Direct and Spillover Effects of Unconventional Monetary and Exchange Rate Policies

Joseph E. Gagnon; Tamim Bayoumi; Juan M. Londono; Christian Saborowski; Horacio Sapriza

doi:10.5089/9781475586237.001.A001

Jump to Content Jump to Main Navigation

Direct and Spillover Effects of Unconventional Monetary and Exchange Rate Policies

Author: Mr. Joseph E. Gagnon¹, Mr. Tamim Bayoumi, Juan M. Londono², Christian Saborowski³, and Horacio Sapriza⁴

¹ 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund
| ² 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund
| ³ 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund
| ⁴ 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

Author’s E-Mail Address: CSabowroski@imf.org

Publication Date:: 13 Mar 2017

eISBN:: 9781475586237

Language:: English

Keywords:: WP; financial asset; capital mobility; quantitative easing; monetary policy announcement; QE measure; countervailing QE

Download PDF (1.7 MB)

This paper explores the effects of unconventional monetary and exchange rate policies. We find that official foreign asset purchases have large effects on current accounts that diminish as capital mobility rises and spill over to financially integrated countries. There is an additional effect through the stock of central bank assets. Domestic asset purchases have an effect on current accounts only when capital mobility is low. We also find that rising US bond yields drive foreign yields, stock prices and depreciations, but less so on days of policy announcements. We develop a theoretical model that is broadly consistent with our results.

Abstract

I. Introduction

Unconventional monetary policy has dominated economic headlines since the Global Financial Crisis of 2008–09. At that time, conventional policy interest rates in major advanced economies approached the zero lower bound, and central banks in these economies searched for alternative ways of shoring up aggregate demand. Policymakers in some emerging markets argued that unconventional policies were causing capital to flow out of advanced economies engaged in unconventional policies and into emerging markets, creating lending booms, pushing up the foreign exchange value of currencies, and threatening to dampen exports. Emerging markets themselves have been key players in the expansion of a better-known form of unconventional policy focused on exchange rates. Led by several Asian economies, a number of countries purchased unprecedented volumes of foreign exchange reserves in the run-up to and during the Global Financial Crisis, and some set up sovereign wealth funds (SWFs), with both policy actions resulting in official capital outflows.

It bears mentioning that asset purchases related to either unconventional exchange rate or monetary policies are not unprecedented or even particularly unusual. What makes these episodes unconventional is the magnitude of the resulting increases in official balance sheets. Moreover, these policy actions have generally not been associated with movements in short-term interest rates. In other words, any macroeconomic impact associated with these policy actions would have to rely on the imperfect substitutability between assets purchased and liabilities issued under the policies.² In that sense, unconventional policies are different from conventional monetary policy, which operates mainly through movements in short-term risk-free interest rates.

Our paper studies the domestic and cross-border spillover effects of unconventional monetary and exchange rate policies. We first explore empirically the impact on the current accounts of the countries engaged in these policies as well as the potential spillovers to current accounts in other countries using annual data. As unconventional policies are a relatively recent phenomenon in the advanced economies, in the second part of our study we complement the evidence from the annual regressions using daily data to explore spillovers from US economic and monetary policy announcements to international financial assets. To rationalize our findings from the annual and daily regressions, in the third part of the paper we develop a macroeconomic model with imperfect asset substitution. The model relates financial asset prices to current account balances, therefore linking the evidence from the daily regressions with that from the annual regressions. Moreover, the model includes channels through which unconventional monetary and exchange rate policies can affect income at home and abroad, helping us to make some qualitative assessments of these broader effects.

In our first empirical contribution, we document the effects of unconventional exchange rates and monetary policies on all countries’ current accounts. To do so, we build on the cross-country annual regressions of Bayoumi, Gagnon, and Saborowski (2015), who in turn built on the seminal work of Chinn and Prasad (2003) and of many others. We proxy unconventional exchange rate policies with net official financial flows (NOF), which are dominated by purchases of foreign exchange reserves and SWF assets. An important contribution of this paper is to improve the set of instruments to control for endogeneity of NOF to exchange rate pressures. We find that NOF has a large and robust positive effect on a country’s current account when capital mobility is low and a moderate effect when capital mobility is high. We also find a large effect of fiscal policy on the current account when capital mobility is high, a new result in this literature. We proxy unconventional monetary policies with central bank acquisitions of domestic assets. For the major advanced economies, this proxy corresponds well with the unconventional policy known as quantitative easing (QE). Interestingly, our dataset includes other episodes of significant increases and decreases in central bank domestic assets in emerging economies. To the extent that these episodes reflect central banks taking risk onto their balance sheet and off private balance sheets, they should be counted as QE even if the specific types of assets that were purchased differ from those in the recent cases of QE in major advanced economies. We find that QE has a small but significant effect on a country’s current account balance when capital mobility is low, but this effect declines to near zero when capital mobility is high. We also find that the spillovers of unconventional policies show up in current account balances of all other countries, with relatively stronger effects in countries that are more integrated with global financial markets.

In our second empirical contribution, we explore the effects of unconventional policies in a higher frequency setting with a focus on announcement effects in the United States. This alternative setting has the advantage that it is not subject to two important concerns with the earlier analysis, namely that there are few observations in annual data and that QE is only a subset of unconventional monetary policy. We find that there are strong spillovers of US bond yields to prices of foreign financial assets immediately following US economic announcements, consistent with the view that good news about future US economic activity is good news for other countries. However, when we focus on changes in US bond yields associated with revisions to monetary policy, the spillovers are much smaller. Smaller spillovers of pure monetary shocks presumably reflect the absence of good news about US economic activity, and thus higher bond yields on these days are not good news for foreign economies. We note, in particular, that monetary shocks have only a small effect on foreign currencies, which is the main channel by which unconventional policy in the United States could reduce current account balances abroad.

We also develop a stylized macroeconomic model with imperfect asset substitution to understand our empirical results. Imperfect asset substitution is key to the model because unconventional monetary policy relies importantly on imperfect substitutability across assets and because observed correlations of asset prices across countries are difficult to explain in standard models with perfect substitution. Our model supports the finding that NOF, our proxy for unconventional exchange rate policy, has a larger and more consistent effect on the current account than QE. Our model also implies that stronger US activity unambiguously raises foreign activity and foreign current account balances, while pure monetary shocks have small and ambiguous effects on current account balances and activity in foreign economies, in line with the evidence from the daily regressions. Imperfect asset substitution creates additional tools that central banks can use to offset unwanted influences from foreign shocks and policies. In addition to their conventional interest rate tool, central banks can use NOF and QE to achieve objectives for the exchange rate, current account, or bond yields.

The remainder of the paper is organized as follows. Section 2 presents the empirical analysis on the effect of unconventional policies on the current account, Section 3 discusses the effect of US UMP on financial markets using daily data, Section 4 provides a stylized model that helps rationalize and relate the empirical findings of the previous sections, and Section 5 concludes.

II. Current Account Effects of Net Official Flows and Quantitative Easing in Annual Panel Data

In this section we document the impact of unconventional exchange rates and monetary policies on all countries’ current accounts. In doing so, we extend the analysis in Bayoumi, Gagnon, and Saborowski (2015) in three ways. First, we use a new set of instruments to identify exogenous variation in NOF in an effort to estimate the link between NOF and the current account. Our results confirm the important effect of official flows on the current account, including the interaction of this effect with capital mobility. Second, we explore the impact of QE on the current account, a potential channel for such policy to affect other countries. Finally, we analyze how policies in each country affect the current accounts of other countries.

A. Empirical Specification

We run our regressions for a sample period of up to 30 years, from 1985 through 2014. The regressions involve up to 2088 observations for 141 countries, including low-income economies. The coefficient standard errors in all regressions, including those using instruments, are robust to heteroskedastic and first-order autoregressive errors. Further information on the data can be found in the Appendix. The following equations present our baseline specifications:

\begin{array}{l} \begin{matrix} \frac{CA Z_{it}}{{GDP}_{it}} = α_{1} (\frac{{NOF}_{it}}{{GDP}_{it}}) + α_{2} (\frac{{NOF}_{it} \times {MOB}_{it - 1}}{{GDP}_{it}}) + β_{1} (\frac{{NOA}_{it - 1}}{{GDP}_{it - 1}}) + β_{2} (\frac{{NOF}_{it - 1} \times {MOB}_{it - 1}}{{GDP}_{it - 1}}) + γ_{1} \frac{{QE}_{it - 1}}{{GDP}_{it - 1}} + \\ γ_{2} (\frac{{QE}_{it - 1} \times {MOB}_{it - 1}}{{GDP}_{it - 1}}) + δ_{1} {SPILL}_{it} + μ_{1} {AUX}_{it} + μ_{2} ({AUX}_{it} \times {MOB}_{it - 1}) + θ_{t} {year}_{t} + u_{i t}; \end{matrix} \\ (\begin{matrix} 2.1 \end{matrix}) \end{array}

\begin{array}{l} \begin{matrix} \frac{{NPFX}_{it}}{{GDP}_{it}} = (α_{1} - 1) (\frac{{NOF}_{it}}{{GDP}_{it}}) + α_{2} (\frac{{NOF}_{it} \times {MOB}_{it - 1}}{{GDP}_{it}}) + β_{1} (\frac{{NOA}_{it - 1}}{{GDP}_{it - 1}}) + β_{2} (\frac{{NOF}_{it - 1} \times {MOB}_{it - 1}}{{GDP}_{it - 1}}) + \\ γ_{1} \frac{{QE}_{it - 1}}{{GDP}_{it - 1}} + γ_{2} (\frac{{QE}_{it - 1} \times {MOB}_{it - 1}}{{GDP}_{it - 1}}) + δ_{1} {SPILL}_{it} + μ_{1} {AUX}_{it} + μ_{2} ({AUX}_{it} \times {MOB}_{it - 1}) + θ_{t} {year}_{t} \end{matrix} \\ \begin{matrix} \begin{matrix} + v_{i t}, \end{matrix} & (2.2) \end{matrix} \end{array}

where CAX and NPFX are the current account and the net private flows, respectively, each excluding net investment income to remove the influence of steady-state differences in stocks of net foreign assets.³ NOF is net official flows, NOA is the stock of net official assets, and MOB is the Aizenman, Chinn, and Ito (2015) measure of legal restrictions on capital mobility, normalized to [0-1], for which a higher value indicates fewer restrictions on private capital flows. ⁴ We define NOF as the acquisition and disposition of assets and liabilities denominated in foreign currency by public-sector institutions in the reporting country.⁵ The auxiliary variables (AUX) include MOB, lagged PPP GDP per capita relative to the United States, the 10-year forward change in the old-age dependency ratio, the lagged real GDP growth rate over the previous 5 years, net energy exports relative to GDP, and the cyclically adjusted fiscal balance relative to GDP.⁶

Equation 2.1 presents the current account as a function of NOF and several control variables. The coefficient α₁ represents the effect of NOF on the current account and the coefficient α₂ allows for a differential effect depending on the level of capital mobility. The coefficient β₁ represents the effect of lagged NOA on the current account while the coefficient β₂ allows for a differential effect with higher capital mobility.

Equation 2.2 is a restatement of the link between the official flows and the current account in Equation 2.1 that takes advantage of the balance of payments (BOP) identity: any effect of NOF on the current account that is less than 1 must show up as a negative effect on net private flows. When NOF has no effect on the current account (CAX) (i.e., α1=0), it must cause a one-for-one reduction of NPFX. Because of errors and omissions in the BOP data, these regressions are not identical. The bias from measurement error in NOF in the estimate of α1 is downward in Equation 2.1 and upward in Equation 2.2, which helps to put a range on its true value, and makes the average of the coefficient estimates from the two equations a convenient statistic to reduce measurement error.

The specifications in Equations 2.1 and 2.2 further include interactions between all our auxiliary variables and the MOB term. We include the additional interaction terms because the effect of some of the auxiliary variables could vary with capital mobility. For example, demographic or fiscal influences on domestic interest rates will lead to higher capital flows, and thus larger current account imbalances, when capital markets are less restricted. Finally, we include our QE measure and a measure of spillovers of aggregate global net official outflows on countries' current accounts, SPILL.

A key empirical issue is the potential endogeneity of official flows to shocks to current account balances and net private flows. On the one hand, endogenous movements are most likely to arise from attempts to stabilize the exchange rate in the face of trade or financial market shocks. On the other hand, examples of exogenous movements in official flows include increasing holdings of foreign assets for precautionary reasons, saving resource revenues for future generations, borrowing for economic development, and achieving economic growth through higher net exports. Gagnon (2012, 2013) shows that endogeneity through stabilization of the exchange rate leads to a positive bias of the coefficient on NOF if current account shocks dominate, and a negative bias if private financial shocks dominate.

We use instrumental variables to address the potential endogeneity of NOF to shocks to current account balances and net private flows. The challenge is to isolate the variation in NOF that is not driven by shocks that simultaneously drive the current account and/or private financial flows. Valid instruments must reflect exogenous motives for reserve accumulation.

A key difference between our regressions and those in Bayoumi, Gagnon, and Saborowski (2015) is the choice of instruments. In this paper, we use two instruments that we believe improve upon those chosen in the previous paper.⁷ The first instrument is the incidence of a financial or currency crisis in the previous three years and the second is the portion of NOF that is not related to foreign exchange reserves. The idea behind the former is to capture a higher propensity to build up reserves for precautionary reasons following a crisis episode. The latter captures SWF-related asset flows as well as development loans. We would expect both SWF flows and development loans to reflect longer-term savings and investment motives and, conditional on the control variables in the second stage, generally not to respond systematically to exchange rate shocks.⁸

In principle, we would like to analyze the impact of all unconventional monetary policies on current accounts. However, unconventional policies include a range of actions, such as guidance over future interest rates, lending to distressed banks and firms, purchases of long-term bonds and other assets, and negative policy interest rates. For tractability, we limit our analysis to QE, which is defined as an increase in central bank domestic assets. This definition measures the extent to which the central bank expands its balance sheet to take risk off the balance sheets of domestic market participants, thus potentially easing financing conditions. Indeed, as shown in Figure 2.1, the measure looks broadly as expected in known QE cases following the Global Financial Crisis. Interestingly, in other countries, large variations in central bank domestic assets often occur during or after financial crises, consistent with the theory of how QE works. We show in section 2.C that our results derive from imperfect substitution across assets and not from the effect of unconventional policies on the overall stance of monetary policy or macroeconomic conditions. We examine important episodes of QE in emerging economies in section 2.D.

The effect of QE on the current account could be either positive or negative because a monetary expansion would be expected to both boost domestic absorption and depreciate the exchange rate. The former increases imports whereas the latter restrains imports and boosts exports. Section 4 presents a model that includes these competing effects.

Importantly, there is a potential source of endogeneity bias when analyzing the impact of QE on the current account. On the one hand, the coefficient on the QE term could be biased down if countries use monetary expansions to combat declines in growth associated with declining current account balances. On the other hand, there may be a positive bias from omitted variables if a (negative) shock to domestic demand causes an expansion of QE at the same time that imports decline and cause the current account balance to increase. To attenuate these biases, we cyclically adjust the QE variable and use the lagged value in the regression.⁹

Finally, we ask how unconventional policy impacts other countries’ current account balances. In principle, any effect of QE or NOF on the current account of the country purchasing the assets must have an equal and opposite effect on the current account of the rest of the world. The magnitude and allocation of the spillovers across countries are one focus of this section. The inclusion of a complete set of time effects in our regressions controls for spillovers of each country’s policies on the assumption that they are distributed equally to all countries as shares of GDP. However, the framework allows us to test alternative assumptions about spillovers. In particular, we test whether spillovers are related to (1) financial integration based on cross-border financial transactions, (2) capital mobility (MOB), (3) financial development based on Sahay et al. (2015), (4) reserve currency shares from the IMF’s COFER database, (5) economic size based on nominal GDP, or (6) a country’s stage of economic development based on PPP GDP per capita. To conduct these tests, we calculate six spillover terms that allocate aggregate global NOF divided by world GDP to countries based on each relevant hypothesis. We do not include QE in our spillover term because it has only a small estimated effect on the current account. We compare the improvement of fit associated with each of these terms to understand which of the hypotheses is the most closely reflected in the data.

In practice, only one of these terms has a statistically significant coefficient. The SPILL term we use in our final regressions is calculated as a measure of financial integration multiplied by aggregate global NOF divided by world GDP. Financial integration is measured as the share of gross private financial transactions in total current and financial transactions in the balance of payments.¹⁰ The motivation for this variable is that NOF from current account surplus countries tend to go to countries that are more financially integrated, who then run current account deficits.

B. Baseline Regression Results

This section describes the results when estimating Equations 2.1 and 2.2 using two-stage least squares. Columns 1 and 2 in tables 2.1a and 2.1b illustrate the results from estimating Equations 2.1 and 2.2 in the absence of the interaction terms with auxiliary variables, QE, and SPILL. Column 1 is based on a regression of the current account excluding investment income (Equation 2.1) and column 2 is based on a regression of net private flows (Equation 2.2). In essence, columns 1 and 2 are equivalent to the model in Bayoumi, Gagnon, and Saborowski (2015). However, in this paper we use a new set of instruments, and we allow for a full interaction between net official flows and the capital mobility index, instead of an interaction with a [0,1] dummy variable. Also, the sample is 2.6 times bigger, adding four years and many low-income countries.

The first-stage results indicate that our instruments are relevant. In each regression, one of the two non-reserves flow instruments is significant. The two crisis instruments are sometimes individually and always jointly significant. The instruments show the expected positive signs in the first-stage regression, and the F-test statistic takes values significantly larger than 10; the null hypothesis that our instruments are irrelevant is thus being rejected in the data. Similarly, the Angrist-Pischke first-stage chi-squared statistic rejects the null that the net official flows term is unidentified. At the same time, the values of the first-stage R²s are around 0.6 and thus do not signal overfitting.

Moving to the second-stage estimation results, we observe that the results for NOF and NOA are remarkably similar to those in Bayoumi, Gagnon, and Saborowski (2015) (columns 1 and 2 of Table 2a in that paper). The coefficients on the auxiliary variables do not differ much from that paper except that relative GDP now carries the expected negative coefficient. The estimated effect of NOF on the current account when capital mobility is lowest (α1) is 0.7 in column 1 and 0.8 in column 2. The overall effect of NOF when mobility is highest is the sum of α1 and α2, or 0.06 in column 1 and 0.52 in column 2. The effect of NOA on the current account is somewhat smaller than what was found in Bayoumi et. al.: it is close to zero under low mobility and rises to around 3 percent under high mobility.

Columns 3 and 4 of Table 1 add interaction terms between all control variables and capital mobility (MOB). The coefficients on NOF and lagged NOA are broadly unaffected by the inclusion of the additional regressors. Many of the auxiliary variables seem to be importantly conditioned by capital mobility, and usually in a sensible way. For example, trend growth and the fiscal balance variable have a larger effect when capital is more mobile, while net energy exports have a larger effect when private capital is less mobile (perhaps reflecting government control of oil export revenues). The coefficient of 0.6 on the fiscal balance when capital is highly mobile is an important finding, as previous studies often struggle to explain coefficients that seem low compared to theoretical predictions.

The final two columns present our full baseline specification. Compared to columns 3 and 4, we now add our measure of QE, the interaction of this variable and capital mobility, and the spillover term. The coefficients on NOF and NOA once again do not change much.

The results in Table 1 illustrate that the QE variable is significant with a coefficient close to 0.25. Its interaction with capital mobility shows a marginally significant negative coefficient of about the same magnitude. This result suggests that QE has a small but significant effect on the current account balance when capital mobility is low, but an effect close to zero as capital mobility approaches its upper bound. The finding is in line with a casual glance at current account balance movements in known episodes of QE in advanced economies (Figure 2.2), in which current accounts did not show a systematic tendency to rise following QE. The finding that QE has a larger effect when capital is less mobile is somewhat surprising. Below, we further examine this result, looking both at robustness checks and the underlying story for influential observations.

Finally, the median value of our financial integration measure is 0.1, so a coefficient of about -20 on SPILL implies that moving from the median to twice the median value of financial integration lowers a country’s current account by twice the value of world NOF divided by world GDP, or around 2 percent of GDP.¹¹

C. Robustness Checks

Table 2.2 presents a number of robustness checks for our baseline regression results. For brevity, the table omits the coefficient estimates for auxiliary variables. The table also shows average coefficients and standard errors between the current account and net private flow regressions.¹² The first column is the baseline result, which is the average of the last two columns in Table 2.1.

The second column is the OLS version of this regression. All coefficients are similar to those in the first column. The bias in the NOF coefficients can go in either direction when NOF is responding endogenously to exchange rate pressures depending on whether the pressures arise from trade or financial shocks. Our results suggest that shocks in both directions are roughly equally important, with biases that cancel out.

The third column replaces the instruments with a full set of country dummies after dropping the few countries that had less than 5 observations with available data. To the extent that current account shocks are more persistent than financial shocks, the country dummies may not fully control for endogeneity from current account shocks, giving rise to an upward bias on the NOF coefficient. Following this reasoning, one would expect the estimated effect of NOF on the current account to be larger. A glance at the results shows that indeed this set of instruments somewhat increases the NOF coefficient. However, it also increases the magnitude of the coefficient on the interaction term, the average effect being only slightly larger for high values of MOB.

The fourth column is a regression weighted by each country’s share of world nominal GDP (so the weights sum to 1 in a given year). This regression is particularly informative about the impact of recent QE by major central banks, given that the euro area, Japan, the United Kingdom, and the United States account for more than half of global GDP. The results are broadly unchanged. In particular, the coefficient on our measure of QE and its interaction with MOB are very similar to the baseline case and imply little or no impact of QE on the current account when capital is highly mobile, as is the case for the major economies.¹³ The only major change in this regression is that the spillover effect is no longer significant. This result may suggest that our spillover term is not doing a good job at explaining spillovers from global NOF to the United States and to some other major economies.¹⁴

The fifth column is based on a regression that aims to reduce the influence of outliers in the data (using the ‘robust’ command in Stata). It uses fitted values of the instrumented variables from separate first-stage regressions. The NOF coefficients are, however, little changed, although the QE and SPILL coefficients are somewhat smaller and less significant.

The sixth and seventh columns explore the importance of the exchange rate regime.¹⁵ The real exchange rate is a key element of the transmission from underlying shocks to the current account. The most obvious effect of exchange rate flexibility is on the speed of adjustment, which should not matter for our estimates due to their long-run nature. However, as discussed in Section 4, the adjustment process under fixed exchange rates may involve different movements in interest rates and other macro variables than under flexible rates, possibly implying a different long-run effect of underlying factors. Although the coefficients are not identical across fixed and flexible rate regimes, the differences are rather small considering the extreme nature of the test.

As discussed above, the QE coefficient is subject to a negative bias if policymakers use monetary policy to offset shocks from net exports and a positive bias if policymakers use monetary policy to offset shocks from domestic demand (which also affect net exports). We expect the bias to be larger for flexible exchange rate regimes than for fixed ones because of the greater ability to use monetary policy in a countercyclical manner. Columns 6 and 7 test this assertion. The QE coefficient is smaller under fixed rates than under flexible rates, and roughly the same under flexible rates as in the baseline regression. This raises the possibility of upward bias. But, even under fixed rates, the QE coefficient is positive, and the difference across regimes is not statistically significant.

The eighth and ninth columns explore the importance of trade openness. Perhaps even more than exchange rate regime, openness to trade might be expected to have an important effect on the coefficients because the trade balances of countries that are more open to trade are more strongly affected by trade-related shocks. Table 2.2 shows that openness has, in fact, only modest effects on the NOF and NOA coefficients. Interestingly, the SPILL coefficient is much larger in open economies than in closed ones, suggesting that most of the spillovers of NOF flow to economies that are open to trade.

Countries with high trade openness are subject to more frequent and larger terms of trade shocks than closed economies. To the extent that countries respond to negative terms of trade shocks by expanding monetary policy, the coefficient on the QE term may be biased down. If this source of bias is important, we may expect that more open economies should have a lower QE coefficient. However, this assertion is not supported by the results presented in columns 8 and 9. Rather, it appears that QE has a larger effect on the current account in countries that are more open to trade. This effect disappears with high capital mobility.

In regressions not shown in Table 2.2, we consider two alternative measures of QE: the current unadjusted and cyclically adjusted changes in central bank domestic assets and the lagged or current changes in the cyclically adjusted monetary base.¹⁶ All alternative measures have smaller (and sometimes slightly negative) coefficients than those in the baseline regressions. We also test for a positive bias from omitted effects of financial crises, which would depress domestic spending and increase the current account at the same time that the central bank might engage in QE. Adding a dummy variable for a current financial crisis has little effect on the QE (or any other) coefficient.

Finally, in unreported results, we test whether the estimated effects of unconventional monetary and exchange rate policy could reflect the overall stance of macroeconomic policy rather than the asset-substitution effects of these policies. To do so, we add alternative measures of the macroeconomic situation in four different regressions. The measures are (1) the growth rate of nominal GDP, (2) the changes in the output gap and GDP deflator, (3) the cyclically adjusted growth rate of the monetary base, and (4) the lagged cyclically adjusted growth rate of the monetary base. None of these alternative regressions has any noticeable effect on the coefficients on NOF, NOA, QE, and SPILL.

D. Influential Observation Analysis

In this section, we focus on specific observations that are influential in identifying the effects of key variables on the current account. The goal is to provide a qualitative assessment of the validity of the underlying theory of how these variables may affect the current account. In particular, we want to confirm that coefficient estimates do not reflect spurious factors or coincidences unrelated to the underlying theory. The analysis also sheds light on the instruments for NOF as measures of exogenous influences on the current account.

We identify influential observations using the “dfbeta” command in Stata, which provides an estimate of the marginal effect of a given observation on a chosen coefficient. The most influential observations are those with the largest marginal effects on (and the same sign as) the estimated coefficient. However, as shown in the robust regression and GDP-weighted regression in Table 2.2, most coefficients are not sensitive to data outliers.

NOF coefficient and associated instruments

Three of the four most influential observations for the NOF coefficient correspond to Azerbaijan in 2008, 2010, and 2011.¹⁷ According to the IMF’s coarse index of exchange rate arrangements, Azerbaijan has a managed exchange rate. It also has a relatively closed capital account. Azerbaijan experienced rapid growth of its net energy exports over the past decade. The government set up a SWF, the State Oil Fund of the Republic of Azerbaijan, in 2001 during the development stage of major new oil fields. As shown in Figure 2.3, the SWF began major financial outflows in 2008, shortly after oil revenues began to grow, which contributed to a notable increase in the current account balance.

Although both reserves and SWF flows appear to have contributed importantly to the current account surplus, they are only weakly correlated. This supports our view that the purposes of these flows are distinct. While SWF flows react to oil revenues and saving-investment objectives, reserves flows are used to manage exchange rate objectives. This is why we exclude reserves flows from our instrument for NOF.

Azerbaijan’s current account is also highly correlated with its net energy exports. It is interesting, in this context, to consider the example of a country with a similar energy export pattern that did not take the decision to save much of the revenues through a SWF. Figure 2.4 displays the current account and financial flows of Nigeria, a country without a SWF during this period.¹⁸ Nigeria’s non-reserves flows are dominated by the repayment of official loans in 2005 and 2006. The Nigerian current account reflects influences of all of these factors. Note, in particular, that the current account declined after 2006 despite a large increase in net energy exports. This comparison supports our regressions, which show that NOF is a more important determinant of the current account than net energy exports.

The second most influential observation is the Republic of Congo 1994. Figure 2.5 shows that non-reserves flows had a noticeable, though not dominant, effect on Congo’s current account in the 1990s. These flows are mainly official development loans. Congo has a fixed exchange rate as part of the CFA franc zone and low capital mobility. The loans have little effect on the nominal exchange rate, but they may affect the real exchange rate to the extent that domestic spending boosts domestic prices. What is more, there is an additional channel to the extent that the government spent the loan proceeds directly on imported materials for development projects. Direct government purchases of imports are similar to a shift in preferences towards imports, which would lower the current account. Given Congo’s exchange rate regime, which is effectively managed by the French finance ministry, it is clear that the non-reserves flows are not responding endogenously to exchange rate pressures.

QE coefficient

The four most important observations (in the positive direction) for the QE coefficient are Malaysia 2008, Angola 2001, and Thailand 1998 and 1999. The QE variable is lagged, so it refers to one year before each of those dates.

In the case of Malaysia, the episode captures a large liquidity injection by the central bank in 2007 amid growing global financial stress. The central bank did not ease conventional monetary policy: it had raised its policy rate by 75 basis points in early 2006; held rates steady in 2007; and began to ease only at the end of 2008 after the collapse of Lehman Brothers.¹⁹ The liquidity injection of 2007 was a response to short-term stress in the banking system and was largely reversed over the course of the following year.

In Angola, the central bank sold off domestic assets to buy foreign assets, with essentially no net change in total central bank assets relative to GDP. The positive QE coefficient suggests that the sale of domestic assets tended to lower the current account. The contemporaneous increase in NOF, in turn, helps explain the increase in the current account despite the decline in QE.²⁰

In Thailand, both years (1997 and 1998) reflect central bank lending to, and recapitalization of, domestic banks during the Asian crisis. This support was important in allowing banks to weather the sudden stop in foreign inflows. The central bank engaged in a quantitative monetary program to ease the effects of changes in the velocity of money. By itself, this policy should not have increased the current account, but it may be a case of bias caused by the same unobserved factor (financial crisis) that both reduced domestic absorption (raising the current account) and caused the central bank to intermediate. However, if the alternative to such easing is fiscal intervention, then there is a case that the quantitative monetary program raised the current account relative to the alternative because it depreciated the exchange rate more. The Thai policy was not associated with any conventional monetary easing, indeed the central bank raised its policy rate 200 basis points to 12.5 percent in 1997 and held it at this level until 1999.

Because some of these observations may raise concerns about a spurious correlation between QE and the current account, we ran a version of our regressions that drops the above four observations as well as the four most important observations in the negative direction.²¹ There was essentially no change in the QE coefficient and it remained statistically significant.

E. Fiscal Balances, Net Official Flows, and Energy Exporters

In the past 15 years, some energy exporting countries have been important contributors to global current account imbalances. A potential question is whether there is a direct effect of energy exports on the non-reserves component of NOF. If non-reserves flows co-move strongly with net energy exports, and if energy exports, in turn, have a large effect on the current account, we would be faced with potential simultaneity bias. In practice, we control for energy exports in the second stage of our annual regressions, implying that non-reserves flows should, conditional on energy exports, still be a valid instrument. To be on the safe side, however, this section analyzes the extent to which energy exports and non-reserves flows co-move in reality. Indeed, we show that there is little direct effect of energy exports on either non-reserves flows or the current account. As a matter of policy, some countries decide to save a large share of energy revenues and others do not. The current account follows net official flows and the fiscal balance, not net energy exports.

Figure 2.6 displays countries where the governments decide to save a large share of net energy exports abroad through net official flows. In two of these countries, Norway and Saudi Arabia, the fiscal balance is almost identical to NOF. In Norway and Saudi Arabia, the current account moves closely with NOF and the fiscal balance, consistent with the sum of the NOF and fiscal coefficients being around 0.8 with high capital mobility and around 0.9 with low mobility.²² In Algeria, which has the lowest capital mobility of these countries, the fiscal balance is noticeably lower than NOF and the current account is closer to NOF than to the fiscal balance, consistent with the larger coefficient on NOF than on fiscal balance when capital mobility is low. The top four rows of Table 2.3 show that NOF, the fiscal balance, and net energy exports (all as shares of GDP) are highly correlated with each other in these countries.

Figure 2.7 displays countries in which governments do not save a large share of net energy exports. These countries have intermediate capital mobility.²³ Table 2.3 shows that the correlation between NOF and the current account is much higher than the correlation between net energy exports and the current account in each of these countries.

On balance, it appears that there is no direct connection between net energy exports and the current account. The strongest connections are between NOF and the current account (when mobility is low) and between the fiscal balance and the current account (when mobility is high). In other words, the government’s choice in using net energy revenues is what matters, not the energy revenues themselves.

F. Dynamics

An issue that is not directly addressed in the regressions displayed in table 2.1 is that of dynamic adjustment or lags. Although the effect of intervention on the exchange rate is expected to be very fast, the effect of the exchange rate on trade and the current account is generally believed to take place gradually over a period of about two years. In our annual data, some of the effect of intervention ought to show up in the same year as the intervention, but some ought to occur in the following year and a small amount might even linger into a third year. The residuals of the regressions in table 2.1 suggest that such dynamics may be important, but they appear to differ across country and across independent variables, which makes it difficult to model them.²⁴ The coefficients are best interpreted as capturing the long-run effect of intervention and other factors, not the immediate effect.

Although oil exports have little long-term effect on the current account, there is a significant temporary effect in many countries, which means that increases in oil revenues boost the current account and SWF flows at the same time. This temporary effect of changes in net oil revenues explains the apparent lack of any lag in the connection between SWF flows and the current account in figure 2.6.

III. Financial Market Effects of US Unconventional Monetary Policy in Daily Data

In the previous section we found that QE, measured by the increase in central bank domestic assets, has a small but significant positive effect on the current account balance when capital is less mobile, and an effect close to zero when capital is highly mobile. In this section, we use high frequency (daily) data to explore further the spillovers of QE and other unconventional monetary policies (UMPs) across markets. This analysis circumvents two limitations of the annual regressions. On the one hand, in advanced economies, UMPs are a relatively recent phenomenon with few observations in annual data. Because GDP and the current account are not observed at a daily frequency, we explore the effects on daily international (non-US) financial asset prices, and, in section 4, we use a theoretical model to link financial and real effects of UMP. On the other hand, in the previous section, we were limited to examining the effects of QE owing to the lack of a good measure of overall UMPs across countries and over time. In this section, our regressions pick up the joint effect of all monetary policy actions. Because conventional monetary policy, embodied in the short-term interest rate, was pinned at the zero lower bound throughout our sample, we infer that our results reflect the effects of UMPs, including, but not limited to, QE. We focus on the cross-border effects of UMP in the United States, as it is the largest economy in the world and its financial markets have an outsized effect on other countries (Bayoumi and Bui 2010, and Bowman, Londono, and Sapriza 2015). Also, the United States was the first country to implement UMP in the context of the Global Financial Crisis.

In the first part of this section, we estimate the effect of daily changes in US long-term (10 year) Treasury bond yields on changes in the prices of financial assets of 40 emerging and advanced economies. The regressions cover the period from November 2008 to July 2015, during which the US Federal Open Market Committee (FOMC) was forced to turn to unconventional policies after its conventional policy instrument reached the zero lower bound. UMP, especially in the United States, primarily operates through yields on long-term bonds (Gagnon, Raskin, Remache, and Sack 2011, and Rogers, Scotti, and Wright 2013, among others). Because the FOMC’s use of UMP is a recent development, daily data provide more observations for analysis than those available in the annual regressions of section 2.

In the second part of this section, we bring the results from the daily regressions closer to the annual regression results of section 2, by exploring the economic determinants of the cross-country variation in the effects of US UMP. In particular, we use a panel-data setting to understand whether each country’s financial depth, capital mobility, sovereign risk, trade linkages with the United States, and exchange rate regime explain the cross-country variation in the baseline daily data results. Our results support the findings of our annual regressions in that the effect of US monetary policy shocks on the US current account is likely to be small because US monetary shocks have only a small effect on exchange rates.

A. Daily Co-Movements, Empirical Specification and Results

Tables 3.1 to 3.3 show the estimated coefficients from individual-country regressions of the daily change in international asset prices on the daily change in the yield of 10-year Treasury bonds. The non-US asset prices, in vector X, are (1) the 10-year sovereign local-currency bond yield, (2) the log of the stock market price index, and (3) the log of the exchange rate (in dollars per unit of non US currency). The benchmark regression setting for country i is the following:

\begin{matrix} Δ X_{i,t} = α_{i} + (β_{1, i} + β_{2, i} D_{Us,t}) Δ Y_{t}^{U S} + u_{i,t}, & (3.1) \end{matrix}

where $Y_{t}^{U S}$ is the yield of 10-year US Treasury bonds and D_US,t is a dummy that takes the value 0 on most days and 1 on days in which the FOMC released either a policy statement or policy minutes or there was a monetary policy speech by the FOMC chair. In separate regressions, D takes the value 1 on days in which data were released on US nonfarm payrolls or the ISM purchasing managers’ index, and 0 otherwise. ²⁵ Thus, the β₁ coefficient in Equation 3.1 measures the average effect of US yields on asset prices on days outside FOMC or economic data events, while the β₂ coefficient reflects the additional response on these special days.

To interpret these coefficients, we note that, on days not associated with news about monetary policy, bond yields tend to rise when there is good news about future US economic activity.²⁶ To be sure, part of the reason for the rise in yields is that market participants expect that faster growth in the US economy will lead the FOMC to tighten policy more than otherwise. But the news element is one of stronger growth and not a change in the FOMC’s policy reaction function. The coefficient β₁ captures the effect of these good news shocks on the rest of the world. On days with an FOMC announcement or FOMC Chair speech, bond yields rise or fall because market participants expect monetary policy to be tighter or looser for a given path of the economy. The sum of the coefficients, β₁ + β₂(FOMC), captures the effect of these monetary shocks on the rest of the world. In reality, there is news about economic prospects, including economic data releases, on FOMC days, so that β₁ + β₂(FOMC) blends the effects of economic news and monetary policy news. But the sign of the β₂(FOMC) coefficient tells us in which direction the effects of a pure monetary shock lie relative to the effects on non-FOMC days.

The evidence in Table 3.1 shows that sovereign yields of almost all advanced economies and of many emerging markets move positively with US sovereign yields; i.e., the estimated β₁ coefficient is positive and significant. However, yields of a few emerging markets, in which default risk is important, are significantly negatively correlated with US yields (significant for Hungary, Russia, and Greece). In these cases, good news about the US economy appears to lower the risk premiums in their sovereign yields. Interestingly, the additional effect of changes in US Treasury yields on FOMC days, β₂(FOMC), usually has the opposite sign as β₁. That is, in most cases, the positive average correlation between non-US yields and US yields is reduced but not eliminated on FOMC days—the estimated β₁+β₂ (FOMC) is closer to zero than the estimated β₁ but remains significant at the same confidence levels for most countries. For countries with an average negative estimated correlation with US Treasury yields, the general negative correlation is reduced or even reversed on FOMC days and becomes not significant at any standard confidence level. Of note, there are interesting variations in the responses of yields in the euro area. In particular, the average response of yields for peripheral euro-area countries is considerably smaller—in some cases not significant—than that of core euro-area countries, while the additional responses on FOMC announcement days, β₂(FOMC), for peripheral countries are smaller and often positive. Thus, tighter US monetary policy appears to increase sovereign risk premiums in these peripheral countries. Overall, our evidence suggests that spillover effects on non-US yields tend to be smaller around monetary policy announcement days, presumably because these changes in US yields are not associated with any boost to demand in foreign economies that would call for higher yields overseas. In contrast, the additional effect of changes in US yields on US economic data release days,β₂(Econ.), although significant for a small number of countries, is generally positive, implying greater co-movement of yields when the underlying news concerns prospects for the US economy.²⁷

The results for stock prices (Table 3.2) are notably uniform across countries. Specifically, stock prices in all countries except Pakistan move positively and significantly with US bond yields. Hence, good news about the US economy boosts international stock prices. The additional effect on FOMC announcement days on non-US stock prices is almost always negative and significant at the 10 percent confidence level for 16 countries. Thus, as for yields, the spillover effects on stock prices are reduced around monetary policy announcement days. For the vast majority of the economies in our sample, the negative additional effect on FOMC announcement days implies a smaller but still positive correlation between changes in non US stock prices and changes in US Treasury yields. The total effect on FOMC announcement days, β₁+β₂ (FOMC), remains significant for 24 countries. Much or all of this remaining positive effect on FOMC days may reflect the impact of economic news on FOMC days that, which is hard to disentangle from our data. The additional effect of changes in US yields on specific US economic data release days is modestly positive but rarely significant and implies no considerable changes in the total effect on economic data release days.

The results for exchange rates (Table 3.3) show that most currencies depreciate with respect to the US dollar when US sovereign bond yields rise. Although the effect of changes in US sovereign bond yields on exchange rates against the US dollar is significant for most currencies, the economic magnitude of the estimated effect is fairly small: a one percentage point increase in US bond yields (which is much larger than a typical daily move) causes a one to three percent depreciation in most currencies against the dollar (which is close to the size of a typical daily move). Not surprisingly, the effect of changes in US bond yields is smaller for currencies pegged to the US dollar, such as the Hong Kong dollar and the Chinese renminbi, and it seems to be positive for traditional funding or safe-haven currencies, such as the Swiss Franc and the Japanese Yen.²⁸ The additional effects of changes in US yields on US FOMC and economic data days are small and rarely significant, but they do reduce the magnitude and significance of the total effect on FOMC and economic data days.

In unreported results, we assess the persistence of the spillover effects by adding lags to the right-hand-side variables in Equation 3.1. Our results show that spillover effects for non-US sovereign yields on regular days, FOMC days, and economic data days are persistent, which suggests that sovereign bonds in most countries tend to co-move with US Treasury yields but less so after monetary policy announcements. Spillover effects for stock prices are also persistent, i.e., stock prices tend to co-move positively with US Treasury yields but the effect is also reduced around monetary policy announcement days, and the overall effect on FOMC announcement days remains positive and significant for almost half the countries in our sample. For exchange rates, adding lags to the effect of FOMC announcements implies a larger positive additional effect on these announcement days. Although the total effect of US yields on exchange rates around FOMC announcement days is still fairly small, it becomes positive for some countries; that is, some currencies appreciate following an increase in US sovereign bond yields on FOMC days. This last result supports the finding that effects of US unconventional monetary policy on exchange rates are small, and thus effects on the US current account are also likely to be small.

We also ran regressions for the period before QE (January 2001 to November 2008 for countries with available data). Many of the results for this sample are similar to those in the baseline daily data regressions. The most notable difference is that the co-movements between non-US asset prices and US Treasury yields on non-announcement days tend to be larger after 2008 than before. In contrast, there is little difference in the co-movements on FOMC or data-release days. This suggests that the impact of US monetary policy on foreign financial conditions is largely unchanged between periods of conventional and unconventional monetary policies, a result consistent with the model presented below.

B. Relating Financial Spillovers to Country Characteristics

In this section, we explore the economic determinants of the cross-country variation in the spillover effects of US QE. Specifically, in Table 3.4, we show the results for the following panel-data regressions:

\begin{matrix} Δ X_{i,t} = α_{i} + (β_{1, 1} + β_{1, 2} * C_{i}) Δ Y_{t}^{U S} + u_{it}, & (3.2) \end{matrix}

\begin{matrix} Δ X_{i,t} = α_{i} + (β_{1, i} + (β_{2, 1} + β_{2, 2} * C_{i}) D_{US,t}) Δ Y_{t}^{U S} + u_{it}, & (3.3) \end{matrix}

where C_i are country-specific variables, each considered in a separate regression to avoid potential collinearity (Hausman and Wongswan, 2011 and Bowman, Londono, and Sapriza, 2015). We consider variables that characterize each country’s degree of capital mobility, depth of the financial system, their exchange rate regime, sovereign risk, and trade integration with the United States.²⁹ In particular, we use the same measure of capital mobility, MOB, that is used in section 2; the ratio of bank assets to GDP to characterize the degree of financial depth; the volatility of the exchange rate with respect to the US dollar to characterize the currency regime; the yield of sovereign bonds to characterize sovereign risk; and the ratio of exports to the United States to domestic GDP to characterize trade linkages between each country and the United States.

We find that sovereign bond yields in economies with high capital mobility or with deeper financial systems co-move more closely with US Treasury yields; i.e., the estimated coefficient β_1,2 is positive and significant at the 1 percent significance level. Also, in line with the evidence in Table 3.1 for the inverse relation between β₁ and β₂(FOMC), the coefficient associated with the additional effect of US sovereign yields on FOMC announcement days is more negative for high-mobility and financially deeper economies. We find that sovereign bond yields in countries with higher sovereign risk (higher average yields) co-move less with US Treasury yields, but the additional spillover effect on monetary policy announcement days is higher for these riskier economies. Our results suggest that bond yields in economies with high-volatility (more flexible) currencies co-move less with US sovereign yields. Finally, trade linkages with the US are positive and significant at explaining the effect of changes in US sovereign yields; i.e., sovereign bond yields in countries with closer trade linkages with the United States co-move more with US sovereign yields.

The results for stocks (Table 3.4b) show that stock prices in high capital mobility economies, economies with deeper financial systems, and economies with more flexible exchange rates co-move more closely with US Treasury yields. This co-movement is weaker in countries whose sovereign bonds are viewed as risky. Counterintuitively, trade linkages with the United States reduce the spillover effects. Overall, country-specific characteristics play only a minor role in explaining the cross-country variations in the additional spillovers around FOMC or US economic data release days.

For exchange rates (Table 3.4c), capital mobility and exports to the United States have little explanatory power for cross-country differences in β₁. Financial depth offsets the negative effect and currency flexibility and sovereign risk increases it. Country-specific factors have little explanatory power for the additional effects seen on FOMC and economic data days, with the possible exception of capital mobility, which offsets the negative effect on exchange rates on FOMC days, although this result is only marginally significant.

IV. Imperfect Asset Substitution in an Open Economy Macro Model

Section 2 of this paper examined the direct and spillover effects of net official flows and quantitative easing on current account balances. Because QE in the form of large-scale purchases of long-term bonds by central banks is a relatively new policy, our annual regressions did not contain many observations of this policy.³⁰ Therefore, section 3 explored further the spillover effects of unconventional monetary policy, regressing changes in various foreign financial prices on changes in US bond yields using daily data during the period of QE in the United States.

This section develops a theory model that relates financial market prices to current account balances, enabling us to draw broader conclusions from the daily regression results. In addition, the model points to some conclusions on the direct and spillover effects of NOF and of QE on income. We use a model with imperfect asset substitution because UMP relies importantly on imperfect substitutability across assets and because correlations of asset prices across countries are difficult to explain in standard models with perfect substitution. It is well-known, for example, that bond yields are often more highly correlated than would be expected and yet open interest rate parity is generally rejected because exchange rates move less than would be expected (or even in the wrong direction).

A drawback of our imperfect substitution model is that we cannot derive closed-form solutions linking asset prices, current accounts, and income. We are able to solve the model numerically for a wide range of parameters, which enables us to draw some qualitative conclusions about links between these variables under various policy regimes. The empirical results of sections 2 and 3 may be interpreted in light of these conclusions.

A. Model Specification

International macroeconomic models of policy transmission typically assume that financial assets are perfect substitutes. Thus, bond yields equal the average of expected future short-term interest rates and exchange rates move to fully offset differences in interest rates across countries. However, when assets are perfect substitutes, NOF and QE policies have no effects. These policies operate through the portfolio balance channel, which assumes imperfect substitution across assets. Our model adds portfolio preferences to a standard macro model, with the standard perfect-substitutes model emerging as a special case in which some of the parameters approach infinity. For tractability, the model ignores inflation.

The equations that comprise our model characterize the demand for bonds and the dynamics of the current account, domestic demand, and monetary policy. In the model, private domestic demand for domestic short-term bonds, BS, follows:

\begin{matrix} BS = a1* RS + b1* (RS - n * RL) + u_{B S}, & (4.1) \end{matrix}

where RS is the short-term domestic interest rate and RL is the long-term domestic interest rate. Foreign demand for domestic short-term bonds, FBS, follows:

\begin{matrix} F BS = a2* (RS - E) + b2 * (RS - n * RL) + c2 * (RS - E − RSF) + u_{F B S}, & (4.2) \end{matrix}

where E is the exchange rate (an increase represents an appreciation), n is the maturity of long-term bonds, and RSF is the foreign short-term interest rate. Private domestic demand for domestic long-term bonds, BL, follows:

\begin{matrix} BL = a3 * n * RL+ b3 * (n * RL - RS) + u_{B L} . & (4.3) \end{matrix}

Foreign demand for domestic long-term bonds, FBL, follows:

\begin{matrix} F BL = a4 * (n * RL - E) + b4 * (n * RL - RS) + c4*(n * RL - E - n*RLF) + u_{F B L} . & (4.4) \end{matrix}

The current account, CA, responds to income (GDP), Y, and exchange rate:

\begin{matrix} CA = - d1 * Y - d2 * E + u_{C A} . & (4.5) \end{matrix}

Absorption or domestic demand, DD, responds to interest rates (IS curve):

\begin{matrix} DD = - e1 * Rs - e 2 * n * RL + u_{D D} . & (4.6) \end{matrix}

Monetary policy follows a Taylor rule and is countercyclical:

\begin{matrix} RS = f1 * Y + u_{R S} . & (4.7) \end{matrix}

As is standard in the literature, we impose a set of identities. First, the demand of short-term bonds, XS, equals its (exogenous) supply:

\begin{matrix} XS = BS + FBS + CBS, & (4.8) \end{matrix}

where CBS represents central bank purchases of short-term bonds. Second, demand of long-term bonds, XL, equals its (exogenous) supply:

\begin{matrix} XL = BL + FBL + QE, & (4.9) \end{matrix}

where QE represents central bank purchases of long-term bonds. Third, current account equals net financial flows:

\begin{matrix} CA = NOF - FBS - FBL, & (4.10) \end{matrix}

where NOF is the central bank purchases of foreign assets. Fourth, income equals spending (GDP identity):

\begin{matrix} Y = DD + CA. & (4.11) \end{matrix}

Finally, the central bank balance sheet identity is the following:

\begin{matrix} MB = CBS + QE + NOF, & (4.12) \end{matrix}

where MB is the growth of monetary base.

In our model, BS, FBS, BL, FBL, CA, DD, Y, RS, RL, E, CBS, and MB are endogenous variables, while RSF, RLF, QE, NOF, XS, XL, and all shocks, u, in equations 4.1 to 4.12 are exogenous. All parameters, ax, bx, cx, dx, ex, and fx, are assumed to be greater than zero.

All variables in our model are expressed as deviations from steady state. Interest rates are in percentage points; the exchange rate is in percent; and other variables are expressed as a proportion of GDP (in percent). Variables are assumed to have been in steady state in period 0 and are expected to return to steady state in periods 2, …, n, where n is the term of the long bond. We are interested in the values of the variables only in period 1, hence, there is no need for time subscripts. The period is at least one year and possibly several years long.

Because all variables are expected to return to 0 in the future, the interest rate parity conditions are RS=E+RSF for short-term bonds and RL=E/n+RLF for long-term bonds. The parameters that operate on deviations from parity would be infinite in a model of perfect substitution and 0 in a model of segmented markets. The pure expectations model of bond yields is RL=RS/n. The long yield rises with the short yield but in inverse proportion to the maturity of the bond. Again this condition is not enforced, so that short-term and long-term bonds are only imperfect substitutes.

The demand for bonds is a function of their own return (expressed in the currency of the investor), their return relative to the returns on other maturities in the same currency, and their return relative to the returns on the same maturity in other currencies. For simplicity and without loss of generality, private domestic residents are assumed to hold only domestic assets.³¹ The central bank is able to purchase foreign assets. Owing to the small country assumption, the type of foreign asset purchased by the central bank does not matter, and it is not tracked in the model.

For simplicity, we define the asset variables as flows to be commensurate with the current account, which is a flow.³² However, it is possible to define a model in terms of stocks that has identical properties under the assumption that central bank purchases of foreign assets in the future exactly equal any non-zero current account in the current period.

The supply of domestic assets is assumed to be exogenous. Equations 4.8 and 4.9 reflect the equilibrium in the bond markets. Equation 4.10 is the balance of payments identity, which states that a current account surplus implies a net acquisition of foreign assets and a current account deficit implies a net incurrence of liabilities to foreigners. The central bank funds its purchases by issuing zero-interest monetary base (Equation 4.12).

The general reduced form of this model is complex and difficult to interpret. However, numerical solutions can provide good understanding of its properties. We set the parameter values of the modelas follows:

The baseline current account parameters (d1, d2) are (0.25, 0.25). These values reflect the median value of domestic content in exports of 25 percent of GDP, and income and price elasticities of trade that are assumed to equal 1.
The baseline absorption parameters (e1, e2) are (0.50, 0.50). These match the effect of monetary policy on US GDP over a two to three year horizon in the Federal Reserve’s FRB/US model. Alternative parameter sets are (0.25, 0.25), (0.25, 0.50), and (1.0, 1.0).
The baseline monetary parameter, f1, is 1. This reflects a coefficient of 0.5 on income in the Taylor rule, plus an implicit effect of higher income on inflation, which also has a coefficient of 0.5 in the Taylor rule in real terms. An alternative parameter is 2.
The ax parameters reflect the response of asset demand to an equal increase in rates of return on all assets. We assume demand changes proportionally for all assets, so that a1=a2=a3=a4. The baseline value is 0.1 and alternative values are 0.01 and 1.
The bx parameters reflect the substitution between short-term and long-term bonds. We assume domestic and foreign residents have the same substitution parameters (b1=b2, b3=b4) and that substitution is symmetric between short-term and long-term bonds (b1=b3, b2=b4). The baseline value is 1 and alternative values are 0.1 and 10.
The cx parameters reflect the substitution between domestic and foreign bonds. For simplicity and without loss of generality, only foreigners can buy foreign bonds. Substitution across currencies is assumed to be equal across short-term and long-term bonds (c2=c4). The baseline values are 0.1 for low capital mobility and 1 for high capital mobility. Under high capital mobility, substitution across currencies is equal to substitution across maturities. We assume alternative values of 0.01 and 10.

Equations 4.1 through 4.12 describe an economy with a flexible exchange rate, E. It is possible to alter the model to that of an economy with a fixed exchange rate in either of two ways. First, the Taylor rule (Equation 4.7) may be dropped and E made exogenous. This is the classic case of giving up domestic monetary autonomy (RS) to fix the exchange rate. Second, NOF can be used to target the exchange rate, effectively making E exogenous and NOF endogenous. This is possible because assets are imperfect substitutes.

We note that the theoretical analysis is limited to examining temporary, albeit persistent, shocks. The linearity of the model is convenient and necessary to obtain workable results. In the real world, nonlinearities may be important at times, particularly when shocks are large. Nevertheless, we believe the model is useful to convey sensible intuition for a wide variety of circumstances.

B. Model Properties and Relation to Annual Regression Results

Effects of NOF

Table 4.2 shows that NOF increases the current account in both flexible and fixed exchange rate regimes. In a flexible regime, this occurs through exchange rate depreciation. In a fixed regime, this occurs because interest rates rise and choke off domestic demand. The rise in interest rates is necessary to maintain the fixed exchange rate when NOF increases. When NOF is used to fix the exchange rate (the final column) it is no longer available as a policy tool. However, the exchange rate could be moved and this would have effects on both NOF and the current account consistent with those under flexible rates.

NOF has a larger effect on the current account when capital mobility is low, regardless of the exchange rate regime. The positive effect of NOF on the current account in all cases, and the larger effect when capital mobility is low, are both supported strongly in the annual regressions.

Table 4.3 shows that NOF increases income under flexible exchange rates and decreases it under fixed exchange rates. The negative effect under fixed rates reflects the need to raise interest rates to maintain the exchange rate.

Effects of QE

When the exchange rate is flexible and when it is fixed using RS, QE has a small and ambiguously signed effect on the current account. When the exchange rate is fixed using NOF, QE has a negative effect. These results hold regardless of capital mobility. Under a flexible exchange rate, there are two offsetting channels: (1) QE lowers long-term interest rates, which depreciates the exchange rate and boosts exports, and (2) lower long-term interest rates increase domestic demand, which boosts imports. When the exchange rate is fixed via NOF, only the second channel operates, implying a negative effect on the current account. When the exchange rate is fixed using RS, RS rises to keep the exchange rate fixed, and this directly offsets the stimulus to domestic demand from QE through long-term rates.

In the annual regressions, QE has essentially no effect on the current account under high capital mobility. This result is consistent with the model. There is a small positive effect under low capital mobility, which also may be consistent with the model.

The model suggests that QE has a positive effect on income under both a flexible exchange rate and an exchange rate that is fixed using NOF. In both cases, the effect is larger when capital mobility is lower. High capital mobility damps the effect of central bank purchases of long-term bonds on bond yields. When the exchange rate is fixed using RS, RS rises to offset downward pressure on the exchange rate and this damps any effect on income.

Spillovers of NOF

Foreign NOF enters the model as a shock to UFBS.³³ In the model, the effects of UFBS can be exactly offset by equal movements in NOF.³⁴ Indeed, under a fixed exchange rate (via NOF) regime, that is what happens, and UFBS shocks have no effect on either the current account or income. When UFBS is not offset by NOF, it has economic effects of equal magnitude and opposite direction to those of NOF.

As this is a model of a single small economy, it has nothing to say on the factors that determine the allocation of NOF across partner countries.

Spillovers of QE and Conventional Monetary Policy

RSF and RLF reflect monetary policy in the rest of the world. Conventional monetary policy operates primarily through RSF but also to some extent through RLF. QE and other unconventional monetary policies operate primarily through RLF, especially when RSF is close to the zero lower bound. Monetary policy is also likely to be associated with movements in aggregate demand in the rest of the world, which in turn causes movements in the exogenous component of the current account, UCA. These will be discussed next in the context of the daily data regressions.

C. Model Properties and Relation to Daily Regression Results

The daily regressions explore the effects of changes in US conditions on each of the other economies. From the perspective of the model, the United States is the foreign economy. The regressions attempt to distinguish between anticipated and unanticipated monetary policy in the United States. Anticipated monetary policy refers to the normal response of policy to changes in economic prospects. News of stronger current or future spending causes long-term bond yields to rise in anticipation of future monetary tightening. However, changes in bond yields in a narrow window around announcements by the monetary authorities (FOMC days) are more likely to contain news about unanticipated monetary policy changes than news about underlying economic prospects.

Effects of “Good News” in the Foreign (US) Economy

A rise in the bond yield on non-FOMC days probably reflects an upward revision to projected US economic activity, which is commonly referred to as “good news.” In the model, good news about the foreign economy implies an increase in demand for domestic exports (UCA) and a rise in foreign bond yields (RLF) (US short-term rates, RSF, were pinned at the zero bound in our sample).

In the model under a flexible exchange rate (as shown in Tables 4.1-4.3), good news about the foreign economy raises the bond yield, raises income, and has an uncertain effect on the exchange rate. Capital mobility increases the effects on yields and income and reduces (or makes more negative) the effect on the exchange rate. The β1 coefficients in the daily regressions confirm these model implications for bond yields and income, if we assume that stock price effects reflect expected income effects.³⁵ The daily regressions are consistent with the model predictions for the exchange rate. Specifically, under flexible exchange rates, increases in RLF depreciate the local currency and increases in UCA appreciate it, so that either result is possible. According to our quantitative results, the effect of RLF dominates.

Model predictions under fixed exchange rates are more uncertain and parameter dependent. The effects of exchange rate flexibility in the daily regressions are broadly sensible. More flexible exchange rates imply less effect of US bond yields on local yields, a greater effect of US bond yields on local stock prices, and a larger (more negative) effect of US bond yields on local exchange rates.

Effects of Tighter UMP in the Foreign (US) Economy

A rise in the US bond yield on a day with an FOMC announcement or speech probably reflects market revisions about future US unconventional monetary policy. In the model, tighter than expected UMP implies an increase in RLF and a decrease in UCA. We focus on the marginal effect on FOMC days (β₂) rather than the overall effect (β₁+β₂) because the overall effect includes the effects of good economic news on FOMC days. Relative to the above results for a good news shock (and assuming a flexible exchange rate), the model implies that tighter foreign UMP raises the local bond yield by less or even reduces it, raises the local stock price by less or even reduces it, and reverses any exchange rate appreciation or makes a depreciation even larger. The first two effects are confirmed by the β2 coefficients in the daily regressions. The exchange rate regressions generally found no significant difference between FOMC days and other days, which is inconclusive but not particularly supportive of the model. The model implies that capital mobility pushes up the local bond yield more and pushes down local income and exchange rate more in response to a foreign UMP tightening. The daily regressions reject this model implication for bond yields and exchange rates, although the latter rejection is only marginally significant (10 percent level. See Table 3.4).

Overall, the daily regressions are broadly consistent with the model. The primary exception may be the lack of any clear additional negative effect of US UMP shocks on foreign exchange rates.

D. Policy Implications

The relatively large co-movements in bond yields and small responses of exchange rates to US bond yields, especially on good news, suggests that many foreign central banks regularly follow the lead of US monetary policy. Indeed, recent research finds that short-term interest rates in many countries respond to US short-term rates more than would be implied by a Taylor rule based on purely local conditions (Carstens 2015, Hofmann and Takats 2015).

An interesting property of the model in this section is that officials have other tools besides conventional monetary policy (RS) with which to combat the effects of foreign shocks. For example, NOF can be used to completely neutralize the effects on all domestic variables of foreign capital inflows into short-term bonds, including spillovers of NOF in foreign countries. Moreover, as shown in the final columns of Tables 4.1 to 4.3, NOF can also insulate a country from shocks to foreign short-term interest rates.³⁶ Another property of the model, not shown in the tables, is that a judicious combination of NOF and QE can be used to fully offset the effects of foreign capital inflows into long-term bonds and changes in foreign long-term interest rates. The existence of these tools allows the authorities to pursue objectives on exchange rates, current accounts, or long-term bond yields without giving up use of monetary policy for stabilization of income.

In addition, other policy tools exist outside the model, such as fiscal policy, capital controls, and macro-prudential measures, which also can influence desired objectives. Pushed to the extreme, these considerations suggest that domestic officials are not as much at the mercy of global shocks as many observers suppose.

Some central banks may be uncomfortable with taking large net positions in long-term bonds and foreign exchange. Political realities may constrain full use of fiscal policy, capital controls, and macro-prudential measures. The model is too simple to fully assess the benefits, costs, and risks involved. Nevertheless, the model does highlight the richness of policy options in a world of imperfect financial substitution.

V. Conclusions

This paper explores the direct effects and spillovers of unconventional monetary and exchange rate policies. Our annual regressions focus on the direct effects and spillovers of official purchases of foreign and domestic assets on the current account balance. Our daily regressions focus on the spillovers to foreign financial market prices of unconventional monetary policy in the United States, which includes, but is not limited to, official purchases of long-term bonds or quantitative easing (QE). Our results are broadly consistent with the predictions of a simple macroeconomic model of a small economy with imperfect substitution of assets across currencies and maturities.

Our findings suggest that official purchases of foreign assets, or net official flows (NOF), have a large direct effect on a country’s current account balance when capital mobility is low. This effect diminishes considerably as capital mobility rises, but it remains significant at high levels of mobility, and there is an important additional effect through the lagged stock of net official assets. At the lowest level of capital mobility, each dollar of NOF raises the current account balance about 75 cents. This effect declines to around 20 cents at the highest level of mobility, but the lagged stock of net official assets (NOA) increases the current account by about 4 cents on the dollar when capital mobility is high. These effects are robust to a wide range of specifications.

Official purchases of domestic assets, or QE, appear to have no significant effect on a country’s current account balance when capital mobility is high. By contrast, there does appear to be a modest but significant positive impact when capital mobility is low. The finding of a significant effect only under low capital mobility is somewhat puzzling, and may reflect unexplained variation between the major advanced economies and the emerging markets that engaged in QE and have low capital mobility. We plan to research this further.

Because the sum of all countries’ current account balances adds up to zero, any effect of official policies on a country’s current account must spill over to other countries. We find that the effects of official flows spill over to other countries in proportion to their degree of international financial integration. This effect is moderately robust, although it does not seem to work as well for the United States and the euro area. The United States, in particular, seems to receive relatively more of the spillovers than is explained by its degree of international financial integration, but the difference is not statistically significant.

Our daily regressions find that US economic developments have important spillovers to foreign economies. In particular, increases in current and expected economic activity in the United States, working in part through expected monetary policy reactions, raise US bond yields and also raise foreign bond yields, increase foreign stock prices, and depreciate foreign currencies. Our theoretical model suggests that stronger US activity unambiguously raises foreign activity and foreign current account balances. These spillovers are stronger when capital mobility is higher and when financial markets are deeper. We find that spillovers are often greater on foreign bond yields than on foreign exchange rates, suggesting that many foreign central banks follow the lead of the Federal Reserve, perhaps out of a desire to stabilize their exchange rates. However, increases in US bond yields associated with tighter than expected future monetary policy (including QE) for a given path of the economy have smaller effects on foreign financial variables, consistent with a roughly neutral effect on foreign activity and foreign current account balances—although there may still be direct contractionary effects on foreign financial markets. Our theoretical model suggests that pure monetary shocks have small and ambiguous effects on current account balances and activity in foreign economies. This result is also consistent with the results of our annual regressions, at least for countries with high capital mobility.

Our theoretical model highlights the potential usefulness of foreign exchange intervention and quantitative easing as responses to the effects of foreign policies of a similar type. In particular, negative spillovers on the current account from the rest of the world can be fully offset by increasing intervention (NOF) at home. Similarly, spillovers onto bond yields from the rest of the world can be offset by countervailing QE at home. However, it is possible that central banks will be uncomfortable with taking large net positions in long-term bonds and foreign exchange. While the model cannot balance all of the benefits, costs, and risks involved, it does provide a rich framework within which to compare conventional and unconventional policy options.

References

Adler, Gustavo, Noemie Lisack, and Rui Mano. 2015. Unveiling the Effects of Foreign Exchange Intervention: A Panel Approach. IMF Working Paper No. WP/15/130. Washington: International Monetary Fund.
- Search Google Scholar
- Export Citation
Aizenman, Joshua, Menzie Chinn, and Hiro Ito. 2015. The Trilemma Indexes. http://web.pdx.edu/~ito/trilemma_indexes.htm (accessed August 17, 2015).
- Search Google Scholar
- Export Citation
Bartolini, Leonard, Linda Goldberg, and Adam Sacarny. 2008. How Economic News Moves Markets. Current Issues in Economics and Finance 14(6): 1–7.
- Search Google Scholar
- Export Citation
Bayoumi, Tamim, and Trung Bui. 2010. Deconstructing the International Business Cycle: Why Does A U.S. Sneeze Give The Rest Of The World A Cold? IMF Working Papers WP/10/239. Washington: International Monetary Fund.
- Search Google Scholar
- Export Citation
Bayoumi, Tamim, and Trung Bui, and Christian Saborowski. 2014. Accounting for Reserves. Journal of International Money and Finance 41: 1–29.
- Search Google Scholar
- Export Citation
Bayoumi, Tamim, Joseph Gagnon, and Christian Saborowski. 2015. Official Financial Flows, Capital Mobility, and Global Imbalances. Journal of International Money and Finance 52 (April): 146–74.
- Search Google Scholar
- Export Citation
Blanchard, Olivier, Gustavo Adler, and Irineu de Carvalho Filho. 2015. Can Foreign Exchange Intervention Stem Exchange Rate Pressures from Global Capital Flow Shocks? IMF Working Papers WP/15/159. Washington: International Monetary Fund.
- Search Google Scholar
- Export Citation
Bowman, David, Juan M. Londono, and Horacio Sapriza. 2014. US Unconventional Monetary Policy and Transmission to Emerging Market Economies. Journal of International Money and Finance 55: 57–59.
- Search Google Scholar
- Export Citation
Carstens, Agustin. 2015. Challenges for Emerging Economies in the Face of Unconventional Monetary Policies in Advanced Economies. Stavros Niarchos Foundation Lecture, April 20. Washington: Peterson Institute for International Economics.
- Search Google Scholar
- Export Citation
Chinn, Menzie, and Eswar Prasad. 2003. Medium-Term Determinants of Current Accounts in Industrial and Developing Countries: An Empirical Exploration. Journal of International Economics 59: 47–76.
- Search Google Scholar
- Export Citation
Gagnon, Joseph. 2012. Global Imbalances and Foreign Asset Expansion by Developing-Economy Central Banks. Working Paper No. 12-5. Washington: Peterson Institute for International Economics.
- Search Google Scholar
- Export Citation
Gagnon, Joseph. 2013. The Elephant Hiding in the Room: Currency Intervention and Trade Imbalances. Working Paper No. 13-2. Washington: Peterson Institute for International Economics.
- Search Google Scholar
- Export Citation
Gagnon, Joseph, Matthew Raskin, Julie Remache, and Brian Sack. 2011. The Financial Market Effects of the Federal Reserve’s Large-Scale Asset Purchases. International Journal of Central Banking 7: 3–44.
- Search Google Scholar
- Export Citation
Hausman, Joshua, and Jon Wongswan. 2011. Global asset prices and FOMC announcements. Journal of International Money and Finance 30 (3): 547–71.
- Search Google Scholar
- Export Citation
Hofmann, Boris, and Elod Takats. 2015. International Monetary Spillovers. BIS Quarterly Review (September): 105–18.
- Search Google Scholar
- Export Citation
International Monetary Fund. 2015. 2015 Spillover Report. July 23. Washington: International Monetary Fund.
- Search Google Scholar
- Export Citation
Laeven, Luc, and Fabián Valencia. 2012. Systemic Banking Crises Database: An Update. IMF Working Paper WP/12/163. Washington: International Monetary Fund.
- Search Google Scholar
- Export Citation
Londono, Juan, and Hao Zhou. 2016. Variance Risk Premiums and the Forward Premium Puzzle. Journal of Financial Economics (forthgoming).
- Search Google Scholar
- Export Citation
Rogers, John, Chiara Scotti, and Jonathan Wright. 2014. Evaluating Asset-Market Effects of Unconventional Monetary Policy: A Cross-Country Comparison. International Finance Discussion Papers No. 1101. Washington: Board of Governors of the Federal Reserve System.
- Search Google Scholar
- Export Citation
Sahay, Ratna, Martin Cihak, Papa N’Diaye, Adolfo Barajas, Ran Bi, Diana Ayala, Yuan Gao, Annette Kyobe, Lam Nguyen, Christian Saborowski, Katsiaryna Svirydzenka, and Seyed Yousefi. 2015. Rethinking Financial Deepening: Stability and Growth in Emerging Markets. IMF Staff Discussion Note 15/08. Washington: International Monetary Fund.
- Search Google Scholar
- Export Citation

Appendix: Data Sources and Definitions

Table 2.1a.

Baseline Regressions—First Stage

Table 2.1b.

Baseline Regressions—Second Stage

* p<0.1, ** p<0.05, *** p<0.01

Table 2.1b.

Baseline Regressions—Second Stage

	CAX	NPFX	CAX	NPFX	CAX	NPFX
MOB, lagged	0	−0.002	−0.007	−0.017	0.011	0.01
	0	0	0	0	0	0
MOB squared and lagged			0.039**	0.061***	0.035*	0.046**
			0	0	0	0
Global Financial Integration					0.140**	0.146*
					0.1	0.1
Interaction with MOB					−0.150**	−0.089
					0.1	0.1
Relative GDP PPP pc, lagged	−0.011	−0.026**	0.02	0.040*	0.012	0.027
	0	0	0	0	0	0
Interaction with MOB			−0.037**	−0.084***	−0.012	−0.054**
			0	0	0	0
Aging	2.165***	2.211***	3.620**	2.165	2.831	1.37
	0.7	0.8	1.7	1.9	1.9	2
Interaction with MOB			−2.989	−1.117	−0.59	0.564
			2.1	2.4	2.3	2.6
Growth, lagged	−0.186***	−0.233***	0.021	0.078	0.101	0.179*
	0.1	0.1	0.1	0.1	0.1	0.1
Interaction with MOB			−0.522***	−0.731***	−0.752***	−0.999***
			0.1	0.2	0.2	0.2
Net Energy Exports	0.171***	0.143***	0.270***	0.253***	0.280***	0.259***
	0	0	0	0	0	0
Interaction with MOB			−0.245***	−0.277***	−0.251***	−0.268***
			0.1	0	0.1	0.1
Fiscal balance	0.392***	0.349***	0.212***	0.128	0.176**	0.132
	0.1	0.1	0.1	0.1	0.1	0.1
Interaction with MOB			0.347***	0.425***	0.421***	0.459***
			0.1	0.1	0.1	0.2
NOF	0.701***	0.780***	0.683***	0.742***	0.724***	0.775***
	0.1	0.2	0.1	0.2	0.2	0.2
Interaction with MOB	−0.643***	−0.26	−0.622***	−0.215	−0.715***	−0.421*
	0.2	0.3	0.2	0.3	0.2	0.3
NOA, lagged	0.007	−0.021**	0.004	−0.026***	0.006	−0.027***
	0	0	0	0	0	0
Interaction with MOB	0.032*	0.03	0.044***	0.048**	0.040**	0.059**
	0	0	0	0	0	0
QE, lagged					0.231**	0.255**
					0.1	0.1
Interaction with MOB					−0.302*	−0.22
					0.2	0.2
SPILL					−18.638***	−24.077***
					5.4	5.8
R-squared	0.457	0.239	0.493	0.308	0.525	0.376
Observations	2088	2088	2088	2088	1745	1745

* p<0.1, ** p<0.05, *** p<0.01

Table 2.1b.

Baseline Regressions—Second Stage

	CAX	NPFX	CAX	NPFX	CAX	NPFX
MOB, lagged	0	−0.002	−0.007	−0.017	0.011	0.01
	0	0	0	0	0	0
MOB squared and lagged			0.039**	0.061***	0.035*	0.046**
			0	0	0	0
Global Financial Integration					0.140**	0.146*
					0.1	0.1
Interaction with MOB					−0.150**	−0.089
					0.1	0.1
Relative GDP PPP pc, lagged	−0.011	−0.026**	0.02	0.040*	0.012	0.027
	0	0	0	0	0	0
Interaction with MOB			−0.037**	−0.084***	−0.012	−0.054**
			0	0	0	0
Aging	2.165***	2.211***	3.620**	2.165	2.831	1.37
	0.7	0.8	1.7	1.9	1.9	2
Interaction with MOB			−2.989	−1.117	−0.59	0.564
			2.1	2.4	2.3	2.6
Growth, lagged	−0.186***	−0.233***	0.021	0.078	0.101	0.179*
	0.1	0.1	0.1	0.1	0.1	0.1
Interaction with MOB			−0.522***	−0.731***	−0.752***	−0.999***
			0.1	0.2	0.2	0.2
Net Energy Exports	0.171***	0.143***	0.270***	0.253***	0.280***	0.259***
	0	0	0	0	0	0
Interaction with MOB			−0.245***	−0.277***	−0.251***	−0.268***
			0.1	0	0.1	0.1
Fiscal balance	0.392***	0.349***	0.212***	0.128	0.176**	0.132
	0.1	0.1	0.1	0.1	0.1	0.1
Interaction with MOB			0.347***	0.425***	0.421***	0.459***
			0.1	0.1	0.1	0.2
NOF	0.701***	0.780***	0.683***	0.742***	0.724***	0.775***
	0.1	0.2	0.1	0.2	0.2	0.2
Interaction with MOB	−0.643***	−0.26	−0.622***	−0.215	−0.715***	−0.421*
	0.2	0.3	0.2	0.3	0.2	0.3
NOA, lagged	0.007	−0.021**	0.004	−0.026***	0.006	−0.027***
	0	0	0	0	0	0
Interaction with MOB	0.032*	0.03	0.044***	0.048**	0.040**	0.059**
	0	0	0	0	0	0
QE, lagged					0.231**	0.255**
					0.1	0.1
Interaction with MOB					−0.302*	−0.22
					0.2	0.2
SPILL					−18.638***	−24.077***
					5.4	5.8
R-squared	0.457	0.239	0.493	0.308	0.525	0.376
Observations	2088	2088	2088	2088	1745	1745

* p<0.1, ** p<0.05, *** p<0.01

Table 2.2.

Robustness Checks

* p<0.1, ** p<0.05, *** p<0.01Coefficients and standard errors are averages across CAX and NPFX regressions

Table 2.2.

Robustness Checks

		Baseline	OLS	Alt Instr	Weighted	Robust	Fixed	Flexible	Tradeopen	Tradeclosed
		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
NOF		0.750***	0.674***	0.970***	0.765***	0.804***	0.83***	0.654***	0.688*	0.781
		0.2	0.1	0.2	0.1	0.1	0.2	0.2	0.2	0.3
	Interaction with MOB	−0.568**	−0.408***	−0.7305**	−0.4135**	−0.751***	−0.981**	−0.3565	−0.52***	−0.5385***
		0.2	0.1	0.3	0.3	0.1	0.4	0.3	0.3	0.4
NOA, lagged		−0.0105	−0.0085	−0.0145	−0.007	−0.0155**	0.004	0.017	−0.0155	0.012
		0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
	Interaction with MOB	0.0495**	0.0455***	0.0505**	0.084***	0.0655***	0.012	0.029	0.0525*	−0.046*
		0.0	0.0	0.0	0.0	0.0	0.1	0.0	0.0	0.0
QE, lagged		0.243**	0.248**	0.269**	0.2165**	0.0965	0.092	0.1985	0.41**	0.0905
		0.1	0.1	0.1	0.1	0.1	0.2	0.1	0.2	0.1
	Interaction with MOB	−0.261	−0.279*	−0.2995	−0.1845	−0.175	0.014	−0.234	−0.495	0.001
		0.2	0.2	0.2	0.1	0.1	0.3	0.2	0.3	0.1
SPILL		−21.4***	−21.5***	−21.2***	1.7	−11.8***	−16.8*	−14.9***	−39.2***	−0.5
		5.6	4.2	5.9	2.9	3.4	8.7	5.8	8.7	3.7
R-squared		0.45	0.37	0.46	0.64	0.50	0.52	0.46	0.51	0.43
Observations		1745	1755	1699	1745	1745	650	1095	873	872

* p<0.1, ** p<0.05, *** p<0.01Coefficients and standard errors are averages across CAX and NPFX regressions

Table 2.2.

Robustness Checks

		Baseline	OLS	Alt Instr	Weighted	Robust	Fixed	Flexible	Tradeopen	Tradeclosed
		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
NOF		0.750***	0.674***	0.970***	0.765***	0.804***	0.83***	0.654***	0.688*	0.781
		0.2	0.1	0.2	0.1	0.1	0.2	0.2	0.2	0.3
	Interaction with MOB	−0.568**	−0.408***	−0.7305**	−0.4135**	−0.751***	−0.981**	−0.3565	−0.52***	−0.5385***
		0.2	0.1	0.3	0.3	0.1	0.4	0.3	0.3	0.4
NOA, lagged		−0.0105	−0.0085	−0.0145	−0.007	−0.0155**	0.004	0.017	−0.0155	0.012
		0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
	Interaction with MOB	0.0495**	0.0455***	0.0505**	0.084***	0.0655***	0.012	0.029	0.0525*	−0.046*
		0.0	0.0	0.0	0.0	0.0	0.1	0.0	0.0	0.0
QE, lagged		0.243**	0.248**	0.269**	0.2165**	0.0965	0.092	0.1985	0.41**	0.0905
		0.1	0.1	0.1	0.1	0.1	0.2	0.1	0.2	0.1
	Interaction with MOB	−0.261	−0.279*	−0.2995	−0.1845	−0.175	0.014	−0.234	−0.495	0.001
		0.2	0.2	0.2	0.1	0.1	0.3	0.2	0.3	0.1
SPILL		−21.4***	−21.5***	−21.2***	1.7	−11.8***	−16.8*	−14.9***	−39.2***	−0.5
		5.6	4.2	5.9	2.9	3.4	8.7	5.8	8.7	3.7
R-squared		0.45	0.37	0.46	0.64	0.50	0.52	0.46	0.51	0.43
Observations		1745	1755	1699	1745	1745	650	1095	873	872

* p<0.1, ** p<0.05, *** p<0.01Coefficients and standard errors are averages across CAX and NPFX regressions

Table 2.3.

Correlations in Energy Exporters, 2000–14 (percent of GDP)

Table 2.3.

Correlations in Energy Exporters, 2000–14 (percent of GDP)

	NOF-Energy	NOF-Fiscal	NOF-CAX	Fiscal-CAX	Energy-CAX
Algeria	0.92	0.90	0.98	0.94	0.91
Norway	0.73	0.90	0.80	0.86	0.58
Saudi Arabia	0.87	0.90	0.94	0.90	0.88
Yemen	0.74	0.86	0.87	0.93	0.89
Colombia	0.37	0.62	0.35	−0.03	−0.34
Indonesia	−0.23	0.41	0.66	0.43	−0.75
Venezuela	0.40	0.42	0.82	0.39	0.45

Table 2.3.

Correlations in Energy Exporters, 2000–14 (percent of GDP)

	NOF-Energy	NOF-Fiscal	NOF-CAX	Fiscal-CAX	Energy-CAX
Algeria	0.92	0.90	0.98	0.94	0.91
Norway	0.73	0.90	0.80	0.86	0.58
Saudi Arabia	0.87	0.90	0.94	0.90	0.88
Yemen	0.74	0.86	0.87	0.93	0.89
Colombia	0.37	0.62	0.35	−0.03	−0.34
Indonesia	−0.23	0.41	0.66	0.43	−0.75
Venezuela	0.40	0.42	0.82	0.39	0.45

Table 3.1.

Spillovers to Sovereign Bond Yields—Daily Regressions

* p <0.1, ** p<0.05, *** p<0.01

Table 3.1.

Spillovers to Sovereign Bond Yields—Daily Regressions

Country	β₁	β₂(FOMC)	β₁+β₂(FOM)	β₂(Econ.)	β₁+β₂(Econ)
Canada	0.613***	−0.061**	0.539***	0.013	0.599***
Australia	0.598***	0.019	0.597***	0.110	0.662***
Hong Kong	0.562***	−0.101*	0.513***	0.076	0.663***
United Kingdom	0.457***	−0.286***	0.179***	0.088	0.502***
Germany	0.434***	−0.228***	0.215***	0.060	0.461***
Euro Area	0.434***	−0.239***	0.211***	0.055	0.461***
Denmark	0.400***	−0.198***	0.183***	0.087	0.425***
Finland	0.386***	−0.217***	0.196***	0.016	0.392***
New Zealand	0.379***	−0.048	0.325***	0.028	0.392***
Netherlands	0.377***	−0.196***	0.200***	0.013	0.375***
Sweden	0.337***	−0.196***	0.147**	0.071	0.376***
Singapore	0.306***	−0.010	0.282***	0.087*	0.363***
France	0.299***	−0.168**	0.171***	0.008	0.318***
Mexico	0.244***	0.179*	0.259***	0.148	0.255***
Norway	0.229***	−0.189***	0.073	0.042	0.271***
Switzerland	0.220***	−0.102**	0.135***	0.002	0.224***
South Korea	0.216***	0.126***	0.313***	0.172***	0.352***
Belgium	0.206***	−0.100	0.163***	−0.072	0.178***
Thailand	0.185***	0.217***	0.352***	−0.009	0.165***
Brazil	0.153**	0.344**	0.303**	0.423**	0.395*
Czech Republic	0.127***	−0.070	0.062	−0.071	0.074
Japan	0.125***	−0.023	0.120***	0.040**	0.169***
Taiwan	0.077***	0.001	0.078**	0.024	0.096***
Malaysia	0.077***	0.070	0.152***	0.008	0.098***
Indonesia	0.074	0.457	0.328	−0.106	−0.182
South Africa	0.055*	0.066	0.053	0.219***	0.224***
Colombia	0.031	0.310**	−0.002	0.428***	0.185***
Poland	0.030	−0.068	−0.039	0.099	0.104*
Chile	0.016	−0.343***	0.017	−0.237*	0.037
China	0.015	−0.020	0.001	0.034	0.044
Turkey	0.004	0.038	0.028	0.184	0.156
Ireland	0.004	−0.038	0.090	−0.305***	−0.161
Spain	−0.009	−0.037	0.068	−0.186*	−0.073
Pakistan	−0.013	−0.010	−0.019	0.257***	0.218***
Italy	−0.046	−0.096	−0.018	−0.140*	−0.065
Philippines	−0.049	0.220**	0.070	−0.069	−0.192*
Portugal	−0.073	0.058	0.097	−0.507***	−0.417**
Russia	−0.164*	−0.375	−0.495	0.326**	0.111
Hungary	−0.187***	0.169	−0.036	−0.046	−0.229
Greece	−0.672***	0.110	−0.267	0.137	−0.247

* p <0.1, ** p<0.05, *** p<0.01

Table 3.1.

Spillovers to Sovereign Bond Yields—Daily Regressions

Country	β₁	β₂(FOMC)	β₁+β₂(FOM)	β₂(Econ.)	β₁+β₂(Econ)
Canada	0.613***	−0.061**	0.539***	0.013	0.599***
Australia	0.598***	0.019	0.597***	0.110	0.662***
Hong Kong	0.562***	−0.101*	0.513***	0.076	0.663***
United Kingdom	0.457***	−0.286***	0.179***	0.088	0.502***
Germany	0.434***	−0.228***	0.215***	0.060	0.461***
Euro Area	0.434***	−0.239***	0.211***	0.055	0.461***
Denmark	0.400***	−0.198***	0.183***	0.087	0.425***
Finland	0.386***	−0.217***	0.196***	0.016	0.392***
New Zealand	0.379***	−0.048	0.325***	0.028	0.392***
Netherlands	0.377***	−0.196***	0.200***	0.013	0.375***
Sweden	0.337***	−0.196***	0.147**	0.071	0.376***
Singapore	0.306***	−0.010	0.282***	0.087*	0.363***
France	0.299***	−0.168**	0.171***	0.008	0.318***
Mexico	0.244***	0.179*	0.259***	0.148	0.255***
Norway	0.229***	−0.189***	0.073	0.042	0.271***
Switzerland	0.220***	−0.102**	0.135***	0.002	0.224***
South Korea	0.216***	0.126***	0.313***	0.172***	0.352***
Belgium	0.206***	−0.100	0.163***	−0.072	0.178***
Thailand	0.185***	0.217***	0.352***	−0.009	0.165***
Brazil	0.153**	0.344**	0.303**	0.423**	0.395*
Czech Republic	0.127***	−0.070	0.062	−0.071	0.074
Japan	0.125***	−0.023	0.120***	0.040**	0.169***
Taiwan	0.077***	0.001	0.078**	0.024	0.096***
Malaysia	0.077***	0.070	0.152***	0.008	0.098***
Indonesia	0.074	0.457	0.328	−0.106	−0.182
South Africa	0.055*	0.066	0.053	0.219***	0.224***
Colombia	0.031	0.310**	−0.002	0.428***	0.185***
Poland	0.030	−0.068	−0.039	0.099	0.104*
Chile	0.016	−0.343***	0.017	−0.237*	0.037
China	0.015	−0.020	0.001	0.034	0.044
Turkey	0.004	0.038	0.028	0.184	0.156
Ireland	0.004	−0.038	0.090	−0.305***	−0.161
Spain	−0.009	−0.037	0.068	−0.186*	−0.073
Pakistan	−0.013	−0.010	−0.019	0.257***	0.218***
Italy	−0.046	−0.096	−0.018	−0.140*	−0.065
Philippines	−0.049	0.220**	0.070	−0.069	−0.192*
Portugal	−0.073	0.058	0.097	−0.507***	−0.417**
Russia	−0.164*	−0.375	−0.495	0.326**	0.111
Hungary	−0.187***	0.169	−0.036	−0.046	−0.229
Greece	−0.672***	0.110	−0.267	0.137	−0.247

* p <0.1, ** p<0.05, *** p<0.01

Table 3.2.

Spillovers to Stock Prices—Daily Regressions

* p <0.1, ** p<0.05, *** p<0.01

Table 3.2.

Spillovers to Stock Prices—Daily Regressions

Country	β₁	β₂(FOMC)	β₁+β₂(FOM)	β₂(Econ.)	β₁+β₂(Econ)
Belgium	0.101***	−0.055**	0.037***	0.125	0.198
Spain	0.088***	−0.035*	0.045**	0.024	0.098***
Euro Area	0.082***	−0.028	0.050***	0.023	0.095***
France	0.082***	−0.030	0.047***	0.017	0.089***
Germany	0.081***	−0.018	0.056***	0.027	0.097***
Russia	0.080***	−0.003	0.077***	0.029	0.103***
Netherlands	0.077***	−0.013	0.059***	0.022	0.090***
Finland	0.075***	−0.028	0.042**	0.028	0.091***
Norway	0.072***	−0.023	0.049**	0.006	0.073***
Brazil	0.071***	−0.068***	0.022	−0.003	0.076***
Sweden	0.069***	−0.024	0.043**	0.019	0.079***
Japan	0.065***	−0.023	0.046**	0.037*	0.097***
Hungary	0.064***	−0.018	0.041**	−0.003	0.055***
Greece	0.063***	0.009	0.065***	0.025	0.080***
United Kingdom	0.061***	−0.020	0.041***	0.009	0.067***
Ireland	0.061***	−0.036**	0.031*	0.003	0.063***
Portugal	0.061***	−0.034**	0.025*	0.002	0.057***
Canada	0.060***	−0.023	0.042*	0.011	0.072***
Mexico	0.056***	−0.034	0.037*	0.006	0.069***
Switzerland	0.054***	−0.022*	0.032***	0.007	0.057***
Czech Republic	0.052***	−0.041**	0.015	0.019	0.071***
Poland	0.051***	−0.031**	0.023*	0.002	0.053***
Denmark	0.049***	−0.029**	0.022*	−0.004	0.043***
Turkey	0.048***	−0.003	0.048**	−0.029	0.025
Italy	0.048***	−0.022	0.021	0.013	0.053***
South Africa	0.041***	−0.028	0.017	0.008	0.049***
Hong Kong	0.038***	−0.039*	0.008	0.029	0.066***
Australia	0.038***	−0.018	0.028*	0.020	0.060***
Philippines	0.034***	−0.029**	0.013	−0.008	0.032**
South Korea	0.033***	−0.026	0.014	0.009	0.043**
Colombia	0.031***	−0.020**	0.013	0.005	0.034***
Taiwan	0.031***	−0.018	0.020	−0.009	0.026
Chile	0.030***	−0.027***	0.009	0.001	0.032***
Indonesia	0.027***	−0.037**	0.000	−0.009	0.020
Singapore	0.025***	−0.026*	0.007	0.017	0.043***
New Zealand	0.020***	−0.007	0.016*	−0.003	0.021**
Malaysia	0.017***	−0.012	0.010	0.003	0.023***
Thailand	0.015***	−0.027*	−0.006	−0.014	0.004
China	0.011*	−0.045***	−0.018	0.003	0.024
Pakistan	0.003	0.018	0.023	−0.016	−0.009

* p <0.1, ** p<0.05, *** p<0.01

Table 3.2.

Spillovers to Stock Prices—Daily Regressions

Country	β₁	β₂(FOMC)	β₁+β₂(FOM)	β₂(Econ.)	β₁+β₂(Econ)
Belgium	0.101***	−0.055**	0.037***	0.125	0.198
Spain	0.088***	−0.035*	0.045**	0.024	0.098***
Euro Area	0.082***	−0.028	0.050***	0.023	0.095***
France	0.082***	−0.030	0.047***	0.017	0.089***
Germany	0.081***	−0.018	0.056***	0.027	0.097***
Russia	0.080***	−0.003	0.077***	0.029	0.103***
Netherlands	0.077***	−0.013	0.059***	0.022	0.090***
Finland	0.075***	−0.028	0.042**	0.028	0.091***
Norway	0.072***	−0.023	0.049**	0.006	0.073***
Brazil	0.071***	−0.068***	0.022	−0.003	0.076***
Sweden	0.069***	−0.024	0.043**	0.019	0.079***
Japan	0.065***	−0.023	0.046**	0.037*	0.097***
Hungary	0.064***	−0.018	0.041**	−0.003	0.055***
Greece	0.063***	0.009	0.065***	0.025	0.080***
United Kingdom	0.061***	−0.020	0.041***	0.009	0.067***
Ireland	0.061***	−0.036**	0.031*	0.003	0.063***
Portugal	0.061***	−0.034**	0.025*	0.002	0.057***
Canada	0.060***	−0.023	0.042*	0.011	0.072***
Mexico	0.056***	−0.034	0.037*	0.006	0.069***
Switzerland	0.054***	−0.022*	0.032***	0.007	0.057***
Czech Republic	0.052***	−0.041**	0.015	0.019	0.071***
Poland	0.051***	−0.031**	0.023*	0.002	0.053***
Denmark	0.049***	−0.029**	0.022*	−0.004	0.043***
Turkey	0.048***	−0.003	0.048**	−0.029	0.025
Italy	0.048***	−0.022	0.021	0.013	0.053***
South Africa	0.041***	−0.028	0.017	0.008	0.049***
Hong Kong	0.038***	−0.039*	0.008	0.029	0.066***
Australia	0.038***	−0.018	0.028*	0.020	0.060***
Philippines	0.034***	−0.029**	0.013	−0.008	0.032**
South Korea	0.033***	−0.026	0.014	0.009	0.043**
Colombia	0.031***	−0.020**	0.013	0.005	0.034***
Taiwan	0.031***	−0.018	0.020	−0.009	0.026
Chile	0.030***	−0.027***	0.009	0.001	0.032***
Indonesia	0.027***	−0.037**	0.000	−0.009	0.020
Singapore	0.025***	−0.026*	0.007	0.017	0.043***
New Zealand	0.020***	−0.007	0.016*	−0.003	0.021**
Malaysia	0.017***	−0.012	0.010	0.003	0.023***
Thailand	0.015***	−0.027*	−0.006	−0.014	0.004
China	0.011*	−0.045***	−0.018	0.003	0.024
Pakistan	0.003	0.018	0.023	−0.016	−0.009

* p <0.1, ** p<0.05, *** p<0.01

Table 3.3.

Spillovers to Exchange Rates (with respect to US dollar)—Daily Regressions

* p <0.1, ** p<0.05, *** p<0.01

Table 3.3.

Spillovers to Exchange Rates (with respect to US dollar)—Daily Regressions

Country	β₁	β₂(FOMC)	β₁+β₂(FOM)	β₂(Econ.)	β₁+β₂(Econ.)
Poland	−0.033***	0.012	−0.013	−0.018	−0.040***
Hungary	−0.033***	0.016	−0.011	−0.001	−0.027
South Africa	−0.030***	0.008	−0.022*	0.014	−0.018
Russia	−0.029***	0.005	−0.015	0.001	−0.019**
Australia	−0.028***	−0.004	−0.027***	0.009	−0.016
Sweden	−0.027***	−0.007	−0.024*	0.001	−0.018
Norway	−0.026***	−0.008	−0.026***	0.014	−0.008
Brazil	−0.026***	−0.009	−0.037***	0.011	−0.019
Czech Republic	−0.023***	0.008	−0.008	−0.013	−0.027**
Canada	−0.023***	0.004	−0.015**	0.006	−0.012
Mexico	−0.022***	0.004	−0.017**	−0.006	−0.025**
Turkey	−0.022***	0.024**	−0.005	0.014	−0.013
New Zealand	−0.020***	−0.004	−0.022**	0.004	−0.016
Chile	−0.017***	−0.007	−0.020***	0.008	−0.007
Colombia	−0.016***	0.010	−0.008	0.007	−0.010
South Korea	−0.014***	0.011	−0.006	0.010	−0.006
United Kingdom	−0.013***	−0.011	−0.018**	−0.005	−0.015
Denmark	−0.011***	0.003	−0.003	0.002	−0.005
Belgium	−0.011***	0.003	−0.003	0.002	−0.005
Euro Area	−0.011***	0.003	−0.003	0.002	−0.005
Finland	−0.011***	0.003	−0.003	0.002	−0.005
France	−0.011***	0.003	−0.003	0.002	−0.005
Germany	−0.011***	0.003	−0.003	0.002	−0.005
Greece	−0.011***	0.003	−0.003	0.002	−0.005
Ireland	−0.011***	0.003	−0.003	0.002	−0.005
Italy	−0.011***	0.003	−0.003	0.002	−0.005
Netherlands	−0.011***	0.003	−0.003	0.002	−0.005
Portugal	−0.011***	0.003	−0.003	0.002	−0.005
Spain	−0.011***	0.003	−0.003	0.002	−0.005
Singapore	−0.010***	−0.002	−0.009*	0.003	−0.005
Malaysia	−0.009***	−0.003	−0.010***	0.000	−0.008*
Indonesia	−0.008**	−0.011	−0.017**	0.000	−0.007
Philippines	−0.005***	0.001	−0.004	−0.002	−0.006
Taiwan	−0.003**	0.001	−0.002	0.002	−0.001
Thailand	−0.003**	0.000	−0.003	0.005	0.001
Pakistan	−0.001	−0.005	−0.006	0.006***	0.004**
China	−0.001*	−0.001	−0.002	−0.001	−0.001
Hong Kong	0.000***	0.000	0.000	0.000	0.000
Switzerland	0.007*	−0.003	0.009	0.006	0.015*
Japan	0.033***	−0.002	0.033***	0.011*	0.043***

* p <0.1, ** p<0.05, *** p<0.01

Table 3.3.

Spillovers to Exchange Rates (with respect to US dollar)—Daily Regressions

Country	β₁	β₂(FOMC)	β₁+β₂(FOM)	β₂(Econ.)	β₁+β₂(Econ.)
Poland	−0.033***	0.012	−0.013	−0.018	−0.040***
Hungary	−0.033***	0.016	−0.011	−0.001	−0.027
South Africa	−0.030***	0.008	−0.022*	0.014	−0.018
Russia	−0.029***	0.005	−0.015	0.001	−0.019**
Australia	−0.028***	−0.004	−0.027***	0.009	−0.016
Sweden	−0.027***	−0.007	−0.024*	0.001	−0.018
Norway	−0.026***	−0.008	−0.026***	0.014	−0.008
Brazil	−0.026***	−0.009	−0.037***	0.011	−0.019
Czech Republic	−0.023***	0.008	−0.008	−0.013	−0.027**
Canada	−0.023***	0.004	−0.015**	0.006	−0.012
Mexico	−0.022***	0.004	−0.017**	−0.006	−0.025**
Turkey	−0.022***	0.024**	−0.005	0.014	−0.013
New Zealand	−0.020***	−0.004	−0.022**	0.004	−0.016
Chile	−0.017***	−0.007	−0.020***	0.008	−0.007
Colombia	−0.016***	0.010	−0.008	0.007	−0.010
South Korea	−0.014***	0.011	−0.006	0.010	−0.006
United Kingdom	−0.013***	−0.011	−0.018**	−0.005	−0.015
Denmark	−0.011***	0.003	−0.003	0.002	−0.005
Belgium	−0.011***	0.003	−0.003	0.002	−0.005
Euro Area	−0.011***	0.003	−0.003	0.002	−0.005
Finland	−0.011***	0.003	−0.003	0.002	−0.005
France	−0.011***	0.003	−0.003	0.002	−0.005
Germany	−0.011***	0.003	−0.003	0.002	−0.005
Greece	−0.011***	0.003	−0.003	0.002	−0.005
Ireland	−0.011***	0.003	−0.003	0.002	−0.005
Italy	−0.011***	0.003	−0.003	0.002	−0.005
Netherlands	−0.011***	0.003	−0.003	0.002	−0.005
Portugal	−0.011***	0.003	−0.003	0.002	−0.005
Spain	−0.011***	0.003	−0.003	0.002	−0.005
Singapore	−0.010***	−0.002	−0.009*	0.003	−0.005
Malaysia	−0.009***	−0.003	−0.010***	0.000	−0.008*
Indonesia	−0.008**	−0.011	−0.017**	0.000	−0.007
Philippines	−0.005***	0.001	−0.004	−0.002	−0.006
Taiwan	−0.003**	0.001	−0.002	0.002	−0.001
Thailand	−0.003**	0.000	−0.003	0.005	0.001
Pakistan	−0.001	−0.005	−0.006	0.006***	0.004**
China	−0.001*	−0.001	−0.002	−0.001	−0.001
Hong Kong	0.000***	0.000	0.000	0.000	0.000
Switzerland	0.007*	−0.003	0.009	0.006	0.015*
Japan	0.033***	−0.002	0.033***	0.011*	0.043***

* p <0.1, ** p<0.05, *** p<0.01

Table 3.4a.

Determinants of Spillover Effects on Sovereign Yields— Panel-Data Regressions

Table 3.4a.

Determinants of Spillover Effects on Sovereign Yields— Panel-Data Regressions

Country characteristics	β₁₂	β₂₂(FOMC)	β₂₂(Econ.)
MOB	21.75***	−16.70***	−9.90*
Bank assets to GDP	3.39***	−3.74**	−1.90
Currency flexibility	−0.11**	−0.01	0.06
Sovereign risk	−4.31***	4.36***	0.31
(Exports-to-US)/GDP	3.78**	1.92	−3.12
* p <0.1, p<0.05, * p<0.01

Table 3.4a.

Determinants of Spillover Effects on Sovereign Yields— Panel-Data Regressions

Country characteristics	β₁₂	β₂₂(FOMC)	β₂₂(Econ.)
MOB	21.75***	−16.70***	−9.90*
Bank assets to GDP	3.39***	−3.74**	−1.90
Currency flexibility	−0.11**	−0.01	0.06
Sovereign risk	−4.31***	4.36***	0.31
(Exports-to-US)/GDP	3.78**	1.92	−3.12
* p <0.1, p<0.05, * p<0.01

Table 3.4b.

Determinants of Spillover Effects on Stock Prices—Panel-Data Regressions

Table 3.4b.

Determinants of Spillover Effects on Stock Prices—Panel-Data Regressions

Country characteristics	β₁₂	β₂₂(FOMC)	β₂₂(Econ.)
MOB	5.05***	−1.50	2.04
Bank assets to GDP	0.76***	−0.28	0.27
Currency flexibility	0.07***	−0.03	0.00
Sovereign risk	−0.24***	0.14	−0.21
(Exports-to-US)/GDP	−0.75***	0.37	−0.18
* p <0.1, p<0.05, * p<0.01

Table 3.4b.

Determinants of Spillover Effects on Stock Prices—Panel-Data Regressions

Country characteristics	β₁₂	β₂₂(FOMC)	β₂₂(Econ.)
MOB	5.05***	−1.50	2.04
Bank assets to GDP	0.76***	−0.28	0.27
Currency flexibility	0.07***	−0.03	0.00
Sovereign risk	−0.24***	0.14	−0.21
(Exports-to-US)/GDP	−0.75***	0.37	−0.18
* p <0.1, p<0.05, * p<0.01

Table 3.4c.

Determinants of Spillover Effects on Exchange Rates—Panel-Data Regressions

Table 3.4c.

Determinants of Spillover Effects on Exchange Rates—Panel-Data Regressions

Country characteristics	β₁₂	β₂₂(FOMC)	β₂₂(Econ.)
MOB	0.39	0.80*	0.47
Bank assets to GDP	0.20***	0.07	0.06
Currency flexibility	−0.06***	0.02*	0.01
Sovereign risk	−0.14***	−0.07	−0.02
(Exports-to-US)/GDP	−0.01	−0.02	−0.23
* p <0.1, p<0.05, * p<0.01

Table 3.4c.

Determinants of Spillover Effects on Exchange Rates—Panel-Data Regressions

Country characteristics	β₁₂	β₂₂(FOMC)	β₂₂(Econ.)
MOB	0.39	0.80*	0.47
Bank assets to GDP	0.20***	0.07	0.06
Currency flexibility	−0.06***	0.02*	0.01
Sovereign risk	−0.14***	−0.07	−0.02
(Exports-to-US)/GDP	−0.01	−0.02	−0.23
* p <0.1, p<0.05, * p<0.01

Table 4.1.

Effects of Selected Exogenous Variables on Bond Yield

Note: The first column is based on equations 1-12. The second column deletes equation 7 and converts E to an exogenous variable. The third column keeps equation 7 but makes NOF endogenous and E exogenous. Equality and inequality signs are placed between cells when the indicated relationship holds for the baseline and all alternative parameters. (Alternatives were tried one at a time and not in combination.)

Table 4.1.

Effects of Selected Exogenous Variables on Bond Yield

	E Flexible		E Fixed Using RS		E Fixed Using NOF
NOF, Low Mobility	±	<	+		n.a.
	V		Λ
NOF, High Mobility	±	<	+		n.a.
QE, Low Mobility	−		±		−
	Λ				Λ
QE, High Mobility	−		±		−
RSF, Low Mobility	±		±		0
	V		V		ǁ
RSF, High Mobility	±		±		0
RLF, Low Mobility	+		+		+
	Λ		Λ		Λ
RLF, High Mobility	+		+		+
UCA, Low Mobility	+	>	−	<	+
					V
UCA, High Mobility	+	>	−	<	+

Table 4.1.

Effects of Selected Exogenous Variables on Bond Yield

	E Flexible		E Fixed Using RS		E Fixed Using NOF
NOF, Low Mobility	±	<	+		n.a.
	V		Λ
NOF, High Mobility	±	<	+		n.a.
QE, Low Mobility	−		±		−
	Λ				Λ
QE, High Mobility	−		±		−
RSF, Low Mobility	±		±		0
	V		V		ǁ
RSF, High Mobility	±		±		0
RLF, Low Mobility	+		+		+
	Λ		Λ		Λ
RLF, High Mobility	+		+		+
UCA, Low Mobility	+	>	−	<	+
					V
UCA, High Mobility	+	>	−	<	+

Table 4.2.

Effects of Selected Exogenous Variables on the Current Account

Table 4.2.

Effects of Selected Exogenous Variables on the Current Account

	E Flexible	E Fixed Using RS		E Fixed Using NOF
NOF, Low Mobility	+	+		n.a.
	V	V
NOF, High Mobility	+	+		n.a.
QE, Low Mobility	±0	±0	>	−
				Λ
QE, High Mobility	±0	±0	>	−
RSF, Low Mobility	+	+	>	0
	Λ	Λ		ǁ
RSF, High Mobility	+	+	>	0
RLF, Low Mobility	+	+		+
	Λ	Λ		Λ
RLF, High Mobility	+	+		+
UCA, Low Mobility	+	+	<	+
	Λ	Λ		V
UCA, High Mobility	+	+	<	+

Table 4.2.

Effects of Selected Exogenous Variables on the Current Account

	E Flexible	E Fixed Using RS		E Fixed Using NOF
NOF, Low Mobility	+	+		n.a.
	V	V
NOF, High Mobility	+	+		n.a.
QE, Low Mobility	±0	±0	>	−
				Λ
QE, High Mobility	±0	±0	>	−
RSF, Low Mobility	+	+	>	0
	Λ	Λ		ǁ
RSF, High Mobility	+	+	>	0
RLF, Low Mobility	+	+		+
	Λ	Λ		Λ
RLF, High Mobility	+	+		+
UCA, Low Mobility	+	+	<	+
	Λ	Λ		V
UCA, High Mobility	+	+	<	+

Table 4.3.

Effects of Selected Exogenous Variables on Income (GDP)

Table 4.3.

Effects of Selected Exogenous Variables on Income (GDP)

	E Flexible		E Fixed Using RS		E Fixed Using NOF
NOF, Low Mobility	+	>	−		n.a.
	V		Λ
NOF, High Mobility	+	>	−		n.a.
QE, Low Mobility	+	>	±0	<	+
	V				V
QE, High Mobility	+	>	±0	<	+
RSF, Low Mobility	+	>	−	<	0
	Λ		V		ǁ
RSF, High Mobility	+	>	−	<	0
RLF, Low Mobility	±0	>	−	<	−
			V		V
RLF, High Mobility	±0	>	−	<	−
UCA, Low Mobility	+	<	+	>	+
	Λ		V		Λ
UCA, High Mobility	+	<	+	>	+

Table 4.3.

Effects of Selected Exogenous Variables on Income (GDP)

	E Flexible		E Fixed Using RS		E Fixed Using NOF
NOF, Low Mobility	+	>	−		n.a.
	V		Λ
NOF, High Mobility	+	>	−		n.a.
QE, Low Mobility	+	>	±0	<	+
	V				V
QE, High Mobility	+	>	±0	<	+
RSF, Low Mobility	+	>	−	<	0
	Λ		V		ǁ
RSF, High Mobility	+	>	−	<	0
RLF, Low Mobility	±0	>	−	<	−
			V		V
RLF, High Mobility	±0	>	−	<	−
UCA, Low Mobility	+	<	+	>	+
	Λ		V		Λ
UCA, High Mobility	+	<	+	>	+

Sources:	IMF, Balance of Payments Statistics version 6 (BOP); IMF, International Financial Statistics (IFS); IMF, Monetary and Financial Statistics; IMF, World Economic Outlook (WEO); United Nations, World Population Prospects 2010 (UN); World Bank, World Development Indicators (WDI); World Bank, Worldwide Governance Indicators (WGI); Norway, Norges Bank; and Singapore, Ministry of Finance. Data on capital controls and exchange rate volatility are from Aizenman, Chinn, and Ito (2015). Daily financial data are from Bloomberg.
Annual Regressions
CAX:	The BOP current account balance minus BOP net investment income.
NPFX:	BOP net financial account flows minus BOP net investment income minus NOF.
NOF:	Based on the BOP data, NOF is the sum of reserves flows and net portfolio investment and other investment flows for central bank and general government except that portfolio liability flows are set at zero for advanced economies because they do not borrow significantly in foreign currency. Nonreserve flows data for Norway are from the Norges Bank website for the Norwegian Pension Fund (Global). Debt forgiveness is removed from NOF but not NOA.
NOA:	Based on the IIP data, NOA is defined as the stock version of NOF. Missing values of liabilities are filled in from the World Bank’s external debt data. In countries in which less data is available for the stock than the flow variable, we use perpetual inventory to project NOA backwards. For Norway, NOA is the sum of reserves and Pension Fund (Global) assets.
GDP:	Nominal GDP in US dollars and in local currency, and real GDP, are from WEO.
MOB:	Capital controls index available at Aizenman, Chinn, and Ito (2015).
QE:	Central bank domestic assets. Source: IMF Monetary and Financial Statistics (MFS) and International Financial Statistics (IFS).
SPILL:	Global Financial Integration multiplied with the sum of NOF across countries and divided by the sum of trend GDP across countries.
Relative PPP
GDP Per Capita: WEO (relative to US level). We set this as missing before 1996 for European transition economies.
Aging:	10-year forward change in ratio of elderly to working age population. Historical elderly ratios through 2010 are from WDI. Ratios for 2020 and 2020 are from UN and are interpolated and extrapolated in order to create 10-year changes for 2001–15.
Growth:	5-year moving average of growth rate of real GDP based on WEO. We corrected an error in Malta real GDP using IFS data. We set real GDP growth as missing for European transition economies before 1996.
Net Energy Exports: Difference between energy production and consumption in tons of oil equivalent (WDI), converted into dollars using Brent oil price (IFS) assuming 7.33 barrels per ton.
Fiscal Balance: General government balance in percent of GDP (WEO) is cyclically adjusted as the residual in a panel regression of the fiscal balance on the level and change of the GDP gap with no country or year effects. The GDP gap is the difference between log real GDP and its 11-year centered moving average using WEO forecasts for 2015–18. A missing value for South Africa in 2005 is interpolated.
Global Financial Integration: Defined as the ratio of BOP private financial account transactions divided by the sum of financial and current account transactions.
Non-reserve Flows: NOF minus reserve assets flow divided by trend GDP.
Crisis:	A dummy that takes the value 1 if the respective country experienced a financial or currency crisis in the previous three years. Source: Laeven and Valencia (2011).
Trade Openness: Exports of goods and services plus imports of goods and services divided by trend GDP.
Exchange rate regime: IMF de facto (coarse) index from www.carmenreinhart.com and Aizenman-Chinn-Ito (2015) rolling measure of ER volatility.
Scaling by trend GDP: When scaling data by GDP, we use the 11-year centered moving average of nominal GDP in US dollars (WEO), including forecast data through 2018.
Daily Regressions
	(Data are from Bloomberg unless otherwise noted.)
Currency flexibility: Sample period daily standard deviation of the exchange rate with respect to the U.S. dollar.
Exchange rate: The log of the local exchange rate (in dollars per unit of non U.S. currency.)
(Exports-to-U.S.)/GDP: Sample average ratio of exports to the United States to domestic GDP based on annual regression data.
FOMC Dummy (D_(U.S.,t)): D takes the value 0 on most days and 1 on days on which the FOMC released either a policy statement, policy minutes, or there was a monetary policy speech by the FOMC Chair. (Constructed by authors)
Econ Dummy (D_(U.S.,t)): D takes the value 1 on days when data were released on U.S. nonfarm payrolls or the ISM purchasing managers’ index, and 0 otherwise. (Constructed by authors)
MOB:	The sample average of MOB from the annual regression data.
Bank assets to GDP: Average from 2008–12 from Helgi Library (http://www.helgilibrary.com/indicators/index/bank-assets-as-of-gdp).
Sovereign bond yield: Daily yield of 10-year constant maturity sovereign local-currency bond.
Sovereign risk: Sample-period average sovereign bond yields.
Stock price: Log of the daily local stock market price index.
U.S. sovereign yield: Daily yield of 10-year constant maturity U.S. Treasury bond.

Sources:	IMF, Balance of Payments Statistics version 6 (BOP); IMF, International Financial Statistics (IFS); IMF, Monetary and Financial Statistics; IMF, World Economic Outlook (WEO); United Nations, World Population Prospects 2010 (UN); World Bank, World Development Indicators (WDI); World Bank, Worldwide Governance Indicators (WGI); Norway, Norges Bank; and Singapore, Ministry of Finance. Data on capital controls and exchange rate volatility are from Aizenman, Chinn, and Ito (2015). Daily financial data are from Bloomberg.
Annual Regressions
CAX:	The BOP current account balance minus BOP net investment income.
NPFX:	BOP net financial account flows minus BOP net investment income minus NOF.
NOF:	Based on the BOP data, NOF is the sum of reserves flows and net portfolio investment and other investment flows for central bank and general government except that portfolio liability flows are set at zero for advanced economies because they do not borrow significantly in foreign currency. Nonreserve flows data for Norway are from the Norges Bank website for the Norwegian Pension Fund (Global). Debt forgiveness is removed from NOF but not NOA.
NOA:	Based on the IIP data, NOA is defined as the stock version of NOF. Missing values of liabilities are filled in from the World Bank’s external debt data. In countries in which less data is available for the stock than the flow variable, we use perpetual inventory to project NOA backwards. For Norway, NOA is the sum of reserves and Pension Fund (Global) assets.
GDP:	Nominal GDP in US dollars and in local currency, and real GDP, are from WEO.
MOB:	Capital controls index available at Aizenman, Chinn, and Ito (2015).
QE:	Central bank domestic assets. Source: IMF Monetary and Financial Statistics (MFS) and International Financial Statistics (IFS).
SPILL:	Global Financial Integration multiplied with the sum of NOF across countries and divided by the sum of trend GDP across countries.
Relative PPP
GDP Per Capita: WEO (relative to US level). We set this as missing before 1996 for European transition economies.
Aging:	10-year forward change in ratio of elderly to working age population. Historical elderly ratios through 2010 are from WDI. Ratios for 2020 and 2020 are from UN and are interpolated and extrapolated in order to create 10-year changes for 2001–15.
Growth:	5-year moving average of growth rate of real GDP based on WEO. We corrected an error in Malta real GDP using IFS data. We set real GDP growth as missing for European transition economies before 1996.
Net Energy Exports: Difference between energy production and consumption in tons of oil equivalent (WDI), converted into dollars using Brent oil price (IFS) assuming 7.33 barrels per ton.
Fiscal Balance: General government balance in percent of GDP (WEO) is cyclically adjusted as the residual in a panel regression of the fiscal balance on the level and change of the GDP gap with no country or year effects. The GDP gap is the difference between log real GDP and its 11-year centered moving average using WEO forecasts for 2015–18. A missing value for South Africa in 2005 is interpolated.
Global Financial Integration: Defined as the ratio of BOP private financial account transactions divided by the sum of financial and current account transactions.
Non-reserve Flows: NOF minus reserve assets flow divided by trend GDP.
Crisis:	A dummy that takes the value 1 if the respective country experienced a financial or currency crisis in the previous three years. Source: Laeven and Valencia (2011).
Trade Openness: Exports of goods and services plus imports of goods and services divided by trend GDP.
Exchange rate regime: IMF de facto (coarse) index from www.carmenreinhart.com and Aizenman-Chinn-Ito (2015) rolling measure of ER volatility.
Scaling by trend GDP: When scaling data by GDP, we use the 11-year centered moving average of nominal GDP in US dollars (WEO), including forecast data through 2018.
Daily Regressions
	(Data are from Bloomberg unless otherwise noted.)
Currency flexibility: Sample period daily standard deviation of the exchange rate with respect to the U.S. dollar.
Exchange rate: The log of the local exchange rate (in dollars per unit of non U.S. currency.)
(Exports-to-U.S.)/GDP: Sample average ratio of exports to the United States to domestic GDP based on annual regression data.
FOMC Dummy (D_(U.S.,t)): D takes the value 0 on most days and 1 on days on which the FOMC released either a policy statement, policy minutes, or there was a monetary policy speech by the FOMC Chair. (Constructed by authors)
Econ Dummy (D_(U.S.,t)): D takes the value 1 on days when data were released on U.S. nonfarm payrolls or the ISM purchasing managers’ index, and 0 otherwise. (Constructed by authors)
MOB:	The sample average of MOB from the annual regression data.
Bank assets to GDP: Average from 2008–12 from Helgi Library (http://www.helgilibrary.com/indicators/index/bank-assets-as-of-gdp).
Sovereign bond yield: Daily yield of 10-year constant maturity sovereign local-currency bond.
Sovereign risk: Sample-period average sovereign bond yields.
Stock price: Log of the daily local stock market price index.
U.S. sovereign yield: Daily yield of 10-year constant maturity U.S. Treasury bond.

First stage for NOF
	CAX	NPFX	CAX	NPFX	CAX	NPFX
	(1)	(2)	(3)	(4)	(5)	(6)
Nonreserve Flows	0.851***	0.851***	0.851***	0.851***	0.868***	0.868***
Interaction with MOB	−0.108	−0.108	−0.108	−0.108	−0.133	−0.133
Crisis	0.004	0.004	0.004	0.004	0.004	0.004
Interaction with MOB	0.004	0.004	0.004	0.004	0.008	0.008
R-squared	0.59	0.59	0.59	0.59	0.59	0.59
	68.5	68.5	68.5	68.5	55.8
F-test	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	55.8 (0.000)
	69.9	69.9	69.9	69.9	58.1
AP Chi-sq test	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	58.1 (0.000)
First stage for NOF interaction
	CAX	NPFX	CAX	NPFX	CAX	NPFX
	(1)	(2)	(3)	(4)	(5)	(6)
Nonreserve Flows	0.032	0.032	0.032	0.032	0.024	0.024
Interaction with MOB	0.709***	0.709***	0.709***	0.709***	0.718***	0.718***
Crisis	−0.001	−0.001	−0.001	−0.001	−0.002	−0.002
Interaction with MOB	0.009*	0.009*	0.009*	0.009*	0.015**	0.015**
R-squared	0.61	0.61	0.61	0.61	0.60	0.60
	56.0	56.0	56.0	56.0	43.3
F-test	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	43.3 (0.000)
	98.6	98.6	98.6	98.6	87.3
AP Chi-sq test	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	87.3 (0.000)

First stage for NOF
	CAX	NPFX	CAX	NPFX	CAX	NPFX
	(1)	(2)	(3)	(4)	(5)	(6)
Nonreserve Flows	0.851***	0.851***	0.851***	0.851***	0.868***	0.868***
Interaction with MOB	−0.108	−0.108	−0.108	−0.108	−0.133	−0.133
Crisis	0.004	0.004	0.004	0.004	0.004	0.004
Interaction with MOB	0.004	0.004	0.004	0.004	0.008	0.008
R-squared	0.59	0.59	0.59	0.59	0.59	0.59
	68.5	68.5	68.5	68.5	55.8
F-test	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	55.8 (0.000)
	69.9	69.9	69.9	69.9	58.1
AP Chi-sq test	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	58.1 (0.000)
First stage for NOF interaction
	CAX	NPFX	CAX	NPFX	CAX	NPFX
	(1)	(2)	(3)	(4)	(5)	(6)
Nonreserve Flows	0.032	0.032	0.032	0.032	0.024	0.024
Interaction with MOB	0.709***	0.709***	0.709***	0.709***	0.718***	0.718***
Crisis	−0.001	−0.001	−0.001	−0.001	−0.002	−0.002
Interaction with MOB	0.009*	0.009*	0.009*	0.009*	0.015**	0.015**
R-squared	0.61	0.61	0.61	0.61	0.60	0.60
	56.0	56.0	56.0	56.0	43.3
F-test	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	43.3 (0.000)
	98.6	98.6	98.6	98.6	87.3
AP Chi-sq test	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	87.3 (0.000)

Topics

Countries

Series

About

Resources

Contributor Notes

Abstract

I. Introduction

II. Current Account Effects of Net Official Flows and Quantitative Easing in Annual Panel Data

A. Empirical Specification

B. Baseline Regression Results

C. Robustness Checks

D. Influential Observation Analysis

E. Fiscal Balances, Net Official Flows, and Energy Exporters

F. Dynamics

III. Financial Market Effects of US Unconventional Monetary Policy in Daily Data

A. Daily Co-Movements, Empirical Specification and Results

B. Relating Financial Spillovers to Country Characteristics

IV. Imperfect Asset Substitution in an Open Economy Macro Model

A. Model Specification

B. Model Properties and Relation to Annual Regression Results

C. Model Properties and Relation to Daily Regression Results

D. Policy Implications

V. Conclusions

References

Appendix: Data Sources and Definitions