Exchange Rate Regime and International Transmission of Business Cycles: Capital Account Openness Matters* Kyunghun Kim† KIEP

Ju Hyun Pyun‡ Korea University Business School January 2018 Abstract

We investigate the role of exchange rate regime in international transmission of business cycles during the global financial crisis. We find that exchange rate regime did not solely draw a distinction in the international transmission of business cycles during the crisis. However, analysis considering capital account openness and countries with currencies pegged to the U.S. dollar indicates that the exchange rate regime plays an important role in shaping business cycle co-movement: adopting a fixed regime with high capital account openness (additionally) increased business cycle co-movement with the United States during the crisis, whereas U.S. dollar peggers with relatively restrictive capital accounts during the crisis were not found to affect business cycle transmission.

Keywords: Exchange rate regime, Business cycle co-movement, Capital account openness, Global financial crisis, Trilemma JEL: E32, E52, F31, F33, F44

*

We are grateful to Alan M. Taylor for his constructive comments. We also thank Hyelin Choi, Makram El-Shagi, Minsoo Han, Kuk Mo Jung, Jung-sik Kim, Jongwook Lee, Frank Shao, Kenneth West and seminar participants in HenU/INFER conference for their valuable comments. All remaining errors and omissions are our own. † Korea Institute for International Economic Policy, Building C, Sejong National Research Complex, 370 Sicheongdaero, Sejong-si 30147, Korea. Email: [email protected] ‡ Corresponding author: Business School, Korea University, 145 Anam-Ro, Seongbuk-Gu, Seoul 136-701, Tel: 822-3290-2610. Email: [email protected]

1

1. Introduction It has long been debated whether fixed and floating exchange rate regimes exert differential influences on business cycle co-movement. Seminal works such as Baxter and Stockman (1989) and Ahmed et al. (1993) find little evidence of systematic differences in business cycles under fixed and floating exchange rate regimes. 1 However, Gerlach (1988) finds that the business cycles observed during the flexible exchange rate regime of the United States and the European OECD countries were more synchronized vis-à-vis those observed during the Bretton Woods period. Subsequent studies such as Artis and Zhang (1997, 1999) and Clark and van Wincoop discuss this issue in the context of the European Exchange Rate Mechanism. They maintain that the role of the exchange rate regime in business cycle transmission in an “open” financial market is closely related to the debate on whether monetary policy coordination against external shocks leads to less output co-movement.2 This study investigates the effect of a country’s exchange rate regime on the international transmission of real business cycles during the global financial crisis (GFC) of 2008–2009, which is akin to a negative shock from the United States to other countries. If the negative shock from the epicenter of the GFC were transmitted to the rest of the world, both the country of origin and the other countries would simultaneously experience economic downturns and their business cycles would become more synchronized (e.g., Kalemli-Ozcan, Papaioannou, and Perri, 2013; Kalemli-Ozcan, Papaioannou, and Peydro, 2013; Pyun and An, 2016). We examine whether there is any distinction in the international transmission of business cycles caused by the 1

However, Baxter and Stockman (1989) use quarterly data of 49 countries and find that the cross-correlations of output generally decreased in the post-1973 period, as compared to the Bretton Woods period. 2 Artis and Zhang (1997, 1999) find that the formation of the European Exchange Rate Mechanism is associated with higher business cycle co-movement among its member countries. However, Clark and van Wincoop (2001) find that in Europe, coordinated monetary policies through a single currency did not significantly increase business cycle synchronization.

2

U.S. GFC shock to 57 other countries with respect to their exchange rate regime. A distinctive feature of this study is that it examines the role of a country’s capital account openness in shaping the effect of exchange rate regime on the international transmission of business cycles. Using a simultaneous equation model that controls for endogeneity as well as various international linkages that transmit the shock of the crisis, we find that exchange rate regime solely did not play a significant role in generating different outcomes of business cycle comovement during the GFC. However, when considering capital account openness and base countries of currency peg, we find that exchange rate regime indeed affects business cycle comovement. That is, countries with a fixed exchange rate regime and high capital account openness were more vulnerable to the negative shock originating in the United States, and that their business cycle co-movements with the United States increased during the GFC. Conversely, the international transmission of business cycle during the GFC were less pronounced in countries with currencies pegged to the U.S. dollar and capital account openness relatively low. Our results are robust, with the inclusion of policy, trade, financial linkages and unobserved country-pair heterogeneity that influence the evolution of the business cycles in countries, independent of the exchange rate regime. Additional analysis considering contemporaneous contagion and time lags of shock transmission using a country’s quarterly gross domestic product (GDP) growth supports our main findings. What is the theoretical rationale for including capital account openness to understand the relationship between exchange rate regime and business cycle co-movement across countries? Our arguments rely on a popular policy trilemma 3 perspective stemming from the Mundell–

3

The trilemma suggests that a government cannot simultaneously opt for an open financial market, exchange rate stability, and monetary autonomy.

3

Fleming model, which pertains to the previous debate on exchange rate regime and international business cycle co-movement in the context of an “open” economy (Artis and Zhang, 1997, 1999; Clark and van Wincoop, 2001). In an integrated financial market, a country’s choice of currency peg is likely to result in a loss of monetary autonomy, keeping the country from proactively stabilizing the fluctuations in business cycles driven by external shocks. As the monetary policy is an important fine tuning tool to smooth a country specific business cycle, imported policy from the base country of currency peg would not perfectly work even though the optimal policy responses to the shock in the country that pegs its currency and its base country are the same. 4 Thus, if the policy trilemma holds, a country in which monetary autonomy is foregone (with a pegged exchange rate and full capital mobility) would be more susceptible to external shocks. In this regard, our empirical finding that fixed regime countries with higher capital market openness had business cycles that were more synchronized with those of the United States during the GFC can be construed as an outcome of a lack of monetary policy independence. In particular, our findings for fixed regime countries are reinforced when controlling for the countries with currencies pegged to the U.S. dollar because the few such countries in the sample i) are mostly emerging and developing countries with “restrictive” capital accounts and ii) experienced the appreciation of the U.S. dollar during the GFC. Thus, countries with currencies pegged to the U.S. dollar were insulated from the negative consequences of the GFC to a certain extent, with some room for independent monetary policy responses via capital controls and low risk of capital reversals. Our finding with regard to these countries is also consistent with that of Rose (2014), who finds that hard fixers and inflation targeters with floating regimes among 4

Monetary autonomy itself does not guarantee monetary policy effectiveness that stabilizes fluctuations. Even in countries with monetary autonomy, policymakers may fail to stabilize business cycle fluctuation. However, skeptics of the European single currency have frequently argued that the inability to respond to country-specific shocks can lead to greater business cycle volatility (Clark and van Wincoop, 2001).

4

emerging and developing countries showed similar phases of business cycles, capital flows, and so on during and after the GFC. Our findings are related to previous studies on external shocks such as financial crisis and their cross-country spillovers. Choudhri and Kochin (1980) find that during the Great Depression, four European countries that opted for fixed exchange rates with the United States suffered severe contractions in both output and prices following the recession in the United States, while countries like Spain, which maintained floating regimes, enjoyed relatively stable output and prices. Mathy and Meissner (2011) show that both trade integration and the gold standard played important roles in the transmission of negative shock during the Great Depression. Hoffmann (2007), using developing country sample, shows that external shocks from the world output or world interest rate changes are less contractionary under floating regimes than fixed regimes. Lane and Milesi-Ferretti (2011) find that fixed regime countries experienced relatively more severe economic downturns during the GFC. These findings indicate that countries with fixed regimes are generally more vulnerable to external shocks, but this study reveals that capital account openness needs to be considered for a better understanding of the role of exchange rate regime in international transmission of business cycles. Our study also contributes to the literature by offering a specific exercise with regard to the effectiveness of the trilemma from a different perspective; it does so by investigating countries that opt for a fixed exchange rate regime and the effect of their remaining choices regarding capital mobility and monetary autonomy on the international transmission of the U.S. business cycle during the GFC. Previous studies empirically confirm the potency of the trilemma

5

(Obstfeld, Shambaugh, and Taylor, 2005; Shambaugh, 2004).5 By introducing a trilemma index for each country to measure a unique combination of three policy choices, Aizenman, Chinn, and Ito (2010, 2013) demonstrate that countries face the trilemma.6 Klein and Shambaugh (2015) find evidence that the trilemma holds, and acquiring monetary autonomy by adhering to midrange policies may be difficult. However, Rey (2015) raises doubt regarding the trilemma empirically.7 Unlike previous studies, which focused on the trilemma configuration, this study examines a sort of “policy outcome” in which the trilemma is embedded, that is, how a country’s choice of currency peg and capital account openness shapes the international transmission of the real business cycle from the United States during the GFC. The remainder of this paper is organized as follows. Section 2 presents our empirical methodology and the data used, while sections 3 and 4 present the empirical results and the results of the robustness tests, respectively. The concluding remarks are provided in section 5.

2. Econometric Methodology 2.1. Empirical Model and Data We introduce the simultaneous equation model for data on country-dyad following previous studies on international transmission of business cycles (An et al., 2017; Davis, 2014; Dees and Zorell, 2012; Imbs, 2004, 2006; Pyun and An, 2016). The data used comprise of 57 countries, including the United States, which is the point of origin of the GFC for the period

5

Bluedorn and Bowdler (2010) evaluate international interest rate transmission from the United States, given the exchange rate regime, and find that the trilemma holds for the identified exogenous shock on U.S. monetary policy. 6 Aizenman, Chinn, and Ito (2010) analyze the effect of the trilemma configuration on economic outcomes and find that countries with lower levels of monetary independence in particular (i.e., choosing the other two policy goals) tend to experience greater output volatility. 7 Rey (2015) argues that the trilemma is broken in the financially globalized world, as the global financial cycle constrains national monetary policies regardless of the exchange rate regime.

6

2001–2013.8 It is crucial to identify the pure effect of exchange rate regime on the international transmission of business cycle during the GFC because there are several possible channels through which the exchange rate can influence the transmission of the business cycle. First, exchange rate movement generated the valuation effect on international financial asset position during the GFC 9 and subsequently, affected the business cycle. In addition, exchange rate variability influenced the transmission of the business cycle through trade linkages, as trade is closely associated with exchange rate movement. Many previous studies discuss business cycle co-movement with respect to financial (Davis, 2014; Imbs, 2004; Kalemli-Ozcan, Sørensen, and Yosha, 2003) and trade integration (Baxter and Kouparitsas, 2005; Calderon, Chong, and Stein, 2007; Clark and van Wincoop, 2001).10 As the aforementioned international linkages along with the exchange rate regime may play an important role in shock transmission, we isolate its effect on business cycle co-movement from that on trade and financial linkages by controlling for these linkages. The simultaneous equations model not only considers the effects of direct and indirect channels of international linkages on business cycle co-movement, but also disentangles the interactions among business cycle co-movement, financial integration, trade integration, and other variables. Our simultaneous equations model consists of five equations, which are as follows: 8

Note that our financial linkage variables sourced from Coordinated Portfolio Investment Survey (CPIS) are available only from 2001 onwards. 9 Bénétrix et al. (2015) examine the valuation impact of currency movements during the GFC and find that the scale of these effects was larger in 2008 relative to the pre-crisis years, and these effects were quite persistent during the period 2008–2012. They also mention that pegging makes it more difficult for a country to experience valuation gains during a crisis. 10 Studies like Clark and van Wincoop (2001) and Baxter and Kouparitsas (2005) support the positive effect of trade on business cycle co-movement. Furthermore, Imbs (2004) and Calderon, Chong, and Stein (2007) show that intraindustry trade pattern plays a role in creating a positive relationship between trade and business cycle co-movement.

7

𝑆𝑌𝑁𝐶𝑖,𝑈𝑆,𝑡 = 𝛼0 + 𝛽1 𝐶𝑈𝑆,𝑡 + 𝛽2 𝐹𝑋𝑅𝑖,𝑈𝑆,𝑡 + 𝛽3 𝑈𝑆𝑃𝐸𝐺𝑖,𝑈𝑆,𝑡 + 𝛽4 𝐶𝑈𝑆,𝑡 ∙ 𝐹𝑋𝑅𝑖,𝑈𝑆,𝑡 +𝛽5 𝐶𝑈𝑆,𝑡 ∙ 𝑈𝑆𝑃𝐸𝐺𝑖,𝑈𝑆,𝑡 + 𝛽6 𝑇𝐼𝑖,𝑈𝑆,𝑡 + 𝛽7 𝐹𝐼𝐷𝐵𝑖,𝑈𝑆,𝑡 + 𝛽8 𝐹𝐼𝐸𝑄𝑖,𝑈𝑆,𝑡 +𝛽9 𝑆𝑖,𝑈𝑆,𝑡 + 𝛽10 𝐹𝑖𝑠𝑐𝑎𝑙𝑆𝑌𝑁𝐶𝑖,𝑈𝑆,𝑡 + 𝐼𝑛𝑡𝑒𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑡𝑒𝑟𝑚𝑠𝑖,𝑈𝑆,𝑡 𝛽 + 𝜀𝑖,𝑈𝑆,𝑡 , 𝑇 𝑇𝐼𝑖,𝑈𝑆,𝑡 = 𝜃0 + 𝜃1 𝐶𝑈𝑆,𝑡 + 𝜃2 𝐹𝐼𝐸𝑄𝑖,𝑈𝑆,𝑡 + 𝜃3 𝐹𝐼𝐷𝐵𝑖,𝑈𝑆,𝑡 + 𝜃4 𝑆𝑖,𝑈𝑆,𝑡 + 𝑋𝑖,𝑈𝑆,𝑡 θ + 𝑢𝑖,𝑈𝑆,𝑡 ,

𝐹𝐼𝐷𝐵𝑖,𝑈𝑆,𝑡 = 𝜙0 + 𝜙1 𝐶𝑈𝑆,𝑡 + 𝜙2 𝐹𝐼𝐸𝑄𝑖,𝑈𝑆,𝑡 + 𝜙3 𝑇𝐼𝑖,𝑈𝑆,𝑡 + 𝜙4 𝑆𝑖,𝑈𝑆,𝑡 +

𝐷 𝑋𝑖,𝑈𝑆,𝑡 ϕ

(1)

+ 𝑤𝑖,𝑈𝑆,𝑡 ,

𝐸 𝐹𝐼𝐸𝑄𝑖,𝑈𝑆,𝑡 = 𝜆0 + 𝜆1 𝐶𝑈𝑆,𝑡 + 𝜆2 𝐹𝐼𝐷𝐵𝑖,𝑈𝑆,𝑡 + 𝜆3 𝑇𝐼𝑖,𝑈𝑆,𝑡 + 𝜆4 𝑆𝑖,𝑈𝑆,𝑡 + 𝑋𝑖,𝑈𝑆,𝑡 λ + 𝜈𝑖,𝑈𝑆,𝑡 , 𝑆 𝑆𝑖,𝑈𝑆,𝑡 = 𝛾0 + 𝛾1 𝐶𝑈𝑆,𝑡 + 𝛾2 𝐹𝐼𝐸𝑄𝑖,𝑈𝑆,𝑡 + 𝛾3 𝐹𝐼𝐷𝐵𝑖,𝑈𝑆,𝑡 + 𝛾4 𝑇𝐼𝑖,𝑈𝑆,𝑡 + 𝑋𝑖,𝑈𝑆,𝑡 γ + 𝑒𝑖,𝑈𝑆,𝑡 , }

where SYNCi,US,t is a measurement of business cycle co-movement between country i and the United States in year t. The measurement of year-by-year business cycle synchronization (SYNC) is the negative divergence, defined as the absolute value of difference between the growth rates of real GDP per capita of countries i and j in year t: SYNCi ,US ,t   (ln Yt i  ln Yt i1 )  (ln YtUS  ln YtUS 1 ) ,

(2)

where 𝑌𝑡𝑖 and 𝑌𝑡𝑈𝑆 denote GDP per capita based on the purchasing power parity (PPP) adjusted by constant 2005 international dollars of countries i and the United States, respectively, both of which are collected from the World Development Indicators (WDI) of the World Bank. Previous studies like Kalemli-Ozcan, Papaioannou, and Perri (2013) and Pyun and An (2016) use this measure to investigate and verify the transmission of shock. This measure does not reflect the volatility of output growth of each country in a country pair, but the co-variation of output growth. Further, the SYNC measure uses time-series information to reflect changes in growth rates as well as the intensity of the output correlation. For example, SYNC increased when a country was negatively affected by the U.S. economy during the GFC. Additionally, section 4 introduces an alternative measurement for business cycle co-movement to verify the robustness of the results. 8

CUS,t is an indicator of the GFC. It is a dummy variable takes the value 1 for the GFC period, 2008–2009, and 0 otherwise. As the alternative crisis variable, we introduce a continuous crisis measure, VIX, which is the Chicago Board Options Exchange Market Volatility Index. FXRi,t is a time-variant binary variable that indicates a country’s choice of fixed exchange rate (or currency peg). This is drawn from Shambaugh (2004) and Klein and Shambaugh (2006). FXRi,t takes the value 1 when a currency either stays within a 2 percent band against the base currency except the U.S. dollar or has zero volatility in all the months except for a one-off devaluation. An additional concern is that the effect of a fixed exchange rate regime on the transmission of business cycle during the GFC may vary with the base country of a currency peg. While a country that pegs its currency to the U.S. dollar is forced to rely upon U.S. monetary policy in the integrated financial market, a country that pegs its currency to that of another currency imports the corresponding country’s monetary policy. Thus, empirical analysis is required to introduce two separate detailed categories of fixed exchange rate regime dummies, namely, countries with currencies pegged to the U.S. dollar and those with currencies pegged to a nonU.S. currency. USPEGi,t is a time-variant binary variable that indicates when a country pegs its currency to the U.S. dollar only. FXRit is mutually exclusive of USPEGit, and FXRit in union with USPEGit together forms a fixed exchange rate regime variable, FXit. In our baseline regression, we first use the FXit variable and then include FXRit and USPEGit separately in the main results. Table 1 provides the list of countries in the sample and their choice of exchange rate regime. Note that for the robustness check, we use a classification of exchange rate regime (the de facto measure) defined by Ilzetzki, Reinhart, and Rogoff (2017). TIi,US,t is the measure for bilateral trade integration between country i and the United States, 9

and it is calculated as the sum of bilateral export and import divided by the sum of the GDP of the two countries. To measure the degree of bilateral financial integration, we use a quantitybased measure following Kalemli-Ozcan, Papaioannou, and Perri (2013). Two financial integration measures, FIEQi,US,t and FIDBi,US,t, are calculated as follows. First, we add the assets of equity (debt) securities issued by the United States and the liabilities of equity (debt) securities held by U.S. investors at time t of a country and divide it by the sum of the GDP of the two countries. We collect these cross-border holding variables based on the market values of positions held and expressed in current U.S. dollars from the Coordinated Portfolio Investment Survey (CPIS) in IMF. Si,US,t is a measurement of the similarity in the production structures of country i and the United States. It is calculated as 𝑆𝑖,𝑈𝑆,𝑡 = ∑𝑁 𝑛 |𝑠𝑛,𝑖,𝑡 − 𝑠𝑛,𝑈𝑆,𝑡 | , where 𝑠𝑛,𝑖,𝑡 is sector n’s share in total value-added in country i, 𝑠𝑛,𝑈𝑆,𝑡 is sector n’s share in the total valueadded of the United States. The sector-based value-added shares are calculated using agriculture, manufacturing, and service industry value-added data from the WDI. Finally, FiscalSYNCi,US,t is a measurement of correlation of the fiscal policies of country i and the United States.11 It is calculated as the absolute value of the difference in the changes in government spending between country i and the United States. [Insert Table 1] Table 2 reports the descriptive statistics of the main variables employed in the analysis. SYNC ranges from –9.960 (most divergent) to –0.002 (most synchronized); and its reported mean is –2.503. The means of equity and debt integration as a percentage of the sum of GDP are 0.53 and 0.41, respectively, but these vary across countries, with their maximum values being 7.31 and 5.51, respectively. 11

This variable can be considered as a proxy for fiscal policy independence.

10

[Insert Table 2]

2.2. Identification The endogenous variables within the system are SYNCi,US,t, TIi,US,t, FIEQi,US,t, FIDBi,US,t, and Si,US,t, which indicates that business cycle co-movement is related to trade and financial integration and the production similarity variable between country i and the United States. We also include the interaction terms between the crisis and trade and financial integration in the first business cycle equation to control for possible channels through which the shock was transmitted, especially during the GFC. In a technical sense, the interaction terms of the crisis variable and trade and financial integration give rise to three more nonlinear endogenous variables, making the number of endogenous variables greater than the number of equations. Following Wooldridge (2010), we add an appropriate set of exogenous variables such as the interaction terms of the crisis variable and other instrumented exogenous variables, instead of adding more equations for the nonlinear endogenous terms. We include XTi,US,t, XEi,US,t, XDi,US,t, and XSi,US,t, all of which are vectors of the exogenous variables, in each equation except the first business cycle co-movement equation for identification (exclusion restrictions). The included variables help explain the relationships among the endogenous variables of country i and the United States. First, a common set of exogenous variables in the equations is formed by the log of the distance from the United States and a border dummy, which influence transactions of real goods and financial assets, and the use of English as an official language, which reflects linguistic and cultural proximity. We also add different sets of exogenous variables to each equation. Two financial integration equations, XEi,US,t and XDi,US,t, contain a common law, as its presence is likely to lead 11

to similar institutions, regulations, and customs in the financial transactions between countries. Moreover, XEi,US,t and XDi,US,t contain the sum of aggregate capital control restriction indices in equities and debt transactions of country i and the United States, sourced from Fernández et al.’s (2016) capital control database. To identify the trade equation, XTi,US,t includes a time-variant regional trade agreement dummy. Finally, in the production similarity equation, XSi,US,t includes the (log) product and the absolute value of the difference between the GDP per capita of the countries. Advanced country dummies and year dummies are included as additional exogenous variables in all equations. Our identification strategy is consistent with that of Pyun and An (2016). To make the results more robust, we include country-pair fixed effects that capture country-pair unobserved heterogeneity in the system as emphasized by Kalemli-Ozcan, Papaioannou, and Peydro (2013). Our main analysis focuses on the effect of the GFC shock on the international business cycle (the transmission of negative shock to the business cycle), which indicates the marginal effect of the crisis variable on business cycle co-movement, ∂SYNCi,US,t/∂CUS,t. In the empirical model, the interaction term of crisis and exchange rate regime allows us to examine the extent to which the international transmission of business cycle during the GFC depends on a country’s exchange rate regime choice. For example, the coefficient of the interaction term, 𝛽4 , in Equation 𝜕𝑆𝑌𝑁𝐶𝑖,𝑈𝑆,𝑡

(1) conceptually implies 𝜕 (

𝜕𝐶𝑈𝑆,𝑡

)⁄𝜕𝐹𝑋𝑅𝑖,𝑡 . If β4 is positive (negative) in Equation (1), the

international transmission of the GFC shock through business cycle (∂SYNCi,US,t/∂CUS,t) is marginally higher (lower) for countries that opt for a fixed exchange rate than for those that opt for a floating exchange rate. Note that our model mainly sheds light on the direct transmission of

12

business cycle from the United States, which is set as the reference country. 12 Furthermore, to examine the role of capital account openness in shaping the effect of exchange regime on business cycle co-movement, we separate our full sample into two subsamples according to the mean value of each country’s averaged de jure financial openness from Chinn and Ito (2008) for the period 2001–2007 just before the GFC and repeat our main analysis. Panels A and B of Table 3 present the countries with high and low capital account openness, respectively. For the robustness of the results, we use another threshold of financial openness: the mean of averaged financial openness of each country in the entire period. [Insert Table 3]

3. Empirical Results 3.1. Exchange Rate Regime and Business Cycle Co-movement Table 4 presents the results of the simultaneous equations model for business cycle comovement (SYNCi,US,t) and the proxies of the international linkages, which are estimated using three-stage least squares (3SLS) and generalized methods of moment (GMM) analysis. Column (1) starts by including a fixed exchange rate regime variable (including all reference currencies, FXi,t ≡ FXRi,t ∪ USPEGi,t) and its interaction term with the crisis indicator. The coefficient of the interaction term of a country choosing a fixed exchange rate and the crisis variable is positive, but statistically insignificant, suggesting that the international transmission of business cycle from the United States during the GFC is not different for fixed and floating exchange rate regimes. 12

The negative shock from the United States can be transmitted to Korea via China, as well as directly to Korea. There might also be inter-linkage transmission channels, such as the United States–China–Korea. However, by excluding the above inter-linkage transmission channels of business cycle co-movement among the other countries, we aim to examine the more direct impacts of the negative shock that originated in the United States.

13

However, in column (2), we categorize the fixed exchange rate variable (FXi,t) as “pegged to the U.S. dollar ” (USPEGi,t) and “pegged to other reference currencies” (FXRi,t). Interestingly, the interaction term of FXRi,t and the crisis variable turns out to be positive and statistically significant at the 10 percent level. This suggests that during the GFC, business cycle comovement was more synchronized between the United States (the origin of the GFC) and countries with fixed exchange rates (except for those with currencies are pegged to the U.S. dollar). In other words, countries with fixed exchange rates were more vulnerable to the negative shock during the GFC after controlling for countries with currencies pegged to the U.S. dollar. In contrast, the estimated coefficient of the interaction of countries with currencies pegged to the U.S. dollar and the crisis variable is negative and significant at the 10 percent level in column (2), implying that spillover effects of the GFC negative shock were decreased in the these countries. Hence, the US dollar pegged countries are likely to contaminate the results for the impact of fixed exchange rate regime on the international transmission of negative shock from the United States during the GFC in column (1). We then repeat our analysis using a divided subsample of a country’s degree of financial openness, to examine the role of exchange rate regime in the context of the trilemma. Our theoretical discussion indicates that the channel through which the currency peg affects the business cycle closely relates to monetary policy coordination in the open financial market. Columns (3) and (4) of Table 4 show that the estimated coefficients of the interaction term of the fixed exchange rate regime (FXRi,t) and the GFC are certainly distinct in terms of capital account openness. Its coefficient is insignificant for countries with low capital account openness, as indicated in column (4), while it is significant and positive for countries with high capital account openness, as is in column (3). This finding suggests that significant transmission of the 14

negative shock via fixed exchange rate regime during the GFC in column (2) is driven by a country’s fixed exchange rate regime in tandem with high financial openness. In addition, subsample analysis of countries with low capital account openness in column (4) shows that the coefficient of the interaction terms of U.S. dollar peg and the crisis variable is significant and negative at the 5 percent level. This suggests that countries with currencies pegged to the U.S. dollar and low capital account openness mainly influence the significant coefficient on the interaction term of U.S. dollar peg and the crisis variable in column (2). Moreover, the coefficients of government spending correlation are significant and positive in column (3), but not in column (4), which indicates that higher fiscal spending correlation with the United States led to business cycle synchronization for countries that were exposed to more flexible capital mobility. The combined findings mentioned above can be understood in the context to the trilemma, as fiscal policy coordination is a remaining valid tool in stabilizing business cycles for countries with a fixed exchange rate regime and high capital account openness. In columns (5) and (6), we repeat our sub-sample analysis using GMM by including country-pair fixed effects. The results of the fixed exchange rate regime and business cycle comovement in columns (5) and (6) are consistent with those in columns (3) and (4). This is not surprising as we already included many time-invariant country-pair characteristics as instrument variables to identify the system of equations. In all columns except (4), the estimated coefficients of the crisis indicator are negative and significant, implying that the U.S.-based shock led to business cycle divergence. This reflects the conventional international business cycle model (e.g., Backus et al. 1992), which emphasizes a risk-sharing mechanism in the integrated financial market. We find that the interaction term with 15

equity integration is positive and that with debt integration is negative, respectively, which is consistent with the findings of Pyun and An (2016). Furthermore, our findings on financial integration indirectly validate our identification of the system of equations. In combining the coefficients of financial integration and their interaction terms, we confirm Davis’s (2014) finding that equity market integration leads to business cycle divergence, while debt market integration leads to business cycle synchronization. The coefficients of trade integration are significant and positive in columns (1) and (2), and its interaction terms with the GFC show significant and negative signs. The results suggest that trade integration leads to business cycle convergence in normal times, but that this effect weakened during the GFC. [Insert Table 4]

3.2. Trilemma? Results of Monetary Autonomy Index As our findings are sufficient in terms of an explanation based on the trilemma, but not necessarily trilemma configuration, one may argue whether a loss of monetary independence certainly leads to our results in Table 4. In this section, we implement a simple experiment to determine the role of monetary policy independence in business cycle co-movement during the GFC. According to the trilemma, countries with a fixed exchange rate regime and low capital account openness have relatively more room to implement their monetary policy, enabling them to reduce spillovers of the GFC negative shock. However, countries with a fixed exchange rate regime and high financial openness can be more vulnerable to external shocks due to their relative lack of monetary policy independence. We first select countries that feature a fixed exchange rate regime and divide them into 16

two subsamples according to the degree of financial openness. We then introduce the monetary policy independence index (MI) as per Aizenman, Chinn, and Ito (2010, 2013) to investigate the extent to which monetary policy autonomy leads to the stabilization of business cycle fluctuation rising from the GFC shock. MI is constructed by computing the correlation of a country’s monthly interest rate (i) with the base country’s interest rate, as follows: 𝑀𝐼𝑖,𝑡 = 1 − [{𝑐𝑜𝑟𝑟𝑡 (𝑖𝑖 , 𝑖𝑗 ) − (−1)}⁄2],

(3)

where the subscript i and j refer to the home country and the base country, respectively. By design, the maximum value is 1 (full independence) and the minimum value is 0 (no autonomy). [Insert Table 5] Table 5 reports the results. First, we directly compare the mean of MI for each sub-sample, fixed exchange rate countries with higher financial openness and those with lower financial openness. As expected, the mean of MI for fixed exchange rate countries with higher financial openness (=0.132) is less than that of those with lower financial openness (=0.404). This indicates that monetary policy independence is likely to be forgone for countries that opt for a fixed exchange rate regime and flexible capital mobility. Moreover, we examine the role of monetary policy independence in the international transmission of business cycle for each subsample. Columns (1) and (2) report the 3SLS results, while columns (3) and (4) report the GMM results with country-pair fixed effects. The coefficients on the interaction terms of MI and the crisis variable are significant and negative in all the columns, indicating that an increase in the monetary policy independence of the base country buffers the negative shock from the United States and leads to business cycle de-synchronization. However, interestingly, the absolute magnitude of the coefficients of the interaction terms with MI for the fixed exchange rate 17

countries with higher openness in columns (1) and (3) is smaller than that of the fixed exchange rate countries with lower openness in columns (2) and (4). This implies that the efficacy of monetary policy during the GFC is weaker for the fixed exchange rate countries with higher capital account openness.

3.3. Why Are Fixed Exchange Rate Countries Distinguished From Countries with Currencies Pegged to the U.S. Dollar? Our results indicate contrasting predictions for countries with a fixed exchange rate regime and those with their currencies pegged to the U.S. dollar during the GFC. The countries with a fixed exchange rate regime and high capital openness were vulnerable to the negative shock from the United State, which is consistent with the trilemma. However, countries with currencies pegged to the U.S. dollar (in general, with restrictive capital account openness) were insulated from the shock from the United States. Then, why did only the countries with currencies pegged to the U.S. dollar indicate distinctive features in the international transmission of business cycle among the fixed exchange rate countries? To answer this question, we provide specific characteristics for countries with currencies pegged to the U.S. dollar and refer to previous studies on the GFC. Table 6 compares business cycle co-movement, GDP growth rates, de jure capital openness, and monetary policy independence index (Aizenman, Chinn, and Ito 2010, 2013) between countries with a fixed exchange rate regime (Panel A) and countries with currencies pegged to the U.S. dollar (Panel B) before and after the GFC. In this simple comparison, we find a stark difference between these two groups of countries. First, the countries with currencies pegged to the U.S. dollar are all emerging and developing countries, except China, Hong Kong, 18

the majority of the fixed exchange rate countries are developed countries (Note that Mainland China is not coded as a country with currencies pegged to the U.S. dollar during the period 2007–2008). Business cycle co-movements for all countries with a fixed exchange rate regime except Malta increased during the GFC. However, business cycle co-movements for countries with currencies pegged to the U.S. dollar indicate different patterns to some extent, that is, only half of the countries had increased business cycle co-movement with the United States during the GFC. GDP growth rates differ between the two groups both before and during the GFC. During the GFC, most of the countries with fixed exchange rate regimes indicate negative growth rates, whereas many US dollar pegged countries continued to exhibit positive growth rates. The difference in the level of economic development between the two groups is also related to the differences in capital account openness, which is generally lower among countries with currencies pegged to the U.S. dollar than that among other fixed regime countries. Consistent with the trilemma, countries with currencies pegged to currencies of countries other than the U.S. dollar have a relatively lower monetary policy autonomy index value (MI) than the countries with currencies pegged to the U.S. dollar. In Panel A, the Eurozone countries exhibit zero degree of monetary autonomy, and their MI values remained unchanged before and during the GFC. However, the countries with currencies pegged to the U.S. dollar show relatively higher MI values than those in Panel A, and indicate even higher monetary independence during the GFC than before it. Table 6 suggests that capital account openness is important in explaining why countries with currencies pegged to the U.S. dollar were less vulnerable to the shocks of the GFC vis-à-vis other fixed regime countries! [Insert Table 6] Furthermore, previous studies maintain that the GFC mostly damaged the developed 19

countries (Imbs, 2010; Lane, 2013). First, Imbs (2010) and Kalemli-Ozcan, Papaioannou, and Perri (2013) suggest that among the advanced countries, financial linkage was more conducive to the transmission of the shock, because the role of multinational banks is more developed in those countries. Lane (2013) argues that while the developed countries experienced negative valuation effects during the GFC owing to their net foreign asset positions, emerging and developed economies benefitted from holding liquid and safe foreign debt assets. Thus, for developed countries, the nature of the GFC exacerbated the negative consequences, but countries with currencies pegged to the U.S. dollar, which largely comprised of emerging and developing countries, were relatively immune to the transmission of the shock of the GFC. During financial crisis, a huge depreciation of domestic currency in a country that shorts foreign currency is likely to induce a capital loss in the international balance sheet (i.e., an adverse balance sheet effect of currency exposure). However, Fratzscher (2009) reports that during the GFC, the U.S. dollar appreciated against all other currencies. Moreover, the repatriation of capital to the United States by U.S. investors—a flight-to-safety phenomenon seen among both U.S. and non-U.S. investors—and an increased need for U.S. dollar liquidity may have played a role in the sharp appreciation trend of the U.S. dollar and deterred any adverse capital reversal from countries with currencies pegged to the U.S. dollar. Thus, in addition to the restrictive capital controls of the countries with currencies pegged to the U.S. dollar, they were insulated from the negative valuation effect and rendered less vulnerable to the negative shock of the GFC. 13

13

Previous study such as Bénétrix et al. (2015) emphasizes that the shock of the GFC became more severe through the valuation effect of foreign currency against the U.S. dollar. The currency induced valuation losses were more severe for countries with a shortage of U.S. dollar and yen during the GFC.

20

4. Robustness Tests 4.1. Alternative Measures for Crisis, Exchange Rate Regime, and Business Cycle Comovement To check the robustness of the results, we first repeat our main regressions in Table 4 by replacing our crisis index with an alternative crisis indicator, VIX, as a proxy for the GFC. Columns (1)–(3) of Table 7 presents the results derived from using the alternative crisis variables. These results are consistent with those in Table 4. Additionally, columns (4)–(6) use different classifications of exchange rate regime as per Ilzetzki, Reinhart, and Rogoff (2017). We choose de facto peg (1 and 2 from their ‘coarse’ classification) as fixed exchange rate. These results are also consistent with our main results in Table 4. Finally, we introduce an alternative measure of co-movement, as per Kalemli-Ozcan, Papaioannou, and Peydro (2013). First, we regress GDP growth on the country fixed effect and time (in years) fixed effect for all countries i, as follows:

ln Yt i  ln Yt i1  i  t  vti , where the residual ( vti ) represents the extent to which output (Y) growth (of country i) deviates from average growth over the estimation. We then construct the business cycle synchronization proxy as the negative absolute value of difference in residuals: SYNC1i ,US ,t   vti  vtUS . This index measures how similarly the output growth rates of two

countries move in any given year. Columns (7)–(8) of Table 5 indicate that the results derived from using SYNC1 are qualitatively consistent with our baseline results from using SYNC. In summary, Table 7 confirms that fixed exchange rate countries with high capital account openness were more vulnerable to the negative shock of the GFC originating in the United States and their business cycles became more synchronized with that of the United States during the GFC. 21

[Insert Table 7]

4.2. Alternative Model: Spillover Effect of Shocks on GDP Growth Rates To make the results more robust in our main model, as well as to consider the possibility of delayed transmission of shocks, we introduce an alternative specification using high frequency data. This is in line with the methods pertaining to cross-country analysis of the spillover of the shock of the GFC used in previous studies (e.g., Lane and Milesi-Ferretti, 2011; Rose and Spiegel, 2010, 2012). Using quarterly data, we set up year-on-year real GDP growth rate (∆𝑦𝑖,𝑡 ) as a new dependent variable 14 . Then, by controlling for other explanatory variables, we reexamine the role of exchange rate regimes in the international transmission of the negative shock of the GFC to other countries. The alternative model specification is as follows: 12

12

12

12

∆𝑦𝑖,𝑡 = ∑ 𝛽1 (𝑠)𝐶𝑈𝑆,𝑡−𝑠 + ∑ 𝛽2 (𝑠)𝐹𝑋𝑅𝑖,𝑡−𝑠 + ∑ 𝛽3 (𝑠)𝑈𝑆𝑃𝐸𝐺𝑖,𝑡−𝑠 + ∑ 𝛽4 (𝑠)𝐶𝑈𝑆,𝑡−𝑠 𝐹𝑋𝑅𝑖,𝑈𝑆,𝑡−𝑠 𝑠=0

𝑠=0

𝑠=0 12

𝑠=0 12

+ ∑ 𝛽5 (𝑠)𝐶𝑈𝑆,𝑡−𝑠 𝑈𝑆𝑃𝐸𝐺𝑖,𝑈𝑆,𝑡−𝑠 + ∑ 𝛽6 (𝑠)𝑊𝑖,𝑈𝑆,𝑡 + 𝐼𝑖 + 𝑑𝑡 + 𝜀𝑖,𝑡 , (4) 𝑠=0

𝑠=0

where 𝐶𝑈𝑆,𝑡 is a crisis indicator. To leverage the benefit of the quarterly data, we prefer to use a continuous crisis measure VIX, which is available at a quarterly frequency, instead of an annual crisis dummy variable. The caveat is that the VIX variable may capture not only the GFC, but also other external global shocks during the sample period. FXRit is a dummy variable that indicates a country with a fixed exchange rate regime except the ones with currencies pegged to the U.S. dollar. USPEGi,t is a time-variant binary variable for countries with currencies pegged to

14

The dependent variable, GDP growth rate, is reported quarterly, but international linkages are constructed on a yearly basis. One-year-lagged trade and financial linkages are used repeatedly for all quarters in the corresponding period.

22

the U.S. dollar. Wi,US,t contains TIi,US,t, FIEQi,US,t, and FIDBi,US,t and their interaction terms with the crisis variable. These are the same variables used in Equation (1). We include the lagged values of FXRi,t, USPEGi,t, Wi,US,t, and their interaction terms and determine their cumulative effects on a country’s GDP growth rates. The number of lagged variables is set to equal twelve (three years). To exploit the advantage inherent to a panel regression, we include the country fixed effect (𝐼𝑖 ), which captures time-invariant unobservable country characteristics such as distance from the United States, border effect, language, and legal and institutional factors. Thus, the control variables used in the system of equations model are absorbed by the country fixed effects. Further, 𝑑𝑡 denotes the time fixed effect. Table 8 presents the sum of the estimated coefficients, which indicates the cumulative effects of the shock of the GFC on output growth via international linkages, and their joint Fstatistics are reported. Columns (1)–(3) of Table 8 correspond to columns (2)–(4) of Table 4. Overall, the results in Table 8 are consistent with those in Table 4. As our dependent variable is GDP growth rate, the negative signs of the interaction terms of a fixed exchange rate regime and the crisis variable indicate that during the GFC, the U.S.-based negative shock had greater spillover effects on the output growth of fixed exchange rate countries. This corresponds to the positive coefficient on the interaction term in Table 4 using SYNC as the dependent variable. In columns (1) and (2), the sum of the estimated coefficients of the interaction term of a fixed exchange rate regime and the crisis variable are negative. This implies that a fixed exchange rate regime amplified the negative shock from the United States. The negative cumulative effect on GDP growth rate is even more pronounced for the countries with higher capital account openness, indicated in column (2), which is consistent with the trilemma as is in 23

our main result. Interestingly, this transmission channel disappears for those with low capital openness in column (3). However, the sum of the estimated coefficients of the interaction terms of the GFC and U.S. dollar pegged status differ from that of a fixed exchange rate regime. The growth rates of countries with currencies pegged to the U.S. dollar during the GFC was greater than average, which implies that the spillover of the negative shock of the GFC was reduced among these countries. This seems to be driven by those with low capital account openness (column (3)). However, no such buffers are found in those with high capital account openness (see column (2) in Table 8). [Insert Table 8] To understand the results in Table 8 more clearly, Figure 1 depicts the impulse responses of the shock of the GFC based on the estimated coefficients of the interaction terms in Table 8. The horizontal axis indicates the quarters after the shock first hit the United States, while the vertical axis indicates the GDP growth rates, which measure growth spillover of the GFC shock. We plot the cumulative impact on GDP growth rates for the corresponding quarters. The black line indicates the results with the full sample. This is divided into countries with high (red line) and low (blue line) capital account openness. The dashed black lines denote 95 percent confidence intervals. Figure 1.A depicts that fixed exchange rate countries are more vulnerable to negative shocks of the GFC, but this growth rate response for the countries look very different across countries with high and low levels of capital account openness. For the countries with high financial openness, the results are similar to those of the full sample, which indicates a greater spillover of negative shocks from the U.S. On the other hand, for the countries with low financial openness, the cumulative effect on GDP growth rates remained positive since the shock first 24

arrived. In Figure 1.B, the cumulative impact on GDP growth for the countries with currencies pegged to the U.S. dollar and low capital account openness ultimately went above zero in the twelfth quarter following the onset of the shock, which implies that there is a time lag during which monetary policy “kicks in” on the real side of the economy. However, the growth rate of the countries with currencies pegged to the U.S. dollar and high capital openness remained negative over the whole period. [Insert Figure 1]

5. Concluding Remarks Previous studies have discussed whether a choice of exchange rate regime influences business cycles and their international transmission, but no conclusive results have been found yet. To deepen our understanding of exchange rate regimes and international business cycle comovement, this study introduces a country’s capital account openness, which closely links exchange rate regime with monetary autonomy, in the context of trilemma. In particular, during a financial crisis, retaining monetary policy independence may be important in alleviating the international transmission of business cycles, so our study provides full insights on the outcome of a country’s choice of exchange rate regime and capital controls in response to external shocks. This study finds the critical role of a country’s capital account openness in shaping the relationship between exchange rate regime and the international transmission of business cycle from the United States to other countries during the global financial crisis (GFC). We find that during the GFC, countries with a fixed exchange rate regime and high capital account openness (monetary autonomy is forgone in the context of the trilemma) were more vulnerable to the negative shock from the United States. Further, their business cycles became more synchronized 25

with that of the United States. However, the countries with currencies pegged to the U.S. dollar were found to be less affected by the shock of the GFC due to their relatively low capital account openness (thereby high monetary autonomy) and the U.S. dollar’s special position during the GFC. Our results are robust when controlling for fiscal policy and possible international linkages through which negative shocks were transmitted during the GFC.

References An, J., Kim, K., and Pyun, J.H., 2017. Does Debt Market Integration Amplify the International Transmission of Business Cycle During Financial Crisis? Unpublished. Ahmed, S., Ickes, B.W., Wang, P., and Yoo, B.S., 1993. International Business Cycles. Am. Econ. Rev. 83, 335-359. Aizenman, J., Chinn, M.D., and Ito, H., 2010. The Emerging Global Financial Architecture: Tracing and Evaluating the New Patterns of the Trilemma’s Configurations. J. Int. Money Finance. 29(4), 615-641. DOI: https://doi.org/10.1016/j.jimonfin.2010.01.005 Aizenman, J., Chinn, M.D., and Ito, H., 2013. The ‘Impossible Trinity’ Hypothesis in an Era of Global Imbalances: Measurement and Testing. Rev. Int. Econ. 21(3), 447-458. DOI: https://doi.org/10.1111/roie.12047 Artis, M.J. and Zhang, W., 1997. International Business Cycles and the ERM: Is There a European Business Cycle? Int. J. Finance Econ. 2, 1-16. DOI: https://doi.org/10.1002/(SICI)1099-1158(199701)2:1<1::AID-IJFE31>3.0.CO;2-7 Artis, M.J., and Zhang, W., 1999. Further Evidence on the International Business Cycle and the ERM: Is there a European Business Cycle? Oxf. Econ. Pap. 51, 120-132. DOI: https://doi.org/10.1093/oep/51.1.120 Backus, David K., Kehoe, Patrick J., and Kydland, Finn E. 1992. International real business cycles. J. Political Econ. 100, no. 4: 745–75. Baxter, M. and Kouparitsas, M.A., 2005. Determinants of Business Cycle Comovement: A Robust Analysis. J. Monetary Econ. 52, 113-157. DOI: https://doi.org/10.3386/w10725 Baxter, M. and Stockman, A.C., 1989. Business Cycles and the Exchange-Rate Regime: Some 26

International Evidence. J. Monetary Econ. 23(3), 377-400. DOI: https://doi.org/10.3386/w2689 Bénétrix, A.S., Lane, P.R., and Shambaugh, J.C., 2015. International Currency Exposures, Valuation Effects and the Global Financial Crisis. J. Int. Econ. 96(1), S98-S109. DOI: https://doi.org/10.1016/j.jinteco.2014.11.002 Bluedorn, J.C. and Bowdler, C., 2010. The Empirics of International Monetary Transmission: Exchange Rate Regimes and Interest Rate Pass-through. J. Money Credit Bank. 42(4), 679713. Calderon, C., Chong, A., and Stein, E., 2007. Trade Intensity and Business Cycle Synchronization: Are Developing Countries any Different? J. Int. Econ. 71(1), 2-21. DOI: https://doi.org/10.1016/j.jinteco.2006.06.001 Chinn, M.D. and Ito, H., 2008. A New Measure of Financial Openness. J. Comp. Policy Anal. 10, 309-322. http://web.pdx.edu/~ito/Chinn-Ito_website.htm. DOI Choudhri, E.U. and Kochin, L.A., 1980. The Exchange Rate and the International Transmission of Business Cycle Disturbances: Some Evidence from the Great Depression. J. Money Credit Bank. 12(4), 565-574. Clark, T.E. and van Wincoop, E., 2001. Borders and Business Cycles. J. Int. Econ. 55(1), 59-85. DOI: https://doi.org/10.1016/S0022-1996(01)00095-2 Davis, J.S., 2014. Financial Integration and International Business Cycle Co-movement. J. Monetary Econ. 64, 99-111. DOI: https://doi.org/10.1016/j.jmoneco.2014.01.007 Dées, S. and Zorell, N., 2012. Business Cycle Synchronisation: Disentangling Trade and Financial Linkages. Open Econ. Rev. 23, 623-643. Fernández, A., Klein, M.W., Rebucci, A., Schindler, M., and Uribe, M., 2016. Capital Control Measures: A New Dataset, IMF Econ. Rev. 64, 2016, 548-574. Fratzscher, M., 2009. What Explains Global Exchange Rate Movements During the Financial Crisis? J. Int. Money Finance. 28(8), 1390-1407. Gerlach, H.S., 1988. World Business Cycles under Fixed and Flexible Exchange Rates. J. Money Credit Bank. 621-632. Hoffmann, M., 2007. Fixed Versus Flexible Exchange Rates: Evidence from Developing Countries. Econ. 74(295), 425-449. DOI: https://doi.org/10.1111/j.1468-0335.2006.00564.x 27

Ilzetzki, E., Reinhart, C., and Rogoff, K., 2017. Exchange Rate Arrangements Entering the 21st Century: Which Anchor will Hold? NBER Working Paper No. 23134. DOI: https://doi.org/10.3386/w23134 Imbs, J., 2004. Trade, Finance, Specialization, and Synchronization. Rev. Econ. Statistics 86, 723-734. DOI: https://doi.org/10.1162/0034653041811707 Imbs, J., 2006. The Real Effects of Financial Integration. J. Int. Econ. 68(2), 296-324. DOI: https://doi.org/10.1016/j.jinteco.2005.05.003 Imbs, J., 2010. The First Global Recession in Decades. IMF Econ. Rev. 58(2), 327-335. https://www.imf.org/external/np/res/seminars/2010/paris/pdf/imbs.pdf Kalemli-Ozcan, S., Sørensen, B.E., and Yosha, O., 2003. Risk Sharing and Industrial Specialization: Regional and International Evidence. Am. Econ. Rev. 93(3), 903-918. DOI: https://doi.org/10.1257/000282803322157151 Kalemli-Ozcan, S., Papaioannou, E., and Perri, F., 2013. Global Banks and Crisis Transmission. J. Int. Econ. 89, 495-510. DOI: https://doi.org/10.1016/j.jinteco.2012.07.001 Kalemli-Ozcan, S., Papaioannou, E., and Peydro, J.L., 2013. Financial Regulation, Financial Globalization, and the Synchronization of Economic Activity. J. Finance. 68, 1179-1228. DOI: https://doi.org/10.1111/jofi.12025 Klein, M.W. and Shambaugh, J.C., 2006. Fixed Exchange Rates and Trade. J. Int. Econ. 70(2), 359-383. DOI: https://doi.org/10.1016/j.jinteco.2006.01.001 Klein, M.W. and Shambaugh, J.C., 2015. Rounding the Corners of the Policy Trilemma: Sources of Monetary Policy Autonomy. Am. Econ. J.: Macroecon. 7, 33-66. DOI: https://doi.org/10.3386/w19461 Lane, P.R. and Milesi-Ferretti, G.M., 2011. The Cross-country Incidence of the Global Crisis. IMF Econ. Rev. 59(1), 77-110. https://www.imf.org/external/pubs/ft/wp/2010/wp10171.pdf Lane, P.R., 2013. Financial Globalization and the Crisis. Open Econ. Rev. 24, 555–580. Mathy, G.P. and Meissner, C.M., 2011. Business Cycle Co-movement: Evidence from the Great Depression. J. Monetary Econ. 58(4), 362-372. DOI: https://doi.org/10.1016/j.jmoneco.2011.07.004 Obstfeld, M., Shambaugh, J.C., and Taylor, A.M., 2005. The Trilemma in History: Tradeoffs among Exchange Rates, Monetary Policies, and Capital Mobility. Rev. Econ. Statistics. 87(3), 28

423-438. DOI: https://doi.org/10.3386/w10396 Pyun, J.-H. and An, J., 2016. Capital and Credit Market Integration and Real Economic Contagion during the Global Financial Crisis. J. Int. Money Finance. 67,172-193. DOI: https://doi.org/10.1016/j.jimonfin.2016.04.004 Rey, H., 2015. Dilemma Not Trilemma: The Global Financial Cycle and Monetary Policy Independence. NBER Working Paper No. 21162. DOI: https://doi.org/10.3386/w21162 Rose, A.K., 2014. Surprising Similarities: Recent Monetary Regimes of Small Economies. J. Int. Money Finance. 49, 5-27. DOI: https://doi.org/10.3386/w19632 Rose, A.K. and Spiegel, M.M., 2010. Cross-country Causes and Consequences of the 2008 Crisis: International Linkages and American Exposure. Pac. Econ. Rev. 15, 340-363. DOI: https://doi.org/10.3386/w15358 Rose, A. and Spiegel, M.M., 2012. The Causes and Consequences of the 2008 Crisis: Early Warning. Jpn. World Econ. 24, 1-16. Shambaugh, J.C., 2004. The Effect of Fixed Exchange Rates on Monetary Policy. Q. J. Econ. 119(1), 301-352. Wooldridge, J.M., 2010. Econometric Analysis of Cross Section and Panel Data, MIT Press, Place.

29

Figure 1. Cumulative Impulse Response Function (IRF) from Table 8 A. Fixed Exchange Rate Regime

B. US Dollar Peggers

8

4

6

2 0

2

%

%

4 0 1 2 3 4 5 6 7 8 9 10 11 12 13

-2

0 0 1 2 3 4 5 6 7 8 9 10 11 12

-2 -4

-4 -6

Quarter

Quarter

C. Equity Market Integration

D. Debt Market Integration

1

2 1.5

0.5

%

%

1 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13

0.5 0

-0.5

-0.5 -1

-1

Quarter

0 1 2 3 4 5 6 7 8 9 10 11 12 13 Quarter

E. Trade Linkage 1.5 1

%

0.5 0 -0.5

0 1 2 3 4 5 6 7 8 9 10 11 12 13

-1 -1.5

Quarter

Note: The black solid line represents the cumulative effect of the global financial crisis originated from the United States on individual country’s GDP growth rates through each linkage estimated based on the full sample. The dashed lines are 95-percent confidence intervals. Red line means the cumulative effect of GFC on GDP growth rates of countries with high capital account openness. Blue line indicates that of countries with low capital account openness.

30

Table 1: Country List of Fixed and Floating Exchange Rate Regime Years of fixed exchange rate regime Fixed exchange rate regime countries 2001, 2006~2007 Argentina* 2001~2013 Austria 2001~2013 Belgium 2004~2006, 2009~2013 Bolivia* 2001~2002, 2004~2013 Bulgaria 2001~2006, 2009~2010, 2012~2013 China, P.R.: Mainland* 2001~2013 China, P.R.: Hong Kong* 2011~2013 Costa Rica* 2001~2013 Cyprus 2003 Czech Republic 2001~2013 Denmark 2002, 2004, 2006~2007, 2009, 2011 Egypt* 2001~2013 Finland 2001~2013 France 2001~2013 Greece 2001~2002 India* 2001~2013 Ireland 2001~2013 Italy 2001~2002, 2008, 2010~2013 Kazakhstan* 2001~2006, 2010~2013 Kuwait* 2005~2007, 2010~2013 Latvia 2001~2013 Lebanon* 2001~2005 Malaysia* 2001, 2003~2013 Malta 2007 Mexico* 2001~2013 Netherlands 2011 Norway 2002~2003, 2005~2007, 2010 Pakistan* 2001~2013 Panama* 2004 Philippines* 2001~2013 Portugal 2005 Russian Federation* 2006, 2009, 2011 Singapore 2005~2013 Slovenia 2001~2013 Spain 2003~2004 Sweden 2002, 2005~2006, 2009, 2012~2013 Switzerland 2013 (coded as a single year peg) United Kingdom 2006 Uruguay* 2006~2009, 2012 Venezuela, RB* Floating exchange rate regime countries Australia Brazil Canada Chile Colombia Germany Hungary Iceland Indonesia Japan Korea, Republic of Mauritius New Zealand Poland South Africa Thailand Turkey Note: * denotes US dollar pegged countries. Lithuania was the US peg country in 2001(†) and fixed exchange rate regime in the rest of the sample period. Source: Shambaugh (2004), and Klein and Shambaugh (2006).

31

Table 2. Basic Statistics N

Mean

SD

Min

Max

p1

p5

p25

p50

p75

p95

p99

SYNCi,US

829

-2.503

2.131

-9.960

-0.002

-8.931

-6.841

-3.690

-1.912

-0.830

-0.140

-0.033

SYNC1i,US

847

-2.432

2.368

-15.672

-0.002

-12.281

-7.021

-3.248

-1.771

-0.705

-0.163

-0.068

Δyi,t

2661

2.804

3.673

-17.64

19.13

-8.7

-3.69

0.99

3.00

5.06

8.1

10.62

FIEQi,US

781

0.527

0.980

0.000

7.310

0.000

0.000

0.010

0.097

0.541

2.517

5.039

FIDBi,US

780

0.413

0.872

0.000

5.514

0.000

0.001

0.015

0.068

0.301

2.439

4.597

TIi,US

857

0.242

0.538

0.000

3.629

0.001

0.002

0.012

0.064

0.223

1.121

3.292

Si,US

814

0.280

0.192

0.008

0.970

0.014

0.048

0.130

0.238

0.380

0.675

0.774

FiscalSYNCi,US

817

-3.456

4.083

-47.228

-0.0004

-16.726

-10.801

-4.441

-2.303

-0.986

-0.182

-0.037

Note: SYNCi,US is the absolute value of real GDP growth differences between country i and the United States. SYNC1i,US is the negative of the absolute differences of residual GDP growth. Δyi,t is a GDP growth rate (percent). FIEQi,US and FIDBi,US are the capital and credit market integration measures, respectively, calculated as the sum of assets and liabilities divided by the sum of the current GDP of the United States and partner country. TIi,US is the trade integration measure, or the sum of exports and imports divided by the sum of the current GDP of the United States and partner country. These are quantity-based measures. Si,US is the measure for similarity in production structure between country i and the United States. FiscalSYNCi,US is the measure for fiscal policy correlation between country i and the United States. Mean, SD, Min, and Max are the mean, standard deviation, minimum value and maximum value of country–year observations, respectively.

32

Table 3. Country List of Capital Account Openness (Chinn and Ito Index) Panel A. Counties with High Financial Openness Kaopen Normalized Kaopen Kaopen Normalized Kaopen (mean) (mean) (mean before 2008) (mean before 2008) Austria 2.421764 1 2.421764 1 Canada 2.421764 1 2.421764 1 Denmark 2.421764 1 2.421764 1 Estonia 2.421764 1 2.421764 1 Finland 2.421764 1 2.421764 1 France 2.421764 1 2.421764 1 Germany 2.421764 1 2.421764 1 Hong Kong SAR, China 2.421764 1 2.421764 1 Ireland 2.421764 1 2.421764 1 Italy 2.421764 1 2.421764 1 Japan 2.421764 1 2.421764 1 Lithuania 2.094048 0.9237298 2.421764 1 Luxembourg ----Netherlands 2.421764 1 2.421764 1 New Zealand 2.421764 1 2.421764 1 Norway 2.421764 1 2.421764 1 Panama 2.421764 1 2.421764 1 Portugal 2.421764 1 2.421764 1 Singapore 2.421764 1 2.421764 1 Spain 2.421764 1 2.421764 1 Sweden 2.421764 1 2.421764 1 Switzerland 2.421764 1 2.421764 1 United Kingdom 2.421764 1 2.421764 1 Uruguay 2.421764 1 2.421764 1 Latvia 2.378069 0.9898307 2.346858 0.9825668 Belgium 2.356221 0.984746 2.309404 0.9738503 Greece 2.333621 0.9794862 2.270661 0.9648334 Egypt, Arab Rep. 2.012792 0.9048192 2.246306 0.9591653 Czech Republic 2.290678 0.969492 2.197044 0.9477005 Mauritius 1.998586 0.9015128 2.070848 0.9183307 Chile 1.87557 0.8728831 2.047231 0.9128341 Hungary 2.203286 0.9491532 2.047231 0.9128341 Israel 2.137743 0.9338992 1.934871 0.8866844 Slovenia 1.657092 0.8220363 1.672697 0.8256682 Bolivia 0.9270861 0.6521406 1.326588 0.7451176 Costa Rica 1.39826 0.7617979 1.19356 0.7141575 Note: Kaopen denotes the capital account openness index from Chinn and Ito (2008). Countries above the average of capital account openness are listed. Country

33

Panel B. Counties with Low Financial Openness (continued) Kaopen Normalized Kaopen Kaopen Normalized Kaopen (mean) (mean) (mean before 2008) (mean before 2008) Australia 1.132745 0.7000041 1.110897 0.6949194 Iceland 0.1584278 0.4732493 1.110897 0.6949194 Indonesia 0.9061967 0.647279 1.110897 0.6949194 Kuwait 1.110897 0.6949194 1.110897 0.6949194 Mongolia 1.205283 0.7168859 1.110897 0.6949194 Romania 1.618028 0.8129448 1.04393 0.6793339 Lebanon 1.022754 0.6744056 0.9597942 0.6597528 Mexico 0.9346103 0.6538917 0.808691 0.6245863 Cyprus 1.28224 0.7347963 0.6568509 0.5892482 Malta 1.215944 0.7193672 0.3546445 0.5189152 Philippines -0.2538763 0.3772929 0.0531748 0.4487535 Bulgaria 1.032339 0.6766363 0.0398924 0.4456623 Venezuela, RB -0.7374324 0.2647539 -0.037227 0.4277141 Slovak Republic 0.5039075 0.5536534 -0.0730207 0.4193838 Brazil 0.0130271 0.4394099 -0.0811437 0.4174933 Korea, Rep. 0.2104095 0.4853471 -0.1173072 0.4090769 Malaysia -0.2793875 0.3713557 -0.1173072 0.4090769 Poland -0.0491755 0.4249333 -0.122283 0.4079189 Thailand -0.5580249 0.3065078 -0.2684104 0.3739104 Russian Federation -0.0453058 0.4258339 -0.3684096 0.3506374 Colombia -0.2793875 0.3713557 -0.5706168 0.3035773 Argentina -0.7900679 0.252504 -0.619512 0.2921978 Barbados -1.17503 0.1629111 -1.17503 0.1629111 China -1.17503 0.1629111 -1.17503 0.1629111 India -1.17503 0.1629111 -1.17503 0.1629111 Kazakhstan -1.17503 0.1629111 -1.17503 0.1629111 Pakistan -1.17503 0.1629111 -1.17503 0.1629111 South Africa -1.17503 0.1629111 -1.17503 0.1629111 Turkey -0.6632779 0.2820121 -1.17503 0.1629111 Bahamas, The -1.875024 0 -1.875024 0 Note: Kaopen denotes the capital account openness index from Chinn and Ito (2008). Countries below the average of capital account openness are listed. Country

34

Table 4. Main Results Dependent Variable: Sample

(1) Full Sample

FXRi,t × CUS,t USPEGi,t × CUS,t FIEQi,US,t × CUS,t FIDBi,US,t × CUS,t TIi,US,t × CUS,t

CUS,t FXRi,t USPEGit FIEQi,US,t FIDBi,US,t TIi,US,t Si,US,t Fiscal SYNCi,US,t Observations

(5) High Openness

(6) Low Openness

GMM w/ country-pair fixed effects

3SLS

Methods FXi,t × CUS,t

SYNCi,US,t (3) (4) High Low Openness Openness

(2) Full Sample

0.217 (0.472) 0.940* (0.541) -1.426* (0.732) 4.746* 5.443** (2.482) (2.512) -3.789 -4.697* (2.390) (2.420) -1.707* -1.500* (0.873) (0.900) -1.640*** -1.641*** (0.443) (0.444) -0.309 0.00396 (0.213) (0.266) -0.599** (0.305) -3.958*** -3.701*** (1.022) (1.012) 4.670*** 4.642*** (1.176) (1.170) 1.165*** 1.099*** (0.345) (0.349) -5.655*** -3.749*** (1.078) (1.190) 0.0758*** 0.0770*** (0.0273) (0.0275) 641 641

1.183** (0.538) -1.336 (0.928) 4.787*** (1.761) -3.873** (1.641) -0.148 (1.200) -1.769*** (0.479) 0.0139 (0.331) -0.527 (0.367) -0.896 (0.753) 1.074 (0.728) 0.372 (0.702) -4.409** (2.134) 0.0746** (0.0339) 368

-0.215 (1.153) -2.178** (1.030) -0.324 (3.592) -1.043 (4.268) -1.018 (0.977) -0.512 (0.680) -0.117 (0.505) -0.650 (0.404) -0.996 (1.301) 3.332* (1.916) 0.448 (0.345) -5.595*** (1.303) 0.0345 (0.0414) 273

0.881* (0.519) -1.441 (1.090) 2.232 (2.233) -1.858 (1.948) 0.971 (1.395) -1.341*** (0.517) 0.310 (0.452) -0.550 (0.800) -0.545 (0.795) -0.675 (1.156) -1.349 (5.373) -4.723 (11.60) 0.0496 (0.0344) 368

0.617 (1.376) -3.782** (1.657) 8.628* (4.694) -13.20** (5.640) 1.364 (1.018) -2.004*** (0.721) -1.585 (1.077) 1.000** (0.510) -11.15*** (2.884) 14.24*** (3.939) 8.048*** (2.137) -17.96** (8.601) 0.0728 (0.0445) 273

Note: Three-stage least squares (3SLS) and two-step GMM are implemented respectively. For the GMM results, robust standard errors are reported in parentheses. *, **, and *** indicate significance at the 10, 5, and 1-percent levels, respectively. SYNC, FIEQ, FIDB, TI, and S are measures for business cycle co-movement, equity market integration, debt market integration, trade integration, and production similarity, respectively, between the United States and individual countries. FXRi,t is a time-variant binary variable that indicates a country’s choice of fixed exchange rate (or currency peg) except for U.S. dollar peg. USPEGi,t is a time-variant binary variable that indicates when a country pegs its currency to the U.S. dollar only. 𝐶𝑈𝑆,𝑡 is a dummy variable coded as 1 if the year is 2008 or 2009. FiscalSYNCi,US is the measure for fiscal policy correlation between country i and the United States. The results of the first equation of the system are reported; those of the other equations are available from the authors upon request. A constant term, advanced country dummy and year fixed effects are included, but not reported.

35

Table 5. Trilemma with Monetary Independence Index Dependent variable: Sample Methods Mean of MI index MIit × CUS,t FIEQi,US,t × CUS,t FIDBi,US,t × CUS,t TIi,US,t × CUS,t

CUS,t MIit FIEQi,US,t FIDBi,US,t TIi,US,t Si,US,t Fiscal SYNCi,US,t Observations

SYNCi,US,t (1) (2) Fixed Regime & Fixed Regime & High Financial Low Financial Openness Openness 3SLS

(3) (4) Fixed Regime & Fixed Regime & High Financial Low Financial Openness Openness GMM w/ country-pair FEs

0.132

0.403

0.132

0.404

-5.436** (2.687) -2.265 (2.999) 2.515 (3.172) 5.929 (3.879) -1.454* (0.860) -2.934*** (0.762) 0.586 (0.724) -0.939 (0.738) 3.437** (1.447) 2.613** (1.214) 0.0476 (0.0358) 206

-14.87** (6.754) 22.26* (12.41) 110.1*** (34.32) 20.09** (9.576) -3.773** (1.730) -1.093 (2.245) -6.246 (4.653) 8.469 (8.595) -0.830 (1.103) -2.913** (1.404) -0.0686 (0.0594) 69

-6.650* (3.452) -6.472 (5.621) 4.777 (5.769) 9.238* (4.780) -1.007 (1.178) -3.160** (1.331) -0.195 (1.071) -0.0602 (2.297) -15.46 (10.00) 10.11 (7.278) 0.0290 (0.0495) 185

-8.276** (3.308) 14.87 (10.05) 160.4*** (27.84) -4.700 (3.141) -5.480*** (0.974) 3.153 (1.964) -12.72*** (4.876) 24.55** (11.60) 4.803** (2.341) -18.39* (10.12) -0.00511 (0.0419) 74

Note: Three-stage least squares (3SLS) and two-step GMM are implemented respectively. For the GMM results, robust standard errors are reported in parentheses. *, **, and *** indicate significance at the 10, 5, and 1-percent levels, respectively. SYNC, FIEQ, FIDB, TI, and S are measures for business cycle co-movement, equity market integration, debt market integration, trade integration, and production similarity, respectively, between the United States and individual countries. FXRi,t is a time-variant binary variable that indicates a country’s choice of fixed exchange rate (or currency peg) except for U.S. dollar peg. USPEGi,t is a time-variant binary variable that indicates when a country pegs its currency to the U.S. dollar only. 𝐶𝑈𝑆,𝑡 is a dummy variable coded as 1 if the year is 2008 or 2009. FiscalSYNCi,US is the measure for fiscal policy correlation between country i and the United States. The monetary independence index (MI) is constructed by the correlation of a country’s monthly interest rates (i) with the base country’s interest rate (Aizenman, Chinn, and Ito 2010, 2013). 𝑀𝐼𝑖,𝑡 = 1 − [{𝑐𝑜𝑟𝑟𝑡 (𝑖𝑖 , 𝑖𝑗 ) − (−1)}⁄2], where the subscript i refers to the home country and j to the base country. The results of the first equation of the system are reported; those of the other equations are available from the authors upon request. A constant term, advanced country dummy and year fixed effects are included, but not reported.

36

Table 6. Fixed exchange rate countries versus US dollar pegged countries Panel A. Fixed regime countries (except for US dollar pegged countries) Capital account Monetary openness Independence Country 2001~ 2008~ 2001~ 2008~ 2001~ 2008~ 2001~ 2008~ 2007 2009 2007 2009 2007 2009 2007 2009 Austria 0.59 0.66 2.23 -1.13 2.42 2.42 0 0 Belgium 0.71 0.86 2.05 -0.84 2.31 2.42 0 0 Bulgaria 0.12 0.88 5.70 0.37 0.20 2.42 0.33 0.39 Cyprus 0.16 0.16 3.89 0.79 0.66 2.42 0.32 0.01 Denmark 0.72 0.95 1.63 -2.91 2.42 2.42 0.05 0.29 Estonia 0.03 0.94 7.72 -10.04 2.42 2.42 0.24 0.55 Finland 0.60 0.83 3.17 -3.78 2.42 2.42 0 0 France 0.61 0.92 1.86 -1.37 2.42 2.42 0 0 Greece -0.14 0.85 4.11 -2.42 2.27 2.42 0.04 0 Ireland -0.56 0.94 4.96 -4.49 2.42 2.42 0 0 Italy 0.09 0.88 1.17 -3.27 2.42 2.42 0 0 Lithuania -0.31 0.93 9.25 -14.81 2.42 1.90 0.21 0.17 Malta 0.33 0.05 1.38 0.55 0.61 2.42 0.26 0.02 Netherlands 0.56 0.55 2.00 -0.61 2.42 2.42 0 0 Portugal 0.37 0.62 1.20 -1.39 2.42 2.42 0 0 Singapore 0.85 0.92 8.86 -0.60 2.42 2.42 0.29 0.58 Slovak Republic -0.83 0.88 5.41 -5.29 0.05 1.36 0.46 0.01 Slovenia 0.26 0.57 5.24 -2.25 2.03 2.03 0.49 0 Spain 0.39 0.93 3.56 -1.23 2.42 2.42 0 0 Switzerland 0.57 0.94 2.40 -2.13 2.42 2.42 0.18 0.48 Note: Czech Republic, Latvia, Luxembourg, and Sweden were dropped due to missing values or the adoption of a different exchange rate regime after 2008. GDP Correlation

GDP growth rates

Panel B. US dollar pegged countries Capital account Monetary openness Independence Country 2001~ 2008~ 2001~ 2008~ 2001~ 2008~ 2001~ 2008~ 2007 2009 2007 2009 2007 2009 2007 2009 Bahamas, The 0.24 0.89 1.76 -3.25 -1.88 -1.88 0.52 0.52 Barbados 0.92 0.91 1.84 -1.90 -1.18 -1.18 0.21 0.36 Bolivia 0.73 0.22 4.46 3.36 1.36 0.58 0.60 0.48 China: Hong Kong 0.58 0.99 4.98 -0.16 2.42 2.42 0.22 0.36 China: Mainland 0.01 0.73 10.26 9.23 -1.18 -1.18 0.42 0.37 Egypt 0.31 0.23 5.10 4.69 2.42 2.16 0.34 0.59 Kazakhstan -0.77 0.95 11.65 3.30 -1.18 -1.18 0.53 0.64 Lebanon -0.05 -0.75 4.53 9.70 0.96 1.11 0.42 0.41 Panama 0.57 -0.04 6.05 6.56 2.42 2.42 0.10 0.29 Venezuela 0.61 0.90 9.31 1.04 -1.09 -1.48 0.52 0.77 Note: Costa Rica, India, Kuwait, Lithuania, Malaysia, Mexico, Mongolia, Pakistan, Philippines, Russia, and Uruguay were dropped due to missing values or the adoption of a different exchange rate regime after 2008. GDP Correlation

GDP growth rates

37

Table 7. Robustness Checks (GMM w/ country pair fixed effects) Dependent Variable:

SYNCi,US,t

SYNC1i,US,t

(1) (2) (3) (4) (5) (6) (7) (8) (9) Full High Low Full High Low Full High Low Sample Sample Openness Openness Sample Openness Openness Sample Openness Openness Alternative GFC variable, VIX index de facto Exchange Rate Regime Alternative dependent variable 0.0696 0.464** -0.262 1.728** 1.517* 1.280 1.382* 1.142* 1.074 FXRi,t × CUS,t (0.210) (0.214) (0.541) (0.798) (0.776) (1.459) (0.789) (0.589) (1.252) -0.112 0.614 0.299 -0.850 -1.166 -1.691 -0.781 -0.862 -2.267 USPEGi,t × CUS,t (0.394) (0.505) (0.619) (0.803) (0.970) (1.314) (1.117) (1.029) (1.992) 1.588* -0.666 2.419 6.770** 3.884 11.42* 7.921** 0.671 10.02** FIEQi,US,t × CUS,t (0.953) (0.816) (2.367) (3.026) (2.388) (5.834) (3.297) (2.276) (4.653) -2.412** 0.115 -5.473* -6.211** -3.653* -17.43*** -7.099** -0.490 -13.51** FIDBi,US,t × CUS,t (1.063) (0.856) (3.224) (2.845) (2.213) (6.718) (3.080) (1.929) (5.406) -0.540 0.722 -0.630 -1.028 1.257 1.918* -1.281 0.671 1.018 TIi,US,t × CUS,t (0.392) (0.996) (0.733) (1.019) (1.374) (1.063) (1.152) (1.383) (1.015) -0.243 -0.409* -0.207 -1.805*** -1.676** -2.280** -1.638*** -1.220** -2.044*** CUS,t (0.179) (0.226) (0.336) (0.639) (0.799) (0.893) (0.474) (0.485) (0.666) 1.058 0.505 -1.088 1.617*** 1.149** 2.106*** 1.082 0.831 -2.572** FXRi,t (0.767) (0.614) (1.265) (0.490) (0.513) (0.547) (0.802) (0.570) (1.251) 0.268 -0.449 0.565 -0.875** 0.818 -1.418*** 0.532 0.0680 1.352*** USPEGi,t (0.389) (0.854) (0.558) (0.393) (1.500) (0.450) (0.446) (0.627) (0.512) -4.801*** -2.360** -11.75*** -1.904* -0.356 -11.76*** -2.149** -0.608 -9.811*** FIEQi,US,t (1.445) (1.187) (3.051) (1.003) (0.753) (3.268) (1.072) (0.791) (2.620) 4.386*** 0.698 15.51*** 1.379 -0.201 13.76*** 1.400 -0.0297 13.17*** FIDBi,US,t (1.688) (1.452) (4.513) (1.242) (1.263) (4.265) (1.328) (1.158) (3.424) 1.580 -10.23** 8.727*** 1.878 -0.397 7.098*** 1.363 -3.061 4.631** TIi,US,t (1.869) (5.067) (2.475) (1.781) (5.131) (2.295) (2.227) (5.552) (2.212) -9.715 -0.517 -9.191 -5.838 1.953 -5.779 -13.87* 0.597 -21.28*** Si,US,t (6.912) (10.90) (9.608) (7.646) (11.36) (8.652) (7.980) (10.93) (8.045) 0.0201 0.0227 0.0868* 0.0481 0.0399 0.0581 0.0526 0.0510 0.0412 Fiscal SYNCi,US,t (0.0312) (0.0405) (0.0495) (0.0310) (0.0350) (0.0441) (0.0414) (0.0554) (0.0513) 641 368 273 641 368 273 648 370 278 Observations Note: Two-step GMM results with country pair fixed effects are reported. Robust standard errors are reported in parentheses. *, **, and *** indicate significance at the 10, 5, and 1-percent levels, respectively. In columns (1)-(3), CUS,t is a continuous VIX index. Columns (4)-(6) use the de facto measure of Ilzetzki, Reinhart, and Rogoff (2017). In columns (7)-(9), we include alternative SYNC variable for the robustness check. SYNC, FIEQ, FIDB, TI, S and Fiscal SYNC are measures for business cycle co-movement, equity market integration, debt market integration, trade integration, production similarity and government spending correlation, respectively, between the United States and individual countries. A constant term, advanced country dummy and year fixed effects are included, but not reported.

38

Table 8. Robustness with Quarterly GDP Growth Rates: Including Time Lags Real GDP growth rate (∆𝑦𝑖,𝑡 , quarterly)

Dependent variable:

(1) Full Sample -2.031**

(2) High Capital Openness -2.519***

(3) Low Capital Openness 5.290

(2.19)

(3.27)

(1.53)

0.228***

-3.583***

1.956***

(11.96)

(5.55)

(54.81)

-0.297

0.144*

0.079

(0.97)

(1.82)

(1.68)

0.713**

-0.201

1.024***

(2.19)

(1.36)

(5.08)

-0.155***

0.731**

0.432***

(2.76)

(2.03)

(9.75)

4.546***

6.631***

15.543**

(8.58)

(3.21)

(2.71)

6.225*

8.544*

-18.714***

(1.76)

(1.79)

(3.71)

-1.428***

27.615***

-8.171***

(10.68)

(12.30)

(65.12)

0.797

-0.600***

-0.526*

(1.22)

(4.77)

(2.03)

-2.098***

0.863*

-2.453**

(2.93)

(1.77)

(2.56)

-0.533

-0.771***

-2.759***

(1.32)

(3.16)

(7.19)

Country Fixed Effects

Yes

Yes

Yes

Time Fixed Effects (quarter)

Yes

Yes

Yes

Observations

1,943

1,202

741

R-square

0.30

0.19

0.27

FXRi,t-s × CUS,t-s (s=0,…,12) USPEGi,t-s × CUS,t-s (s=0,…,12) FIEQi,US,t-s × CUS,t-s (s=0,…,12) FIDBi,US,t-s × CUS,t-s (s=0,…,12) TIi,US,t-s × CUS,t-s (s=0,…,12)

CUS,t-s (s=0,…,12) FXRi,t-s (s=0,…,12) USPEGi,t-s (s=0,…,12) FIEQi,US,t-s (s=0,…,12) FIDBi,US,t-s (s=0,…,12) TIi,US,t-s (s=0,…,12)

Note: The results show the sum of the estimated coefficients of all lagged variables. We include contemporaneous and lagged values of all explanatory variables up to three years (The number of lagged variables is set to equal twelve). F-statistics are reported in parentheses. *, **, and *** indicate significance at the 10, 5, and 1-percent levels, respectively. ∆𝑦𝑖,𝑡 is a dependent variable, and is year-on-year GDP growth rates (percent) for country i at t. FXRi,t-s is a time-variant binary variable that indicates a country’s choice of fixed exchange rate (or currency peg) except for U.S. dollar peg. USPEGi,t-s is a time-variant binary variable that indicates when a country pegs its currency to the U.S. dollar only. CUS,t-s is quarterly frequency VIX index. FIEQ, FIDB, and TI are measures for equity market integration, debt market integration, and trade integration, respectively, between the United States and individual countries.

39