Real exchange rate and international spillover effects of US technology shocks

Abstract

The paper presents new empirical evidence on the international effects of surprise and anticipated technology shocks in the US. We employ the proxy-instrumental variable approach to identify structural vector autoregressions in a panel setting and empirically study the transmission of US technology innovations to the G7 countries. Both unanticipated and anticipated exogenous technology improvements lead to a strong and persistent real appreciation (from the point of view of the US), along with an expansionary effect on US macroeconomic aggregates, except for hours worked which initially decline. Internationally, there is a strong and precisely estimated positive spillover on foreign output, consumption, and hours worked in the case of surprise shocks, and a weaker but still mostly non-negative effect in the case of technology news shocks. We show that the empirical evidence is qualitatively compatible with the predictions of a New Keynesian international business cycle model with imperfect financial markets, traded and non-traded goods and imported intermediate inputs in production.

Introduction

What are the effects of technology shocks in open economies with strong cross-border linkages? In this paper, we estimate structural vector autoregressions (VAR) using macroeconomic data on a panel of countries to uncover the dynamic effects of US technology shocks on the G7 countries (other than the US, henceforth called G6). We make two key contributions. First, we use recently developed proxy-instrumental variable methods to identify both unanticipated, or surprise shocks on US productivity, as well as anticipated shocks representing news of future technology improvements. We assess the effects of both shocks on a wide range of variables that are important for an understanding of the international transmission process of exogenous changes in productivity. Most importantly, we present evidence on the responses of real exchange rates and on the international transmission of shocks in the US on the other countries. Second, we show that most of the evidence is qualitatively consistent with a version of a New Keynesian international business cycle model, and show the influence of some key parameters on the ability of the model to generate empirically plausible results.

A growing recent literature analyzes open economy aspects of technology shocks with VAR models relying on different identification assumptions. However, so far there is substantial disagreement concerning the international transmission of technology shocks and their impact on real exchange rates. While some studies find that an exogenous unanticipated increase in productivity leads to a real appreciation (e.g., Enders et al., 2011; Corsetti et al., 2014), others provide evidence in favor of a real depreciation (e.g., Miyamoto and Lan Nguyen, 2017). Moreover, some studies find that surprise and anticipated technology shocks are associated with a different exchange rate behavior: appreciation following anticipated and depreciation following surprise technology shocks (e.g., Nam and Wang, 2015, Levchenko and Pandalai-Nayar, 2020).

We contribute to this debate using the proxy-instrumental variables (proxy-VAR for short) methodology (Stock and Watson, 2012; Mertens and Ravn, 2013) that identifies structural shocks through instrumental variables that are arguably correlated with a particular economic shock. In contrast to existing empirical identification approaches, this method has the advantage that neither controversial parameter restrictions, nor assumptions on the contribution of a shock to the forecast error variance of particular variables, nor a priori knowledge of the signs of impulse responses are required. Specifically, we use Fernald's (2014) series of utilization and markup adjusted total factor productivity (TFP) as an instrument that is arguably correlated with a surprise technology shock in the US. For the identification of anticipated or news shocks, we follow Miranda-Agrippino et al. (2020) and Cascaldi-Garcia and Vukotic (2020) and use the non-forecastable part of patent applications as an instrument.

Our central novel finding is that both surprise and anticipated domestic technology shocks (we take the US to be the domestic economy, henceforth) lead to a strong and persistent real exchange rate appreciation that is robust across specifications and precisely estimated. We argue that this common real exchange rate response to both types of technology shocks is in line with economic theory, as this is precisely what is predicted by standard New Keynesian two country models with imperfect financial markets, tradable and non-tradable goods and a price elasticity of tradable goods of the size of recent empirical estimates. Thus, the empirical effects of properly identified surprise and anticipated US technology shocks are not puzzling from a theoretical point of view.

Furthermore, we find that both shocks lead to a real output increase both domestically and abroad. However, the positive output effect in the foreign economies is weaker than in the US, such that we observe generally positive but quantitatively limited international business cycle spillovers. Similar observations pertain to private consumption and investment. In the US, a surprise technology shock initially lowers hours worked, as predicted by sticky-price models. In the case of anticipated shocks, the productivity increase is delayed for several periods, while domestic consumption increases already on impact and domestic investment decreases in the short-run. The foreign effects of anticipated US productivity shocks are generally weaker and less precisely estimated, though. Further, US technology shocks tend to lower real domestic export prices, and at the same time increase the real prices of services that are likely to contain a large non-tradable component. We interpret this as evidence that technology shocks mainly originate from the traded sector of the US economy.

On the theoretical side, we demonstrate that a two-country New Keynesian business cycle model with capital accumulation, imperfect financial markets, traded and non-traded goods and a productive role for imported intermediate inputs is capable of explaining most of the empirical results if technology shocks take place in the sector producing tradable goods. The key parameter restrictions for this outcome are a relatively high but empirically realistic price elasticity of tradable goods, and complementarity between non-traded and traded goods. Given these, the model's prediction that domestic productivity increases coincide with positive spillovers and a real appreciation is well in line with the empirical findings. Moreover the model implies that the productivity improvement leads to an initial employment decline at home together with productivity and employment increases abroad, and a positive response of foreign output, consumption, and investment, and a decline in domestic tradable prices associated with an increase in non-tradable prices, much as we find empirically.

The paper is related to various strands of literature. The result of a real appreciation following a favorable home surprise technology shock is also found by Enders et al. (2011) and Corsetti et al. (2014) using sign-restricted VARs. In contrast, Nam and Wang (2015), Levchenko and Pandalai-Nayar (2018) and Miyamoto and Lan Nguyen (2017) find a real depreciation of the US currency relative to an aggregate of other countries or Canada using recursiveness or long-run restrictions. Compared to these papers, ours is complementary in that it uses the proxy-instrumental variables method as an alternative identification scheme, and in considering a broad set of empirical macro variables for the panel of G7 countries that comprise several of the largest trading partners of the US.¹ News shocks have originally been discussed in closed-economy contexts (Beaudry and Portier, 2006; Schmitt-Grohe and Uribe, 2012). The open economy dimension of anticipated shocks has recently been explored by Kamber et al. (2017) who use a combination of a short-run zero and sign restrictions, without reporting results for real exchange rates. Nam and Wang (2015), and Levchenko and Pandalai-Nayar (2020) identify an anticipated technological shock as the one that maximizes the forecast error variance of TFP over a fixed horizon, as in Barsky and Sims (2011), and find a domestic real appreciation after positive technology news. In contrast to these studies, the patent based indicator of anticipated technological change that we use in the present paper as an instrumental variable allows us to identify news shocks without restricting the short-run or long-run response of productivity a priori.

Concerning the international spillovers of technological shocks, Miyamoto and Lan Nguyen (2017) and Levchenko and Pandalai-Nayar (2020) find positive effects of US shocks on Canadian output, and Kamber et al. (2017) find similar results with respect to three further countries. Corsetti et al. (2014) and Nam and Wang (2015) estimate an increase in relative output in the US relative to an aggregate of other countries following a positive US technology shock, but do not report how the relative response is distributed between the actual output levels of both countries or aggregates. In the present paper, in contrast, we report the responses of domestic and foreign output levels, as well as those of other macroeconomic aggregates and international relative price measures, for both the US and the G6 panel of foreign economies.

The rest of the paper is organized as follows. Section 2 contains the empirical results. We explain the empirical method in Section 2.1, discuss the data used in Section 2.2, and present results in Section 2.3. Section 3 introduces the theoretical model, Section 4 shows the model impulse responses and compares them to the empirical evidence, whereupon Section 5 concludes.