RDP 2019-03: Explaining Monetary Spillovers: The Matrix Reloaded 1. Introduction

It is well established that interest rates co-move across countries. The extent of this co-movement, and the underlying drivers, are more uncertain but are important for many reasons. The greater the co-movement of a country's interest rates with foreign rates that is unrelated to domestic conditions, the weaker the control by the central bank over domestic financial conditions, diminishing its ability to achieve its policy objectives. Interest rate co-movement is also an important channel through which financial shocks can propagate internationally. In addition, co-movement may diminish the diversification opportunities available to international investors in fixed income markets.

Concerns regarding co-movement have been particularly prominent surrounding quantitative easing (QE) and its reverse, as ‘quantitative tightening’ gathers momentum. The exceptionally large expansions in major central banks' balance sheets in the wake of the financial crisis and thereafter depressed domestic yield curves. QE policies are also commonly believed to have spilled over to very easy financial conditions and low yields in other countries, which may not have been warranted given domestic economic conditions in those economies. However, a pertinent policy question remains over whether major central banks' eventual balance sheet wind-down will spillover to other countries' yield curves in a symmetric manner as macroeconomic and financial conditions are very different now to when these policies were first implemented.

While many papers have documented some co-movement of interest rates internationally, extant work often struggles to cleanly identify whether the co-movement stems from spillovers in a causal sense or rather from common drivers. In this paper, we improve on the existing literature by using cleanly identified monetary policy shocks from high-frequency interest rate changes to precisely estimate the spillovers from one country's interest rates to others.[1]

We identify three components to a monetary policy shock: (i) a ‘target’ policy rate shock, (ii) a shock to the expected ‘path’ of policy, and (iii) a ‘term premium’ shock. This set-up encompasses the wide range of information contained in central bank announcements, and allows us to use a sample that covers both the period of ‘normal’ interest rate policies prior to the financial crisis and the period of ‘zero’ policy rates that followed in the QE period.

Our study uses a rich set of data in the time-series and cross-sectional dimensions. Using high-frequency data to measure the interest rate change to the originating economy's monetary policy announcement ensures exogeneity and thus enables us to pin down the direction of spillovers in a causal sense. We perform this analysis for monetary policy shocks originating from 7 advanced economies. We look beyond the ‘matrix’ of monetary policy spillovers among these 7 economies, to consider an even larger matrix of responses of 47 advanced and emerging market economies. We test for spillovers for short- and long-term interest rates. This approach provides more power for the analysis in the cross-sectional dimension, to better shed light on the nature and extent of interest rate spillovers.

Another key feature of our work is to thoroughly test through which channels interest rate spillovers occur. We propose three alternative channels: (i) domestic economic conditions (including economic linkages), (ii) FX regime, and (iii) the impact of bond risk premia (and financial factors more broadly).

We use a comprehensive set of financial and economic data for our broad panel of economies, encompassing bilateral and aggregate economic and financial links as well as economy-specific factors. With these data at hand, we explore the economic and financial conditions that lead to stronger (or weaker) interest rate spillovers.

We find that there are strong spillovers originating from Federal Reserve monetary policy announcements, leading to a swift repricing of fixed income markets globally. Notably, however, the Fed is not the sole originator of spillovers. We also present evidence of significant spillovers from ECB policies, albeit to a lesser extent. However, spillovers from other advanced economy central banks, including from the Bank of Japan and the Bank of England, are mild.

The spillovers we document are much more prevalent for long-term interest rates, while short rates do not consistently respond to foreign monetary policy news. This suggests that central banks have been able to retain a significant degree of autonomy in their interest rate policies (consistent, for example, with Obstfeld (2015)), despite the forces of the global financial cycle.[2] One may argue, however, in line with Rey (2013) that it is particularly longer-term rates that determine financial conditions. Our results are thus consistent with the view that the independence of central banks to determine financial conditions is limited by the presence of spillover effects. And, somewhat surprisingly, we find that such spillover effects are larger to advanced economies (that are well-integrated in global capital markets) than they are to emerging markets.

We obtain a clear picture regarding the factors explaining different intensities of spillovers across economies. There is no empirical support for a macroeconomic channel in explaining the strength of spillovers. Neither trading linkages nor general economic openness are related to the sensitivity of interest rates to policy shocks in other currency areas. There is partial support for a channel related to exchange rates. In support of the bond risk premium spillover channel, financial openness unambiguously emerges as the strongest factor in explaining the extent of the sensitivity of an economy's interest rates to monetary policy shocks in major advanced economies. In explaining interest rate sensitivity, ‘financial openness’ is best captured by bilateral portfolio equity flows and the amount of the economy's debt denominated in the currency of the spillover originator economy, although the results are robust to using many alternative measures of financial openness.

The remainder of the paper is structured as follows. In Section 2 we outline the channels through which policy in one country can spill over (in the broad sense of the word) to other countries' interest rates and discuss the related literature. In Section 3 we provide a road map of our methodology for detecting spillovers and testing the different spillover channels. In Section 4 we outline the detailed data we use to first identify spillovers and then to test the channels. We then present our results on global spillovers and their main drivers in Sections 5 and 6, respectively. We then conclude.


Note that throughout this paper, we use the term ‘spillovers’ in a broad sense to encompass changes in an economy's interest rate that are in direct response to those in another economy's interest rate. [1]

Miranda-Agrippino and Rey (2015) suggest US monetary policy is a key driver of the global financial cycle. See, for example, Cerutti, Claessens and Rose (2017) for new evidence and a sceptical view regarding the existence of a global financial cycle, as conditions in the core do not explain a large share of global capital flows. [2]