The Rise of Dollar Credit in Emerging Market Economies and US Monetary Policy

The Rise of Dollar Credit in Emerging Market

Economies and US Monetary Policy

Introduction

The surge of dollar credit in the emerging market economies (EMEs henceforth) has

drawn close attention from the international financial market participants and the

policymakers due to its potential role in the monetary and financial stability of the global

economy. The data from the Bank of International Settlements (BIS) shows that the

outstanding US dollar credit to non-bank borrowers in the EMEs has risen from 1.69

trillion dollar in 2008Q4 to 3.25 trillion dollar in 2015Q4. The cyclical nature of EME

dollar credit

could also amplify the overall credit condition in the domestic economy and

increase the vulnerability of these the EMEs in dealing with negative economic shocks.

For example, during the expansionary phase, the dollar credit floods into the EMEs, which

raises the challenge to stabilize the domestic monetary base. In the contractionary phase,

the dramatic outflow of dollar credit also creates difficulty to stabilize the exchange rate

and asset prices (Avdjiev et al., 2012). This was very evident during the ’taper tantrum’ of

2013. On the other hand, however, it has also been argued that the increasing role of dollar

credit in the EMEs may signify a more integrated global financial market and a greater

degree of risk sharing across different countries.

The growing academic interest in the literature on examining the dollar credit in the

EMEs has also coincided with the recent emphasis on the role of US monetary policy in

boosting global liquidity. Global liquidity refers to the global factor that drives cross-

border spillover in financial conditions and credit growth. Shin (2013) proposes two

phases of global liquidity after the Millennium. The first phase of global liquidity

transmission (2002-2008) is more associated in form of bank loans through global

banking system. The international bond market gradually took over the share of bank

loans in the second phase of global liquidity transmission starting in 2009. Since then, the

large scale of bond purchases as a result of of Quantitative Easing (QE) programs had led

to the portfolio re-balancing effect

in the international bond market. A few papers and

policy studies have provided narrative evidence (Borio, et al. 2011; McCauley, et al., 2015)

in support of the hypothesis that the abundance of global liquidity was one of the causes

of the boom of the dollar credit in the EMEs.

Although much attention has been given to the rise of dollar credit and its apparent

relationship with US monetary policy and the valuation of the US dollar, not much work

has been done to systematically disentangle the short-run and the long-run relationships

among these variables. It is perfectly plausible to think that the short-run relationship

between dollar credit in emerging markets and the stance of monetary policy in the US

may be very different from its long-run relationship. The long-run dollar credit in the

emerging market may be driven more by its long-run absorptive capacity instead of the

short-run increase in liquidity or the weakness of the US dollar. Therefore, it is very

important to decompose the overall EME dollar credit and isolate the cyclical variations

from its long-term trend, by taking into account the information on US interest rate and

the dollar valuation. In addition to estimating the short-run and the long-run correlation

between the shocks to each series, this approach will also yield us a quantitative estimate

of how big the cyclical component in dollar credit was at different points in time during

the last few years.

To examine the long-run and short-run relationship among these variables, we

propose to use a correlated multivariate unobserved component (UC hereafter) model

which is a multivariate counterpart of the correlated univariate UC model as outlined in

Morley, Nelson and Zivot (2003).

This model allows us to decompose the movements in

dollar credit, interest rate and dollar index

into a slow-moving trend component and a

cyclical component simultaneously. The slow-moving trend captures the long-run

evolution of these variables and the cyclical component captures the short-run

movements. This model allows us to not only estimate the permanent and transitory

movements, but also provides us a measure of correlation of the long-run movements and

the short-run movements among dollar credit, interest rate and dollar index.

For

example, if the recent narrative about the role of exceptionally low interest rate in

boosting EME dollar credit is correct, then we would observe a negative correlation

between the shock to the transitory component of dollar credit and the shock to transitory

component of the US interest rate. To take into account the zero lower bound problem

associated with the federal funds rate and the short-term interest rate in the recent time

period, we use the shadow interest rate as proposed by Wu and Xia (2016) as a proxy for

US monetary policy stance. The shadow interest rate takes into account the impact of

unconventional monetary policy on interest rate and unlike the federal funds rate, is

allowed to fall below zero.

To understand the intuition behind the structure of our model, one could think of the

credit activities in the EMEs as a form of international investment. From the investors’

perspective, the return of lending dollar in the EMEs depends on the interest rate paid

from these credit instruments. The US interest rate, the risk-free rate in the international

credit market, is a major factor in pricing the interest rates on the international bank loans

and the corporate bond yields. From the EME borrowers’ perspective, the dollar valuation

in the currency market is also critical in determining their real external debt burden,

besides the US monetary policy rate. The expectation of domestic currency appreciation

will lower the expected external debt burden in the future and vice versa. To summarize,

we can think of this three-variable dynamic system as the application of interest rate

parity in international credit activities.

We find interesting and economically meaningful results from the estimated

multivariate correlated unobserved component model. The maximum likelihood

estimates of our correlated multivariate UC model suggest that there is a strong negative

correlation between the transitory shock to dollar credit and the transitory shock to the

US interest rates. This suggests that a temporary decline in interest rate below its long-

run level is associated with an increase in dollar credit above its long-run level. We also

find a very high negative correlation between the transitory shocks to dollar credit and

the transitory shocks to the dollar index, implying an appreciation of US dollar is

associated with a decline in dollar credit in the EMEs in the short-run. These results

support the anecdotal and narrative evidence on the strong relationship between dollar

credit and the US interest rates and also between dollar credit and the strength of the US

dollar. We also find that the trend-cycle decomposition of EME dollar credit from our

multivariate correlated unobserved component model captures the recent boom and bust

behavior and compares favorably to a univariate trend-cycle decomposition benchmark.

In particular, our results suggest that the dollar credit before the taper tantrum was 10%

above its long-run trend in the emerging market economies.

The rest of the paper is structured as follows. The second section provides a literature

review about global liquidity transmission and discusses the associated monetary policy

spillover effect. The third section introduces the data used in this study and the setup of

the correlated multivariate UC model. The fourth section interprets and presents the

results from the model. The last section concludes the paper.

2.2 Literature Review

The literature on dollar credit in the EMEs is nascent. Few papers have tried to address

this issue from different perspectives. In understanding the phenomenon that dollar

credit outside the US behaves differently from the US domestic dollar credit, Borio,

McCauley and McGuire (2011) take a look at the recent behavior of international credit

and associate it with the overall credit conditions. Their descriptive analysis reveals the

fact that US dollar credit in some countries has been outgrowing the overall credit during

the credit booms. A formal analysis later on from Avdjiev, McCauley and McGuire (2012)

regresses the cross-sectional change in credit-to-GDP ratio on the change in international

credit during the credit boom phase (2002-2008) and regresses the credit growth in the

EMEs on the change in international credit during the credit bust phase (2008-2011).

Their results suggest that international credit amplifies the overall credit cycles in the

EMEs. While in these studies, the authors often use the share of international credit in the

overall credit to measure the cyclical variation of international credit, it is worthwhile to

recognize that the trend of international credit and the trend of the overall credit can be

driven by different underlying factors, therefore they do not need to share the common

trend in the numerator and denominator. For instance, the long-term trend of

international credit could be explained by the integration of international financial

market, but the long-term trend of the overall credit condition may respond more to the

domestic economic fundamentals. Thus without isolating the long-term trend from these

two credit series, it is very difficult to perform a clean analysis about the cyclical

comovement between the international credit and the overall credit condition.

Since the recent global financial crisis of 2008-2009, the outstanding US dollar credit

to the non-bank borrowers in the EMEs has roughly doubled within the past seven years.

The commentators associate this surge to US monetary policy spillover, mainly because of

the ultra low interest rates in the US. He and McCauley (2013) survey a number of studies

on the transmission of monetary policy of the major advanced economies to East Asia and

conclude that policy rates, bond yields and exchange rates are three price channels in the

transmission process. McCauley, McGuire and Suchko (2015) also link the US monetary

policy, leverage and flow into bond funds to explain the dollar credit extended to non-US

borrowers. In this paper, since our goal is to understand the behavior of the overall dollar

credit condition, regardless of the credit instruments, we focus on the two price

channels—interest rate and currency appreciation/depreciation, an analytical framework

which can be interpreted as based upon international interest rate parity theory.

It should be pointed out that we focus on EME dollar credit, a subject we believe to be

ideal to study the monetary policy spillover effect in global liquidity transmission. This is

because, the composition of bank loan and bond issuance depends on country-specific

contexts, for instance, the regulatory emphasis of capital control. Tighter regulation on

international banking practice could push the domestic borrowers to issue bonds

overseas and vice versa (Caballero et al., 2015). Since our goal is to understand the trend

and cycle of the dollar credit in emerging market economies, we purposely ignore the

credit breakdown based on types of financial instruments. Furthermore, we choose to

study EME dollar credit, instead of Euro, Yen or other currency credit, is because different

currency credit could be sensitive to different monetary policy rates and dollar credit

dominates other currencies in the currency breakdown of international credit (Borio, et

al., 2011). The last but not the least, we work with the dollar credit in the EMEs, because

the dollar credit outside US may respond to factors in different ways when it comes to the

EMEs compared to other advanced economies. (Borio, et al. 2011; McCauley, et al., 2015)

2.3 Data and Empirical Model

2.3.1 Data Description

Our sample period spans from the first quarter of 2000 to the last quarter of 2015. The

sample period is based on data availability. Measuring dollar credit can be a challenging

exercise. Fortunately, the Bank of International Settlements (BIS) website provides global

liquidity indicators to measure the ease of financing in global financial markets.

Among

various global credit aggregates, we use the US dollar credit to non-bank sector in the

EMEs in this study. This measure aggregates all the maturities of credit instruments. The

original data is measured in trillions of US dollar and we take the natural log of this series

and use the log-transformed series in the model estimation.

The second variable in our exercise is US interest rate that proxies the stance of US

monetary policy. The ideal candidate would have been the federal funds rate if there was

not a zero lower bound (ZLB) problem. The ZLB issue arises during and after the Great

Recession, as the Fed quickly lowered the federal funds rate close to zero and also

implemented the unconventional monetary policies, including large-scale purchases of

financial assets from private financial corporations. These unconventional monetary

policies helped inject more liquidity into the market than implied by the federal funds

rate, which basically had no room to be lowered further. The overall monetary policy

stance thus can not simply be captured by the variations in the federal funds rate itself.

To take this ZLB problem into account, Wu and Xia (2016) proposed to use the shadow

interest rate to provide a comprehensive measure to summarize the overall stance of

monetary policy while the federal funds rate stuck at the ZLB environment. The shadow

interest rate measure originated from the idea in Black (1995) to price the interest rate

as an option. Wu and Xia (2016) conducted an analytical approximation for the forward

rate in the Shadow Rate Term Structure Model (SRTSM), by linearizing the state-space

model representation of the three-factor SRTSM. Then they used the estimated

parameters and decomposed three unobserved factors to compute the shadow interest

rate. The likelihood ratio test could not reject the hypothesis that the parameters relating

the shadow interest rate to key macroeconomic variables under the ZLB environment are

the same as those that related the federal funds rate to those variables before the Great

Recession. Because of its intuitive appeal, Wu and Xia (2016) shadow interest rate

measure has been gaining widespread attention.

Because of the above mentioned reasons, we use the shadow interest rate proposed

by Wu and Xia (2016) as a proxy for the Fed’s overall monetary policy stance. The reason

we choose to use this measure in our benchmark estimation over some alternative

measures, for instance, the longer-term interest rates which suffer less from the ZLB issue,

is because the term premium in the long-term interest rates could potentially contaminate

the monetary policy stance. Nevertheless, we still perform a robustness test with 1-year

treasury bill rate, given the fact that our dollar credit measure aggregates all the credit

instruments regardless of their term to maturity. The third variable in our exercise is the

US dollar exchange rate in the global currency market. We use the broad trade-weighted

US dollar index to measure the dollar valuation in the currency market. The larger the

index, the stronger the US dollar is, or equivalently, the more US dollar appreciates, and

vice versa.

2.3.1 Empirical Model

In this paper, we use a trivariate unobserved component model to model the dynamics

in EME dollar credit (Yt), interest rate (It) and dollar index (Dt). This model is multivariate

extension of the model proposed by Morley, Nelson and Zivot (2003). To understand the

intuition behind the structure of our model, one could think of the credit activities in the

EMEs as a form of international investment. From the investors’ perspective, the return of

lending dollar in the EMEs depends on the interest rate paid from these credit

instruments. The US monetary policy rate, the risk-free rate in the international credit

market, is a major factor in pricing the interest rates on the international bank loans and

the corporate bond yields. From the EME borrowers’ perspective, the dollar valuation in

the currency market is also critical in determining their real external debt burden, besides

the US monetary policy rate. The expectation of domestic currency appreciation will lower

the expected external debt burden in the future and vice versa. To summarize, we can

think of this three-variable dynamic system as the application of interest rate parity in

international credit activities. Moreover, the interest rate parity

//www.frbatlanta.org/cqer/research/shadowrate.aspx?panel = 1

has to hold in both the short-run and the long-run market equilibrium. Therefore, the

dynamics of interest rate and exchange rate will provide useful information in explaining

the dynamics of dollar credit in the EMEs.

Our model takes the following form:

EME Dollar Credit:

Yt = τyt + cyt (1)

τyt = µy + τyt−1 + ηyt,ηyt ∼ iidN(0,σηy2 ) (2)

) (3)

Interest Rate:

It = τit + cit (4)

τit = µi + τit−1 + ηit,ηit ∼ iidN(0,σηi2 ) (5)

cit = φ1icit−1 + φ2icit−2 + εit,εit ∼ iidN(0,σεi2 ) (6)

Dollar Index:

Dt = τdt + cdt (7)

τdt = µd + τdt−1 + ηdt,ηdt ∼ iidN(0,σηd2 ) (8)

cdt = φ1dcdt−1 + φ2dcdt−2 + εdt,εdt ∼ iidN(0,σεd2 ) (9)

Each series is decomposed into a stochastic trend component (τjt,i = Y,IorD) and a cyclical

component (cjt,i = Y,IorD) implying an I(1) process for all the variables. The non-

stationarity of these variables are confirmed by the unit root tests where we do not reject

the null of unit root for all the variables.

We also do not impose the common trend

restriction, i.e., all three variables have their own trend and cycle components and these

components are allowed to have a certain degree of correlation based on the economic

intuition we will discuss later. In fact, we do test for cointegration among these three

variables and do not find evidence to support this for the sample period under study.

Secondly, we specify the dynamics of trend and cycle components. The cyclical

component in each series is assumed to follow an AR (2) process. This assumption

captures the auto correlation structures as observed in the correlogram and provides rich

dynamics in the data series to enable us to identify all the parameters under the state-

space model framework (Morley, Nelson and Zivot, 2003). The trend components are

assumed to follow a random walk process with a drift, and as mentioned above, we do not

impose a common trend among these three variables.

Thirdly, we assume the shocks to the trend and cycle components follow a white noise

process, but allow for non-zero cross-correlation across series. The shocks to the trend

components (ηjt,i = Y,I,orD) have a long-run effect on the trend because the trend is

assumed to follow a random walk process. The shocks to the cyclical component (εjt,i =

Y,IorD) have a short-run effect on the cycle because the cycle follows a stationary

autoregressive process with two lags. The shocks to each trend component are allowed to

be correlated across each other, so are the shocks to the cyclical components. However,

we impose the zero correlation between the shocks to the trend component and the

shocks to the cycle component within and between series. That is to say, we assume that

the shocks that generate a long-run effect are different from the shocks that generate a

short-run effect. This assumption for example, isolates the monetary policy shocks, which

often are considered neutral in the long term, from the productivity shocks, which has a

persistent effect in the real economy.

Below is the correlated multivariate unobserved component model setup based on the

previous discussion. It should be pointed out that, in the variance-covariance matrix of

the shocks to the trend and cycle, σηyηi, σηyηd, and σηiηd are the pairwise covariance of the

shocks to the trend of EME dollar credit, monetary policy rate and dollar index. σεyεi, σεyεd,

and σεiεd are the pairwise covariance of the shocks to the cycle of EME dollar credit,

monetary policy rate and dollar index. The estimates of correlation coefficients, instead

of covariances, will be reported in Table 1. We estimate the model using the classical

maximum likelihood via the Kalman filter.

Measurement Equation:

Variance-Covariance Matrix of the Shocks to Trend and Cycle:

2.4 Results and Interpretation

We present the results of this model in the next three subsections. In the first

subsection, we discuss the parameter estimates and also provide interpretation for the

correlation of the shocks to the trend and cycle among the three variables. In the second

subsection, we discuss the evolution of the estimated trend and cycle of each series and

examine whether the trend and cycle components capture the long-run trend and the

cyclical variation during the sample period. In the third subsection, we look at both the

long-term and short-term comovement among these three variables.

2.4.1 Dynamic Relationship among EME dollar Credit, US Interest

Rate and the Dollar Index

Table 1 provides the maximum likelihood estimates of all the parameters. The

corresponding standard errors are in the parentheses. The results suggest that the cyclical

shocks dominate the variation in the overall EME dollar credit. The standard deviation of

the shocks in the trend of EME dollar credit is 0.014

, smaller compared to the standard

deviation of the shocks in the cycle, 0.017. Additionally, for the shadow interest rate, the

standard deviation of the shocks in the trend is 0.274, almost twice as large as the

standard deviation of the shocks in the cycle. The results seem to suggest that variations

in slow moving component in the shadow interest rate is dominant. However, the

variations in cyclical component are also significant. For the dollar index, the standard

deviation of shocks to the trend and cycle components are 0.015 and 0.016 respectively,

implying similar importance of transitory and permanent variation in dollar index. As far

as the estimated parameters of the cyclical components are concerned, we find that the

cyclical component of all the variables are persistent implying a shock to the cycle, though

transient, persists for a while.

The correlation analysis of the shocks to the cycles among these variables suggests that

the cyclical variation among EME dollar credit, monetary policy rate and dollar credit are

strongly correlated. The correlation coefficient between the transitory shock to dollar

credit and the transitory shock to US interest rate is -0.97. The correlation coefficient

between the transitory shock to EME dollar credit and the transitory shock to dollar index

is -0.94. And the correlation coefficient between the transitory shock to US interest rate

and the the transitory shock to dollar index is 0.99. The standard errors of these estimated

parameters confirm that these correlation coefficients are significantly different from

zero.

The strong negative correlation between the transitory shock to US interest rate and

the transitory shock to dollar credit in the EMEs suggests that, a temporary decrease in

US interest rate below its long-run trend leads to a temporary increase in EME dollar

credit as it lowers the cost of borrowing. In addition to the lower cost, the transitory

decline in interest rate may also reflect temporary abundance of liquidity as witnessed

during the QE programs. The negative correlation between transitory shock to dollar

index and transitory shock to dollar credit in the EMEs is also very intuitive. An increase

in dollar index implies an appreciation of dollar index above its long-run trend and these

movements make the dollar financing activities in the EMEs less appealing temporarily.

We can motivate the strong positive correlation between the transitory shock to US

interest rate and the transitory shock to US dollar index by considering an open economy’s

short-run equilibrium model. At the initial output level and given the sticky price level in

the short-run, an increase in US money supply pushes down the US interest rate. Since the

US monetary change is temporary and does not affect the expected future exchange rate,

so to preserve interest rate parity, the exchange rate must depreciate immediately to

create the expectation that the US dollar will appreciate in the future. Therefore, in the

short run, a negative transitory shock to US monetary policy rate is predicted to associate

with a negative transitory shock to US dollar index, which is exactly what we identify in

our model estimation.

The correlation analysis of the shocks to the trends among these variables confirm our

earlier finding that these three variables do not share a common trend and hence, are not

cointegrated.

The correlation coefficient between the permanent shocks to EME dollar

credit and the permanent shocks to US interest rate is 0.67. The correlation coefficient

between the permanent shock to US interest rate and the permanent shocks to dollar

index is -0.68. The standard errors of these estimated parameters confirm that these

correlation coefficients are significantly different from zero. The correlation coefficient

between the permanent shock to EME dollar credit and the permanent shock to dollar

index is insignificant and low. Therefore, it is not likely that these variables share the

common trend, which supports the conjecture that we need to decompose the trend and

cycle components from the series to understand the business cycles of these variables.

The negative correlation between the permanent shock to US interest rate and the

permanent shock to dollar index seems puzzling at the first glance. However, it can be

explained using the forward looking behavior of these variables. If an unexpected

permanent increase in interest rate provide the signal that inflation is expected to go up

in future, then nominal exchange rate may respond instantaneously in response to this

Table 1: Maximum Likelihood Estimates: A Correlated Multivariate UC Model

Description

Parameter

Estimate

(Standard Error)

Log likelihood value

llv

EME dollar credit

295.8723

S.D. of permanent shocks to the EME dollar credit

σηy

0.0139

(0.0041)

S.D. of temporary shocks to the EME dollar credit

σεy

0.0168

(0.0033)

the EME dollar credit drift

µy

0.0274

(0.0025)

EME dollar credit 1st AR parameter

φ1y

1.3851

(0.1399)

EME dollar credit 2nd AR parameter

φ2y

shadow interest rate

-0.4770

(0.1251)

S.D. of permanent shocks to shadow interest rate

σηi

0.2738

(0.0312)

S.D. of temporary shocks to shadow interest rate

σεi

0.1549

(0.0327)

Shadow interest rate drift

µi

-0.0792

(0.0481)

Shadow interest rate 1st AR parameter

φ1i

1.9072

(0.0355)

Shadow interest rate 2nd AR parameter

φ2i

US dollar index

-0.9462

(0.0363)

S.D. of permanent shocks to US dollar index

σηd

0.0148

(0.0029)

S.D. of temporary shocks to US dollar index

σεd

0.0158

(0.0027)

US dollar index drift

µd

-0.0026

(0.0021)

US dollar index 1st AR parameter

φ1d

1.3585

(0.1144)

US dollar index 2nd AR parameter

φ2d

Cross-series correlations

-0.5726

(0.0980)

Correlation: Permanent dollar credit/

Permanent interest rate

ρηyηi

0.6669

(0.2118)

Correlation: Permanent dollar credit/

Permanent dollar index

ρηyηd

0.0842

(0.3396)

Correlation: Permanent interest rate/

Permanent dollar index

ρηiηd

-0.6864

(0.2532)

Correlation: Transitory dollar credit/

Transitory interest rate

ρεyεi

-0.9719

(0.1157)

Correlation: Transitory dollar credit/

Transitory dollar index

ρεyεd

-0.9406

(0.2205)

Correlation: Transitory interest rate/

Transitory dollar index

ρεiεd

0.9941

(0.0372)

Figure 1: Trends from Correlated Multivariate Unobserved Component Model

The figures below report the trend components of EME dollar credit, shadow interest rate and

dollar index. The dollar credit in the EMEs is measured in log trillions of USD and multiplied by

100. The shadow interest rate is measured in percentage points. The log of US dollar index has

been multiplied by 100.

news about higher than expected inflation. Similarly, the forward-looking behavior can

explain the positive correlation between permanent shock to interest rate and permanent

shock to dollar credit. An unexpected permanent increase in interest rate may provide

information about higher than expected inflation, which in turn may lead to a depreciation

of the dollar as argued before. This will be associated with an unexpected permanent

increase in dollar credit in emerging market economies. Overall, the results from the

correlation analysis clearly suggests value in examining the short-run and the long-run

relationships separately. The results suggest that fundamentals seem to matter more in

the long-run, whereas the short-run boom may be associated with temporary phases of

monetary policy as well as movements in dollar index.

2.4.2 Trend-Cycle Decomposition

In this subsection, we decompose the trend and cycle of EME dollar credit, interest rate

and dollar index using the correlated multivariate unobserved component model. The

stochastic trend in the multivariate UC model captures the long-run evolution in

EME dollar credit and also reflects the effect of recent global financial crisis (Figure 2). In

the long-run, there is an increasing trend in EME dollar credit, due to the global financial

integration. There was a downward shift in the trend during the financial crisis. The effect

of this negative shock on the trend of EME dollar credit persisted for few years.

The cyclical component from our multivariate UC model captures the evolution of the

dollar credit and its dynamic relationship with US interest rate and the dollar index really

well (Figure 2). During the initial part of our sample, the dollar credit cycle was negative

implying lower than potential credit in these countries. After the collapse of Bear Stearns,

emerging market economies provided a sanctuary for the international capital, mainly

because the investor community assumed that the EMEs are decoupled from the

developed markets.In the third quarter of 2008, Lehman Brother filed bankruptcy that led

to financial panic. These events were associated with a global contraction in credit

lending, which brought the dollar credit back down to its long-term trend. During Figure

2: Trend/Cycle Decomposition from Correlated Multivariate UC Model

This figure reports the trend component and the cyclical component of EME dollar credit.

The dollar credit in the EMEs is measured in log trillions of USD and multiplied by 100. The UC

trend is the trend component of EME dollar credit decomposed using the correlated multivariate

unobserved component (UC) model. The HP trend is the trend component of EME dollar credit

decomposed using Hodrick and Prescott (HP) filter.

this time period, the Federal Reserve lowered the policy rate significantly and dumped

liquidity into the financial market through several runs of the QE programs. The portfolio

re-balancing effect quickly directed the investors to the credit market in the EMEs to chase

for yield. Shin (2013) denotes the year 2009 as the beginning of the second phase of global

liquidity, because of the rapid capital flow into the EMEs in the form of nonfinancial firm

bond issuance. Our cyclical component from the multivariate UC model indicates the

expansionary cycle of EME dollar credit during 2009-2014, a phenomenon widely

observed and agreed upon among the financial market observers and researchers.

Starting from 2015, the persistent recovery of US economy and the slowdown in the EMEs

overturned the dollar credit flow and created the concern about the instability of the EME

financial markets and the feedback loops between the EMEs and advanced

economies.

2.4.3 Co-movement among the Cyclical Components

The estimated cyclical components from our model are displayed in Figure 3. Both the

qualitative and quantitative measures of the cycles are consistent with intuition and they

follow the dynamics of dollar credit, interest rates and dollar index very well. We observe

that cyclical variation in dollar index is strongly negatively associated with the cyclical

variation of EME dollar credit. Before 2008 financial crisis, when dollar was relatively

strong, the overall EME dollar credit was increasing but was still below its trend,

consistent with the idea that stronger dollar works against capital inflow in the EMEs. The

correlation between the cyclical components of EME dollar credit and the dollar index

suggests that the weak dollar could potentially contribute to the surge in the credit flow

into the EMEs. When the financial crisis hit the US economy, the dollar depreciated and

the investors with dollar assets were looking for higher yields in alternative asset classes.

During this recessionary phase in the US, EME dollar credit witnessed an expansionary

period during the 2009-2014 sample period. The negative correlation between cyclical

components of EME dollar credit and the dollar index also suggests that the strong dollar

could potentially slow down the credit flow into the EMEs. Since 2015, the recovery of the

Figure 3: Cycles in the Correlated Multivariate Unobserved Component Model

This figure reports the cyclical components of the three variables EME dollar credit, shadow

interest rate and the dollar index. The dollar credit in the EMEs is measured in log trillions of

USD and multiplied by 100. The shadow interest rate is measured in percentage points. The US

dollar index is in logs and has been multiplied by 100.

US economy has played a role in strengthening of the US dollar and as a result financial

markets started anticipating a hike in the interest rates. EME dollar credit seemed to have

entered a contractionary phase during this time period. Note that the overall EME dollar

credit could still increase due to the underlying increasing trend. The dollar index is also

relevant for the market participants to estimate the overall return when US policy rate

hits the zero-lower bound. The risk of international carry trade, a typical cyclical

investment behavior, mainly comes from the exchange rate risk under this environment.

Therefore, it is not surprising that the dollar index can provide useful information in

understanding the EME dollar credit conditions.

The plot of the cyclical component of interest rate suggests that interest rates were

slightly above its long-run trend prior to the great recession. In the earlier part of the

sample, the results suggest that the interest rate were below its trend suggesting an

accommodative stance of monetary policy. This is consistent with the arguments

proposed in the literature that has argued that the interest rates were too low in the

prefinancial crisis period (Taylor, 2007). The unconventional monetary policy tools

utilized by the Federal Reserve during the great recession led to a decline in the interest

rate below its long-run average in our model during and after the great recession. This

period was also associated with surge in dollar credit above its long-run trend as shown

in Figure 3. The plot also shows that interest rate started moving towards its long-run

trend at the beginning of 2015 and this is also associated with the decline in the cyclical

component of dollar credit.

2.5 Robustness Check

In this section, we check the robustness of our results by substituting Wu and Xia

(2016) shadow interest rate with other proxies for the stance of monetary policy. We also

compare the estimated trend-cycle from our approach with univariate trend-cycle

decomposition using different methods.

2.5.1 An Alternative Interest Rate Measure

As explained earlier, there are several possible proxies that can capture the stance of

monetary policy. We use shadow interest rate because it does not suffer from ZLB

problem, and is also not contaminated by term premium . The shorter term interest rates,

for instance, federal funds rate or three-month treasury bill rate, suffer seriously from the

ZLB issue during post-2008 sample period. While the longer term interest rates raise less

concern from this problem, they are more easy to be contaminated by the variation in the

term premium. Since our EME dollar credit measure aggregates all the credit instruments,

regardless of the term to maturity, by taking the middle ground, we use one-year Treasury

bill rate in the model to check the robustness of our result. The results from this exercise

is shown in Figure 4. The plot clearly shows the robustness of our estimation to the use of

1-year interest rate as a measure of monetary policy stance, as the estimated cycle closely

resembles the one with the shadow rate as a measure of monetary policy stance.

Figure 4: The Cyclical Component from Correlated Multivariate UC Model

This figure reports the cyclical component of EME dollar credit. The dollar credit in the

EMEs is measured in log trillions of USD and multiplied by 100. The UC cycle (Shadow Rate) is

the cyclical component of EME dollar credit decomposed using the shadow interest rate as the

measure of US monetary policy stance in the correlated multivariate unobserved component

(UC) model. The UC cycle (1-year T-bill Rate) is the cyclical component of EME dollar credit

decomposed using the one-year treasury bill rate as the measure of US monetary policy stance in

the same UC model setup.

2.5.2 Comparison with Univariate Trend-Cycle Decomposition

In addition to the estimation of the correlation between shocks to the permanent and

transitory component, the use of multivariate model in theory should also provide us a

superior measure of trend and cycle as compared to the univariate model. To examine this

hypothesis, we also perform trend-cycle decomposition using the univariate models

(Figure 5). The univariate models include a linear trend model, a HP filter model and a

univariate UC model. The linear trend model decomposes the EME dollar credit series into

a linear-trend component and a cycle component. The HP filter method uses an algorithm

to smooth the original data series to estimate the trend component and the difference

between them is the cyclical component. The parameter value λ is set at 1600 as suggested

by Hodrick and Prescott for the quarterly data. The univariate UC model only uses the

series of EME dollar credit to decompose a stochastic trend component and a cyclical

component with the same specification as in the multivariate UC model.

Figure 5: Comparing EME Dollar Credit Cycles

This figure reports cyclical components of EME dollar credit based on alternate decompositions.

The results presented above clearly demonstrates that the estimate of trend and cycles

obtained from the the multivariate UC model is better able to capture the dynamics of

these three variables. The linear trend model, assuming a constant slope in the trend

component, is not appropriate since it assumes no shock to the trend. The recent crisis is

considered as the most severe financial crisis after the Great Depression (1929-1933) and

we observe a clearly big negative shock to EME dollar credit series which the linear trend

model is unable to capture. Other univariate models, without assuming a linear trend, fail

to generate realistic trend and cycle series by ignoring the relationship between EME

dollar credit and its price channels. The HP cycle seems to mimic the counterpart from

multivariate UC model very well before the crisis but diverge afterwards. The cycle from

the univariate UC model, instead, is close to the multivariate counterpart after the crisis.

Taking into account all the historical events as mentioned in the cycle interpretation, the

outflow of dollar credit slowed down before the crisis, due to the domestic boom, but the

magnitude of the negative cycle did not seem to be as large as suggested by the univariate

UC model. The HP cycle fails to capture the credit boom since 2009, while it performed

reasonably well before the crisis. Overall, it is clear from the analysis presented above that

there is valuable pay-off in utilizing information from other variables that are useful in

explaining EME dollar credit if one is interested in extracting its permanent and transitory

component.

2.6 The Effects of US Monetary Policy Shocks

With global financial integration and dollar as international currency, US monetary

policy plays a critical role in global liquidity transmission. Both the Fed and the EME

authorities are concerned about the dynamic impact of US monetary policy changes on

the international financial system. It would be an interesting exercise, therefore, to

examine the dynamic impact of monetary policy shock on dollar credit in the EMEs using

our multivariate unobserved component model.

Figure 6 plots the impulse responses of EME dollar credit and US dollar index when

there is a transitory Fed interest rate hike. Based on our model, the cyclical variation of

the three variables are correlated through the contemporary shocks in the error terms.

Given a positive transitory shock to US monetary policy, we would expect a negative

contemporaneous transitory shock to EME dollar credit and a positive contemporaneous

transitory shock to dollar index. The effects of the transitory shock would not disappear

immediately after the current period because of the persistent nature of the cyclical

components as it depends on the lagged values. From our impulse response analysis,

Figure 6: Impact of a Transitory Increase in Interest Rate

This figure reports the impulse response functions of EME dollar credit and the dollar index to to

a temporary increase in US interest rates

we find a hump shaped response of US dollar index and a U-shaped response of dollar

credit to a contractionary monetary policy shock. The transitory nature of the shock

suggests that these effects slowly disappear over time. The results suggest that the US

dollar index and EME dollar credit do not move to the new equilibrium immediately after

the transitory shock to US monetary policy. From policy maker perspective, the US interest

rate hike above it long-run trend would strengthen US dollar and induce dollar credit

outflow for several months. Although the effect eventually disappears, however, the

transition process may come with international financial instability. On one hand, the

process creates the challenges in the balance of payment for the EME authorities. They

need to equip with enough official reserves to manage the rapid outflow of capital and

stabilize the foreign exchange rate. On the other hand, the appreciation of US dollar and

the outflow of dollar credit leave the EME borrowers pressured to pay back the dollar-

denominated debt.

In order to be prepared for the transmission of international financial risk, in this case

through the rise in US interest rate above its trend, policy makers should adopt both

macro- and micro-prudential policies to deal with the monetary policy spillover effect.

Traditional macro-prudential policies focus on the soundness of financial corporations,

however, after 2009, non-financial firms engage heavily in the carry trade activities and

serve as surrogate financial intermediaries. The associated financial risks create new

challenges to the existing regulatory framework. For example, Hoffman (2014) provides

evidence that the very low world funding interest rates are associated with a rise in

volatile capital flows and asset market bubbles in fast-growing emerging markets.

Furthermore, although the integration of financial markets promote risk sharing in the

long run, in the short run, the external liquidity shocks and the interconnectivity of

financial markets make the international financial risk transmission easier and faster. In

other words, the global financial system may become more fragile in the short run, due to

the externalities in the market.

Internalizing the costs and benefits require macro-

prudential policy and international coordination.

2.7 Conclusions

In this paper, we use a correlated multivariate unobserved component model to

examine the hypothesis about the role of ultra low US interest rates in the dollar credit

boom in the emerging market economies. In doing so, we also decompose the movements

in dollar credit in emerging markets, US interest rate and the dollar index into a

permanent and transitory component. The correlations among the cyclical components

support the idea that the rise of dollar credit in the EMEs is associated with US interest

rate and the US dollar index below its long-run trend. The estimated permanent and

transitory component from our model captures the dynamic features of EME dollar credit

series and performs better than univariate benchmarks in capturing the boom and the

boost during the last few years. The strong cyclical correlations among dollar credit in the

EMEs, US interest rate and the dollar index suggest that the policymakers may need to

take into account the US monetary policy spillover effect on domestic credit conditions of

the EMEs, by observing the stance of the US monetary policy and the behavior of the US

dollar in the foreign exchange market. Macro-prudential policies and international

coordination may be justified and needed, along with micro-prudential policies, as a

consequence of global liquidity transmission and the implied international financial

instability.

Chapter 3

Corporate Overseas Debt Issuance in the

Context of Global Liquidity Transmission

3.1 Introduction

The recent surge in non-financial corporate

(hereafter ”corporate” ) overseas debt

issuance after 2007-2009 financial crisis has started drawing attention from

macroeconomic researchers, as it plays a critical role in the conduct of international

capital flow activities in the emerging markets. This surge in the overseas debt issuance

is also referred to as the second phase of global liquidity (Shin, 2013).The first phase

(2003-2007) of global liquidity is associated with a rapid increase in cross-border

international bank loans. The international banks lose the market share to international

bond markets in the cross-border activities substantially after the global financial crisis,

partly because of the strengthened financial system regulation. This fall in cross-border

lending by international banks was followed by a rise in the overseas debt issuance of the

non-financial corporate sector. The relative importance of corporate overseas debt

issuance can be gauged from the fact that more than half of the net ”external” financing of

emerging economies in 2012 took place through the issuance of international debt

securities (Turner, 2014).

Given its importance for the stability of the global financial system, it is important to

understand the behavior and determinants of corporate overseas debt issuance. The

purpose of this paper is to fill this gap in the literature. In particular, we want to examine

three hypotheses related to the overseas debt issuance of these corporate firms. First, is

there an evidence of price arbitrage on the part of these firms? Traditionally, a

textbookversion corporate only issues bonds overseas because of foreign currency

exposures. Think about the case when an exporting firm expects to receive a payment in

foreign currency. This firm should issue foreign currency liability to match the foreign

currency asset, in order to hedge foreign currency exposure. This behavior is considered

as a typical corporate risk management practice to help this company focus on the main

operating activities. In other words, they are not supposed to be interested in doing price

arbitrage in foreign exchange markets. Nevertheless, in recent years, many studies

suggest that corporate firms, especially large firms in emerging markets, may behave like

financial intermediaries in overseas debt issuance activities. [Black and Munro (2010),

Bruno and Shin (2015), Caballero et al (2015), Shin and Zhao (2013)]. Secondly, we also

examine whether the firms get around capital control measures enacted by the countries

and act more like a financial intermediary. This hypothesis is motivated by the recent

behavior of the firms in the emerging markets where we observe a surge in debt issuance

even in the presence of capital controls. Thirdly, we also examine the recent debate about

the transmission of the U.S monetary policy to the global financial system by examining

the link between overseas debt issuance and risk premium. To examine these hypotheses,

we utilize a recently developed database on international debt securities by the Bank of

International Settlement and perform a panel study of 32 countries for the 1993-2015

sample period.

Overall our results are consistent with the idea that non-financial firms in emerging

economies have been acting like financial intermediaries. Firstly, we find evidence in

support of price arbitrage hypothesis in case of emerging economies where we find

significant negative impact of level and volatility of exchange rate on changes in overseas

debt issuance. This implies that the corporate firms issue debt overseas in expecting that

domestic currency will appreciate against the US dollar. We also find that capital control

on bond market are positively correlated with corporate overseas debt issuance in

emerging economies whereas this relationship has opposite pattern in the advanced

economies. This difference in response to capital control across border reflects that

corporate firms in emerging markets have strong incentive to walk around capital control

to tap into international bond market, whereas corporate firms in advanced economies

typically follow the regulation to reduce cross-border financial activities. We also find

strong evidence between a measure of risk premium in the U.S. and overseas debt

issuance in emerging economies implying that overall credit conditions in the U.S. do play

a significant role. For advanced economies, however, we don’t find a significant

relationship between risk premium and debt issuance by its corporate firms implying that

the non-financial firms in the advanced economies do behave very differently than the

firms in the emerging market economies.

The remainder of this paper is structured as follows. Section 2 reviews the literature

on international debt securities. Section 3 presents our conceptual framework and

econometric methodology followed by a discussion. In section 4 and 5 we interpret and

check the robustness of the results. Section 5 concludes the paper.

3.2 Literature Review

The surge in corporate overseas debt issuance plays an important role in the second

phase of global liquidity. Turner (2014) suggests that declining in term premium in 10-

year US treasuries may have implication on greater sensitivity to global long-term interest

rates in emerging economy bond markets. McCauley et al (2014) also suggests that term

premium compression in US treasuries has significantly stimulate offshore

dollardenominated bond issuance. Shin (2013) points out, furthermore, the transmission

of financial condition across borders has taken the form of ’reaching for yield’. The

compositional shift in asset managers’ portfolios increases the demand for emerging

market corporate bonds, which leads to the decline of risk premium for these debt

securities. Meanwhile, the issuances of international debt securities explode in response

to the compression in risk premium and the declining capital cost in overseas bond

market. Chung et al (2014) identify a positive relationship between domestic money

growth and capital flow to the non-bank sector in emerging markets. Large corporates

borrow money overseas and hold short-term instruments in home country, such as

deposits and other liquid assets. This behavior essentially enables small corporates to

borrow money overseas and weakens the independence of monetary policy aiming at

domestic liquidity control. These papers describe the paths in monetary policy spillover

and the role of corporates in global liquidity transmission in the context of strict

regulation imposed on international banking system. We borrow Figure 1 from Shin and

Zhao (2013) to visualize the role of corporate debt issuance in global liquidity

transmission. In the first phase of global liquidity, domestic household depositors and

international investors (mostly international banks) supply liquidity to domestic financial

system to finance the final corporate borrower production projects. However, in the

second phase of global liquidity, domestic large corporates, rise to serve as surrogate

intermediaries to facilitate liquidity transmission into domestic financial system,

presumably due to the restriction on direct lending from international banks. The large

corporates headquartered in emerging markets may use their overseas subsidiaries to

issue bonds in international financial markets and receive the proceeds through intra-

company transactions (The subsidiaries can pay for operation costs for the headquarters,

for instance). In this way, the proceeds flowing into emerging markets are treated as a

form of foreign direct investment and do not appear in the residency-based external debt

positions. This conjecture worries policy makers about the effectiveness of capital control

policies at the border and the creation of systematic risks outside traditional international

financial regulation framework.

Under this backdrop, researchers investigate the determinants of corporate debt

issuance behavior by exploiting micro evidence from firm-level overseas debt issuance

and financial accounting information. Black and Munro (2010) examines the

onshore/offshore bond issuance decision by non-government residents of five Asia-

Pacific countries. Price arbitrage is identified as the most important motivator to issue

offshore, for both financial and non-financial corporates. Market completeness and

liquidity are also estimated to drive issuance decisions, i.e. firms seeks for more complete

financial markets to issue Figure 7: Transmission of Global Liquidity across Borders

Shin and Zhao (2013)

larger-size, longer-maturity bond at a lower capital cost. Nevertheless, this study uses

residency-based international bond issuance data, which may not be able to capture the

real volume of cross-border issuance, especially given the fact that non-financial firms use

intra-company transactions to avoid capital regulation across border. As depicted by

Bruno and Shin (2015), the difference widen significantly between nationality-based and

residency-based amount of external debt outstanding: the nationality-based external debt

position reaches roughly twice as much as the residency-based measures in 2014. (Figure

2) Shin and Zhao (2013) use firm-level financial accounting information based on

consolidated balance sheet and debt issuance data to further investigate the role of

corporates as surrogate financial intermediaries. Their results also suggest that

corporates in emerging markets behave like financial intermediaries in the sense that the

correlation between financial assets and financial liabilities has a positive sign, which is

supposed to be an accounting feature for financial corporates instead of non-financial

corporates.

Figure 8: Corporate Overseas Debt Issuance in Emerging Markets

Bruno and Shin (2015)

Two recent micro studies, instead, focus on testing carry trade hypothesis. Both papers

support the positive correlation between corporate overseas debt issuance and firm cash

holding, which essentially indicates a carry trade position. Bruno and Shin (2015),

furthermore, point out this phenomenon is more prevalent for emerging market firms

during favorable carry trade periods. Caballero et al (2015) suggests that there is evidence

for carry trade activities in countries with higher levels of capital controls. In our paper,

we also find that corporate overseas debt issuance behaviors are positively correlated

with capital controls at the border, especially in emerging markets where overall capital

control levels are much higher than in advanced economies.

To summarize, the literature provides some evidence to support the idea that corporates

dramatically increase overseas debt issuance to conduct price arbitrage and serve as

surrogate financial intermediaries to facilitate capital flow across borders in emerging

markets or countries with strict capital controls. Put it differently, corporates may step

into the vacuum whereas financial sectors are blocked by international capital control

regulation. This paper, from our knowledge, is the first paper using macro nationality-

based corporate overseas debt issuance data, to tackle several hypotheses in the literature

and to explain the determinants of corporate overseas debt issuance behavior.

3.3 Empirical Models

3.3.1 Price Arbitrage VS. Risk Management

To test the hypotheses mentioned above, we build three empirical models to study the

determinants of corporate overseas debt issuance behavior. In first model, we try to test

price arbitrage hypothesis against risk management hypothesis based on the

contradictive implication in response to exchange rate variables from these two

hypotheses. If corporates behave more like price arbitragers, they will issue less debt

when domestic currency depreciates/ when the exchange rate is volatile. Whereas, if

corporates behave like what textbook suggests, they will issue more debt when domestic

currency depreciates because when domestic currency depreciates, export increases so

as the foreign currency exposure.

In order to hedge the foreign currency exposure, they should issue more debt securities.

In addition, when exchange rate is more volatile, corporates are expected to have stronger

incentive to hedge larger portion of foreign currency asset exposure. In our model, the

exchange rate is computed based on direct quote against US dollar, which means the

exchange rate number is interpreted as the amount of domestic currency one US dollar

can purchase. Therefore, an increase in the exchange rate number implies depreciation of

domestic currency.

To be explicit, if price arbitrage hypothesis dominates risk management hypothesis,

then we would expect that β1 > 0 and β2 > 0; if the other way around, then β1 < 0 and β2 <

0. Besides exchange rate and exchange rate volatility, we also control for relevant

economic fundamentals. In this case, we control for both domestic real GDP growth rate

and current account balance. In the most complete specification, we also control for

government foreign exchange market intervention, by adding the growth rate of official

reserve. This specification allows us to see the impact of exchange rate and exchange rate

volatility on corporate overseas net debt issuance while government intervention is in

place.

(10)

3.3.2 The Effectiveness of Capital Control Policies

In the second model, we are interested to study the effect of capital control on

corporate overseas debt issuance. The capital control policies imposed by governments

are designed to control for the amount of international capital flow across borders.

Therefore, if the capital control policies were effective and well-designed, we would

expect the reduction of all types of international capital flow, implying the reduction of

corporate debt issuance. If, instead, we found the effect of capital control was the rise in

corporate debt issuance, then the effectiveness of capital control policies could be in

doubt. Explicitly speaking, the capital control policies may impose a binding constraint on

financial sectors, whereas the corporates may gain comparative advantage to take the role

as financial intermediaries. As explained previously, corporates are able to use intra-

company transactions to avoid the international capital control regulation.

Thereupon, the positive sign of capital control coefficient suggests the lack of effectiveness

of capital control on corporates, whereas the negative sign indicates the effectiveness of

capital control on both financial and non-financial sectors. As you may have noticed, we

do not control both exchange rate variables and capital control variables simultaneously

in one specification. This is due to the fact that capital control policies and exchange rate

stability are strongly dependent on each other. According to international monetary policy

trilemma, if the government imposes capital control at the border, then the country will

gain the ability to stabilize the exchange rate and the independence of domestic monetary

policy. If, instead, the government is willing to let exchange rate float according to market

forces, then the capital are free to flow across borders and monetary authorities still

maintain the independence of domestic monetary policies. Hence, the bottom line is, we

are not able to identify the clean effect of capital control and exchange rate variables

separately, by putting both of them into one specification.

(11)

3.3.3 Advanced Economy Monetary Policy Spillovers: An Indirect

Test based on Corporate Risk Premium

As explained in the introduction, there may exist advanced economy monetary policy

transmission effect. In this study, we perform an indirect test based on the effect of

corporate risk premium. If β1 is negative, which suggests a decrease in risk premium will

increase corporate debt issuance, provides a piece of supportive evidence in the process

of monetary policy spillovers. We test this hypothesis by adding back all the variables in

the previous regressions and are able to show that risk premium indeed causes the rise

in corporate overseas debt issuance.

(12)

We control for both country fixed effect and year fixed effect in panel regressions in all the

model estimations. The standard errors we report in our paper are Driscoll and

Kraay (1998) robust standard errors. Driscoll and Kraay (1998) propose a nonparametric

covariance matrix estimator that produces heteroscedasticity- and

autocorrelationconsistent standard errors that are robust to general forms of spatial and

temporal dependence. Because the nonparametric technique of estimating standard

errors place no restrictions on the limiting behavior of the number of panels, the size of

cross-sectional dimension in finite sample does not constitute a constraint on feasibility.

These features make Driscoll and Kraay (1998) standard error the most suitable

candidate in our models. Our sample includes 32 countries and quarterly data

observations spanning across 1993-2015. Clearly, we have limited cross-sectional

dimensions but relatively large time series dimensions. Since our conjectures are mostly

based on the stylized fact in emerging markets, we split the sample countries into

emerging market subsample (20 countries) and advanced economy subsample (12

countries).

3.4 Data Description

Our sample includes 32 countries

during the period 1993Q3-2015Q1. The key

variable we are interested in, net debt issuance based on nationality of corporate issuers,

is from Bank of International Settlement (BIS) website. Luckily, we are also able to find a

dataset, just available recently, about capital control measures in various financial

markets, constructed based on IMF annual reports. This dataset allows us to disentangle

the effect of the capital control policies in each financial sector on corporate overseas debt

issuance separately. Table 2 lists all the data sources we use in this study.

Table 2: Data Sources

Variable

Data Source

Net Debt Issuance (Millions of USD)

Bank for International Settlements (BIS): http://www.bis.org

Exchange Rate (Quarterly Average)

OANDA: http://www.oanda.com

Exchange Rate Volatility

OANDA: http://www.oanda.com

GDP Growth Rate (%)

FRED: https://fredqa.stlouisfed.org; IFS: http://www.imf.org

Current Account Balance (Millions of USD)

IFS: http://www.imf.org

Foreign Reserve Growth Rate (%)

IFS: http://www.imf.org

Risk Premium (Junk Spread) (%)

FRED: https://fredqa.stlouisfed.org

Capital Control Measures (range: [0,100])

NBER: http://www.nber.org/data/international-finance/

The variables used in this paper are constructed as described below.

Net Debt Issuance: We remove the seasonality in the international debt security amount

outstanding, which are issued by non-financial corporates and categorized based on

nationality of issuers. Then first difference these series to get net debt issuance in millions

of US dollars for each country.

Exchange Rate:measured in direct quote, i.e. in domestic currency per unit of US dollar.

We take the average of the daily closing rate in the quarter to serve as quarterly average

exchange rate.

Exchange Rate Standard Deviation: The standard deviation of exchange rate within the

quarter based on the daily closing rate.

Exchange Rate Volatility: = (ExchangeRateStandardDeviation/ExchangeRate) ∗ 100. It

can be interpreted as percentage deviation from the quarterly average. This measure of

exchange rate volatility gets rid of unit of measures, therefore is comparable across

currencies.

Real GDP Growth Rate: We take log difference of seasonality-adjusted real GDP to get the

quarterly GDP growth rate.

Current Account Balance: We remove seasonality in current account balance data and

convert series to be measured in millions of US dollars.

Foreign Reserve Growth Rate: We take log difference of official reserve assets which are

measured in US dollars.

Risk Premium: We use BAA corporate bond rate minus 10-year Treasury bond rate to

measure risk premium in corporate bonds.

Capital Control Measures: We use international capital control indexes on the money

market/ bond market/ equity market/ real estate market/ foreign direct investment

separately. These measures are continuous variables ranging from 0 to 100. This variable

is only available between 1995 and 2013.

Table 3: Summary statistics

Variable

Mean

Std. Dev.

Min.

Max.

Net Debt Issuance (Millions of USD)

358.199

2077.493

-22934.971

23020.229

2720

Exchange Rate

356.624

1599.789

0.014

10000

2738

Exchange Rate Standard Deviation

2.82

35.909

1151.724

2738

Exchange Rate Volatility

1.7

1.956

48.979

2738

Real GDP Growth

0.88

1.816

-12.074

42.294

2637

Current Account Balance (Millions of USD)

2127.202

11555.506

-39157.066

108339.003

2603

Foreign Reserve Growth

2.743

10.293

-95.652

85.048

2815

BAA minus 10 Treasury Bond Yield

2.406

0.794

1.37

5.58

2855

Capital Control: Money Market

38.035

38.612

100

2432

Capital Control: Bond Market

36.397

39.083

100

2176

Capital Control: Real Estate Market

47.862

36.247

100

2432

Capital Control: Direct Investment

41.612

40.304

100

2432

Capital Control: Equity Market

37.87

39.116

100

2432

3.5 Results and Interpretation

3.5.1 Price Arbitrage Hypothesis VS. Risk Management Hypoth-

esis

Price arbitrage hypothesis implies that domestic currency depreciation and volatile

exchange rate against US dollar has a negative impact on corporate overseas net debt

issuance. Whereas, risk management hypothesis suggests that domestic currency

depreciation will stimulate export and create larger currency exposure position needed

to be hedged by issuing overseas liabilities, and the more volatile the exchange rate

against US dollar, the stronger the incentive for corporates to hedge the exposures. These

ideas implies a positive impact on overseas debt issuance from domestic currency

depreciation and exchange rate fluctuation, from the risk management perspective.

Table 4 reports the results from the first model that tests price arbitrage against risk

management hypothesis. In all specifications, as the exchange rate increases, i.e. domestic

currency depreciates, corporate overseas debt issuance will decrease; as exchange rate

becomes more volatile, the less the corporate will issue debt securities overseas. This

result is in line with price arbitrage hypothesis. The corporates issue debt overseas in

expecting that domestic currency will appreciate against US dollar. From model

specifications (1) to (4), we add real GDP growth and current account balance to control

for the economic fundamentals. Interestingly, current account balance is insignificant in

explaining corporate overseas debt issuance, which confirms the result that corporate

overseas debt issuance does not strongly associate with hedging foreign currency

receivable exposure. In the last column, by adding the growth rate of official reserve, a

proxy for government foreign exchange market intervention, we find that government

intervention stabilizes corporate debt issuance by mitigating the effect of exchange rate

volatility, while the exchange rate variables still play a significant role in explaining the

corporate overseas debt issuance behavior.

Table 5 reports the same regressions using advanced economy subsample. No

significant impacts are found for exchange rate and exchange rate volatility, which

suggests that corporates in advanced economies are not sensitive to exchange rate

variables in overseas bond issuance activities. Together with insignificant effect from

current account balance, we could not support either of the hypothesis in advanced

economies. Our conjecture is that corporates in advanced economies could issue bonds

overseas in domestic currency so that these firms are less sensitive to exchange rate

variables. Another interpretation could be, if countries fall into more flexible exchange

rate regimes, there is less opportunity to conduct price arbitrage in the foreign exchange

markets. Most emerging markets impose much stronger capital control at the border to

stabilize the exchange rate and maintain monetary policy independence. Thereupon, the

exchange rates in emerging markets may not reflect market expectation about the ’true’

exchange rate against US dollar. It will take much longer time to arbitrage away these zero-

risk opportunities because capitals need to find a way to walk around the capital control

regulation at the border. Whereas, in advanced economies, domestic financial markets are

well integrated into international financial markets and the overall capital control levels

are much lower at the border for these counties. Corporates headquartered in advanced

economies presumably on average face a smaller interest rate gap at the border, due to

financial market integration. Moreover, they do not have comparative advantage,

compared to financial corporates, to rise as financial intermediaries because of less

regulation at the border. Overall, there is no evidence to support the price arbitrage

hypothesis for the corporates in advanced economies.

To0

,,,

900

OTO

od,

sasayjualeduisoysneys

69VL69VTV9OST6C9T67Z9

TSUONEAIasSq

SAKSAKSIASAASAAIJJA

POX

ICI,

SAX

SAXSAASAXASAAVAIJqpPaxl

AUN’)

(60°21)

(vo'zt)(t6'zZ)

(Test)

(6st)

exe

COOT

zee

LOOT

weve

LOC

xxv?

OES

L’STGjuejysu0)

(Ze)

TEESE

(T-YUMoOLHaAIISayeIUJO

(Zs'T)(09'T)

60700'00€700'0(T-3)a.urTegjUNosdyIUeLIND

(s0'T-)(90'T-)(S6'0-)

£9°81-PE'8T-76

91-(T-Ymmorndayeeu

(98'T-)(S6'T-)(66'T-)(8E°Z-)

8SS0°0-

8S0'0-

1IS90'0-

18Z0'0-(T-ayeyasueyoxy

(8t'z-)(v9'2Z-)(Zz'Z-)(0Z'Z-J

1'Z-)

v8'ST-

6T-

wLV’8T-

LEVI

6T'ET-(T-YAnefojaieyosueyoxg

DIUENSS]}G9q19

N2UeNSS]1

G99

N2IUENSS]}G9q}aN2UENSS]}1G9q1aN2IUENSS]}G9aqION

(s)(v)(e)(Z)

(1)

"SUONEUITJSO|[EUIpoy

Oda

lolesIo.Iopiepueys

3snqo1(g66T)Aeet[pueloosiiq‘aptsoddoayyo1paidsisayjodAyjuawaseueUulyslLisyjseatloyM

‘sAesaualeA}I

e[OAS}

puea}elasueyoxajo

sJUdTYJaooy

OgGs}IpaidsisayjodAyasesjg1esdL1g'aye.lasueyoxajoasejaAeA]19}1en

bayWo1jUONeAapasejuadJodayjsepainseaulsiAye[oA

ayelasueyoxy

uoMedaidapAdua1InddINsawiopo}useAInbasia}e1as

uUeYIxXaUlaseaIOUIULAIOJaJay

aseyoinduesie[jopspnjoywunsadAduasin9d

IsawWopjoyUNoWwesy}y'9'1'9j0nbya.lIpsulsnpa.inseawis{a}e.1adsuUeYyoOX

19}.

1en

bay

WMIIUENss]qapseas.1gA0}aUud}e.1OdIOOSIa[qeLiea

quapuadapay

Ne[OAa}e1aSULYOXa

pUb|}e.1OSULYOXIUOUOT

CIT

AUILBATOIPe.UODaYUOpaseqsasayjOdAYOMy9Y}S9}0}S19[qeysiy,L,

Saxe

PSUISIEWy:sisayjodAH

UsWeseuepLysly

/ase.IGIVsdLg

Pagel,

100

arts

900

OL0

sasayjuareduisonsyeys

6V66V6E0OTTTLOTT

LSuoneAtesg(¢

SIA

SAA

SAASAASIAVIJJAPOXT

ICI,

SOX

S9AS9X

SAA

VaJJqpoxty;AUNG)

(st'o)(eT'0)(T0'T-)(Z8'0-)

(hr'Z-)

G8'VLIZ'ESS'LTE-0'SLZ-

98E-JUeIsUO)

(66'0-)

810°9-

(T-YIMoOINeALASaY

[EO

(Zr'0-)(Z¥'0-)

Z010°0-8010'0-(T-3)e29ueTegqunosoy}ueLIND

(ot'0)(70'0)(6T'0-)

096'PZPT'ZS98'8-(T-

Mong

TeaY

(62z'0-)(9Z'0-)(Sz'0-)(O¥F'0-)

66£'V-SL6'E-Z18'€-S86'S-(T-aieyasueyoxy

(19'T-)(19'T-)(9Z'T-)(EET-)(ZE'T-)

9'SST-Z'VST-8'bZ1-9

6ZT-F

T-(T-

Anne

foAayeyasueyoxy

ddUPNSS]}G9(19

NeIUeNSS]}1G9qJeNeDUeNSS]3GQaq1e

NODUENSS]GeqJeNeoUeNSS]IGIQION

(s)(p)(e)(Z)()

‘SUOT]EUINSA|[eUIps}1odatates101lap.iepue}s

ysnqo1(g66T)Aeerypueloosiiq‘ayisoddoay

paidsisayjodAy

UsWiaseUeULYSLIay}sealayM

‘sANesaUaIeAT

e[OASIpue|}elasUeYyoXaJO

SJUSTDYJa0y4}0qGS1IIpaidstsayjodAyasesjg1eadLig‘a}e.1lasueyoxajoase.1aAeA]

bay}

Wol1UONelAapasejUuadJodayjsepainseaulsiAy]e[OA

ayelasueyoxy

uoedaidepAdua.LInddNsawopo}useAInbasia}e1ad

UeYIXaUlaseaIOUIUeAIOJa1ay

aseyoinduesieljopsnjoywunsadAduasind

INsawopjoyUNowesy}'a'1ajonbypa.lIpsulsnpa.inseauls{a}e.1adsuUeYoOX

19}.

JeNbay

UMIIUENSS]qapseas.19A0}aUd}e.1Od1OISIa[qeLiea

quapuadapay

Al[Ne[OAa}e1asULYOXapUb|}e1OSULYOX9UOUONedT[CUIIDAIOIPe.QUOIaYUWOpasegsasayOdAYOMy9Y}S9}0}S19[qeysiy,L,

SaTWOU0DTpsdUeApyY:sIsay

OdAP

UsWaseUL|STY

/ASe.IGIVSILIg:Ga|qeL

To summarize, the results from the first model support the conjecture that price

arbitrage incentive dominates corporates overseas debt issuance behavior in emerging

markets. This assessment is in line with the micro evidence from Black and Munro (2010),

which concludes that price arbitrage is the most important incentive for corporates to

issue bonds overseas, although the evidence for financial corporates is even more

prevalent. On the other hand, there is no such evidence found among the counterparts in

advanced economies. Shin and Zhao (2013) suggests that corporates in advanced

economies behave more like textbook-version corporates.

3.5.2 The Effectiveness of Capital Control in Emerging Markets

Typically we would expect a reduction in cross-border activities from all market

participants when facing a strengthened international regulation. This intuition implies a

negative impact of capital control on corporate debt issuance overseas. Table 6 reports

the effect of capital control on corporate debt issuance in emerging markets. Surprisingly,

we find that capital control on bond market are positively correlated with corporate

overseas debt issuance. This result suggests that corporates may exercise their

comparative advantage as surrogate financial intermediaries, while financial sectors face

strict regulation at the border. This result, together with Caballero et al (2015), illustrates

the importance of corporate overseas debt issuance as surrogate financial service in

countries where strict international capital flow regulation is in place.

Using advanced economy data in the same specifications (Table 7), we see the opposite

pattern: capital controls in financial markets are negatively correlated with corporate

overseas debt issuance. This difference in response to capital control across border

reflects that corporates in emerging markets have strong incentive to walk around capital

control to tap into international bond market, whereas corporates in advanced economies

typically follow the regulation to reduce cross-border financial activities. Especially

among our sample countries, many advanced economies have international financial

centers in their home countries or they are by themselves financial centers (e.g. Hong

Kong and Singapore). Corporates in these countries do not have strong incentive to issue

bonds overseas, when they face more strict capital control at the border.

TOO

,,,

900

QTO

od,

sasayjuatedulsonsne}s

COETCOETCOELOLZLC6E

TSUONPAIOSqQ

(16'0-)(S9'0)(19'0-)(Z9'T)(04'0)

(€S'T)

6L9'Z

(2Z'0)

(09'T)

(ST’€)

(€S'0)

(s3'T)(vz'z)

(912)

(zE'z)(S0'Z)

ZS00°0

~28500'0

66500'0

~1Z900'0

(0z'0-)(19'0-)(99'0-)(¢6'0-)(79'0-)

SALSAKSIASAASAAIDAJJA

PAXIL

JICIA

SAASAASAAS9AWVAJJAPOXLJATUNO)

8ITT-2o

LSOT

0S-8'07TSZZEqWesUO)

([-2)JUeUNsSeaAUT{I9IIG:[0.19U0DeHIde7)

OTE

O(T-)IOyITeWeIeIsAeoy:[oUODeydeD

LLL

(1a)

eyseWAINDY:;o.QUoDeydey

L6L'2(T-1)e41eWpuog:joUODfeyWdeD

229

O(T-3)exIeWASUOW:[O.UODTeydeD

'8S00'0(T-e0ue[egqyUNOD2y}UaLIN)

ZEEL-SLTL-bb

ZI-10'6T-27'7T-(1-

Molndayleeu

SIUENSSIG9}9

NeIUeNSS]}q9q}9

NOIUENSS]}

G00

IUeNSS]}qQoqJaNooueNss}}qaqieN

(s)(y)(E)(Z)

‘SUOTCUINSa][eUIPs}.10da.10.1es.10.1lapilepueys}sngol(g66T

)Aeelypue][OosLiq

aouenssiyqapseas.1aA0e}e.1od.1ossulule[dxaulefo.r1e

ske[dAn1[0djo.1uosTeyWdesyITyM4S9}0}.1a

p.1OUL10}Ias

[eI

ULULIYINadseuoxa

puljo.UOI[eyde.aUOppeaM

‘UOTeIYDadsyoeauy'a}e1y

MoI

dday[ee1ey)

puesoueryleqyunosoejywaLind'a

s|eyUaWepUNJIIUIOUOIAaIOJ[O.1]U0IEM

‘SUO}eI

Dads

|[eu]isa}1enbaywuIyMessuenssngep

S¥9S19A0}9U9}e.10d.109SI9[qeLIeA}Ua

pudadapay

AOLAeYaqoaIUeNSsSqapseas.19aA09}e.10d.109U0]O.UOITeNdedJ0DaJJaayApNjso}jsi9[qejay,L

Saye

SUIs.1dW

:[0.1]U0)

[ede

j)JossauaAnIayjqey

L:99[GeL

TOO

,,,

900

QTO

od,

sasayjuatedulsonsne}s

LZ8TZ8TZ866Z2T

Z8Su0neAlesqg

SALSAKSIASAASAAIDAJJAPAXLJICIA

SAASAASAAS9AWVAJJAPOXL]AUNO)

(so'v-)(Z8'r-)(v8'r-)(TE'E-)(SZ'7-)

90TT-

eal

OLZT-

B8BIT-

exi0

TOL-

wae

T'Z9TT-queysuos

(vI'T-)

LS9'L-

([-)JUeUNs9aAUIID9a.1IG:[0.19U0D[e}Ide’)

(Z€'0)

960°

T(T-ieyTeEWeareIsyesy:[oUODeydey

(Zo'e-)

SZr'6-(T-)eyeWAINDYJoNUODeydey

(L6T-)

L19°S-(T-3)1

Wpuog:jo.UODleydey

(6T'€-)

wee

GT'ZT-(T-3)yeyIeWAuop]:jonuogleydey

(zz'0)(Sz'0)(0¢

(ZT'0-)(vE'0)

Z9S00'0L£900'079Z00'0£9£00'0-TS800'0(T-1)29ue[egiUNODDyqWA.LIND

(t¢'0)(8¢'0)(Sz'0)(¢s'0)(ZZ'0)

9L'OLTP'0ZZS'ETSE

OE8T'

(T-umoindayeed

SIUENSSIG9}9

NeIUeNSS]}q9q}9

NOIUENSS]}

G00

IUeNSS]}qQoqJaNooueNss}}qaqieN

(s)(y)(E)(Z)

"SUOTJEUTS9[[CUIPs}OdaJales1oLiapslepueys}sngo.1(g66T

)Aee.y

puel[OosLIq'adueNssqapseastaAoa}e1Od1oosulUle[dxaulaforesAe[dAstjod

JO1QUODTeVI

ded

YIIYM}S9}0}.

POULSUINYIea1OWas|[eINULULOYINadseuoxa

puljo.UoOo[eydesauOppeaM

‘UOHeoyDadsyoeauy'aye.1y

Mois

ddy)[ee1ey}puessueyeqjunos9eUaLInd'a'I's]e}UaWe

pUNJIIWOUOIaIOJ[O.1]U0IaM

‘SUO}eI

ads]

[eulsej1enbayjuiyywMaouenssngqap

SB9S19A0}9U9}L.10d.1ODSI9[qeLIeAjUapUsdapay

IOLAeEYaqaoUeNSsTqapseas.19A09}e.10d.109U0[O.1}U0I

[EU

AeIJOWaJJaayIApNyso}jsia[qeyoy_L,

SaIWIOUOIApadueApY:[0.]U0)eEUdejJOssaUsANIaTJ

ZAGeL

3.5.3 Advanced Economy Monetary Policy Transmission and an

Indirect Test Based on Risk Premium

There exists a common sense that compression in corporate risk premium originates

from expansionary monetary policy in advanced economies, among researchers and

market participants. To test monetary policy spillover effect, one of the key factors in the

spillover chain is the response of corporate overseas debt issuance from compression in

risk premium. In the third model, we test this broad hypothesis by providing some

supportive evidence from the impact of risk premium on corporate overseas debt

issuance behavior. Table 8 provides strong evidence to support that corporates issue more

bonds overseas in response to compression in risk premium in emerging market

economies.

No significant effect in advance economies (Table 9) suggests that corporates in these

countries do have different incentive in conducting cross-border activities, compared to

the emerging market counterparts. In advanced economies, domestic financial markets

are well connected in the international financial market. Compression in risk premium in

international bond market also imply compression in risk premium in domestic bond

market. Therefore, these firms have no strong incentive to go abroad to issue bonds,

whereas the corporates in emerging markets face a segregation between domestic bond

market and international bond market.

These results offer some support to the idea of advance economy monetary policy

spillover effect. Although the less integration of emerging markets in international

financial system, monetary policy in advanced economies do push international investors

to crack through border barriers to chase for yield; and meanwhile market participants in

emerging markets also try to walk around the regulation to arbitrage the return across

the borders. This phenomenon raises the concern about the relevant liquidity measures

for policy makers, even for those policy makers in countries where impose tight capital

regulation at the borders. They may also need to put an eye on the global liquidity

measure, as it helps explain anomalies in domestic liquidity supply. (Chung et al, 2014)

In the most complete regression estimation, we include all the three sets of variables

together with real economic fundamentals. The effect of exchange rate variables

disappears. This result may imply that the effect of capital control can dominate the effect

of exchange rate variables because they are strongly interdependent and it is hard to tease

apart the marginal effect if we try to regress them simultaneously in one estimation.

Without controlling for capital control policies, we found the consistent results as in

previous model specifications.

TOO

,,,

S00

QTO

od,

sasayjuaieduisonsyeys

9LZTE9VT697T

COST

Z6S

LOOZTSUOHeAIesq(

SOASAASAASAASAASIAWVIJJAPOXLJeI_,

SOASOKSALSAASAASAAVIJqpoxtjAUNO)

(sz'v)(7z'6)(€6'8)

(BEF)

(ES'7) (85'S)

6T0T

L6Z1

VIZI

EVOL

..0'S66

ves

L'6ySqueysuo)

(to's)

PVL'C

(T-1)exTeWpuog:o.UODfeydey

(z¢'T-)(1é'T-)

Z0€0'0-

9LS0'0-

(T-Jareyesueupxy

(et

T-)(Zr'2Z-)(0£'Z-)

€89'9-

_LV8t-

26°

ST-(T-

[OAaTeyasueyoxy

(or'z)(99'T)(08'T)(Z8'T)

79900'0ZEF00'0

JZ¥00'0

7L¥00

0(T-)a0ueTegjunossyquaLIN)

(ss'T-)(Zp'T-)(os'T-)(90'T-)(¥6'0-)

€£°9Z-1Z'7ZZ-98

02-SL9T-ZS'ST-(T-YmMolngay

eeu

(11T'e-)(S9°Z-)(99°2-)(89°2Z-)(99'2-)(Fr'Z-)

ff)

69T-

Tbt-

CVi-

LUT-

LET-

S'6LT-(T-umnimaigysny

JIUPNSS]IQIQI]9N2IUENSS]Q9

G19

N2*uUeNSS]Q9IqQ

JPNsouenss]}qoa

qje

N2uUenss]}qoqjeNesuenss

192°d79N

(9)(s)(F)(E)(Z)

"‘SUONeUINSa][eUIpeiodaJa1es10.1lapiepuejysjsngol

(8661,

)Aeerypurl[oosLiq

uintwia1dys11a}e.10d.109sa.njdesysiym

‘pjatApuogAinseal

ieaX¢

[

snulwiplatApuoqyygAqpa.nseaulsi

WINIWaIdysty

pajjolaaolidsAd

A1eyauOWIsy}.10ddnsojaouaplAsaWOssla1ayNeYepn[IuoIaM

‘adueNssyqapseas.iaA0e}e.10d.109

ulaseaDUTUeAGpaMoy][OjsiuMniws1dyslLulesea.oa

prey

19}.

1enbayjuiy

Maouenssiyqepjeuseas.1aA0e}e.10d.109s1a[qelieAjuapuadap

wntwaidysiiaye1od.1osuodnpaseqiajjataAo[IdsAa1[odA1eyauUOWAWOU0Ia

padUPAPEUO}sS9a}{Ia1IPUIURSULIOJ.1ada[qeysiy

(axe

P[SUISIEW

)ISaTOeIIPU[UY:UOISsTUIsUeL]

ADT[OgAIeBUOW:8alqeL,

Too

G00

QTo

od,

sasayuatedulsoysneys

96L6V66VE66T6EDNTOZOTSUOHEAIeSq¢

SOX

SIASOASOASAASOAVIJJA

POX

IeI_

SAASOASAASAASAASAAVIIJJqpoexLyAUNO’

()(zeT-)(er

T-)

(rr

T-)(so't)(eo'T)

O£'€S6-S

OLOT-9'SENT-L'9E9TE'909

TIUeISUOD

(¥Z°0-)

CCC

C-

yaye|Wpuog:jo.1U07/eude)

(95°0-)(92'0-)

8V'0T-S06'E-

(T-)aieyasueyoxy

tort

T-)

8'9ST-S'9ST-9'9ST-(T-a)Anejojayeyasueyoxg

(ZT'0-)(Z7'0-)(8¥'0-)(T7'0-)

TE700'0-8010'0-0T

L0'0-1S600'0-(T-2)aouelegjunoss

yuan

(sr'0)(80'0)(TT'0)(Z0'0)(TT'0-)

€6'7Z29S

EBL

SSZ8'£659'S-(T-a)

Monga

yeeu

(sz'0)(60'0)(60'0)

(Tz'0-)(TT'0-)

(oT

'0-)

TE'868Z'ZEL9'EES7

PL-PL'BE-SP

EE-(T-YunIwelgysy

SIUENSS]IG

9739

NODUENSS]1G9(39

NOdUENSS]}q0q}9

NOIUENSS]}qo(}9NOoUeNSS]}qQoqJONooUeNSS]}qoqION

(9)(s)

(He)

(ZT)

‘SUOQPUINSA[PUIPs}odaJales1o.1iapiepuejs}snqol

(866

)Aeerypurl[oosLiq

wuntwasidysiieye10d.109sa.njdesysiym

‘pjatApuogAinseal]

1eaXQ

[

snulwiplatApuoqyygAqpa.nseaulsi

UINIWa1dyshypajja.1aao][dsAdT[odA1eyaUO

}.10ddnsojaduapAVaWOSSIa.1ayeYepn[IuoIaM

‘BIUENSsSH

qapsess.1aA0a}e.10d.109

UlaseaOUTUPAGpaMO][OJSIWINIWd.1dysLulased.1da

prey]

1a}

enbayuIyLMadueNssn

qapjeuseas.1aA09}e.10d.

109s

la[qeLieAjuapuadap

ay,

wntuiaidysLieye1od.1osuodnpaseqiajja1aAo[IdsAd1[odAreyauUOWAWOUOIa

psdUPAPEUO}S9}I91IPUIURSULIOJ.1ada[qeysiy

(Awou0d

paoueApy)}sa]aIIPUlUY:UOIssIuIsuel]

ADT[OgAIeUOW:69a[GEL

3.6 Robustness Check

So far, our main results show that the compression in risk premium increases

corporate overseas debt issuance and the stronger regulation policy makers impose at the

border, the more bonds corporates issue overseas. These features seem to suggest the role

of corporates as surrogate financial intermediaries. If this interpretation is solid, by

adding the interaction between risk premium and capital control measures, we should see

the effect of this interaction term is negative in the second phase of global liquidity. The

reason is as follows. If corporates are indeed surrogate financial intermediaries, they have

stronger incentive to issue overseas when both risk premium is lower and financial

corporates face more strict capital control at the borders. Given a constant level of capital

control, the lower the corporate risk premium is, the more corporate bond issuance. Given

a constant level of risk premium, the tighter the capital control is, the less corporate bond

issuance. This is because, corporate capital cost from issuing bonds is the risk free interest

rate plus corporate risk premium. A constant level of risk premium implies a constant

level of capital cost in the bold part. Given a constant level of capital cost, the corporates

should have less incentive to serve as surrogate financial intermediaries when facing more

strict capital regulation at the borders. Therefore, the effect of the interaction term in the

second phase of global liquidity is expected to be negative if corporates indeed serve as

surrogate financial intermediaries.

We provide the robustness test results in Table 10 below. To tease apart the effect of

capital control and risk premium, we incorporate them separately in different regressions

and also split the whole sample based on the timing of the second phase of global liquidity.

The first two columns report the effect of capital control on corporate overseas debt

issuance. The effect of capital control in the 2007-2013 subsample is positive and three

times as much as the counterpart in the 1993-2006 subsample, suggesting that

strengthened international capital control policies indeed stimulate corporates to act as

financial intermediaries across borders. The middle two columns report the effect of risk

premium on corporate overseas debt issuance. Corporates were not sensitive to corporate

risk premium before 2007. However, since 2007, one percentage decrease in corporate

risk premium lead to more than 100 million US dollar more corporate bond issuance

within the following quarter. The last two columns provide further evidence to support

the corporate role as surrogate financial intermediaries in the second phase of global

liquidity. The effect of interaction term between capital control and risk premium is

insignificant before 2007, while in the second phase of global liquidity, the coefficient of

the interaction is negative and significant. Based on the intuition described in the last

paragraph, the data favor the conjecture about corporates behaving like financial

intermediaries. It is worth to point out, in the last two regressions, we control for linear

time trend instead of time fixed effect because our capital control measures are in annual

frequency. There will not exist meaningful variation in the interaction term if we control

for annual time fixed effect. Therefore, we instead use annual time trend to control for the

variation over time.

To sum up, we perform a robustness check to verify our interpretation about

corporates serving as financial intermediaries, with further evidence by exploiting

information in subsamples and allowing for the interaction between variables.

Too

900

OTO

>a,

sasayjuaredulsonsyeys

LSS6TLV09868Z2ZSS6T

ZSUONeAIISqQ

wed

EOSEST-

axa

ITV

£0'0V

06S0000'0

CLSG9ZC-

(v2'S)(8Z'T)

LEET

(99'2-)(8€'0)

6LT'2-¥8Z'0

sud

ITT

9V'STS87'7

OS'ZE-TSGZ

(92'S-)(ZZ'T-)(0S°9)C)(60'0-)

(OFT)

(v2'0)(SZ'E-)(66'E-)(TZ'0)

(st's)(zr't)(Zz'

e)(vr'F)

+xBCS'6

one

(to'0)(¢0'z-)(So'0)

(TE'Z-J(TT'0)

(6T'Z-)

69800'0-TZ7000'0

1Z600'0-T0T00'0

(69°T-)(s9°0)

(Th'2Z-)(Z9'0)(ZT'T-)(Z9'0)

ONONSOXSAAS9XS8AIIAJJAPOXI

CIA

SAA

SOXSAA

SOA

SAASOAVIJJq

PaxtyAQunNo)

88ET087

8T-S8

BZ1uesUOy

leah

(T-umnrmeigysty,(puog)jonuopeyde)

0'02Z-E€E

TI(T-untua1gysry

106

'Z(T-3)394eWpuog:jonuoDleydey

90600'0-(T-3)e2ueTeg

UNOS.

YIUAaLINY

'6£-079

PL0'LZ-08T

L(T-uMoin

day

ead

aoUeNSss]1qeq12

NeDUeNsspqeqjeaNeouenssy]}qaq19

NedUeNSs]1q9q1e

NeduUeNsspqeqjeaNeouensspjqeqieNn

(N(NwlE)(Z))

‘SUOTEUINSI][ePUIpe}.1ode1

31es.10.laprepuejs}sngo.i(g66

)Aee

pueljoosiiq

Sul|[eystwMniwaidysiipuepausyysuetsslo.uosjeyidesy}oquaymeasea.1ulo}

peyedxasleouenssiseas.1aA0e}e.10d.10daYJasneseaq

‘SaLIeIpeul1ajUleDUeUIJe}esO.LINSses[o.19}e10d.10994

UMeUT]SIwNTWaidystipue

[O.QUODTE}IdedDJOULIaUOTIeIIUIBYYo

sap

YJaooayYyoustseaynesauay

Aypinbiy[eqo[sjooseydpuosasayyovsejsayjsepadlsoiod

sSeMUSYM‘/00Z1eyepueslojaqaouenssyqapseas.1aA0a}e10d1oduOUINIWIaIdysTIpuejo.Uuos[eydesjo}ajjaay}sapAoidayqeyay

AleIpsUlilayU][eMueuUl,jayesoO.LINSsea[oyajze.10d.107)ay

:ssauIsnqoy:9

La[qel,

3.7 Conclusions and Policy Implication

This paper studies the determinants of corporate overseas debt issuance in 32

countries during the period 1993-2015. The results provide some macro evidence to

support the conjecture that corporates in emerging markets serve as financial

intermediaries at the border to facilitate global liquidity transmission. Corporates hold a

carry trade position, in other words, borrowing liabilities in foreign currency and holding

assets in domestic currency, during the periods when domestic currency is expected to

appreciate against US dollar and the exchange rate is less volatile. The rise in corporate

overseas debt issuance can be explained as the product of advanced economy monetary

policy spillover and capital control policies at the border. As corporate risk premium

compresses, corporates have incentive to serve as surrogate financial intermediaries

across border, especially in countries where domestic financial sector faces strict

international capital flow regulation.

Policy makers should carefully evaluate the potential side effect from international

capital control policies. Ill-designed these policies reduces the effectiveness of cross-

border capital control. Furthermore, these policies may create the systematic risk outside

the traditional framework and makes it harder for policy makers to monitor and manage

international capital flow activities. Additionally, policy makers should be aware the

international financial risk transmission through either monetary policy shocks in

advanced economies or financial risk materialization in emerging market corporates. The

last but not the least, domestic currency depreciation and volatile exchange rate against

US dollar may add uncertainty in the capacity for emerging markets corporates to borrow

and rollover the existing debt.

Chapter 4

What is Driving Debt Dollarization in

Global Economy? A Dynamic Factor

Analysis

4.1 Introduction

The takeoff of US dollar-denominated debt issuance

has been one of the most eye-

catching phenomena in the global financial market since the Great Recession. Many

studies suggest that loose US monetary policy may factor in the global liquidity

transmission during this period. (Chung et al., 2015) While Basel III tightened the cross-

border capital flow regulation, the border-control policy loopholes push international

capital flow towards the global bond market from the global banking system. In short, the

rapid growth of dollar debt seems to be closely related to the super-low policy rates and

quantitative easing policies from the Fed. Nevertheless, it is possible that some other

factors, such as, the fast economic growth of emerging market economies, European

Sovereign Debt Crisis and some country-specific factors, could play a critical role in

explaining the second phase of global liquidity. (Shin, 2014)

In this paper, we want to understand the relative role of these factors in explaining the

dollar-denominated debt growth in the global bond market. This research question

becomes more important nowadays in the sense that market participants and policy

makers are more aware of international liquidity transmission as a potential threat to

global financial stability. The conventional wisdom in the literature suggests that domestic

business cycle and monetary policy determine the liquidity condition within one country.

This was mainly due to lack of global financial market integration. Domestic condition was

more relevant for policy makers at that time to implement policies to offset economic

fluctuation. With global trade and financial integration, the impact of the external shocks

on domestic market has been increasing over time. It is important for market participants

and policy makers to watch the external economic environment, especially the most

influential player in global financial market - Federal Reserve System of United States (the

Fed).

Taper Tantrum in 2013 is a good example of the Fed’s monetary policy influence

on foreign exchange rates and asset prices. Meanwhile, regional economic integration, for

instance, the European Monetary Union (EMU), also have an impact on foreign currency

loans and deposit in domestic banking system. (Kishor and Neanidis, 2015)

This paper decomposes the movements in dollar debt in a set of 12 countries into

global, emerging market and idiosyncratic factors using a dynamic factor model. This

approach helps us in examining the relative importance of different factors in evolution of

dollar debt growth in these countries. This also helps us in examining the hypothesis of

loose US monetary policy as the primary source of the boom in dollar debt growth.

We find that the global factor accounts for most of the variations of the dollar debt

growth in these countries. The emerging market factor also plays a secondary role in

explaining the rapid growth of the outstanding dollar debt balance among emerging

market economies. The country-idiosyncratic factor, however, is not as important

compared to these external factors. The results suggest that the global financial market

integration plays an important role in the cross-border liquidity transmission. From the

time series of three decomposed factors, we find that the global factor has been

accelerating since 2009. This pattern coincides with the extremely loose monetary policy

environment in the US after Great Recession. This result further enhances the message to

the policy makers that major country monetary policy spillover effect can have a

significant impact on the cross-border capital flow of other countries. We also find that

financial market uncertainty, VIX, is highly negatively correlated with the level of global

factor implying global increase in the level of debt dollarization in case of a decline in

financial market uncertainty.

The rest of the paper are structured as follows. The second section provides a

literature review about global liquidity transmission and discusses the associated

monetary policy spillover effect. The third section introduces the data used in this study

and the setup of the dynamic factor model. The fourth section interprets and presents the

results from the model. The last section concludes the paper.

4.2 Literature Review

The recent surge in non-financial corporate (hereafter ”corporate” ) overseas debt