 2 and 3 show the fraction of economies whose interest rates are significantly affected by the policy shocks originating from our seven major advanced economy central banks for short-term and long-term rates, respectively. To simplify the exposition, we show the strength of spillovers to recipient economies grouped by world regions and split into advanced economies and emerging market economies.

Figure 2: Global Spillover Matrix for Short Rates

2011–15 sample

Figure 3: Global Spillover Matrix for Bond Yields

2011–15 sample

5.1 Spillovers to Short-term Interest Rates

A first key finding is that there are hardly any meaningful spillover effects to rates at the short-end of the yield curve. Spillovers to 1- and 6-month interest rates, Figure 2, display quite a bit of noise. While some of the estimated effects are intuitive, for example the ECB has the greatest spillover to emerging market Europe, others are not.^[21] What is clear is that no central bank triggers widespread short-rate spillovers; for 1-month rates not even the Fed generates statistically significant spillovers to more than 20 per cent of economies in any given region. Furthermore, the pattern of measured spillovers to 6-month interest rates bears little resemblance to those to 1-month rates. Overall, for short-term interest rates it is diffcult to distinguish any economically significant spillovers from noise.

5.2 Spillovers to Long-term Interest Rates

The spillover matrices for bond yields show much clearer, and economically meaningful, patterns, as depicted in Figure 3. These are even clearer for 10-year yields than they are for the 2-year yields. There are more significant spillovers from monetary policy shocks originating from the Fed and the ECB. For most regions well over half of economies' 10-year yields have a significant response to Fed monetary policy news. Interestingly, there are significant spillovers from the ECB to three-quarters of advanced economies outside of Europe, but there are no significant spillovers to emerging market economies, including those in Europe.

It is also notable that there are also significant spillovers to the non-European advanced economies from the Bank of Japan and even the other four central banks for which we measure monetary spillovers (Reserve Bank of Australia, Bank of Canada, Bank of England and Swiss National Bank). In contrast, these central banks have little consistent impact on emerging market economies. A potential reason for the smaller spillovers from the ECB, Bank of Japan and Bank of England could be the smaller use of their currencies in trade invoicing, as argued by Zhang (2018).

Given that spillovers are much stronger and more consistent originating from the Fed and ECB, and to longer-term government bond yields, we focus on these in our following deeper analysis of spillover channels. Moreover, the observation that spillovers are more prevalent for long-term rates than short-term rates suggests a relatively minor role of the channel operating via domestic economic conditions, as this channel is likely to present through spillovers of short rates. That said, we explore the validity of this channel in more depth in Section 6 based on observable proxies.

5.3 Panel Regressions

We move from our originator-recipient specific regressions and adopt a panel regression specification to understand the drivers of spillovers to long-term rates.^[22] The panel regression restricts the unconditional spillover strength to be the same across different economies. We first present the baseline regression with only using monetary policy shocks as regressors (top half of Table 3).

Table 3: The Baseline Spillover Regression
	Target	Path	Premium	10-year US Treasury yield	VIX	R² (%)
ECB	−0.12	0.32	0.32			1.5
ECB	(−1.29)	(4.03)	(2.61)			1.5
Fed	0.32	0.28	0.36			2.8
Fed	(2.24)	(3.44)	(4.67)			2.8
ECB	−0.09	0.24	0.16	0.18	0.07	2.4
ECB	(−1.00)	(3.06)	(1.29)	(4.26)	(2.23)	2.4
Fed	0.32	0.27	0.38	0.30	0.10	5.5
Fed	(2.57)	(3.73)	(5.50)	(6.73)	(2.83)	5.5
Notes: The table rows report the results of panel regressions as given by Equation (2), which in turn serve as baseline specification for our analysis; the dependent variable is the daily change in 10-year bond yields in our set of 47 recipient economies; as regressors, besides the monetary shocks for the ECB and the Fed, some specifications also consider the daily change in the US Treasury yield and the VIX as global controls; the reported coefficients correspond to ${\hat{β}}_{j}$ and ${\hat{θ}}_{j}$ in Equation (2); t-stat from panel-corrected standard errors (PCSE) are given in parentheses; cells coloured red (blue) indicate statistically significant positive (negative) coefficients at a 10 per cent level

The estimated coefficients and panel-corrected standard errors corroborate the existence of significant monetary spillover effects from both the Fed and ECB. The coefficients on monetary policy shocks are all significant, with the exception of the target shock from the ECB. We also add two global controls to the regression – the daily change in 10-year US Treasury yields and the VIX.^[23] Both variables are significant for the Fed and ECB regressions. These global factors are intended to capture other drivers independent of monetary policy shocks that would drive co-movements of interest rates globally. Yet, also after controlling for these global factors, most monetary policy shocks remain significant. An exception is the risk premium shock from the ECB which loses its significance once the global controls are added to the regression. This specification including the two global variables serves as our baseline regression for the following analysis on the determinants of spillovers.

These effects are not only statistically, but also economically significant. Our results suggest that a 100 basis point ‘target’ shock from the Fed on (average) translates into around a 30 basis point change in 10-year government bond yields globally. At 38 basis points, Fed-induced bond risk premium shocks have the largest global effects, whereas path shocks still account for a sizable 27 basis point spillover effect. The pass-through is smaller for ECB shocks (also estimated with less statistical confidence), yet ECB shocks still account for an economically sizable 20 basis point global spillover effect on average.

Footnotes

Depending on data availability, for 1-month or 6-month interest rates, we used OIS rates, government bill rates, interbank rates or deposit rates. Please see the Online Appendix for details. [20]

For instance, many short rates in Latin America respond to RBA announcements as they do to Fed announcements. [21]

Given that strong and consistent spillovers only emerge from shocks originating from the Fed and ECB to long-term interest rates in recipient economies, all panel regressions focus on shocks from the Fed and ECB. The data sample spans from 2004 to 2015 for the Fed shocks, and from 2006 to 2015 for the ECB shocks. [22]

The daily change in the 10-year US Treasury yield controls for any spillovers to global yields outside of our event window. For regressions with the shocks originating from the Fed, the daily change is orthogonalised relative to the shocks to avoid collinearity. [23]