Has globalization changed the international transmission of U.S. monetary policy?

Boeck, Maximilian; Mori, Lorenzo Working Paper Has globalization changed the international transmission of U.S. monetary policy? Working Paper, No. 15/2023 Provided in Cooperation with: Norges Bank, Oslo Suggested Citation: Boeck, Maximilian; Mori, Lorenzo (2024) : Has globalization changed the international transmission of U.S. monetary policy?, Working Paper, No. 15/2023, ISBN 978-82-8379-305-5, Norges Bank, Oslo, https://hdl.handle.net/11250/3166793 This Version is available at: https://hdl.handle.net/10419/310389 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Working Paper Has Globalization Changed the International Transmission of U.S. Monetary Policy? Norges Bank Research Authors : Maximilian Boeck Lorenzo Mor i K eywords: Monetary policy, international Spillovers, TVP-VARs. 15 | 2023 Norges Bank Working Paper Working papers fra Norges Bank, fra 1992/1 til 2009/2 kan bestilles på e-post: [email protected] Fra 1999 og senere er publikasjonene tilgjengelige på www.norges-bank.no Working papers inneholder forskningsarbeider og utredninger som vanligvis ikke har fått sin endelige form. Hensikten er blant annet at forfatteren kan motta kommentarer fra kolleger og andre interesserte. Synspunkter og konklusjoner i arbeidene står for forfatternes regning. Working papers from Norges Bank, from 1992/1 to 2009/2 can be ordered by e-mail: [email protected] Working papers from 1999 onwards are available on www.norges-bank.no Norges Bank’s working papers present research projects and reports (not usually in their final form) and are intended inter alia to enable the author to benefit from the comments of colleagues and other interested parties. Views and conclusions expressed in working papers are the responsibility of the authors alone. ISSN 1502-8143 (online) ISBN 978-82-8379-305-5 (online) Has Globalization Changed the International Transmission of U.S. Monetary Policy?∗ Maximilian Boeck†Lorenzo Mori‡ December 2023 Abstract We estimate a time-varying parameter vector autoregression to examine the evolution of international spillovers of U.S. monetary policy in light of increasing globalization in real and financial markets. We find that the adverse international effects of a U.S. tightening have substantially increased over the past three decades, peaking during the Great Recession. Based on a cross-country analysis and counterfactual simulations, we argue that such amplification can primarily be attributed to the surge in trade integration, while the role of rising financial integration in explaining the time-variation is limited. Keywords: Monetary Policy; International Spillovers; TVP-VARs. JEL Codes: C32, E32, E52, F42, F62. ∗This working paper should not be reported as representing the views of Norges Bank. The views expressed are those of the authors and do not necessarily reflect those of Norges Bank. A previous version of this paper circulated under the title “U.S. Monetary Policy and Globalization: A Time-Varying Perspective”. We thank, without implying endorsement, Knut Are Aastveit, Juan Antolin-Diaz, Drago Bergholt, Alessandro Bucciol, Giovanni Caggiano, Fabio Canova, Jin Cao, Efrem Castelnuovo, Luca De Angelis, Timo Dimitriadis, Alessandro Dovis, Zeno Enders, Francesco Furlanetto, Francesco Fusari, Luca Gambetti, Georgios Georgiadis, Christian Glocker, Eleonora Granziera, Pia Heckl, Marie Hoerova, Florian Huber, Ragnar Juelsrud, Thomas A. Lubik, Massimiliano Marcellino, Ralf R. Meisenzahl, Gernot Müller, Pascal Paul, Gert Peersman, Giovanni Pellegrino, Giovanni Ricco, Thomas Zörner, and participants at various research seminars and conferences for their valuable comments. Global industrial production data for the project were kindly provided by Valerie Grossman, Enrique Martínez-García, and Braden Strackman. Part of this research was conducted during Boeck’s visiting period at the University of Tübingen and Mori’s visiting period at Ghent University, whose kind hospitality is gratefully acknowledged. Maximilian Boeck gratefully acknowledges financial support by the European Union - NextGenerationEU, in the framework of the GRINS - Growing Resilient, INclusive and Sustainable project (GRINS PE00000018 – CUP B43C22000760006). The views and opinions expressed are solely those of the authors and do not necessarily reflect those of the European Union, nor can the European Union be held responsible for them. †Università Bocconi. E-mail: [email protected]. ‡University of Padova and Norges Bank. E-mail: [email protected]. 1 1. Introduction Two well-known stylized facts in international economics – portrayed in Figure 1 – read as follows. First, the global macro-financial system is centered around the U.S. dollar, which represents the dominant currency in trade invoicing and issuance of financial assets (Gopinath, 2015; Rey, 2016). This implies that U.S. hikes affect global outcomes by depressing global trade and financial conditions.1Second, globalization led to a massive increase in global trade and financial integration, which could have substantially modified the global transmission of U.S. shocks. Motivated by the interaction of these two facts, this paper examines the implications of globalization for the international transmission of U.S. monetary policy. We estimate a proxy-SVAR model that allows for time-varying parameters to capture the possible change in the impact of policy disturbances on the global financial and real cycles. We do so because – as shown once again in Figure 1 – globalization has materialized at a changing pace over time, with a change in direction after the Great Recession, a phenomenon known as “slowbalization”. We document that globalization has led to substantial time-variation in the international ramifications of policy shocks: U.S. policy hikes generate stronger recessionary effects over time, with a flattening out of effects after the Great Recession. Whereas there is vast evidence showing that a monetary hike engineered by the Federal Reserve (Fed) generates global recessionary effects (e.g., Dedola et al., 2017; Iacoviello and Navarro, 2019; Degasperi et al., 2020; Georgiadis and Schumann, 2021; Breitenlechner et al., 2022; Kalemli-Özcan and Unsal, 2023; and Bräuning and Sheremirov, 2023), our results on the relevance of the time-variation is novel. To study the effects of exogenous U.S. monetary policy shocks and their international spillovers over time, we use a time-varying parameter vector autoregression (TVP-VAR). We analyze the transmission of U.S. monetary policy mediated by the reaction of U.S. industrial production, U.S. prices, and global economic, trade, and financial indicators. To gauge the effects on global outcomes, we rely on the global (excluding U.S.) production and trade indices and international asset prices. We estimate our time-varying model using Bayesian techniques (Primiceri, 2005; Paul, 2020) and achieve shocks’ identification using the high-frequency instrument of Miranda-Agrippino and Ricco (2021), which directly controls for the information channel of monetary policy. The effects of U.S. monetary policy are transmitted globally through trade and financial channels. The former refers to the traditional Mundell-Fleming effects, in which the current account adjusts because monetary policy affects both aggregate demand (of foreign goods) and the currency. A domestic policy tightening generates recessionary effects at home, which depresses demand for imports (rest of the world - RoW - export). This demand effect should be partly counteracted 1The pivotal role of the dollar as an international currency brought many observers to view the Federal Reserve as a world banker, see for instance Gourinchas and Rey (2007), Gopinath (2015), Rey (2016), Gopinath and Stein (2018), Gourinchas et al. (2019) and Ilzetzki et al. (2019). 2 Figure 1: Motivating Evidence: Global Trade and Financial Integration as Share of GDP. Notes: Global (dollar) trade integration is defined as the sum of global exports and imports (invoiced in dollar) as a percentage of GDP (World Bank national accounts data and Boz et al. (2022)). Global (dollar) financial integration is defined as the sum of global external assets and liabilities (denominated in dollar) as a percentage of GDP (data from the External Wealth of Nations database (Lane and Milesi-Ferretti (2018)) and Bénétrix et al. (2019)). through a revaluation of the currency, which leads to expenditure-switching from home goods to foreign goods. However, since U.S. imports are mainly invoiced in dollar, a revaluation of the currency neither boosts nor dampens the competitiveness of foreign goods; thus, the expenditureswitching channel is of minor importance for the U.S. Still, the U.S. dollar is not only the dominant invoicing currency for U.S. trade but for global trade in general (Goldberg and Tille, 2008; Gopinath, 2015; Gopinath et al., 2020; Boz et al., 2022). Hence, expenditure-switching plays a substantial role in trade between non-U.S. countries that use the dollar as their invoicing currency. The prevalence of dollar invoicing in global trade leads to both inflation spillovers via a widespread surge in import prices as well as negative output spillovers in response to the appreciation of the dollar (Gopinath and Neiman, 2014; Gopinath et al., 2020; Georgiadis and Schumann, 2021; Cook and Patel, 2023).2 Monetary policy also operates globally through a financial channel by affecting asset prices, which either relaxes or binds balance sheet or leverage constraints and thus shapes investors’ risk aversion (see, inter alia, Farhi and Werning, 2014; Bruno and Shin, 2015; Rey, 2016). This, in turn, 2Georgiadis and Schumann (2021) argue that asymmetries in dollar invoicing shares between countries and imports/exports lead to higher output spillovers. In the case of full dominant currency paradigm (all trade is invoiced in dollar), third-country expenditure-switching effects would nullify (in terms of output). Asymmetries then cause expenditure-switching between and within countries. Cook and Patel (2023) argue that global value chains lead to asymmetric adjustments in trade of the dominant-currency economy compared to regional economies. 3 has significant effects on cross-border capital flows, funding costs of agents, and eventually feeds back into asset prices. This causes a strong case for an international risk channel of monetary policy (Miranda-Agrippino and Rey, 2020). Particularly, the existence of the global financial cycle (see Rey, 2015, 2016) acts as a transmitter for the financial channel. Again, the currency plays an important role, because the dollar dominates the global financial system and is an important transmitter of global risk shocks (Georgiadis et al., 2021, 2023). Against the evidence presented, this paper is particularly interested in understanding whether and how the relevance of the trade and financial channel has changed over time due to globalization. As indicated by the motivating evidence in Figure 1, it is plausible that both channels have gained strength during globalization and disentangling them is ultimately an empirical matter. We find strong evidence of growing global spillovers of U.S. monetary policy. The negative response of RoW industrial production has increased significantly over the last decades of rising globalization. The spillover sizes only stabilized with the onset of the Great Recession, which is consistent with the slowdown in trade and financial integration that we see in the data. The observed time-variation is substantial and statistically significant. While a one percentage point (pp) hike in monetary policy leads to about a −0.6% contraction in RoW industrial production in 1993, a samesized shock in 2008 results in a downturn of about −3.2%. When we look at the transmission via the trade and financial channel, we find that monetary policy shocks generate global trade contractions and financial frictions. However, we find a disconnect between the two channels when looking at the time-variation. While the effects on RoW trade are significantly time-varying and track well the pattern we find for RoW industrial production, the ones on global financial conditions are relatively constant. Finally, rising spillovers in real activity are somehow mirrored by a strengthening of domestic effects on U.S. prices and output. Greater international spillovers imply a greater potential for these effects to spillback to the U.S. economy. Based on previous estimates by Breitenlechner et al. (2022), we argue that spillover-spillback loop effects are the likely driver of the rising effects of U.S. policy shocks on the U.S. economy that we find in the data. This calls for incorporating spillback effects in the calibration of U.S. policy decisions today more than in the past. Given the disconnect in the evolution of effects on global trade (highly time-dependent) and financial conditions (relatively constant over time), our estimations suggest that trade integration is the primary factor driving the variability in spillover effects. However, in the presence of a large and time-varying global financial multiplier, even minor shifts in the impact on financial conditions could explain a significant proportion of the variations in the reactions of RoW industrial production. We dig deeper into this aspect by performing a heterogeneity analysis and counterfactual simulations. Regarding the heterogeneity analysis, we generally find stronger spillover effects for emerging markets (EMEs) than advanced (AE) economies. Given that EMEs are more impacted by U.S. disturbances and their share of world economic activity have increased from approximately 20% 4 to 40% over the past three decades (Lane, 2019), part of the observed time-variation is explained mechanically by composition effects. We then construct a (balanced) panel of 22 countries and estimate the response of country-specific industrial production to U.S. monetary policy shocks over time. We analyze the outcomes of this analysis in conjunction with country-specific financial and trade integration data. This enables us to explore the correlation between the growth of spillovers and the increasing economic integration and understand the nature of this relationship. Overall, we find that countries exhibiting greater historical levels of trade and financial integration tend to undergo more pronounced economic downturns in the aftermath of contractionary U.S. policy shocks. Both channels are active on average. However, when it comes to explaining the time-variation in effects, our estimations again suggest that the trade dimension holds greater significance. While countries that have substantially enhanced their trade integration tend to experience exacerbated recessionary impacts as time progresses, no such association emerges in relation to heightened financial integration. We also find evidence that rising financial integration is not responsible for time-variation in real spillovers by running counterfactual simulations. To do so, we first identify a global financial shock following the approach of Gilchrist and Zakrajšek (2012), and then simulate counterfactual scenarios in which U.S. monetary disturbances do not impact global financial conditions (Sims and Zha, 2006; Antolín-Díaz et al., 2021; McKay and Wolf, 2023). By shutting off the financial channel in the transmission of U.S. monetary policy shocks, we obtain two results. On the one hand, the financial channel is important on average, accounting for about 30 −40% of the overall response of RoW industrial production. When it comes to explaining the time-variation in the RoW industrial production response, on the other hand, its role is again found to be limited. The paper contributes to the vast literature that uses linear models to document the negative effects of U.S. monetary policy hikes on the global and real financial cycles.3We are closely related to the contributions of Miranda-Agrippino and Rey (2020), Dedola et al. (2017), Iacoviello and Navarro (2019), and Degasperi et al. (2020). While the first paper extensively investigates the impact of U.S. monetary policy on the global financial cycle, the others focus on examining the heterogeneity of spillovers in real economic activity. Differently, we use a time-varying model and document the increasing international spillovers of U.S. policy shocks.4 3Canova (2005), Maćkowiak (2007), Georgiadis (2016), Feldkircher and Huber (2016), Dedola et al. (2017), Iacoviello and Navarro (2019), Degasperi et al. (2020), Georgiadis and Schumann (2021), and Ca’Zorzi et al. (2023) analyze the foreign responses of real activity to U.S. policy decisions. See Rey (2016), Gerko and Rey (2017), Jordà et al. (2019), Habib and Venditti (2019), Dées and Galesi (2021), and Miranda-Agrippino and Rey (2020) for the transmission to international financial conditions. Obstfeld (2020) provides a comprehensive overview on the global dimension of U.S. monetary policy. 4The time-varying nature of monetary policy shocks on domestic outcomes has been documented in many studies, see for instance Cogley and Sargent (2005), Primiceri (2005), Boivin and Giannoni (2006), Canova and Gambetti (2009), Galí and Gambetti (2015) and Aastveit et al. (2023). 5 Additionally, we contribute to the literature examining the time-varying dimension of international monetary policy. Our paper connects closely to Liu et al. (2022), who estimate a time-varying parameter model to jointly model monetary policy decisions in the U.S., U.K., and Euro area. While their study highlights time-varying network structures in central banks’ decisions, we investigate the global spillovers of U.S. shocks and their underlying drivers. Ilzetzki and Jin (2021) compare the international transmission of U.S. monetary policy shocks prior to and after the 1990s. They find that, before the 1990s, world industrial production declines in response to a U.S. monetary tightening, while during the period 1990-2007, U.S. contractions are expansionary abroad. In contrast to their paper, we model the changes in international transmission channels using a time-varying model (vs. sample-splitting strategy) and focus on the dynamics within the post-1990s period. Furthermore and contrary to their findings, we find evidence in favor of a growing (negative) role of U.S. policy tightenings for global economic activity. The remainder of the paper proceeds as follows. Section 2 discusses the empirical strategy and specification. In Section 3, we present the empirical results, including various extensions, the heterogeneity analysis, the counterfactual exercises, and a battery of sensitivity checks. Finally, Section 4 concludes. 2. Empirical Methodology We empirically examine the international spillovers of U.S. monetary policy using a medium-scale TVP-VAR model that allows for time-variation in the parameters. On the domestic level, we include the federal funds rate, U.S. consumer price index, and U.S. industrial production. This information set allows us to track the domestic transmission channels to U.S. monetary policy shocks. Given our interest in international spillovers, we further include indicators for global economic activity, trade, and the financial cycle. We proxy the global real and trade cycles using rest of the world (RoW, i.e., excluding the U.S.) industrial production and export indices constructed by the Federal Reserve Bank of Dallas (Grossman et al., 2014).5The global financial cycle index is derived from a dynamic factor model constructed from a comprehensive panel of risky asset prices traded worldwide (Miranda-Agrippino and Rey, 2020), summarizing global financial conditions.6 Consistent with the literature, we stationarize the variables before estimating the TVP-VAR. We use the indicators for U.S. prices, U.S. industrial production, RoW industrial production, and RoW exports in log-differences to compute the growth rate and keep the remaining variables in levels. Our monthly dataset covers the time span from 1980M1 to 2017M12, which we split in two parts. 5We proxy global trade with RoW exports since it completely excludes U.S. produced goods. However, we find very similar results when considering RoW imports. 6Aldasoro et al. (2023) show that the global financial cycle index as a price-based global factor is remarkably similar to a quantity-based global factor based on cross-border capital flows. 6 Figure 3 confirms the findings obtained in the linear setting, with a monetary policy tightening being followed by conventional negative demand-type effects in the U.S. economy. A domestic contraction in industrial production goes along with a decline in prices. The magnitudes of these effects are consistent with the linear specification. Some comments, however, are in order. First, to obtain same-size shocks, the impact effect of the shocks on the policy rate itself diminishes over time. This pattern is consistent with a progressive decline of the long-run trend of the U.S. interest rate, which reaches its trough with the zero lower bound period (2009-2015).16 Second, time-dependent patterns arise in the responses of U.S. aggregates. Peak contractions in domestic industrial production aggravate over time (rising from −1.9% in 1993 to −4.4% in 2008) and only stabilize at their lowest level with the onset of the Great Recession (this evolution is in line with the findings of Paul (2020), who consider a different specification). A similar pattern emerges for U.S. prices, which contract stronger over time. The impact in the initial periods of the sample is about −0.6%, which grows in magnitude and reaches −1.1% during the Great Recession. Figure 4 presents the dynamic responses of RoW exports, RoW industrial production, and the global financial cycle. Consistent with an amplified role of the international transmission channel of U.S. monetary policy shocks, we find an increase in the (recessionary) effects of U.S. shocks on global trade, production, and the financial cycle. Throughout the period considered, all variables react negatively (and most of the time statistically significant so) to a monetary policy shock. The peak effect on RoW real activity strongly increases over time, increasing from −0.6% in 1993 to −3.2% in 2008.17 In this regard, the linear VAR seems to capture well the mean effect over time, masking though the time-specific heterogeneity. Similarly, the response of RoW exports (as a measure of trade) is strongly growing in magnitude (from −3.6% to −11.4%).18 The time-variation in the impulse response of the global financial cycle is more limited but yet non-negligible from an economic point of view, with an increase over time from −0.63 to −0.90. Although the magnitude of the global effects has been growing since the beginning of the sample, the pace of this growth notably accelerates in the early 2000s. This intensification of international spillovers coincides precisely with a period characterized by factors such as the trade boom, relaxed financial regulation and supervision of banks (Shin, 2012), and a sharp rise in 16 Figure C1 reports the evolution of the long-run trend along with the on impact response of the U.S. policy rate underlying our specification. Accordingly, the long-run trend is relatively stable in 1990s, but starts decreasing in the 2000s, and becomes even negative during the zero lower bound period. See Appendix C for more details. 17 Given that the shocks are normalized to have the same magnitude, the effects over time are comparable. However, the shape of the policy rate’s responses different. To eliminate any concern, we report in Figure D1 the ratios of the peak responses for RoW vs. U.S. industrial production (see Hofmann and Peersman (2017) for a similar use of ratio impulse response functions). In each 𝑡, the shock hitting the two variables is the same. We obtain an increase in the ratio response over time, which again points towards significant time-variation in global spillovers. 18 In unreported checks available upon request, we find extremely similar results when considering RoW imports. This is consistent with a symmetric contraction of RoW export and import in response to U.S. policy shocks (Gopinath et al., 2020; Degasperi et al., 2020). 13 Figure 4: Time-Varying Impulse Responses Functions of Global Variables. Notes: Responses of domestic variables to a contractionary U.S. monetary policy shock that induces a one percentage point (1 pp, 100 basis points) increase in the federal fund rate in 1993M1. Left column reports the evolution over time of the median impulse-response functions. The right column reports the peak effects over time for each variable with 68% posterior credible sets. Grey dotted horizontal lines: (constant) peak effects in the linear VAR of a same-sized shock (which consists of a 0.76 pp increase in the policy rate in the linear specification). 14 dollar-denominated cross-border positions held by international actors (Rey, 2016). All these factors indicate a heightened role of international linkages in transmitting U.S. monetary policy decisions abroad, which is evident in the data. This downward trend stabilizes only with the Great Recession. Several explanations could account for this stabilization. It may be attributed to the slowdown in trade and financial integration resulting from the financial crisis, the effectiveness of international macro-prudential policies (which reduced banks’ risk-taking propensity and their relevance in intermediation), or the presence of the zero lower bound period, which could potentially impact our estimates. Our empirical model appears to effectively capture this economic narrative. The stronger global effects of U.S. monetary policy shocks can also trigger spillover-spillback loop effects. This means that international recessionary effects spillback to the domestic economy and affect its economic aggregates. Evidence for an active spillback mechanism is provided by Breitenlechner et al. (2022), who find that spillbacks account for nearly half of the overall effect of U.S. monetary policy on domestic real activity using counterfactual simulations. In our estimations, such spillover-spillback loops are the likely explanation of the time-varying effects that we find in the response of U.S. domestic variables. Finally, we investigate whether time-variation is statistically significant by focusing on particular episodes in the sample. We consider the periods 1993M1 and 2008M8. We report in Figure 5 the impulse response functions for the two time periods (left column) and the posterior distribution of the differences (right column). The primary distinction between the two examined periods lies in the effects on RoW industrial production. On the one hand we find no evidence in support of a global downturn in real activity following monetary policy contractions in 1993, which is somehow in line with the results in the linear model. On the other hand, such negative effects are clearly present in 2008. A similar pattern arises for RoW export. In addition, as Figure 5 indicates, differences for RoW exports and RoW industrial production between these time periods are statistically significantly different from zero (right column). Time-variation is a relevant pattern for global real spillovers. In sharp contrast, the evidence for the global financial cycle is weak: While U.S. disturbances generate significant financial frictions in both periods, the difference in the impulse-response functions is not statistically different from zero. 3.3 Wider Propagation Channels To get a better understanding of the time-varying transmission of U.S. monetary policy shocks, we analyze the effects on a range of relevant macroeconomic and financial variables. To compute the impulse responses, we augment the baseline VAR by one variable at a time, which results in specifications with a total of seven variables. Since the state-space would become too large to estimate sensible results, we reduce the number of lags to two. Estimation and prior specification are kept unchanged (we refer to Appendix A for the exact variable definitions and transformations). 15 Figure 5: Differences in Impulse Responses: 1993M1 vs. 2008M8 Notes: Left column: median impulse-response functions and 68% posterior credible sets for the variables considered at 1993M1 vs. 2008M8. Vertical axis: percentage change; horizontal axis: impulse response horizon in months. Right column: difference in impulse responses in such periods (median and 68% posterior credibility intervals are reported). We report the results of these extensions in Figure 6 and Figure 7. On the domestic level, we find that U.S. monetary tightenings are followed i) by abrupt increases in U.S. corporate credit spreads - proxied by the excess bond premium (EBP) of Gilchrist and Zakrajšek (2012); ii) an appreciation of the U.S. dollar effective exchange; and iii) a rise in the U.S. export import ratio. The sign of these effects are as expected. First, Caldara and Herbst (2019) highlight the role of financial conditions in 16 Figure 6: Impulse Responses of Extensions to the Baseline Specification. Notes: Responses of additional variables to a contractionary U.S. monetary policy shock that induces a one percentage point (1 pp, 100 basis points) increase in the federal fund rate in 1993M1. Left column reports the evolution over time of the median impulse-response functions. The right column reports the peak effects over time for each variable with 68% posterior credible sets. transmitting monetary policy shocks. Similar to their findings, an increase in the EBP is associated with a tightening of financial conditions as expected through the (domestic) risk-taking channel of monetary policy.19 Second, the appreciation of the dollar is expected by the uncovered interest rate parity. Third, the positive response of the U.S. export import ratio implies no discernible expenditure-switching channel, as expected in the dominant currency paradigm (Gopinath et al., 19 As mentioned earlier, and in contrast to Caldara and Herbst (2019), our specification remains robust to the inclusion of credit spreads due to the choice of the instrument (Miranda-Agrippino and Ricco, 2021; Miranda-Agrippino and Ricco, 2023). 17 Figure 7: Impulse Responses of Extensions to the Baseline Specification. Notes: Responses of additional variables to a contractionary U.S. monetary policy shock that induces a one percentage point (1 pp, 100 basis points) increase in the federal fund rate in 1993M1. Left column reports the evolution over time of the median impulse-response functions. The right column reports the peak effects over time for each variable with 68% posterior credible sets. 2020). While U.S. exports decline in response to less aggregate demand, U.S. imports do not outweigh this force. The dollar appreciation leads in principle to an increase in the competitiveness of foreign goods and to a boost in imports. However, if most of these imports are already priced in dollar, this counteracting force vanishes (see e.g. Degasperi et al., 2020 for similar results). We uncover that these responses, while being statistically significant throughout the whole sample, exhibit very mild evidence of time-variation. 18 We now turn to the global responses in Figure 7. In line with the dominant currency paradigm, we find that U.S. tightenings result in some inflationary pressures in the RoW (via the revaluation of the dollar and a widespread surge in import prices). Particularly interesting is the decline in inflation spillovers over time (from about +2.1% to +0.1% at peak), which can be attributed to the rise of global value chain participation (Georgiadis et al., 2019).The inflationary pressures in the RoW are tackled with an endogenous increase in interest rates from the major central banks, proxied by the policy rate indicator for RoW economies.20 The gradual decrease in the policy rate response that we find in the data is consistent with i) the increasing effects on RoW production and ii) the diminishing inflation spillovers. Finally, we look at the responses of the global stock market, measured through the RoW MSCI index. The indicator shows an abrupt decline in response to U.S. disturbances, with peak responses being stable over time (about −15%). These findings are in line with the response of the global financial factor of Miranda-Agrippino and Rey (2020) in our benchmark specification: Again, we do not find evidence of time-variation in the effects of U.S. shocks on global financial markets. 3.4 Heterogeneity Analysis We have documented a significant time-variation in the international spillovers of U.S. monetary policy. In this and the following section, we conduct additional exercises to explain the findings and link it to different channels. As highlighted by Kalemli-Özcan (2019), De Leo et al. (2022), and Ca’Zorzi et al. (2023), monetary policy spillovers are quite asymmetrical between advanced and emerging economies (AEs and EMEs). Given that EMEs have increased their share of world economic activity from approximately 20% to 40% over the past three decades, part of the observed time-variation could be explained by mechanical composition effects. Hence, we look into differences of spillovers to AEs and EMEs. To start disentangling the channels at play, we are interested in finding cross-sectional variation. To do so, we estimate country-specific spillovers to industrial production. In the next section, we will finally combine these estimates with country-specific trade and financial data to evaluate the evolution of the transmission channels. To look into the difference between AEs and EMEs, we adapt the baseline specification by replacing the RoW industrial production indicator and RoW export indicator with the respective indicator for AEs or EMEs.21 The results, shown in Figure 8, reveal more pronounced recessionary 20 The comprehensive RoW policy rate indicator published by the Federal Reserve of Dallas displays explosive patterns in the 1980s and 1990s, driven by the merging economies’ data. This makes the estimation infeasible. Hence, we consider the index for RoW advanced economies as a proxy for RoW policy response. 21 Data is again taken from the Database of Global Economic Indicators of the Federal Reserve Bank of Dallas. The industrial production series for EMEs is available from 1987M1. To estimate the model starting from 1980M1, we assume that, from 1980M1 to 1986M12, the growth rate in industrial production of EMEs is equal to the one in the comprehensive RoW series (we always observe the actual EMEs series in the estimation sample). See the exact transformations and list in Appendix A. 19 Figure 8: Comparison of Advanced and Emerging Market Economies. Notes: Responses to a contractionary U.S. monetary policy shock that induces a one percentage point (pp) increase in the federal fund rate in 1993M1. Upper panel reports the evolution over time of the median impulse-response functions. The lower panel reports the peak effects over time for each variable with 68% posterior credible sets. effects in emerging economies - consistent with existing studies. The peak effect on emerging (advanced) countries’ industrial production is −1.6% (−0.5%)in 1993 and −3.2% (−2.4%)in 2008. In those years, the peak response of RoW industrial production, which combines both emerging and advanced economies, is −0.6% and −3.2%, respectively. While in the beginning of the sample period the response of RoW industrial production is strongly tilted towards the one of advanced economies (emerging economies have little relevance in the overall index), the response in 2008 aligns more with the effect in EMEs (of course, estimation uncertainty must be take into account). Furthermore, the dynamic responses reveal that the response of AEs’ industrial production 20 returns back to the zero line relatively quickly, while the contraction in EMEs’ industrial production is far more persistent. Since the composition of RoW industrial production between 1993 and 2008 has strongly changed (in favour of EMEs), composition effects can account for part of the rising spillovers that we observe in the data. However, the time-variation in the responses of both emerging and advanced economies’ industrial production signals that mechanical composition effects cannot fully explain the rising spillovers, which must depend on other factors. In order to make progress on this issue, we are interested in retrieving more cross-sectional heterogeneity with respect to international spillovers. Therefore, we break down the response of RoW industrial production into its country-specific components. We re-estimate our benchmark VAR replacing the aggregate RoW industrial production with national-level indices. This allows us to examine the response at a more granular level. We consider a total of 22 countries in our analysis. These countries were selected based on two criteria: i) they are included in the aggregate RoW production measure of Federal Reserve Bank of Dallas and, ii) monthly data is available from 1980 onwards. Our sample includes: Austria, Belgium, Brazil, Canada, Chile, Colombia, France, Germany, Greece, India, Italy, South Korea, Japan, Malaysia, Mexico, Netherlands, Peru, Portugal, South Africa, Spain, Sweden, UK.22 To save space, we report each country’s median time-varying impulse response functions in Appendix E (Figure E1-Figure E4). We also report the evolution of country-specific mean effects over time in Figure 9. While we have so far considered peak effects as summary statistics of impulse response functions, we now shift to mean effects (i.e., we report the mean effect from ℎ=0, ..., 24 for each 𝑡) to account for the heterogeneity in the shapes of the effects. Given that we observe a positive short-run response of industrial production in a significant share of countries - which would not be considered when focusing on peak effects, mean effects seem to be a more adequate measure of the effects in a given country. (However, the results are very similar when considering peak effects.) The findings in Figure 9 align with our benchmark estimations and reveal a consistent pattern of increasing spillover effects. A U.S. policy tightening generates recessionary effects in most countries considered, especially after the early 2000s. These effects tend to intensify over time until the Great Recession, after which they stabilize. The magnitudes of the economic downturns are in the ballpark of our estimates for the aggregate RoW production. Additionally, two sources of cross-sectional heterogeneity among countries arise. Firstly, the average relevance of recessionary effects varies across countries: Certain countries, such as Canada and Mexico, historically exhibit a greater susceptibility to U.S. shocks, whereas others like the UK are less affected. This dimension can be summarized by taking the average over time of the peak effect in country 𝑖. Secondly, the increase in recessionary effects can be more or less substantial, and this dimension can be 22 We thank the authors of Grossman et al. (2014) for kindly sharing their data with us. Data sources are described here: https://www.dallasfed.org/research/international/dgei#tab2$ 21 Figure 9: Mean Effects of Country-Specific Industrial Production. Notes: Mean effects in the responses of country-specific (non-U.S.) industrial production indices to a contractionary U.S. monetary policy shock that induces a one percentage point (pp) increase in the federal fund rate in 1993M1. summarized by comparing the recessionary effects in country 𝑖at the end of the sample with those at the beginning. Time-variation is pervasive in Japan, Germany, and Spain, while it is more limited in the Netherlands and UK. 3.5 Evaluation of the Channels Our empirical analysis again points towards significant time patterns in global spillovers. But what are the main drivers of such dynamics? We shed light on this aspect by establishing a connection between country-specific effects and country-specific information on trade and financial integration. We define financial integration as the ratio of countries’ (dollar-denominated) external assets and liabilities to GDP, while trade integration is determined by the ratio of total trade to GDP (for a similar choice, see Ca’Zorzi et al., 2023).23 Financial and trade integration are then computed 23 We retrieve the data for financial integration from Bénétrix et al. (2019). Data for trade integration is instead taken from the World Bank (World Development Indicators). Given that country-specific shares of dollar invoiced trade 22 References Aastveit, K A, Furlanetto, F and Loria, F (2023) ‘Has the Fed Responded to House and Stock Prices? A Time-Varying Analysis’. Review of Economics and Statistics Vol. 105(5), pp. 1314– 1324. Abbate, A, Eickmeier, S, Lemke, W and Marcellino, M (2016) ‘The Changing International Transmission of Financial Shocks: Evidence from a Classical Time-Varying FAVAR’. Journal of Money, Credit and Banking Vol. 48(4), pp. 573–601. Aldasoro, I, Avdjiev, S, Borio, C and Disyatat, P (2023) ‘Global and Domestic Financial Cycles: Variations on a Theme’. International Journal of Central Banking Vol. 5, pp. 49–98. Amir-Ahmadi, P, Matthes, C and Wang, M.-C (2023) Understanding Instruments in Macroeconomics – A Study of High-Frequency Identification. Working paper. Antolín-Díaz, J, Petrella, I and Rubio-Ramírez, J F (2021) ‘Structural Scenario Analysis with SVARs’. Journal of Monetary Economics Vol. 117(C), pp. 798–815. Bauer, M and Swanson, E T (2023a) ‘A Reassessment of Monetary Policy Surprises and HighFrequency Identification’. NBER Macroeconomics Annual Vol. 37, pp. 87–155. Bauer, M D and Swanson, E T (2023b) ‘An Alternative Explanation for the “Fed Information Effect”’. American Economic Review Vol. 113(3), pp. 664–700. Boivin, J and Giannoni, M P (2006) ‘Has Monetary Policy Become More Effective?’ The Review of Economics and Statistics Vol. 88(3), pp. 445–462. Boz, E, Casas, C, Georgiadis, G, Gopinath, G, Helena Le Mezo and, A M and Nguyen, T (2022) ‘Patterns of Invoicing Currency in Global Trade: New Evidence’. Journal of International Economics Vol. 136(103604), pp. 1–16. Breitenlechner, M, Georgiadis, G and Schumann, B (2022) ‘What Goes Around Comes Around: How Large Are Spillbacks from US Monetary Policy?’ Journal of Monetary Economics Vol. 131, pp. 45–60. Bruno, V and Shin, H S (2015) ‘Cross-Border Banking and Global Liquidity’. The Review of Economic Studies Vol. 82(2). Bräuning, F and Sheremirov, V (2023) ‘The Transmission Mechanisms of International Business Cycles: Output Spillovers through Trade and Financial Linkages’. IMF Economic Review Vol. 71((1)), 300–325. 29 Bénétrix, A, Gautam, D, Juvenal, L and Schmitz, M (2019) Cross-Border Currency Exposures. New Evidence Based on an Enhanced and Updated Dataset. IMF Working Paper No. 2019/299. Caldara, D and Herbst, E (2019) ‘Monetary Policy, Real Activity, and Credit Spreads: Evidence from Bayesian Proxy SVARs’. American Economic Journal: Macroeconomics Vol. 11(1), pp. 157–92. Canova, F (2005) ‘The Transmission of US Shocks to Latin America’. Journal of Applied Econometrics Vol. 20(2), pp. 229–251. Canova, F and Gambetti, L (2009) ‘Structural Changes in the US Economy: Is there a role for monetary policy?’ Journal of Economic Dynamics and Control Vol. 33(2), pp. 477–490. Carter, C K and Kohn, R (1994) ‘On Gibbs Sampling for State Space Models’. Biometrika Vol. 81(3), pp. 541–553. Ca’Zorzi, M, Dedola, L, Georgiadis, G, Jarocinski, M, Stracca, L and Strasser, G (2023) ‘Making Waves: Monetary Policy and Its Asymmetric Transmission in a Globalized World’. International Journal of Central Banking Vol. 19(2), pp. 95–144. Cogley, T and Sargent, T J (2005) ‘Drifts and Volatilities: Monetary Policies and Outcomes in the Post WWII US’. Review of Economic Dynamics Vol. 8(2), pp. 262–302. Coibion, O (2012) ‘Are the Effects of Monetary Policy Shocks Big or Small?’ American Economic Journal: Macroeconomics Vol. 4((2)), pp. 1–32. Cook, D and Patel, N (2023) ‘Dollar Invoicing, Global Value Chains, and the Business Cycle Dynamics of International Trade’. Journal of International Economics Vol. 145, p. 103839. De Leo, P, Gopinath, G and Kalemli-Özcan, (2022) Monetary Policy Cyclicality in Emerging Economies. NBER Working Paper No. 30458. Dedola, L, Rivolta, G and Stracca, L (2017) ‘If the Fed Sneezes, Who Catches a Cold?’ Journal of International Economics Vol. 108, pp. 23–41. Dées, S and Galesi, A (2021) ‘The Global Financial Cycle and US Monetary Policy in an Interconnected World’. Journal of International Money and Finance Vol. 115, p. 102395. Degasperi, R, Hong, S and Ricco, G (2020) The Global Transmission of US Monetary Policy. CEPR Discussion Paper No. 14533. Degasperi,RandRicco,G (2021) ‘Information and Policy Shocks in Monetary Surprises’. Working Paper. 30 Del Negro, M and Primiceri, G E (2015) ‘Time Varying Structural Vector Autoregressions and Monetary Policy: A Corrigendum’. The Review of Economic Studies Vol. 82(4), pp. 1342–1345. Farhi, E and Werning, I (2014) ‘Dilemma not Trilemma? Capital Controls and Exchange Rates with Volatile Capital Flows’. IMF Economic Review Vol. 62(4), pp. 569–605. Feldkircher, M and Huber, F (2016) ‘The International Transmission of US Shocks — Evidence from Bayesian Global Vector Autoregressions’. European Economic Review Vol. 81, pp. 167– 188. Frühwirth-Schnatter, S (1994) ‘Data Augmentation and Dynamic Linear Models’. Journal of Time Series Analysis Vol. 15(2), pp. 183–202. Galí, J and Gambetti, L (2015) ‘The Effects of Monetary Policy on Stock Market Bubbles: Some Evidence’. American Economic Journal: Macroeconomics Vol. 7(1), pp. 233–57. Georgiadis, G (2016) ‘Determinants of Global Spillovers from US Monetary Policy’. Journal of International Money and Finance Vol. 67(2), pp. 41–61. Georgiadis, G, Gräb, J and Khalil, M (2019) Global Value Chain Participation and Exchange Rate Pass-Through. ECB Working Paper Series No. 2327. Georgiadis, G, Müller, G J and Schumann, B (2021) Global Risk and the Dollar. CEPR Discussion Paper No. 16245. Georgiadis, G, Müller, G J and Schumann, B (2023) Dollar Trinity and the Global Financial Cycle. CEPR Discussion Paper No. 18427. Georgiadis, G and Schumann, B (2021) ‘Dominant-Currency Pricing and the Global Output Spillovers from US Dollar Appreciation’. Journal of International Economics Vol. 133, p. 103537. Gerko, E and Rey, H (2017) ‘Monetary Policy in the Capitals of Capital’. Journal of the European Economic Association Vol. 15(4), pp. 721–745. Gertler, M and Karadi, P (2015) ‘Monetary Policy Surprises, Credit Costs, and Economic Activity’. American Economic Journal: Macroeconomics Vol. 7(1), pp. 44–76. Gilchrist, S and Zakrajšek, E (2012) ‘Credit Spreads and Business Cycle Fluctuations’. American Economic Review Vol. 102(4), pp. 1692–1720. Giraitis, L, Kapetanios, G and Yates, T (2018) ‘Inference on Multivariate Heteroscedastic Time Varying Random Coefficient Models’. Journal of Time Series Analysis Vol. 39(2), pp. 129–149. 31 Goldberg, L S and Tille, C (2008) ‘Vehicle Currency Use in International Trade’. Journal of International Economics Vol. 76(2), pp. 177–192. Gopinath, G (2015) The International Price System. National Bureau of Economic Research Working Paper 21646. Gopinath, G, Boz, E, Casas, C, Díez, F J, Gourinchas, P.-O and Plagborg-Møller, M (2020) ‘Dominant Currency Paradigm’. American Economic Review Vol. 110(3), pp. 677–719. Gopinath,GandNeiman,B(2014) ‘Trade Adjustment and Productivity in Large Crises’. American Economic Review Vol. 104((3)), pp. 793–831. Gopinath, G and Stein, J C (2018) Banking, Trade, and the Making of a Dominant Currency. NBER Working Paper No. 24485. Gourinchas, P.-O and Rey, H (2007) ‘International Financial Adjustment’. Journal of Political Economy Vol. 115(4), pp. 665–703. Gourinchas, P.-O, Rey, H and Sauzet, M (2019) ‘The International Monetary and Financial System’. Annual Review of Economics Vol. 11, pp. 859–893. Grossman, V, Mack, A and Martínez-García, E (2014) ‘Database of Global Economic Indicators (DGEI): A Methodological Note’. Journal of Economic and Social Measurement Vol. 39(3), pp. 163–197. Gürkaynak, R S, Sack, B P and Swanson, E T (2005) ‘Do Actions Speak Louder than Words? The Response of Asset Prices to Monetary Policy Actions and Statements’. International Journal of Central Banking Vol. 1(1), pp. 55–93. Habib, M M and Venditti, F (2019) The Global Capital Flows Cycle: Structural Drivers and Transmission Channels. ECB Working Paper Series No. 2280. Hofmann, B and Peersman, G (2017) Monetary Policy Transmission and Trade-offs in the United States: Old and New. BIS Working Papers 649. Iacoviello, M and Navarro, G (2019) ‘Foreign Effects of Higher US Interest Rates’. Journal of International Money and Finance Vol. 95, pp. 232–250. Ilzetzki, E and Jin, K (2021) ‘The Puzzling Change in the International Transmission of US Macroeconomic Policy Shocks’. Journal of International Economics Vol. 130, p. 103444. Ilzetzki, E, Reinhart, C M and Rogoff, K S (2019) ‘Exchange Arrangements Entering the TwentyFirst Century: Which Anchor Will Hold?’ The Quarterly Journal of Economics Vol. 134(2), pp. 599–646. 32 Jarociński, M and Karadi, P (2020) ‘Deconstructing Monetary Policy Surprises – The Role of Information Shocks’. American Economic Journal: Macroeconomics Vol. 12(2), pp. 1–43. Jordà, Ò, Schularick, M, Taylor, A M and Ward, F (2019) ‘Global Financial Cycles and Risk Premiums’. IMF Economic Review Vol. 67(1), pp. 109–150. Kalemli-Özcan, S (2019) ‘U.S. Monetary Policy and International Risk Spillovers’. Jackson Hole Symposium Proceedings. Kalemli-Özcan, Şebnem and Unsal, F (2023) ‘Global Transmission of FED Hikes: The Role of Policy Credibility and Balance Sheets’. Brookings Papers on Economic Activity (conference draft). Karlsson, S (2013) ‘Forecasting with Bayesian Vector Autoregression’. Handbook of Economic Forecasting Vol. 2, pp. 791–897. Kim, S, Shephard, N and Chib, S (1998) ‘Stochastic Volatility: Likelihood Inference and Comparison with ARCH Models’. The Review of Economic Studies Vol. 65(3), pp. 361–393. Lane, P R (2019) ‘Globalisation and Monetary Policy’. Speech at the University of California, Los Angeles, 30 September. Lane, P R and Milesi-Ferretti, G M (2018) ‘The External Wealth of Nations Revisited: International Financial Integration in the Aftermath of the Global Financial Crisis’. IMF Economic Review Vol. 66(189), pp. 189–222. Liu, L, Matthes, C and Petrova, K (2022) ‘Monetary Policy across Space and Time’. Advances in Econometrics (Essays in Honor of Fabio Canova) Vol. 44 Part B. Maćkowiak, B (2007) ‘External Shocks, U.S. Monetary Policy and Macroeconomic Fluctuations in Emerging Markets’. Journal of Monetary Economics Vol. 54(8), pp. 2512–2520. McKay, A and Wolf, C (2023) ‘What Can Time Series Regressions Tell Us About Policy Counterfactuals?’ Econometrica. Melosi, L (2017) ‘Signalling Effects of Monetary Policy’. The Review of Economic Studies Vol. 84(2), pp. 853–884. Mertens, K and Ravn, M O (2013) ‘The Dynamic Effects of Personal and Corporate Income Tax Changes in the United States’. American Economic Review Vol. 103(4), pp. 1212–47. Miranda-Agrippino, and Ricco, G (2023) ‘Identification with External Instruments in Structural VARs’. Journal of Monetary Economics Vol. 135, pp. 1–19. 33 Miranda-Agrippino, S and Rey, H (2020) ‘US Monetary Policy and the Global Financial Cycle’. The Review of Economic Studies Vol. 87(6), pp. 2754–2776. Miranda-Agrippino, S and Ricco, G (2021) ‘The Transmission of Monetary Policy Shocks’. American Economic Journal: Macroeconomics Vol. 13(3), pp. 74–107. Montiel Olea, J L, Stock, J H, Watson, M W et al. (2021) ‘Inference in Structural Vector Autoregressions Identified with an External Instrument’. Journal of Econometrics Vol. 225(1), pp. 74–87. Mumtaz, H and Petrova, K (2023) ‘Algorithms for Proxy SVAR models with time-varying parameters’. Journal of Money, Credit and Banking Vol. 55((2-3)), pp. 635–654. Nakamura, E and Steinsson, J (2018) ‘Identification in Macroeconomics’. Journal of Economic Perspectives Vol. 32(3), pp. 59–86. Noh, E (2019) ‘Impulse-response analysis with proxy variables’. Working Paper. Obstfeld, M (2020) ‘Global Dimensions of U.S. Monetary Policy’. International Journal of Central Banking, pp. 73–132. Omori, Y, Chib, S, Shephard, N and Nakajima, J (2007) ‘Stochastic Volatility with Leverage: Fast and Efficient Likelihood Inference’. Journal of Econometrics Vol. 140(2), pp. 425–449. Paul, P (2020) ‘The Time-Varying Effect of Monetary Policy on Asset Prices’. Review of Economics and Statistics Vol. 102(4), pp. 690–704. Primiceri, G E (2005) ‘Time Varying Structural Vector Autoregressions and Monetary Policy’. The Review of Economic Studies Vol. 72(3), pp. 821–852. Rey, H (2015) Dilemma not Trilemma: The Global Financial Cycle and Monetary Policy Independence. NBER Working Paper No. 2162. Rey, H (2016) ‘International Channels of Transmission of Monetary Policy and the Mundellian Trilemma’. IMF Economic Review Vol. 64(1), pp. 6–35. Shin, H S (2012) ‘Global Banking Glut and Loan Risk Premium’. IMF Economic Review Vol. 60(2), pp. 155–192. Sims, C A and Zha, T (2006) ‘Does Monetary Policy Generate Recessions?’ Macroeconomic Dynamics Vol. 10(2), pp. 231–272. Stock, J H and Watson, M W (2012) Disentangling the Channels of the 2007-2009 Recession. NBER Working Paper No. 18094. 34 Wu, J C and Xia, F D (2016) ‘Measuring the Macroeconomic Impact of Monetary Policy at the Zero Lower Bound’. Journal of Money, Credit and Banking Vol. 48(2-3), pp. 253–291. 35 A. Data All series were gathered from the sources listed below, particularly the FRED and several papers. In Table A1, we define the exact transformations of the variables used in the estimation of the TVP-VAR. Note that we use month-on-month growth rates. In specific cases (indicated by stars ∗in the Table A1), we are missing data points in the presample period. In order to estimate the TVP-VAR on the same estimation sample (for which all data points are available), we pursue the following strategy. If we have information on a similar series for the same time period, we assume the same growth scenario. Specifically, we observe industrial production of EMEs only from 1987M1 to 2017M12. Hence, we assume that industrial production in EMEs grows as the corresponding series in the RoW during the time frame 1980M1 to 1986M12. This allows us to extend the series backwards. 36 Table A1: Variable Definitions. Variable Transformation in TVP-VAR Details 𝑓 𝑓 𝑟𝑡FFR𝑡FFR𝑡is the federal funds rate (Fred), replaced with the shadow rate of Wu and Xia (2016) during the ZLB period (2008M12-2015M12). 𝑐𝑝𝑖𝑈𝑆 𝑡100 ×hln CPI𝑈𝑆 𝑡−ln 𝐶𝑃𝐼𝑈𝑆 𝑡−1i CPI𝑈𝑆 𝑡is U.S. consumer price index (Fred). 𝑖𝑝𝑈𝑆 𝑡100 ×hln IP𝑈𝑆 𝑡−ln IP𝑈𝑆 𝑡−1i IP𝑈𝑆 𝑡is U. S. industrial production (Fred). 𝑥𝑖𝑚𝑈𝑆 𝑡XIM𝑈𝑆 𝑡XIM𝑈𝑆 𝑡is the ratio between U.S. export and import in goods and services (Fred), interpolated from quarterly to monthly using a shapepreserving piecewise cubic interpolation. 𝑒𝑏𝑝𝑈𝑆 𝑡EBP𝑈𝑆 𝑡EBP𝑈𝑆 𝑡is the U.S. excess bond premium by Gilchrist and Zakrajšek (2012). 𝑔 𝑓 𝑐𝑡GFC𝑡GFC𝑡is the global financial cycle indicator by Miranda-Agrippino and Rey (2020). 𝑟𝑒𝑒𝑟𝑡100 ×ln REER𝑈𝑆 𝑡REER𝑈𝑆 𝑡is the U.S. real effective exchange rate from the BIS. 𝑐𝑝𝑖𝑅𝑜𝑊 𝑡100 ×hln CPI𝑅𝑜𝑊 𝑡−ln CPI𝑅𝑜𝑊 𝑡−1i CPI𝑅𝑜𝑊 𝑡is RoW consumer prices (excluding the U.S.) Grossman et al. (2014). 𝑖𝑝𝑅𝑜𝑊 𝑡100 ×hln IP𝑅𝑜𝑊 𝑡−ln IP𝑅𝑜𝑊 𝑡−1i IP𝑅𝑜𝑊 𝑡is the RoW industrial production (excluding the U.S.) by Grossman et al. (2014). 𝑚𝑠𝑐𝑖𝑅𝑜𝑊 𝑡100 ×hln MSCI𝑅𝑜𝑊 𝑡−ln MSCI𝑅𝑜𝑊 𝑡−1i MSCI𝑅𝑜𝑊 𝑡is RoW MSCI (excluding the U.S.). 𝑒𝑥𝑅𝑜𝑊 𝑡100 ×hln EX𝑅𝑜𝑊 𝑡−ln EX𝑅𝑜𝑊 𝑡−1i EX𝑅𝑜𝑊 𝑡is RoW exports (excluding the U.S.) by Grossman et al. (2014). 𝑖𝑝𝐴𝐸 𝑡100 ×ln IP𝐴𝐸 𝑡−ln IP𝐴𝐸 𝑡−1IP𝐴𝐸 𝑡is the advanced economies’ industrial production (excluding the U.S.) by Grossman et al. (2014). 𝑒𝑥𝐴𝐸 𝑡100 ×ln EX𝐴𝐸 𝑡−ln EX𝐴𝐸 𝑡−1EX𝐴𝐸 𝑡is AE exports (excluding the U.S.) by Grossman et al. (2014). 𝑖𝑝𝐸 𝑀𝐸 𝑡100 ×ln IP𝐸 𝑀𝐸 𝑡−ln IP𝐸𝑀 𝐸 𝑡−1IP𝐸𝑀 𝐸 𝑡is the emerging market economies’ industrial production by Grossman et al. (2014).∗ 𝑒𝑥𝐸 𝑀 𝐸 𝑡100 ×ln EX𝐸 𝑀𝐸 𝑡−ln EX𝐸𝑀 𝐸 𝑡−1EX𝐸𝑀 𝐸 𝑡is the emerging market economies exports by Grossman et al. (2014). Notes: RoW is short for rest-of-world. All variables are available over the pre-sample and estimation sample period, ranging from 1980M5 to 2017M12. Exceptions is: EME IP𝑡1987M1 to 2017M12. 37 B. TVP-VAR: Prior Settings and Estimation We describe the details on the prior density choice and hyperparameter calibration. We have to set prior densities for the initial values of 𝜽, which we denote with 𝜽0. For all these initial values, we specify Gaussian distributions. We also have to specify prior densities for the covariance matrices 𝑸, where we use an inverse-Wishart prior density.26 Last, we also need a prior distribution for the covariance matrix of the VAR 𝚺, which is again following an inverse-Wishart distribution. To calibrate the prior distributions, we use the first 13 years as a training sample. This results in a training sample ranging from 1980M1 to 1992M12 of length 𝜏=156. We obtain estimates using Ordinary Least Squares (OLS) from this training sample and assume the following prior densities for the coefficients in the TVP-VAR 𝜽0∼ N ˆ 𝜽𝑂𝐿𝑆,4∗𝑉(ˆ 𝜽𝑂𝐿𝑆), 𝑸∼𝑖𝑊 𝜅2 𝑄∗𝜏∗𝑉(ˆ 𝜽𝑂𝐿𝑆, 𝜏), (B.1) where the subscript 𝑂𝐿𝑆 refers to the OLS estimator of the respective coefficient. For the initial values, we use the OLS point estimates and four times its variance. For the covariance matrix 𝑸, the scaling matrix is chosen to be a fraction of the corresponding OLS estimates (multiplied with the corresponding degrees of freedom). Finally, we have to choose a value for the hyperparameter 𝜅2 𝑄, governing the time-variation in the state equation. In particular, we assume 𝜅2 𝑄=0.015 (following Paul, 2020). We use a rather conservative value for this hyperparameter such that the time-variation is not inflated by our prior. An additional note is in order: We observe the high-frequency surprises not until 1991M1. Following the procedure of Paul (2020), we plug in zeros for the observations prior to this period. This should not cause a bias in OLS as long as those zeros are from a random sample. This should be indeed the case for the monetary policy surprises. Nevertheless, we estimate those coefficients with less precision. Last, we discuss the prior density on the covariance matrix 𝚺, which is defined as follows 𝚺∼𝑖𝑊 (𝑰𝑀, 𝑀 +1), where the scaling matrix is set to an identity matrix and the degrees of freedom are set to 𝑀+1, as recommended by Karlsson (2013). Regarding the estimation procedure, we set up the Gibbs sampler along the lines of Del Negro and Primiceri (2015) to obtain posterior distributions. In particular, we use Kalman filtering techniques to obtain the unobservable states in 𝜽𝑇=(𝜽′ 1, . . . , 𝜽′ 𝑇)′. This results in a Gaussian state space model, where standard Bayesian methods for the Kalman filter can be applied (Carter and Kohn, 1994; Frühwirth-Schnatter, 1994). The remaining posterior quantities are rather standard and inference is conducted via an MCMC algorithm. 26 We denote by 𝑖𝑊 (𝑆, 𝑑)an inverse-Wishart distribution with degrees of freedom 𝑑and scale matrix 𝑆. 38 Figure E3: Country-Specific Industrial Production Impulse Responses. Notes: Responses to a contractionary U.S. monetary policy shock that induces a one percentage point increase in the federal fund rate in 1993M1. Countries: Spain, South Africa, Brazil, Chile, Columbia, Mexico. 45 Figure E4: Country-Specific Industrial Production Impulse Responses. Notes: Responses to a contractionary U.S. monetary policy shock that induces a one percentage point increase in the federal fund rate in 1993M1. Countries: Peru, South Korea, Malaysia, India. 46 F. Additional Results of the Sensitivity Analysis Figure F1: Robustness: Zero Lower Bound. Notes: Responses to a contractionary U.S. monetary policy shock that induces a one percentage point increase in the federal fund rate in 1993M1. Baseline model in blue; red lines correspond to the model excluding the ZLB period (ending in 2008M11). 47 Figure F2: Robustness: Stochastic Volatility. Notes: Peak Responses to a contractionary U.S. monetary policy shock that induces a one percentage point increase in the federal fund rate in 1993M1. Baseline model vs. extended model that allows for stochastic volatility specification. Details on stochastic volatility: As in our benchmark analysis and in Primiceri (2005), we use OLS estimates on a pre-sample (1980M1-1992M12) to calibrate the prior distributions. In addition, a prior belief on the extent of time-variation in 𝑨−1 𝑡and ℎ𝑖𝑡 (𝑖=1, . . . , 𝑛)must be specified. Using the notation of Primiceri (2005), this boils down to a selection choice on three parameters: 𝜅𝑄(governing time-variation of autoregressive coefficients), 𝜅𝑊(variance of the residuals), 𝜅𝑆(covariance of the residuals). We set 𝜅𝑄=0.015 (as before), 𝜅𝑊=0.001, and 𝜅𝑆=0.001. The value of 𝜅𝑊 is among the ones considered by Primiceri (2005), while our 𝜅𝑆is relatively tighter (to avoid ill behaviors in our relatively shorter sample). We estimate the stochastic volatility model as in Kim et al. (1998) but with the refinement of Omori et al. (2007). 48 Figure F3: Robustness: Different Monetary Policy Instruments. Notes: Responses to a contractionary U.S. monetary policy shock that induces a one percentage point increase in the federal fund rate in 1993M1. Baseline model in blue (instrument of Miranda-Agrippino and Ricco, 2021); red lines correspond to the estimates using the high-frequency monetary policy instruments of Gertler and Karadi (2015) (GK) and Jarociński and Karadi (2020) (JK). Figure F4: Robustness: Prior Calibration. Notes: Responses to a contractionary U.S. monetary policy shock that induces a one percentage point increase in the federal fund rate in 1993M1. Baseline model in blue; red lines corresponds to the estimates using a tighter (𝜅2 𝑄=0.01) and wider (𝜅2 𝑄=0.02) prior against the baseline prior (𝜅2 𝑄=0.015). 49