NBER WORKING PAPER SERIES

SELF-ORIENTED MONETARY POLICY, GLOBAL FINANCIAL MARKETS ANDEXCESS VOLATILITY OF INTERNATIONAL CAPITAL FLOWS

Ryan BanerjeeMichael B. DevereuxGiovanni Lombardo

Working Paper 21737http://www.nber.org/papers/w21737

NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue

Cambridge, MA 02138November 2015

The views expressed here are our own and do not reflect those of the Bank for International Settlements.Part of Michael B. Devereux’s contribution to this work was supported by the Economic and SocialResearch Council [grant number ES/I024174/1] , and the Social Science and Humanities ResearchCouncil of Canada. The views expressed herein are those of the authors and do not necessarily reflectthe views of the National Bureau of Economic Research.

NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies officialNBER publications.

© 2015 by Ryan Banerjee, Michael B. Devereux, and Giovanni Lombardo. All rights reserved. Shortsections of text, not to exceed two paragraphs, may be quoted without explicit permission providedthat full credit, including © notice, is given to the source.

Self-Oriented Monetary Policy, Global Financial Markets and Excess Volatility of InternationalCapital FlowsRyan Banerjee, Michael B. Devereux, and Giovanni LombardoNBER Working Paper No. 21737November 2015JEL No. E3,E5,F3,F5,G1

ABSTRACT

This paper explores the nature of macroeconomic spillovers from advanced economies to emergingmarket economies (EMEs) and the consequences for independent use of monetary policy in EMEs.We first empirically document the effects of US monetary policy shocks on a sample group of EMEs.A contractionary monetary shock leads a retrenchment in EME capital flows, a fall in EME GDP,and an exchange rate depreciation. We construct a the- oretical model which can help to account forthese findings. In the model, macroeconomic spillovers are exacerbated by financial frictions. Weassess the extent to which domestic monetary policy can mitigate the negative spillovers from foreignshocks. Absent financial frictions, international spillovers are minor, and an inflation targeting rulerepresents an ef- fective policy for the EME. With frictions in financial intermediation, however, spilloversare substantially magnified, and an inflation targeting rule has little advantage over an exchange ratepeg. However, an optimal monetary policy markedly improves on the performance of naive inflationtargeting or an exchange rate peg. Furthermore, optimal policies don’t need to be coordinated acrosscountries. Under the specific set of assumptions maintained in our model, a non-cooperative, self-orientedoptimal policy gives results very similar to those of a global cooperative optimal policy.

Ryan BanerjeeMacroeconomic AnalysisBank for International SettlementsCentralbahnplatz 2CH-4002 BaselSwitzerlandryan.banerjee@bis.org

Michael B. DevereuxDepartment of EconomicsUniversity of British Columbia997-1873 East MallVancouver, BC V6T 1Z1CANADAand NBERmbdevereux@gmail.com

Giovanni LombardoBank for International Settlementsgiannilmbd@gmail.com

1. Introduction

In recent years, the global economy has seen dramatic examples of volatility in capitalflows to emerging market countries. Following the global financial crisis and the subsequentrapid monetary easing in the US and other advanced economies, there was a period of largecapital inflows into many fast growing EMEs such as China, India and Brazil. In 2013,the threat of a US monetary ‘taper’ led to an abrupt reversal of inflows to many emergingeconomies. The defining characteristic of these two episodes is that capital flows were drivento a large degree by macroeconomic and financial conditions in the advanced economies,especially those in the US. Although the size of the US economy relative to world GDP hasfallen in recent decades, the US still plays an outsized role in the global financial system(e.g. Fischer, 2014), one reason being the overwhelming predominance of the US dollar as afunding currency for global capital flows.

There is substantial empirical evidence linking international capital flows to US assetprices and US monetary policy. Rey (2013), Miranda-Agrippino and Rey (2014), and Brunoand Shin (2015) describe a ‘global financial cycle’ in which capital flows to many countriesare highly positively correlated and closely tied to US monetary policy. Miranda-Agrippinoand Rey (2014) find that a tightening of US monetary policy leads to a spike in global riskaversion, a fall in cross border lending, and a fall in asset prices at a global level. Theyidentify a single global factor that can explain a large part of the movement in cross bordercredit flows, as well as domestic credit growth. Moreover, this factor can be related tochanges in US policy rates.

A major policy question arising from these events is whether US monetary policy impartsa global ‘externality’ through spillover effects on world capital flows, credit growth and assetprices. Many policy makers in emerging markets (e.g. Rajan, 2014) have argued that theUS Federal Reserve should adjust its monetary policy decisions to take account of the excesssensitivity of international capital flows to US policy. This criticism questions the view that a‘self-oriented’ monetary policy based on inflation targeting principles represents an efficientmechanism for the world monetary system (e.g. Obstfeld and Rogoff, 2002), without theneed for any cross-country coordination of policies.

A related question is whether EMEs that find themselves excessively affected by capitalflow volatility need more policy tools besides interest rate and exchange rate adjustment.Rey (2013) argues that for small open countries in present day global financial markets, theclassic policy ‘trilemma’ which states that independent policy may be followed provided theexchange rate is flexible, in fact collapses to a ‘dilemma’, since exchange rate adjustmentcannot easily insulate against large reversals in capital flows. The ‘dilemma’ is one whereemerging market countries can either maintain an open capital account but remain vulnerableto the global financial cycle, or choose to impose capital controls in order to achieve a greaterdegree of macro policy independence.

Our paper provides empirical as well as theoretical analysis of international spillovers,their consequence for the design of monetary policy, and the desirability of ‘self-oriented’monetary policy. We add to the empirical literature on macroeconomic spillovers by ex-amining the impact of US monetary policy shocks on EMEs using a panel of 16 emergingmarket countries. US monetary policy shocks are identified as in Coibion et al. (2012), whoprovide an update to the Romer and Romer (2004) shocks. We examine the response of EME

2

policy rates, exchange rates, GDP, inflation, and gross capital flows to these shocks. Theestimation is based on the local projection method proposed by Jorda (2005). We find thatan unexpected US monetary policy tightening leads to a fall in EMEs’ economic activity, arise in policy rates, an exchange rate depreciation, and a retrenchment of capital flows; thatis, a fall in both capital inflows to EMEs and outflows from EMEs.

In the theoretical analysis, we develop a simple core-periphery DSGE model driven bymonetary policy and financial shocks in the core country whose currency dominates the flowsof financial capital across borders. Our model is based on the relationship between financialinstitutions in a large financial centre (global banks or asset managers) and borrowing banksor financial institutions in an emerging market country. We find that when these financialinstitutions face agency constraints which restrict the growth of their balance sheets, thenmonetary policy shocks or financial shocks in the centre country can produce many of thefeatures of international capital flows described above. A monetary contraction in the centreleads to a sharp decline in capital inflows to the peripheral country, a fall also in outflowsfrom the periphery, a real exchange rate depreciation in the periphery, and a rise in interestrate spreads which precipitates a coordinated downturn in real economic activity.

We find that, for the baseline calibration of our model, the response of asset prices andinterest rate spreads in emerging economies to a monetary contraction in the centre countrycan in fact be larger than the direct responses of these variables in the centre country itself.Thus, sudden reversals in the monetary policy stance of the centre country can generatewhat looks like excessive responses in the financial markets of emerging economies. This isthe case even if the emerging economy allows its exchange rate to adjust freely.

The key mechanism in the model is the magnification effect of shocks to the balance sheetsof global lenders compounded with those of local emerging market borrowers. A monetarytightening in the centre country raises interest rates and funding costs for global lenders.This erodes their net worth, requiring them to reduce lending to local emerging marketborrowers. In addition, EMEs experience an immediate real exchange rate depreciation.The combination of increased borrowing costs and unanticipated depreciation, which raisesthe costs of servicing existing debt, leads a sharp decline in net worth for emerging marketborrowing institutions. This leads to a rise in spreads in emerging markets. We find that thespreads rise significantly more in the emerging market country than in the centre country,since they are subject to a ‘double agency’ effect. In contrast to a basic core-peripheryDSGE model without constrained financial institutions, a simple inflation targeting rule(like a Taylor rule) does not insulate the peripheral economy from international monetaryspillovers.

We go on to explore the implications of alternative policy and financial structures on thenature of financial and real spillovers. We ask how the nature of spillovers would differ if theemerging market were able to borrow in its own currency. This would eliminate the directdeterioration of balance sheets coming from exchange rate depreciation. We find in this casethat the contraction in lending and the rise in spreads is mitigated somewhat, so that theimpact on the real economy is smaller. But despite this, the emerging economy is still highlyvulnerable to the cutback in direct capital flows and the increase in funding costs comingfrom the centre country, so that the overall magnitude of spillovers is still very large.

With frictionless financial markets, a pegged exchange rate magnifies the response of realvariables to the external shock, as it curtails the required adjustment in the real exchange

3

rate. But when global and local financial firms are subject to agency constraints, the mag-nitude of spillovers differs little between an exchange rate peg and an inflation targetingmonetary policy.

We find similar results when we allow drivers of capital flows other than core countrymonetary shocks. Direct shocks to the financial system in the core country triggers many ofthe same features as those of the monetary shock described above.

These results would seem to support the ’dilemma’ view. But in fact, we show thatthis conclusion does not follow when we study optimal monetary policy responses. A globalcooperative monetary response to a financial downturn can largely eliminate the negativeimpact of the capital flow spillovers. For this response to work however, it is essential thatthe periphery country exploit the flexibility of its exchange rate. Thus, when an optimalcooperative monetary rule is considered, the policy ‘trilemma’ becomes relevant again.

Practically speaking, monetary policy is set at the national level, and especially forthe countries at the financial centre, national considerations alone will dictate policy re-sponses. Although naive inflation targeting monetary rules have poor properties in dealingwith international spillovers in the presence of agency distortions in international financialintermediation, this does not mean that any self-oriented monetary policy is ineffective. Weshow that in a model with non-cooperative (Nash) optimal monetary policy game whereboth core and peripheral countries independently choose an optimal monetary policy. Wefind that, under the specific assumptions of our model, the outcome of the non-cooperativegame is very similar to the optimal cooperative solution.

Our paper is related to a large literature on capital flows and the macroeconomics ofEMEs. Many recent papers have documented the empirical features of capital flows andspillovers from advanced economies to emerging market economies. In particular Ahmedand Zlate (2014), Chen et al. (2015), Bowman et al. (2015), and Chen et al. (2014) examinethe impact of the recent US unconventional monetary policy on emerging markets. Exceptfor Chen et al. (2015), who look at the response of emerging market inflation and real GDP,these papers restrict their focus to the impacts on interest rates, asset prices and capitalflows. Gilchrist et al. (2014) examine the effect of US monetary policy shocks including theperiod before unconventional monetary policies, but focus on the response of sovereign bondyields. Fratzscher et al. (2014) examine the impact of ECB unconventional monetary policieson emerging market asset prices. In addition, as discussed above, Rey (2013), and Miranda-Agrippino and Rey (2014), examine spillovers in capital markets using factor analysis toidentify global shocks. In a recent work Dedola et al. (2015) provide empirical evidence ofstrong international spillovers of US monetary policy shocks. Using a BVAR, they identifymonetary policy shocks by means of sign restrictions. Bluwstein et al. (2015) study spilloversfrom the euro area to nine non-euro European economies using a mixed-frequency BVAR.Among their findings: counties with more integrated financial markets and larger shares ofbanks react more strongly to ECB unconventional policies. The credit-channel plays a minorrole in the transmission.

Our paper differs from this previous literature in a number of respects. In particular, wefocus on US monetary policy shocks before the advent of the zero lower bound, we focus onthe general macroeconomic response in a group of emerging market countries, including theresponse of interest rates, real GDP, inflation and capital flows, and rather than employinga restrictive VAR specification, we follow Jorda (2005) in using a local projection method

4

to construct impulse response functions to US monetary shocks. In addition, we explore theresponse of both gross capital inflows and outflows to EMEs. 1

The conceptual framework employed in our paper is similar to that of Bruno and Shin(2014), although our structural model and analysis is very different from their paper. Insome respects our modelling strategy is close to the works by Devereux and Yetman (2010),Dedola and Lombardo (2012), Dedola et al. (2013), Ueda (2012), Kollmann et al. (2011),Kolasa and Lombardo (2014), Choi and Cook (2004), Perri and Quadrini (2011) Korinek(2014), and Nuguer (2014). These authors study various positive and normative aspectsof international spillovers due to financial frictions. Our paper builds on these ideas toaddress the specific questions highlighted above. A closely related investigation is carriedout by Agenor et al. (2014), who model a small-open-economy DSGE model with two-layersof financial intermediation. Their main focus is on financial market regulation and macro-prudential policy.

The rest of the paper is organized as follows. Section 2 discusses the empirical evidenceand provides novel estimates of the response of EMEs’ variables to US policy shocks. Section3 develops the theoretical model. Section 4 discusses the parametrization of the model.Section 5 discusses monetary policy shocks, while Section 6 provides results on financialshocks. Sections 7 and 8 introduce the optimal (Ramsey) cooperative policy and the optimalnon-cooperative policy, respectively. Section 9 compares the results under the different policyarrangements. Finally Section 10 offers some concluding remarks.

2. Capital Flows to Emerging Markets: Some recent evidence

In 2013 and 2014, emerging market economies experienced significant volatility in grossand net capital flows. Observers have attributed much of this to actual or prospectivechanges in monetary policy in advanced economies. But in fact, highly volatile capitalflows are a fact of life for emerging market countries. Figure 2.1 illustrates net flows intoemerging market portfolio funds for a group of emerging markets since 2009. Following thehighly accommodative monetary policies of advanced countries in 2009-2010, there was asignificant uptick in net inflows to emerging markets. This continued with some volatilityuntil 2013, when the proximate cause of the US ‘taper’ announcement led to large outflowsfrom EME countries, both in bonds and equity assets.

Figure 2.2 shows the currency composition of emerging economies net issuance of debtsecurities over the past four years. A significant fraction of new issues remain denominatedin foreign currencies, with the US dollar still representing the major share of these. Theright hand panel of Figure 2.2 shows that the US dollar comprises about 90 percent of theoutstanding stock of debt securities for this representative group of EMEs.

1Alberola et al. (2012) look at the response of gross foreign outflows as well as gross inflows to EMEsduring episodes of financial crises. As in our empirical and theoretical analysis, they show that financialcrises may be associated with a retrenchment of capital flows - there is a fall in both inflows to EMEsand outflows from EMEs. They show however that this response is critically related to the size of foreignexchange reserve holdings - EME economies with higher FX reserves tend to see a greater fall in capitaloutflows during financial crises. Our formal analysis abstracts from the importance of FX reserves. Wefurther discuss the implication of explicitly allowing a role for FX reserves below.

5

Figure 2.1: Capital Flows to EMEs

–40

–30

–20

–10

0

10

20

30

2009 2010 2011 2012 2013 2014 2015

BondsEquities

Figure 2.2: Currency Exposure for EMEs

Net issuance, major currenciesshare of total debt securities Outstanding stock, major currencies

–10

0

10

20

30

2012 2013 2014 2015

USD EUR JPY

0.0

0.4

0.8

1.2

1.6

2012 2013 2014 2015

USD EUR JPY

All issuers, all maturities, by nationality of issuer. Countries: Argentina, Brazil, Chile, China, Colombia, theCzech Republic, Hong Kong SAR, Hungary, India, Indonesia, Israel, Malaysia, Mexico, Peru, the Philippines,Poland, Russia, Saudi Arabia, Singapore, South Africa, South Korea, Thailand and Turkey. Data up to 13April 2015.

Sources: Dealogic; Authors’ calculations.

6

Figure 2.3: High Correlation of Spreads.

Source: BofA Merrill Lynch – option-adjusted spreads – retrieved from FRED, Federal Reserve Bank of St. Louis.

Our theoretical analysis of spillovers depends in a central way on the correlation of interestrate spreads across countries. Figure 2.3 illustrates the path of interest rate spreads in theUS domestic economy, in Asia, Latin America and Emerging Markets generally. The USdomestic issue on average has the lowest risk spreads, but clearly there is an extremely highcorrelation between all the spreads. We also see that the jump in spreads in EMEs duringthe 2008-9 financial crisis far exceeded the analogous increase in the US.

2.1. Empirical estimates of US monetary policy shocks spillovers to EMEs

As a backdrop to our theoretical analysis, we wish to document some general featuresof macroeconomic spillovers to emerging market economies. We do this by focusing on theresponse to one particular shock; a US monetary policy shock. In order to empiricallyexamine how US monetary policy spills over to EMEs, we estimate the response of a numberof EMEs’ economic variables to an unexpected US monetary policy contraction. We identifymonetary policy shocks following Coibion et al. (2012), who update the Romer and Romer(2004) estimates of US monetary policy shocks. Unlike the recent literature on the impactsof unconventional US monetary policy, our policy shocks are focused on the experience beforethe zero interest rate period.

Rather than explicitly specifying a VAR model, we following (Jorda, 2005) in using localprojection techniques to estimate the impact of US monetary policy shocks across differenteconomic states in a panel of emerging economies. We estimate the impact of the policyshock in period t labeled MSt, on the variable of interest yi,t+h in country i at time t + hfrom the following local projection

7

yi,t+h − yi,t−1 = αh + θhMSt + γhwi,t−1 + εi,t+h forh = 0, 1, ..., H (2.1.1)

where wi,t−1 is a vector of control variables known prior to the US monetary policy shocks.Assuming the conditional mean can be linearly approximated, θh estimates the averagetreatment effect of a period t US monetary policy shock in period t + h. Thus, impulseresponses are computed as a sequence of the θh (h = 0, 1, ..., H) estimated in a series ofsingle regressions for each horizon. Standard errors are clustered by country to account forwithin country correlation.

Our estimates are based on quarterly data between Q1 1989 and Q2 2007. We constructa quarterly measure of US monetary policy shocks, MSt, by aggregating the Coibion et al.(2012) US monetary policy shocks from each meeting in a given quarter. The emergingeconomies included in the unbalanced panel are Argentina, Brazil, Chile, China, Colombia,India, Indonesia, Korea, Malaysia, Mexico, Peru, the Philippines, Russia, South Africa,Thailand and Turkey. The dependent variables considered are the log bilateral exchange ratewith the US dollar, log real GDP, inflation, the domestic policy rate, and both portfolio debtoutflows and inflows as a share of GDP. The vector of control variables wi,t−1 consist of twolags of output growth, export-weighted GDP growth of country i’s major trading partners,inflation, the domestic policy rate, the log change in the US dollar bilateral exchange rate,the change in domestic long-term bond yields, the change in the ratio of foreign exchangereserves to short-term external borrowing, both portfolio debt inflows and outflows as a shareof GDP, past US monetary policy surprises and US 10-year Treasury yields. The HP-filtereddomestic output gap in period t− 1 is also included as a control variable.

The top-left panel of Figure 2.4 shows that there is some persistence in US monetarypolicy surprises as measured by the Romer and Romer shocks, but they essentially die outafter two quarters. Following a 100 basis point US monetary policy surprise (annualized),quarterly real GDP in emerging economies decline by around 0.5% after one quarter but thisinitial negative effect reverses. Bilateral exchange rates initially depreciate by around 4%(quarterly). The inflation rate initially increases by around 250 basis points (annualized),which could be due to strong exchange rate pass-through, but the subsequent decline issuggestive of a price puzzle, similar to that found in advanced economies. Domestic policyrates are initially tightened by nearly 200 basis points (annualized) in response to the 100basis point US monetary policy tightening, but this initial tightening is quickly reversedafter one quarter. In terms of the impact on gross capital flows, the impulse responses showa retrenchment of capital flows in response to US monetary policy tightening. There is apersistent fall in portfolio debt inflows to emerging economies which decline by nearly 3% ofGDP (quarterly), and remain around 1% lower after four quarters. In addition, (quarterly)portfolio debt outflows also fall following US monetary policy tightening. This evidence ofthe retrenchment of gross capital flows is similar to that in Alberola et al. (2012) who findfalls in both capital in- and out-flows from emerging economies following periods of financialstress.

We now go on to develop a DSGE model of macroeconomic spillovers which can be usedto construct a theoretical counterpart to these empirical results.

8

Figure 2.4: EME impulse responses to US monetary policy shocks (90% confidence intervals)

Romer and Romer shock Log real GDP Log exchange rate against USD

Inflation rate Domestic policy rate Portfolio debt inflow/GDP

Portfolio debt outflow/GDP

0.0

0.5

1.0

−1.0

−0.5

0.0

0.5

1.0

−8

−4

0

4

−5.0

−2.5

0.0

2.5

−6

−3

0

3

−4

−3

−2

−1

0

−1

0

1

0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5

0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5

0 1 2 3 4 5

9

3. The Global Model

Our results are structured around a two country core-periphery model. The centre/corecountry is assumed to be large relative to the peripheral country. We denote the emergingeconomy with the superscript ‘e’ and the centre country with the superscript ‘c’.

The schemata for our model is described in Figure 3.1. In the centre country there arehouseholds, global financiers (banks or asset managers2), capital goods producers, productionfirms, and a monetary authority. There is a global capital market for one-period risk freebonds. In the emerging market country there are also households, local borrowers (banks orfinancial managers), capital goods producers, production firms, and a monetary authority.The centre country households make deposits with global financiers at the centre countryrisk free rate, and can hold centre country one-period nominal government debt, which mayalso be traded on international capital markets. The global banks receive deposits fromhouseholds in the centre country, and invest in risky centre country technologies, as wellas in emerging market banks. Along the lines of Gertler and Karadi (2011), banks in bothcountries finance purchases of capital from capital goods producers, and rent this capital togoods producers. The borrowing banks in the emerging market economy are funded throughloans from global banks/financiers.3 There are two levels of agency constraints; global banksmust satisfy a net worth constraint in order to be funded by their domestic depositors, andlocal EME banks in turn must have enough capital in order to receive loans from globalbanks. In both countries, the production goods firms use capital and labour to producedifferentiated goods, which are sold to retailers. Retailers are monopolistically competitiveand sell to final consuming households, subject to a constraint on their ability to adjustprices. This set of assumptions constitutes the minimum arrangement whereby capital flowsfrom advanced economies to EMEs have a distinct directional pattern, financial frictionsact to magnify capital slow spillovers, and (due to sticky prices) the monetary policy andexchange rate regime may have real consequences for the nature of spillovers and economicfluctuations.

The emerging country is essentially a mirror image of the centre country, except thathouseholds in the emerging country do not finance local banks, but instead engage in inter-temporal consumption smoothing through the purchase and sale of centre currency denom-inated nominal bonds.4 Banks in the emerging market use their own capital and financingfrom global financiers to make loans to local entrepreneurs. The net worth constraints onbanks in both the emerging market and centre countries are motivated along the lines of

2In the remainder of the paper, to simplify the discussion, we will refer to capital goods financiers in boththe centre and peripheral countries as banks. It should be noted however that the key thing that distinguishesthem is that they make levered investments, and are subject to contract-enforcement constraints. In thissense, they need not be literally banks in the strict sense.

3This assumption is meant to capture the feature that within-country financial intermediation betweensavers and investors is more difficult in EMEs than in advanced economies. See e.g. Mendoza Quadrini andRios-Rull 2009. We could relax the extreme assumption that EME households did not directly finance EMEbanks. Under the reasonable assumption that frictions in intermediation within EMEs exceeded those in thecentre country, our qualitative results would remain unchanged.

4We assume that the market for centre country nominal bonds is frictionless. Adding additional frictionsthat limit the ability of emerging market households to invest in centre country nominal bonds would justexacerbate the impact of financial frictions that are explored below.

10

Figure 3.1: The world economy

Architecture of model

!"#$%"&'('&

&&'('&

!"#$%)*&

+)#,&

&&&&&'('&

-./0"1.*2&

3.4)*&

5)#,&6*.5)*&

5)#,&

!&78%9&'('&78%9&

!):&

;..20&

!):&

;..20&

&&&&!&

!"#$%)*&

+)#,&

&&&&&&&&!&

-./0"1.*2&

!):8$)*&9)%,"$0&

3.)#0&

<":.08$0&

=%.2/4>.#&

78#)#4"&

11

Gertler and Karadi (2011).

3.1. The Emerging Market Economy (EME)

A fraction n of the world’s households live in the emerging economy. Households consumeand work, and act separately as bankers. A banker member of a household has probability θof continuing as a banker, upon which she will accumulate net worth, and a probability 1−θ ofexiting to the status of a consuming / working household member, upon which all net worthwill be deposited to her household’s account. In every period, non-bank household membersare randomly assigned to be bankers so as to keep the population of bankers constant.

While EME households don’t have access to the local financial market, they can trade ininternational bonds (Be

t ) with foreign agents.5 These bonds can be thought of as T-bills of thecore country (rebated directly to core-country households), deposits at core-country financialintermediaries, or simply bonds traded directly with core-country households. Under theassumptions of our model these three alternatives are equivalent.6

Households in the EME have preferences over (per capita) consumption Cet and labor He

t

supply given by:

E0

∞∑t=0)

βt

(Ce(1−σ)t

1− σ− H

e(1+ψ)t

1 + ψ

)where consumption is broken down further into consumption of home (Ce

et) and foreign (Cect)

baskets as

Cet =

(ve

1ηC

e1− 1η

et + (1− ve)1ηC

e1− 1η

ct

) 11−η

Here η > 0 is the elasticity of substitution between home and foreign goods, ve ≥ n indicatesthe presence of home bias in preferences,7 and we assume in addition that within each basket,goods are differentiated and within country elasticities of substitution are σp > 1.

Given this, the true price index for EME households is

P et =

(veP 1−η

et + (1− ve)P 1−ηct

) 11−η

Then the household budget constraint is described as follows

P et C

et + StB

et = W e

t Het + Πe

t +R∗t−1StBet−1

Households purchase dollar (centre country) denominated debt (Bet ). St is the nominal

5This is clearly an extreme case. In reality EME households’ saving does reach domestic firms viathe local banks too. Since, in our model, EME households can lend to domestic firms indirectly, via theinternational financial market, our assumption emphasizes the strong influence that core-country financialconditions exert on EME financial markets.

6In particular we are not assuming a special role for government debt, nor asymmetries in the degree towhich the contract between depositors and core-country banks can be enforced.

7Home bias is adjusted to take into account of country size. In particular, for a given degree of opennessx ≤ 1, ve = 1− x(1− n), and a similar transformation for the centre country home bias parameter.

12

exchange rate (price of centre country currency). They consume home and foreign goods.W et is the nominal wage, and Πe

t represents profits earned from banks and firms, net of newcapital infusion into banks. R∗t is the centre country rate on bonds. Households have thestandard Euler conditions and labor supply choices described by

EtΛet+1

R∗tπet+1

St+1

St= 1

W et

P et

= Ceσt H

eψt

where Λet+1 ≡ β

(Cet+1

Cet

)−σ, and πet+1 ≡

P et+1

P et.

Given two-stage budgeting, it is straightforward (and omitted here) to break down con-sumption expenditure of households into home and foreign consumption baskets.

3.2. Capital goods producers

Capital producing firms in the EME buy back the old capital from banks at price Qet (in

units of the consumption aggregator) and produce new capital from the final good in theEME economy subject to the following adjustment cost function:

P et I

et

(1 + ζ

(P et It

P et−1It−1

− 1

)2)

where Iet represents investment in terms of the EME aggregator good.EME banks then finish the capital goods and rent them to intermediate goods producers.8

Ket = Iet + (1− δ)Ke

t−1

where Ket is the capital stock in production.

3.3. EME banks

EME banks begin with some bequeathed net worth from their household, and continueto operate with probability θ, as described above. We also follow Gertler and Karadi in thenature of the incentive constraint. Ex ante, EME banks have an incentive to abscond withborrowed funds before the investment is made. Consequently, conditional on their net worth,their leverage must be limited by a constraint that ensures that they have no incentive toabscond.

At the end of time t a bank i that survives has net worth given by N et,i in terms the

EME good. It can use this net worth, in addition to debt raised from the global bank, toinvest in physical capital at price Qe

t in the amount Ket+1,i. Debt raised from the global

bank is denominated in centre country currency (although later we will experiment with

8Equivalently, we could assume that the bank provides risky loans to intermediate goods producers,who use the funds to purchase capital. The only risk of this loan concerns the (real) gross return on theunderlying capital stock.

13

local-currency denomination). In real terms (in terms of the centre country CPI), we denotethis debt as V e

t,i. Thus, EME bank i′s balance sheet is given by

N et,i +RERtV

et,i = Qe

tKt+1,i (3.3.1)

where RERt =StP ctP et

is the real exchange rate.

Bank i′s net worth is the difference between the return on previous investment and itsdebt payments to the global bank.

Nt,i = Rek,tQ

et−1K

et,i −RERtRct−1V

et−1,i

where Rct−1 is the ex-post real interest rate received by the global bank, equal to the prede-termined nominal interest rate adjusted by ex-post inflation in the centre country, and Re

k,t

is the gross return on capital.Because it has the ability to abscond with the proceeds of the loan and its existing net

worth, the loan from the global bank must be structured so that the EME bank’s continuationvalue from making the investment exceeds the value of absconding. Following Gertler andKaradi (2011), we assume that the latter value is κet times the value of existing capital ( κetis a random variable, and represents the stochastic degree of the agency problem). Hencedenoting the bank’s value function by Jet,i, it must be the case that

Jet,i ≥ κetQetK

et+1,i. (3.3.2)

This is the incentive compatibility constraint faced by the bank.Once the bank has made the investment, at the beginning of period t + 1 its return is

realized.The problem for an EME bank at time t is described as follows:

Max Jet,i [Ket+1,i,V

et,i]

= EtΛet+1

[(1− θ)(Re

k,t+1QetK

et+1,i −RERt+1RctV

et,i) + θJet+1,i

]subject to (3.3.1) and (3.3.2).

The full set of first order conditions for this problem are set out in the Appendix.The evolution of net worth averaged across all EME banks, taking account that banks

exit with probability 1− θ, and that new banks receive infusions of cash from households atrate δT times the existing value of capital, can be written as:

N et+1 = θ

((Re

kt+1 −RERt+1

RERt

Rc,t)QetK

et +

RERt+1

RERt

Rc,tNet

)+ δTQ

etK

et−1

The first term on the right hand side captures the increase in net worth due to survivingbanks, given their average return on investment. The second term represents the ‘start-up’financing given to newly created banks by households.

Firms in the EME hire labour and capital to produce retail goods. Since a central aimof our analysis is to explore the role of monetary policy and the exchange rate regime forcapital flows and macroeconomic spillovers, we assume that firms in both countries haveCalvo-style sticky prices with Calvo re-set parameter 1 − ς. The representative EME firm

14

has production function given by:

Y et = AetH

e(1−α)t K

e(α)t−1

Given this, then we can define the aggregate return on investment for EME banks (aver-aging across idiosyncratic returns) as

Rekt+1 =

Rezt+1 + (1− δ)Qe

t+1

Qet

where Rzt+1 is the rental rate on capital and δ is the depreciation rate on capital.The representative EME firm chooses labour and capital so as to minimize costs. We can

then define the EME firm’s real marginal cost implicitly by the conditions

MCet(1− α)AetHe(−α)t K

e(α)t−1 = W e

rt (3.3.3)

MCetαHe(1−α)t K

e(α−1)t−1 = Re

zt

The Calvo pricing formulation implies the following specification for the PPI rate ofinflation πppiet in the EME. Here Π∗et denotes the inflation rate of newly adjusted goods prices,Fet and Get are implicitly defined, and σp

σp−1 represents the optimal static markup of price

over marginal cost:

Π∗et =σp

σp − 1

FetGet

πppiet (3.3.4)

Fet = YetMCet + Et[βςΛe

t,t+1πppiet+1

ηFet+1

](3.3.5)

Get = YetPet + Et[βςΛe

t,t+1πppiet+1

−(1−η)Get+1

](3.3.6)

πppiet1−η = ς + (1− ς) (Π∗et)

1−η (3.3.7)

3.4. Monetary policy

In the baseline specification, the central bank follows a flexible inflation targeting policy,here captured by a generalized Taylor rule, i.e.

logRt = λr,e logRt−1 + (1− λr,e)(λπ,e log

(πetπess

)+ λy,e log

(Y et

Y ess

))+ εer,t. (3.4.1)

where εer,t is a monetary policy shock.In analysis below we will also consider the design of optimal monetary policies. We will

focus on Ramsey optimal policy with commitment (or ‘open loop’ optimal policy), wherethe path of interest rates is adjusted in order to implement an optimal policy. We explorethe nature of optimal policy under two strategic assumptions about policy making. In thefirst, we will assume that central banks cooperate in order to implement the Ramsey optimalallocation. In this case they choose the allocation that maximizes the size-weighted averageof households’ welfare across countries, subject to the competitive equilibrium conditions.Following this, we define and analyze a Ramsey optimal non-cooperative monetary policy,

15

in which each central bank follows its own optimal monetary rule, taking that of the othermonetary authority as given.

We will also consider the case of an exchange rate peg. In this case we implement theEME central bank’s policy in terms of the simple targeting rule ∆St = 0. The details areprovided in Appendix A.9

3.5. The centre country

The centre country households have similar preferences to those of the EME, and itsproduction firms sell to the emerging market country households. The centre country’sfinancial institution (the global bank) receives deposits from the households and guaranteesthem the risk-free interest rate in return. The global bank then invests in the centre countrytechnology as well as the EME bank debt.

Centre country representative household preferences are:

E0

∞∑t=0

βt

(Cc(1−σ)t

1− σ− H

c(1+ψ)t

1 + ψ

)and their budget constraint is given by:

P ct C

ct +Bc

t = W ctH

ct + Πc

t +R∗tBct + T ct

Centre country households make deposits in the banking system, and receive returns R∗t .They receive profits Πc

t from (financial and non financial) firms, net of capital infusions intothe new banks.

The definition of centre country CPI’s, and bond and labour supply choices for thecentre country households are exactly analogous to those of the EME country, so we omitthem here. Likewise, the specification for capital producing firms and the dynamics of theaggregate capital stock for global banks is identical to that described for the EME economy.See Appendix A for details.

9We do not explicitly model the central bank balance sheet or the consequences of alternative exchangerate regimes for the accumulation of international reserves. This is justified on the same grounds as that ofWoodford (2003) who models monetary policy in ‘cashless’ economies abstracting from monetary aggregatesor the central bank balance sheet. In following this strategy, we are attempting to focus specifically on therole of the exchange rate and interest rates as monetary policy levers for EMEs. A more extended analysiscould be done by explicitly modelling the effect of foreign exchange rate reserves. In particular, if as inBacchetta et al. (2013), we assumed the EME residents had limited access to international financial markets,then central bank reserves, or foreign currency swaps with central banks, could play an effective role inproviding risk sharing. Alternatively, following Gertler and Karadi (2011) and Dedola et al. (2013), we couldallow for EME central banks to provide foreign currency loans to EME banks, thereby introducing a rolefor unconventional monetary policy in EMEs. We abstract away from the first option in this paper, sincethe consequences of limited capital mobility have been extensively examined elsewhere. The second option,direct foreign currency loans or foreign currency swaps, could be an interesting alternative policy tool. So,while in this paper we focus on standard policy interventions, we leave the second option for future research.Nuguer (2015) constructs a multi-country model with a global interbank market in a similar fashion to ourmodel, and explores the effects of unconventional credit policies.

16

3.6. Centre country banks

A representative global bank j has a balance sheet constraint given by

V ejt +Qc

tKcjt = N c

jt +Bct

where V ejt is investment in the EME bank, and Qc

tKcjt is investment in the centre country

capital stock. N ejt is the bank’s net worth, and Bc

t are deposits received from households.All variables are denominated in real terms, (in terms of the centre country CPI).

The global bank’s value function can then be written as:

J cjt = Et maxKcj,t+1,V

ejt,B

ct

Λct+1

[(1− θ)(Rc

kt+1QctK

cjt +RctV

ejt −R∗tBc

t ) + θJ cjt+1

]Here, Λt+1 is the stochastic discount factor for centre country households, Rct is, as

described above, the return on the global bank’s loans to the EME bank, and R∗t is therisk-free rate paid to domestic depositors.

The bank faces the no-absconding constraint:

Jjt ≥ κct(V ejt +QctK

cjt

)where, as in the EME case, κct measures the degree of the agency problem, and as before weassume it is subject to exogenous shocks.

We describe the first order conditions for the global bank in detail in the appendix. Asin the case of the EME banks, we can describe the dynamics of net worth for the globalbanking system by averaging across surviving banks, and including the ‘start-up’ fundingprovided by centre country households. We then get the following law of motion for networth

N ct+1 = θ

((Rckt+1 −R∗t

)QctK

ct + (Rc,t −R∗t )V e

t +R∗tNct

)+ δTQ

ct+1K

ct−1. (3.6.1)

Again, the details of the production firms and price adjustment in the centre country areidentical to those of the EME economy, so we leave the description to the appendix.

3.7. Monetary Policy

The central bank of the centre country, in our baseline specification, follows a Taylor ruleof the type described above (see equation 3.4.1).

4. Calibration

Our aim is to use the model to provide a general qualitative assessment of the empiricalevidence discussed above. We do not attempt to find the best fit of the theoretical modelwith the data. Therefore, we take parameters values that fall generally in line with therelated literature, leaving a more quantitative assessment of the model to future research.Table 1 summarizes the values assigned to the parameters. In particular, we set the opennessparameters νe and νc in line with the trade shares of the US with a group of emerging markets,and the trade shares of the same group of EMEs with the US using the IMF DOT statistics

17

Table 1: Parameter ValuesLabel Valuen 0.15σp 6ς 0.8λy,c 0.2λπ,c 1.2λr,c 0.85β 0.99δ 0.025δT 0.004

Label Valueζ 1.728η 1.5ψ 0.276θ 0.96α 0.3

νe = νc 0.96σ 1.02

κc = κe 0.38

(average shares since 2000). Given that the two shares are very similar in the data we setboth to νc = νe = 0.96. The intertemporal elasticity of substitution is set at approximatelyunity, so that σ = 1.02. The Armington elasticity of substitution between home and foreigngoods (η) is 1.5, while the micro elasticity of substitution σp is 6. The discount factor is setat β = 0.99, while the Frisch elasticity of labour supply 1

1+ψis set at 0.8. The parameters of

production are standard; the share of hours in production, 1−α is .7, while the depreciationrate δ is set at 0.025 (at quarterly frequency), and the parameter in the adjustment costtechnology is 1.73. From Gertler and Karadi (2011) we take the banking sector parametersso that θ = 0.96 (the survival rate of banks), κe = κc = 0.38 (i.e. the steady-state value ofthe incentive compatibility parameter), and δT = 0.004 (i.e. the transfer rate to new banks).10 The probability of changing prices in a quarter (1− ζ) is 0.2.

We will focus on shocks to monetary policy and to ‘financial shocks’, represented byshocks to the parameter κct , the fraction of investment that can be obtained by an abscondingglobal bank. We assume that monetary policy shocks are i.i.d. with a 1 percent standarddeviation. Shocks to κct are AR(1) processes with persistent 0.9 and standard deviation of 1percent also. The Taylor rule coefficients are chosen at standard levels (see Table 1).

5. The impact of monetary policy on capital flows and international transmission

We first explore the impact of centre country monetary shocks on global GDP, capitalflows, asset prices, leverage, and interest rate spreads. The main set of questions we areinterested in is how is the impact of monetary tightening in the centre country is affectedby the presence of financial frictions. In addition, how does the relationship between globalbanks and local banks affect the spillover effects of monetary policy shocks, and how dothese spillovers compare to the effect of a monetary policy shock in a standard multi countryDSGE model without financial frictions?

10As in Gertler and Karadi (2011), these parameters target the steady state interest spread of 100 basispoints, the horizon of bankers of about 10 years, and the steady state leverage ratio of 4. Although clearlyleverage ratios differ across jurisdictions, we do not attempt to match EME ratios, as there is wide variabilityacross different countries. A higher leverage ratio in the EME would magnify our results.

18

In addition, we wish to go beyond the question of transmission with financial frictions toaddress the question of how important is the monetary policy response in the EME country.Does the exchange rate policy followed by the EME significantly affect the internationaltransmission mechanism in the presence of financial frictions? A closely related question isto what extent does the currency of denomination of nominal liabilities affect the transmissionproperties of the model in response to centre country monetary tightening. Does ‘liabilitydollarization’ significantly exacerbate the cross country transmission of monetary shocks?

Figure 5.1 illustrates the effect of a monetary policy tightening in the centre countryin the case without financial frictions.11 The monetary shock is scaled to represent a 1%innovation to the policy rule.12 Without financial frictions, and under a flexible exchangerate (plain line) the shock is almost wholly absorbed within the centre country. The EMEcountry’s real economy is well insulated from the monetary policy shock. The EME policyrate rises only slightly, and there is a sharp real depreciation of the EME currency, butalmost no effect on EME GDP, investment, or asset prices. In the centre country itself,there is a sharp fall in GDP, investment and asset prices. We note also that, in the absenceof financial frictions, the monetary policy tightening leads to an increase in bank lending tothe EME, and an increase in capital outflows from the EME. This pattern of gross capitalflows goes against the empirical evidence described above.

The minimal degree of international transmission of monetary policy in the absence offinancial frictions is in line with traditional models, and supports the theoretical presumptionof an important role for flexible exchange rates in the response to external shocks. The Figurealso illustrates the effect of the same shock, but assuming that the EME central bank choosesan exchange rate peg (but again without financial frictions, crossed line). In this case, the realexchange rate depreciation is dampened significantly, the EME short term policy rate risessharply, and there is a significant fall in real GDP and investment in the EME. Interestinglyhowever, in this case, bank lending to the EME still rises, relative to the initial steady state.

When we introduce financial frictions in the form described in our model however, theresults are dramatically different. Figure 5.2 shows that in the baseline case, with financialfrictions (solid-plain line), the monetary tightening in the centre country precipitates a largeand persistent fall in capital inflows to the EME. The fall in bank loans causes a sharpfall in asset prices, an increase in bank leverage13 , and a rise in interest rate spreads inthe EME. There is a general fall in investment and GDP of similar orders of magnitudein both the centre country and the EME. We also see a fall in capital outflows from the

11In the IRFs, NFA E denotes the aggregate net foreign assets of the EME in terms of steady-state GDPin units of the consumption aggregator. Households NFA, denotes the net foreign assets of the householdssector, in terms of their steady-state value. Note that EME-bank debt is also in terms of its steady-state value.The latter is equal to households’ net foreign assets in the steady state (implying a zero net foreign assetposition in the steady state). ∆FX is the change in nominal foreign exchange, while “spread” measures theex-ante difference between the gross return on capital and the policy rate: a measure of financial inefficiency.

12Due to the endogeneity of the policy rate, the latter moves by less.13The question of whether leverage is procylical or countercyclical has been debated in the literature.

Bruno and Shin 2014b, argue bank leverage is pro cyclical, but Gertler (2012), points out that book valueand market value leverage may move in different directions over the cycle. In particular, in a model similarto ours, he shows that book value leverage may be procyclical while market value leverage is countercyclical,as in our impulse responses.

19

−0.

0528

−0.

0391

−0.

0253

−0.

0116

0.00

22G

DP

E

−0.

0538

−0.

0403

−0.

0269

−0.

0134

0.00

01G

DP

C

−0.

1166

−0.

0823

−0.

048

−0.

0137

0.02

05In

vest

men

t E

−0.

1166

−0.

0822

−0.

0479

−0.

0135

0.02

08In

vest

men

t C

0.01

46

1.64

55

3.27

64

4.90

72

6.53

81x

10−

3P

olic

y R

ate

E

0.37

6

1.91

65

3.45

7

4.99

76

6.53

81x

10−

3P

olic

y R

ate

C

0

0.01

11

0.02

23

0.03

34

0.04

46R

eal E

xcha

nge

Rat

e

0.01

02

0.01

77

0.02

52

0.03

28

0.04

03H

ouse

hold

s N

FA

0.00

98

0.01

72

0.02

47

0.03

22

0.03

97B

anks

Deb

t E

−0.

0374

−0.

0276

−0.

0179

−0.

0082

0.00

16B

anks

Tot

al A

sset

s C

0.15

43

0.22

34

0.29

26

0.36

17

0.43

09Le

vera

ge E

0.18

49

0.25

05

0.31

62

0.38

18

0.44

74Le

vera

ge C

−1

−0.

50

0.51

spre

ad E

−1

−0.

50

0.51

spre

ad C

−0.

0537

−0.

0395

−0.

0253

−0.

0111

0.00

31A

sset

Pric

e E

−0.

0537

−0.

0395

−0.

0253

−0.

0111

0.00

31A

sset

Pric

e C

48

1216

2024

28−

10.4

254

−7.

0405

−3.

6557

−0.

2708

3.11

41x

10−

3In

flatio

n E

48

1216

2024

28−

11.7

768

−8.

7029

−5.

6291

−2.

5552

0.51

86x

10−

3In

flatio

n C

48

1216

2024

28−

0.00

56

0.01

06

0.02

69

0.04

32

0.05

95∆

FX

48

1216

2024

28−

0.19

59

−0.

0753

0.04

53

0.16

59

0.28

65N

FA

E

Figure 5.1: Monetary Policy Shock in centre country: no financial frictions. Plain line=Flexible exchangerate; Crossed line=Peg.

20

EME, so that there is a general retrenchment in gross capital flows. The contrast with thecase without financial frictions is highlighted even more when we look at the comparisonof the quantitative effects on leverage, asset prices and spreads across the two countries.Even though the monetary tightening is precipitated by the shock in the centre country, theresponse of spreads, leverage and asset prices is greater in the EME. This is associated witha much greater fall in investment spending in the EME than in the centre country itself.

These results are consistent with the observation that emerging markets are highly sensi-tive to sudden reversals of capital flows, especially those associated with monetary tighteningin advanced economy markets. The international transmission in this model is critically tiedto the financial amplification mechanism coming from the linkage between bank’s net worthand their asset valuation. A monetary tightening reduces aggregate demand and invest-ment, which leads to a fall in the price of capital. This leads to a fall in bankers net worthin the centre country, amplifying the fall in investment. At the same time, the fall in centrecountry bank net worth leads to fall in capital flows to the EME, reducing investment andasset prices in the EME, generating a further fall in EME net worth. As a result interestrate spreads rise in both countries. In contrast to the case without financial frictions, we seethat monetary tightening in the centre country leads to a substantial and persistence fall inglobal bank lending to the EME.

We can again ask how the exchange rate regime affects the international transmission inthe case of financial frictions. Here the results are very different from the conventional DSGEmodel. With financial frictions and bank-balance sheet linkages, there is relatively littledifference between the baseline case and the EME monetary policy with pegged exchangerates (dashed line). The exchange rate peg does limit the EME real depreciation. Thismagnifies the fall in real GDP, since there is less compensating expenditure switching towardsEME goods. But the fall in capital inflows and outflows, the rise in leverage and spreads, thefall in asset prices, and the fall in EME investment is almost identical to that in the baselinecase with flexible exchange rates. Thus, these results tend to support the argument that inthe presence of financial frictions in capital flows, there is only a limited role for nominalexchange rate adjustment in insulating the economy from external shocks. We will see thiseven more clearly in the case of a financial shock in the analysis below.

How do the results depend on the denomination of bank lending? The baseline caseassumes that all borrowing is done in centre country currency (e.g. US dollars). Hence,the centre country monetary shocks precipitates an unanticipated depreciation in the EMEcurrency that has a direct negative impact on the EME bank’s net worth. This negativeeffect of ‘liability dollarization’ on balance sheets has been much discussed in the literatureon emerging market crises and exchange rate adjustment (Bruno and Shin, 2014). Figure 5.2illustrates the case where debt is denominated in domestic currency (dotted line). In thatalternative specification, an unanticipated centre country monetary shock still generates areal exchange rate depreciation for the EME country, but there is no direct negative valua-tion effect on the EME banks balance sheet. The impulse responses show that under localcurrency denomination of liabilities the transmission effect of the centre country monetarycontraction is lessened. There is a smaller spike in the EME spread relative to the baselinecase. EME leverage rises by less, and the asset price falls by less. Consequently the fall ininvestment and GDP is reduced by about 30% at their trough. But even without the directvaluation effect of the exchange rate change, the effect of the fall in centre country capital

21

−0.

1456

−0.

1098

−0.

0741

−0.

0383

−0.

0026

GD

P E

−0.

0741

−0.

0559

−0.

0378

−0.

0196

−0.

0015

GD

P C

−0.

8419

−0.

5968

−0.

3516

−0.

1065

0.13

86In

vest

men

t E

−0.

3144

−0.

2233

−0.

1321

−0.

041

0.05

02In

vest

men

t C

−4.

6739

−2.

0475

0.57

9

3.20

54

5.83

19x

10−

3P

olic

y R

ate

E

−0.

3327

1.20

85

2.74

96

4.29

07

5.83

19x

10−

3P

olic

y R

ate

C

0.00

22

0.01

84

0.03

46

0.05

09

0.06

71R

eal E

xcha

nge

Rat

e

−0.

1707

−0.

1302

−0.

0896

−0.

0491

−0.

0085

Hou

seho

lds

NF

A

−0.

0779

−0.

0583

−0.

0388

−0.

0192

0.00

04B

anks

Deb

t E

−0.

128

−0.

1006

−0.

0733

−0.

0459

−0.

0186

Ban

ks T

otal

Ass

ets

C

0.02

45

0.26

26

0.50

07

0.73

87

0.97

68Le

vera

ge E

−0.

0119

0.10

37

0.21

94

0.33

5

0.45

07Le

vera

ge C

0.00

01

0.01

29

0.02

57

0.03

84

0.05

12sp

read

E

−0.

0008

0.00

48

0.01

04

0.01

61

0.02

17sp

read

C

−0.

3109

−0.

226

−0.

1411

−0.

0563

0.02

86A

sset

Pric

e E

−0.

1367

−0.

0999

−0.

063

−0.

0262

0.01

07A

sset

Pric

e C

48

1216

2024

28−

19.5

074

−13

.795

6

−8.

0837

−2.

3718

3.34

x 10

−3

Infla

tion

E

48

1216

2024

28−

14.7

075

−10

.673

9

−6.

6403

−2.

6067

1.42

68x

10−

3In

flatio

n C

48

1216

2024

28−

0.00

74

0.01

4

0.03

53

0.05

66

0.07

79∆

FX

48

1216

2024

28−

1.07

48

−0.

8008

−0.

5268

−0.

2528

0.02

12N

FA

E

Figure 5.2: Monetary Policy Shock in centre country. Solid=baseline; dashed=peg; dots=local currencydebt.

22

flows still leads to a large balance sheet deterioration and a fall in real activity. Relative tothe case without financial frictions, there is still a large negative impact on EME investmentand real GDP.

These results would seem to underscore the message of Rey (2013) and others, suggest-ing that despite having flexible exchange rates, emerging market countries are extremelyvulnerable to volatile capital flows related to US monetary policy shocks. Under a conven-tional monetary policy rule, exchange rate adjustment can then only play a limited role inmitigating the impact of shocks, and it suggests the need for other direct forms of capitalrestrictions or macro-prudential policies that directly target the balance sheets of banks.

We should note of course that the monetary rule described above is an ad-hoc speci-fication. An optimal monetary policy response can be designed that will do much betterin response to the centre country monetary shock. In the case of an optimal cooperativemonetary rule, this statement becomes trivial, because then it is always optimal to directlyoffset the monetary shock itself, and the impact of the monetary shock is entirely eliminated.But a more interesting question arises when the EME must respond unilaterally. We explorethis response in the section on non-cooperative monetary policy and financial shocks below.

Finally, we note that the impact of shocks in our model is extremely asymmetric. Fig-ure 5.3 reports the effect on both the EME and the centre country of a monetary policycontraction of similar magnitude to that of Figure 5.2 but now coming from the EME. Theimpact on the centre country real activity is negligible. This is to be expected, since theEME is small relative to the world economy. There is a fall in GDP in the EME, since themonetary contraction leads to an immediate real exchange rate appreciation. But remark-ably, we find that the contraction in real activity in the EME is now smaller than in theresponse to the centre country shock. The critical feature is that the monetary shock in theEME does not directly impact on the EME bank’s balance sheet. In fact, there is a smallboost to the bank’s net worth, coming from the unanticipated real appreciation. But theeffects on spreads, leverage and asset prices is small, and as a consequence, investment fallsby substantially less than in response to an external monetary tightening.

6. Financial Shocks

In this section we discuss the spillover effects of financial shocks originating in the centrecountry.14 Figure 6.1 shows impulse responses for a 1% increase in the incentive compat-ibility constraint parameter κct . The first noticeable effect of this shock is the relativelystrong comovement across countries. As discussed by Devereux and Yetman (2010) andDedola and Lombardo (2012), financial shocks in one single economy, in a world charac-terized by financial integration and financial frictions, can generate highly synchronized

14The 2008-2009 financial crisis has motivated considerable research on the role of credit shocks. Jermannand Quadrini (2012) show, in a model with financial constraints, that financial shocks can explain the 2008-2009 US recession as well as other previous episodes. Helbling et al. (2011) provide empirical evidence onthe role of financial shocks in driving global recessions. Boivin et al. (2013) shed light on the macroeconomicconsequences of financial shocks for the US economy using a large set of macro and financial variables.Christiano et al. (2014) estimate a DSGE model with financial frictions a la Bernanke et al. (1999) and showthat financial shocks (the idiosyncratic shock to financially constrained borrowers) are the most importantshock driving the business cycle.

23

−0.

0728

−0.

0547

−0.

0366

−0.

0185

−0.

0004

GD

P E

−0.

101

1.43

42

2.96

95

4.50

47

6.04

x 10

−3

GD

P C

−0.

3408

−0.

2398

−0.

1389

−0.

0379

0.06

31In

vest

men

t E

−0.

0108

0.00

28

0.01

64

0.03

0.04

36In

vest

men

t C

−0.

1592

1.33

35

2.82

63

4.31

9

5.81

18x

10−

3P

olic

y R

ate

E

−2.

149

−0.

5913

0.96

63

2.52

39

4.08

16x

10−

4P

olic

y R

ate

C

−0.

0448

−0.

0322

−0.

0196

−0.

007

0.00

56R

eal E

xcha

nge

Rat

e

−0.

123

−0.

0923

−0.

0616

−0.

0309

−0.

0002

Hou

seho

lds

NF

A

−0.

032

−0.

0214

−0.

0109

−0.

0003

0.01

03B

anks

Deb

t E

−1.

7591

2.04

58

5.85

08

9.65

57

13.4

606

x 10

−3 B

anks

Tot

al A

sset

s C

0.00

04

0.11

69

0.23

35

0.35

0.46

65Le

vera

ge E

−0.

0555

−0.

0412

−0.

0269

−0.

0126

0.00

17Le

vera

ge C

−0.

0003

0.00

51

0.01

05

0.01

58

0.02

12sp

read

E

−9.

5418

−6.

2936

−3.

0454

0.20

28

3.45

1x

10−

4sp

read

C

−0.

1401

−0.

1022

−0.

0644

−0.

0265

0.01

13A

sset

Pric

e E

−2.

127

0.05

16

2.23

02

4.40

88

6.58

75x

10−

3A

sset

Pric

e C

48

1216

2024

28−

14.7

904

−10

.758

8

−6.

7272

−2.

6955

1.33

61x

10−

3In

flatio

n E

48

1216

2024

28−

2.73

97

2.34

01

7.42

12.4

998

17.5

797

x 10

−4

Infla

tion

C

48

1216

2024

28−

0.04

99

−0.

036

−0.

0222

−0.

0083

0.00

55∆

FX

48

1216

2024

28−

0.42

36

−0.

2684

−0.

1132

0.04

2

0.19

72N

FA

E

Figure 5.3: Monetary Policy Shock in EME. Solid=baseline; dashed=peg; dots=local currency debt.

24

responses across countries. As in the case of the response to centre country monetary policyshocks, we see that impulse responses are essentially invariant to the exchange rate regimeor the currency denomination of liabilities. The synchronization of credit spreads (measuredas wedges between the return on capital and the domestic policy rate) is the dominant factorin generating business cycle movements. Again, the contraction in the emerging economy ismarkedly larger than that in the centre country (the epicentre of the shock), mainly due tothe asymmetric size of the two economic regions.15 In terms of capital flows the consequenceof the centre-country financial shock is a “retrenchment” of international capital. As before,gross inflows into EME banks fall as do EME outflows from households. This adjustment isreminiscent of the capital flows observed during the 2008-2009 financial crisis (e.g. Broneret al., 2013).

7. Optimal cooperative monetary policy

So far we have documented that financial market integration can generate disproportionaleffects on EME of shocks originating in the centre country, quite independently of the ex-change rate regime. These results, therefore, seem to provide theoretical support to the ideaof the changed nature of the monetary policy problem in a world characterized by financialmarket integration: even under flexible exchange rates, free capital mobility is incompatiblewith an independent domestic monetary policy (Rey, 2013).16 Nevertheless, in this sectionwe show that a more appropriate description of the effects of financial integration on the pol-icy problem is that capital flows exacerbate the policy trade-offs and, thus, require differentpolicy strategies. It is well known that openness, in general, affects the optimal monetarypolicy response to shocks (e.g. Corsetti et al., 2010, Faia and Monacelli, 2008, Devereux andSutherland, 2007, Devereux and Engel, 2003, Lombardo and Ravenna, 2014 and Kolasa andLombardo, 2014). In particular a monetary policy strategy that seems appropriate under aparticular mix of shocks ceases to be attractive under a different mix of shocks. Financialintegration not only changes the trade-offs faced by central banks, it also changes the typeof shocks that the economy is likely to experience. To illustrate this point, we study theresponse of the two economies to financial shocks under the Ramsey cooperative optimalpolicy.17.

The optimal cooperative policy solves the following problem

maxYt

E0

∞∑i=0

βiCB(nU

(Cet+i, H

et+i

)+ (1− n)U

(Cct+i, H

ct+i

))(7.0.1)

subject to all the equations characterizing the decentralized equilibrium (i.e. FOC of theprivate agents and resource constraints, see Appendix A for details), where βCB is thediscount factor of the central bank (which we take to be identical to the discount factor of

15Note that the “financial wedges” move almost identically. The double layer of financial frictions defacto faced by the EME bank does not generate larger spreads than in the centre country.

16Georgiadis and Mehl (2015) provide some GVAR-based evidence that financial integration might actu-ally have increased monetary policy effectiveness, due to valuation effects.

17Fujiwara et al. (2015) provides theoretical support to the idea that cooperation in the presence offinancial frictions is welfare improving.

25

−6.

0725

−4.

5159

−2.

9593

−1.

4026

0.15

4x

10−

3G

DP

E

−18

.420

1

−11

.865

1

−5.

31

1.24

5

7.80

01x

10−

4G

DP

C

−0.

0395

−0.

0274

−0.

0153

−0.

0031

0.00

9In

vest

men

t E

−11

.025

1

−6.

545

−2.

065

2.41

51

6.89

52x

10−

3In

vest

men

t C

−2.

4711

−1.

3732

−0.

2753

0.82

26

1.92

05x

10−

4P

olic

y R

ate

E

−7.

544

−3.

0783

1.38

75

5.85

33

10.3

19x

10−

5P

olic

y R

ate

C

0.51

81

0.89

54

1.27

26

1.64

99

2.02

72x

10−

3 Rea

l Exc

hang

e R

ate

−16

.172

7

−12

.415

8

−8.

6589

−4.

9021

−1.

1452

x 10

−3

Hou

seho

lds

NF

A

−3.

4938

−2.

3867

−1.

2796

−0.

1724

0.93

47x

10−

3B

anks

Deb

t E

−5.

9954

−4.

4412

−2.

8871

−1.

3329

0.22

12x

10−

3 Ban

ks T

otal

Ass

ets

C

0.00

16

0.01

16

0.02

16

0.03

15

0.04

15Le

vera

ge E

−0.

0119

−0.

0044

0.00

31

0.01

06

0.01

81Le

vera

ge C

−0.

03

0.59

5

1.22

1.84

5

2.47

x 10

−3

spre

ad E

−0.

1248

0.44

6

1.01

68

1.58

76

2.15

84x

10−

3sp

read

C

−13

.668

2

−9.

8888

−6.

1094

−2.

3299

1.44

95x

10−

3A

sset

Pric

e E

−6.

6961

−4.

8283

−2.

9604

−1.

0926

0.77

53x

10−

3A

sset

Pric

e C

48

1216

2024

28−

5.91

68

−3.

9106

−1.

9043

0.10

19

2.10

82x

10−

4In

flatio

n E

48

1216

2024

28−

1.25

25

−0.

681

−0.

1095

0.46

2

1.03

35x

10−

4In

flatio

n C

48

1216

2024

28−

2.01

11

2.36

68

6.74

47

11.1

227

15.5

006

x 10

−4

∆ F

X

48

1216

2024

28−

0.08

74

−0.

0666

−0.

0457

−0.

0248

−0.

0039

NF

A E

Figure 6.1: Financial shock in centre country. Solid=baseline; dashed=peg; dots=local currency debt.

26

the households) and Yt is the vector containing all the endogenous variables of the model. Insolving for the optimal policy we follow the “timeless” perspective advocated by Woodford(2003).

Figure 7.1 compares the baseline case with the optimal cooperative policy. The optimalpolicy reduces considerably the effect of financial shocks on both economies. In particularthe EME spread increases only modestly and less than the spread in the centre country.This is reflected in a considerably smaller asset-price decline and, thus, in a smaller fall ininvestment. Leverage and spreads co-move much less across countries than under the Taylorrule. In order to achieve this allocation, the Ramsey policy-maker needs to depart from theinterest rate adjustment observed under the Taylor rule. In particular nominal interest ratesfall markedly in both countries and inflation is allowed to increase, albeit only temporarily.The real exchange rate in the EME appreciates on impact, providing extra relief to thebalance sheet of EME banks. The optimal policy strongly mitigates capital “retrenchment”,thus preventing the strong credit contraction that the Taylor rule brings about.

The implication of these results is that domestic monetary policy matters, in a non-trivial way, even under financial integration and financial frictions in international capitalflows. Nevertheless, it also shows that financial integration, and the spillovers of foreignfinancial shocks that come with openness, do exacerbate the trade-offs faced by centralbanks. The objective of inflation stabilization cannot be achieved to the same extent as in africtionless international capital market. Stabilization of financial market variables becomesan important objective of policy, suggesting room for macro-prudential interventions (e.g.Farhi and Werning, 2013).

8. Non-cooperative monetary policy

The previous sections showed that while a naive Taylor rule has little advantage overan exchange rate peg, an optimal cooperative discretionary monetary policy could play aneffective role in dealing with spillovers in international financial markets. But an obviousobjection to this is that cooperative monetary policy is an unrealistic ideal. Practicallyspeaking, monetary policy is set at the national level based on domestic objectives. Howwould our results differ if we allow for optimal monetary policy, but recognizing that policyis set by each country separately? In this section we analyse the effects of non-cooperative,self-oriented monetary policies. In particular we solve our model for the open-loop, Nashequilibrium and compare it with the baseline Taylor-type rule and the globally optimalRamsey policy.

As is well known, there is no single non-cooperative optimal policy, as the solution de-pends on the strategy space underlying the game. Also, defining the strategy space in termsof policy rates does not yield a saddle-path equilibrium. Alternatively one could choose tosolve for the non-cooperative equilibrium in terms of explicit instrument rules (closed-loopNash equilibrium). Nevertheless, in this case too there is some degree of arbitrariness bothin the choice of the instrument and in the choice of the feedback variables and lags. Mindfulof these issues, we proceed by providing results for a baseline case where the strategy spaceis defined in terms of PPI inflation (as for example in Benigno and Benigno, 2006).

27

−5.

0123

−3.

676

−2.

3396

−1.

0033

0.33

3x

10−

3G

DP

E

−1.

6954

−0.

61

0.47

55

1.56

1

2.64

65x

10−

3G

DP

C

−0.

0362

−0.

0251

−0.

014

−0.

0029

0.00

83In

vest

men

t E

−9.

7296

−5.

5801

−1.

4307

2.71

88

6.86

83x

10−

3In

vest

men

t C

−3.

969

−2.

4966

−1.

0242

0.44

81

1.92

05x

10−

4P

olic

y R

ate

E

−3.

3839

−2.

5123

−1.

6407

−0.

7692

0.10

24x

10−

3P

olic

y R

ate

C

−2.

0851

−1.

057

−0.

0289

0.99

91

2.02

72x

10−

3 Rea

l Exc

hang

e R

ate

−10

.626

4

−8.

0059

−5.

3854

−2.

7649

−0.

1444

x 10

−3

Hou

seho

lds

NF

A

−3.

4588

−2.

5876

−1.

7165

−0.

8453

0.02

59x

10−

3B

anks

Deb

t E

−5.

422

−3.

8506

−2.

2792

−0.

7077

0.86

37x

10−

3 Ban

ks T

otal

Ass

ets

C

−0.

0078

0.00

41

0.01

6

0.02

8

0.03

99Le

vera

ge E

−0.

0126

−0.

0054

0.00

18

0.00

9

0.01

61Le

vera

ge C

−0.

1539

0.49

01

1.13

41

1.77

81

2.42

21x

10−

3sp

read

E

−1.

2329

4.36

31

9.95

92

15.5

552

21.1

513

x 10

−4

spre

ad C

−11

.975

4

−8.

6859

−5.

3964

−2.

1069

1.18

26x

10−

3A

sset

Pric

e E

−6.

3884

−4.

6002

−2.

8121

−1.

0239

0.76

42x

10−

3A

sset

Pric

e C

48

1216

2024

28−

3.24

49

−1.

9488

−0.

6527

0.64

34

1.93

95x

10−

4In

flatio

n E

48

1216

2024

28−

1.12

49

−0.

3778

0.36

94

1.11

65

1.86

37x

10−

4In

flatio

n C

48

1216

2024

28−

2.39

76

−1.

0515

0.29

47

1.64

08

2.98

7x

10−

3∆

FX

48

1216

2024

28−

0.08

6

−0.

0623

−0.

0386

−0.

0149

0.00

87N

FA

E

Figure 7.1: Financial shock in centre country. Solid=baseline; dashed=Ramsey

28

8.1. The non-cooperative policy problem

Each central bank chooses the allocation that maximizes the intertemporal welfare of thehouseholds living in its jurisdiction, i.e.

maxYet

E0

∞∑i=0

βiCBU(Cet+i, H

et+i

)(8.1.1)

subject to all the equations characterizing the decentralized equilibrium (i.e. FOC of theprivate agents and resource constraints, see Appendix for details), where Yet is the vectorcontaining all the endogenous variables of the model except the foreign variable that co-defines the strategy space.18 A similar problem is solved by the center-country central bank.

The set of first order conditions of the two monetary authorities jointly defines the set ofbest-responses and, thus, the implicit policy rules followed by the central banks. As in theoptimal cooperative monetary policy case, we follow the “timeless” perspective advocatedby Woodford (2003).

9. Comparison

Figure 9.1 compares the (cooperative) Ramsey policy with the Taylor-type rule used inour baseline specification as well as with the non-cooperative policy described above, underthe assumption that the strategy space is defined in terms of domestic PPI inflation rates.

Under this strategy the non-cooperative policy generates dynamics in the main macrovariables that are essentially identical to the fully cooperative one.

The lessons that we can draw from this results are the following. First, monetary policy iscrucial in determining the response of the economy to shocks (compare the arbitrary Taylor-type rule with the two optimal rules). Second, an appropriately chosen monetary policycan considerably mitigate the spillovers from foreign shocks. Third, a self-oriented monetarypolicy response can achieve the same outcome of the fully cooperative policy.

So, while financial frictions and financial globalization could generate scope for coopera-tion in other policy dimensions (e.g. regulation), our analysis suggests that, to first order,the response of the economies under the two regimes (cooperation and non-cooperation)are virtually identical. The intuition behind this similarity is the following. Self-orientedstabilization requires smoothing of credit-spread volatility. This policy is beneficial for allcountries. The spillover from foreign stabilization, while positive, is not strong enough togenerate free-riding incentives (e.g. see Dedola et al., 2013), since, in particular, financialstructures are not symmetric. The classic negative spillovers due to price rigidities are alsosufficiently small (e.g. Obstfeld and Rogoff, 2002). The net result entails negligible gainsfrom cooperation.

As a caveat, it should be noted that we are not considering exceptionally large shocksleading to a liquidity-trap equilibrium. Looking at the coordination gains of unconventional

18Thus, if the PPI inflation rates πppi,et and πppi,ct define the strategy space, Yet excludes πppi,ct and Yctexcludes πppi,et . The policymakers recognize that the counterpart can freely choose one variable: it has onedegree of freedom.

29

policies when policy rates hit the zero lower bound is beyond the scope of the presentanalysis.19

10. Conclusions

The post crisis years have exposed substantial strains in the workings of the internationalfinancial system. Excessive volatility in international capital flows have raised questionsabout the efficacy of self-oriented monetary policies and the benefits of flexible exchangerates under inflation targeting.

We offer some further empirical evidence on the international transmission of US mon-etary policy shocks to emerging market economies. We do so by applying local projectionmethods to a large panel of countries and conditioning on monetary policy shocks identifiedusing standard techniques. We find that US monetary policy shocks tends to generate acontraction in emerging markets, as well as a fall in both in- and out-flows of capital: aglobal retrenchment.

We then develop a simple template which allows both for an understanding of the sourcesof excess volatility of capital flows to emerging markets, as well as an evaluation of policyresponses to capital flows. We show that, like in the data, under financial frictions a core-country monetary policy shock generates a contraction in the emerging economy, alongsidea global capital retrenchment. We show that financial frictions play a crucial role both inour positive as well as in our normative analysis. Absent financial frictions, and with flexibleexchange rates, the spillovers to the emerging economy are minimal.

Our results indicate that the simple prescriptions about the benefits of flexible exchangerates and inflation targeting are very unlikely to hold in a global financial environment domi-nated by the currency and policy of a large financial centre, such as the current situation withthe US dollar and US monetary policy. Our preliminary analysis does suggest however thatan optimal monetary policy can substantially improve the workings of the international sys-tem, even in the absence of direct intervention in capital markets through macro-prudentialpolicies or capital controls. Moreover, under the specific assumptions maintained in thispaper, this outcome can still be consistent with national independence in policy, or in otherwords, a system of ‘self-oriented’ monetary policy making.

19See Caballero et al. (2015) for a recent analysis of monetary policy spillovers at the ZLB.

30

48

1216

2024

28−

5.01

23

−3.

6614

−2.

3104

−0.

9595

0.39

14x

10−

3G

DP

E

48

1216

2024

28−

1.69

54

−0.

5701

0.55

52

1.68

05

2.80

59x

10−

3G

DP

C

48

1216

2024

28−

0.03

62

−0.

0251

−0.

014

−0.

0029

0.00

83In

vest

men

t E

48

1216

2024

28−

9.72

96

−5.

5801

−1.

4307

2.71

88

6.86

83x

10−

3In

vest

men

t C

48

1216

2024

28−

3.96

9

−2.

4966

−1.

0242

0.44

81

1.92

05x

10−

4P

olic

y R

ate

E

48

1216

2024

28−

3.54

11

−2.

6302

−1.

7193

−0.

8085

0.10

24x

10−

3P

olic

y R

ate

C

48

1216

2024

28−

2.11

9

−1.

0825

−0.

0459

0.99

06

2.02

72x

10−

3 Rea

l Exc

hang

e R

ate

48

1216

2024

28−

10.6

264

−7.

9964

−5.

3663

−2.

7363

−0.

1062

x 10

−3

Hou

seho

lds

NF

A

48

1216

2024

28−

3.45

88

−2.

5615

−1.

6642

−0.

7669

0.13

03x

10−

3B

anks

Deb

t E

48

1216

2024

28−

5.42

2

−3.

8414

−2.

2608

−0.

6801

0.90

05x

10−

3 Ban

ks T

otal

Ass

ets

C

48

1216

2024

28−

0.00

99

0.00

25

0.01

5

0.02

74

0.03

99Le

vera

ge E

48

1216

2024

28−

0.01

27

−0.

0055

0.00

17

0.00

89

0.01

61Le

vera

ge C

48

1216

2024

28−

0.25

21

0.41

64

1.08

5

1.75

35

2.42

21x

10−

3sp

read

E

48

1216

2024

28−

1.23

29

4.36

31

9.95

92

15.5

552

21.1

513

x 10

−4

spre

ad C

48

1216

2024

28−

11.9

754

−8.

6859

−5.

3964

−2.

1069

1.18

26x

10−

3A

sset

Pric

e E

48

1216

2024

28−

6.38

84

−4.

6002

−2.

8121

−1.

0239

0.76

42x

10−

3A

sset

Pric

e C

48

1216

2024

28−

3.24

49

−1.

9488

−0.

6527

0.64

34

1.93

95x

10−

4In

flatio

n E

48

1216

2024

28−

1.20

11

−0.

4176

0.36

6

1.14

95

1.93

3x

10−

4In

flatio

n C

48

1216

2024

28−

2.43

52

−1.

028

0.37

92

1.78

64

3.19

36x

10−

3∆

FX

48

1216

2024

28−

0.08

6

−0.

0622

−0.

0385

−0.

0147

0.00

91N

FA

E

Figure 9.1: Policy responses to a financial shock in centre country; dot-dashed=Taylor, dashed=Nash,solid=Ramsey.

31

References

Agenor, P.-R., Alper, K., Pereira da Silva, L. A., 2014. Sudden floods, macroprudentialregulation and stability in an open economy. Journal of International Money and Finance48, 68–100.

Ahmed, S., Zlate, A., 2014. Capital flows to emerging market economies: a brave new world?Journal of International Money and Finance 48, 221–248.

Alberola, E., Erce, A., Serena, J. M., 2012. International reserves and gross capital flows:dynamics during financial stress. Globalization and Monetary Policy Institute WorkingPaper 110, Federal Reserve Bank of Dallas.

Bacchetta, P., Benhima, K., Kalantzis, Y., 2013. Capital controls with international reserveaccumulation: Can this be optimal? American Economic Journal: Macroeconomics 5 (3),229–62.

Benigno, G., Benigno, P., 2006. Designing Targeting Rules for International Monetary PolicyCooperation. Journal of Monetary Economics 53, 473–506.

Bernanke, B. S., Gertler, M., Gilchrist, S., March 1999. The Financial Accelerator in a Quan-titative Business Cycle Framework. In: Taylor, J. B., Woodford, M. (Eds.), Handbook ofMacroeconomics. Elsevier.

Bluwstein, K., Canova, F., et al., 2015. Beggar-thy-neighbor? the international effects of ecbunconventional monetary policy measures. Tech. rep., CEPR Discussion Papers.

Boivin, J., Giannoni, M. P., Stevanovi, D., 2013. Dynamic effects of credit shocks in adata-rich environment. FRB of New York Staff Report (615).

Bowman, D., Londono, J. M., Sapriza, H., 2015. Us unconventional monetary policy andtransmission to emerging market economies. Journal of International Money and Finance55, 27–59.

Broner, F., Didier, T., Erce, A., Schmukler, S. L., 2013. Gross capital flows: Dynamics andcrises. Journal of Monetary Economics 60 (1), 113–133.

Bruno, V., Shin, H. S., 2014. Cross-border banking and global liquidity. The Review ofEconomic Studies.

Bruno, V., Shin, H. S., 2015. Capital flows and the risk-taking channel of monetary policy.Journal of Monetary Economics 71, 119–132.

Caballero, R. J., Farhi, E., Gourinchas, P.-O., October 2015. Global imbalances and currencywars at the zlb. Working Paper 21670, National Bureau of Economic Research.

Chen, J., Mancini-Griffoli, T., Sahay, R., Dec. 2014. Spillovers from United States Mone-tary Policy on Emerging Markets: Different This Time? IMF Working Papers 14/240,International Monetary Fund.

32

Chen, Q., Filardo, A., He, D., Zhu, F., Mar. 2015. Financial crisis, US unconventional mon-etary policy and international spillovers. BIS Working Papers 494, Bank for InternationalSettlements.

Choi, W. G., Cook, D., 2004. Liability dollarization and the bank balance sheet channel.Journal of International Economics 64 (2), 247–275.

Christiano, L. J., Motto, R., Rostagno, M., 2014. Risk shocks. American Economic Review1 (104), 27–65.

Coibion, O., Gorodnichenko, Y., Kueng, L., Silvia, J., Jun. 2012. Innocent Bystanders?Monetary Policy and Inequality in the U.S. NBER Working Papers 18170, National Bureauof Economic Research, Inc.

Corsetti, G., Dedola, L., Leduc, S., October 2010. Optimal monetary policy in openeconomies. In: Friedman, B. M., Woodford, M. (Eds.), Handbook of Monetary Economics.Vol. 3 of Handbook of Monetary Economics. Elsevier, Ch. 16, pp. 861–933.

Dedola, L., Karadi, P., Lombardo, G., 2013. Global implications of national unconventionalpolicies. Journal of Monetary Economics 60, 66–85.

Dedola, L., Lombardo, G., 2012. Financial Integration, Financial Frictions, and the Interna-tional Propagation of Shocks. Economic Policy 27, 319–359.

Dedola, L., Rivolta, G., Stracca, L., June 2015. If the Fed sneezes, who gets a cold? Mimeo.

Devereux, M., Sutherland, A., 2007. Monetary Policy and Portfolio Choice Choice in anOpen Economy Macro Model. Journal of the European Economic Association 5 (2-3),491–499.

Devereux, M., Yetman, J., 2010. Leverage constraints and the international transmission ofshocks. Journal of Money, Credit and Banking 42, 71–105.

Devereux, M. B., Engel, C., 2003. Monetary Policy in the Open Economy Revisited: PriceSetting and Exchange-Rate Flexibility. Review of Economic Studies 70 (4), 765–783.

Faia, E., Monacelli, T., 2008. Optimal Monetary Policy in a Small Open Economy withHome Bias. Journal of Money, Credit and Banking 40 (4), 721–750.

Farhi, E., Werning, I., 2013. A theory of macroprudential policies in the presence of nominalrigidities. NBER Working Paper (w19313).

Fischer, S., 2014. The federal reserve and the global economy. Speech given at the PerJacobsson Foundation Lecture, 2014 Annual Meetings of the International Monetary Fundand the World Bank Group, Washington, D.C.URL http://www.federalreserve.gov/newsevents/speech/fischer20141011a.htm

Fratzscher, M., Lo Duca, M., Straub, R., 2014. ECB unconventional monetary policy actions:Market impact, international spillovers and transmission channels. International MonetaryFund.

33

Fujiwara, I., Teranishi, Y., et al., 2015. Financal frictions and policy cooperation: a casewith monopolistic banking and staggered loan contracts. Tech. rep., Federal Reserve Bankof Dallas.

Georgiadis, G., Mehl, A., Jan. 2015. Trilemma, not dilemma: financial globalisation andMonetary policy effectiveness. Globalization and Monetary Policy Institute Working Paper222, Federal Reserve Bank of Dallas.

Gertler, M., 2012. Comment. NBER Macroeconomics Annual.

Gertler, M., Karadi, P., 2011. A model of unconventional monetary policy. Journal of mon-etary Economics 58 (1), 17–34.

Gilchrist, S., Yue, V., Zakrajsek, E., 2014. The response of sovereign bond yields to USmonetary policy. Tech. rep., Mimeo.

Helbling, T., Huidrom, R., Kose, M. A., Otrok, C., 2011. Do credit shocks matter? a globalperspective. European Economic Review 55 (3), 340–353.

Jermann, U., Quadrini, V., February 2012. Macroeconomic Effects of Financial Shocks.American Economic Review 102 (1), 238–71.

Jorda, O., March 2005. Estimation and Inference of Impulse Responses by Local Projections.American Economic Review 95 (1), 161–182.

Kolasa, M., Lombardo, G., 2014. Financial Frictions and Optimal Monetary Policy in anOpen Economy. International Journal of Central Banking 10 (1), 43–94.

Kollmann, R., Enders, Z., Muller, G., 2011. Global banking and international business cycles.European Economic Review 55 (3), 407–426.

Korinek, A., 2014. International Spillovers and Guidelines for Policy Cooperation: A WelfareTheorem for National Economic Policymaking. Mimeo.

Lombardo, G., Ravenna, F., 2014. Openness and optimal monetary policy. Journal of Inter-national Economics 93 (1), 153–172.

Miranda-Agrippino, S., Rey, H., 2014. World asset markets and the global financial cycle.London Business School, mimeo.

Nuguer, V., 2014. Financial intermediation in a global environment. Banco de Mexico, Work-ing Paper No 2015-05.

Obstfeld, M., Rogoff, K., 2002. Global Implications of Self-Oriented National MonetaryPolicy Rules. Quarterly Journal of Economics 117, 503–535.

Perri, F., Quadrini, V., Jul. 2011. International recessions. NBER Working Papers 17201,National Bureau of Economic Research, Inc.

34

Rajan, R. G., 2014. Competitive monetary easing: Is it yesterday once more? Remarks givenat the Brookings Institution.URL http : //www.rbi.org.in/scripts/BS SpeechesV iew.aspx? id = 886

Rey, H., 2013. Dilemma not trilemma: the global financial cycle and monetary policy inde-pendence. In: Jackson Hole Economic Symposium.

Romer, C. D., Romer, D. H., 2004. A new measure of monetary shocks: Derivation andimplications. American Economic Review 94 (4), 1055–1084.

Ueda, K., 2012. Banking globalization and international business cycles: Cross-borderchained credit contracts and financial accelerators. Journal of International Economics86 (1), 1–16.

Woodford, M., 2003. Interest and Prices: Foundations of a Theory of Monetary Policy.Princeton U.P., Princeton, NJ.

35

Appendix A. Full list of equations used in the simulation

In this appendix we list all the equations of the model used in the derivation of the numerical results.All variables preceded by exp are in expressed in logs (i.e. we take a log-approximation). A ss subscript

indicates the value of that variable in the non-stochastic steady state. For simplicity we omit the expectationoperator, with the understanding that variables dated t + 1 are unknown at time t, so that the relatedexpression implicitly involves the expectation operator

Objective of the policymaker(s):

(A.1)U cbt = uu hn

(exp (Cet )

1−σ

1− σ− χ exp (He

t )1+ψ

1 + ψ

)+ uu f (1− n)

(exp (Cct )

1−σ

1− σ− χ exp (Hc

t )1+ψ

1 + ψ

)

where uu h and uu f are {0, 1} weights defining the policy problem (both equal 1 for cooperation and eitherequal to zero for Nash).

Price dispersion measures:

(A.2)exp (Det ) = ς exp

(Det−1

)exp

(πppi,et − χπ πppi,e,ss

)σp

+ (1− ς) exp(π∗,et

)(−σp)

(A.3)exp (Dct ) = ς exp

(Dct−1

)exp

(πppi,ct − χπ πppi,c,ss

)σp

+ (1− ς) exp(π∗,ct

)(−σp)

Optimal price set by price-changing firms:

(A.4)exp(Π∗,et

)=exp (F et )

exp (Get )

(A.5)exp(Π∗,ct

)=exp (F ct )

exp (Gct)

where(A.6)exp (F et ) = exp (Y et ) exp (MCet ) + ς exp

(Λet+1

)exp

(πppi,et+1 − χπ πppi,e,ss

)σp

exp(F et+1

)(A.7)exp (F ct ) = exp (Y ct ) exp (MCct ) + ς exp

(Λct+1

)exp

(πppi,ct+1 − χπ πppi,c,ss

)σp

exp(F ct+1

)

(A.8)exp (Get ) =exp (Y et ) exp (P et )

σp

σp−1

+ ς exp(Λet+1

)exp

(πppi,et+1 − χπ πppi,e,ss

)σp−1

exp(Get+1

)

(A.9)exp (Gct) =exp (Y ct ) exp (P ct )

σp

σp−1

+ ς exp(Λct+1

)exp

(πppi,ct+1 − χπ πppi,c,ss

)σp−1

exp(Gct+1

)where MC is the marginal cost.

Relative PPI price dynamics:

(A.10)1 = ς exp(πppi,et − χπ πppi,e,ss

)σp−1

+ (1− ς) exp(Π∗,et

)1−σp

(A.11)1 = ς exp(πppi,ct − χπ πppi,c,ss

)σp−1

+ (1− ς) exp(Π∗,ct

)1−σp

Euler equation for foreign bonds:

(A.12)

exp(Λet+1) exp(R

∗t )

exp(πct+1)

exp (RERt+1)

exp (RERt)− 1 = 0

36

Centre country consumption Euler equation:

(A.13)exp

(Λct+1

)exp (R∗

t )

exp(πct+1

) − 1 = 0

Gross return on capital:

(A.14)exp(Rk,et

)exp

(Qet−1

)=(exp (MCet ) exp (Aet ) α exp (He

t )1−α

exp(Ket−1

)α−1+ (1− δ) exp (Qet )

)

(A.15)exp(Rk,ct

)exp

(Qct−1

)=(α exp (MCct ) exp (Act) exp (Hc

t )1−α

exp(Kct−1

)α−1+ (1− δ) exp (Qct)

)Aggregate price indexes (in units of consumption aggregator):

(A.16)1 = νe exp (P et )1−ηp + (1− νe) (exp (P ct ) exp (RERt))

1−ηp

(A.17)1 = νc exp (P ct )1−ηp + (1− νc)

(exp (P et )

exp (RERt)

)1−ηp

CPI inflation measures:

(A.18)exp (P et ) exp (πet ) = exp(P et−1

)exp

(πppi,et

)

(A.19)exp (P ct ) exp (πct ) = exp(P ct−1

)exp

(πppi,ct

)Labor market clearing condition:

(A.20)exp (Aet ) exp (MCet ) (1− α) exp (Het )

(−α)exp

(Ket−1

)αexp (Cet )

(−σ)= χ exp (He

t )ψ

(A.21)exp (Act) exp (MCct ) (1− α) exp (Hct )

(−α)exp

(Kct−1

)αexp (Cct )

(−σ)= χ exp (Hc

t )ψ

Accumulation law for capital:

(A.22)exp (Ket ) =

(exp (Iet ) + exp

(Ket−1

)(1− δ)

)(A.23)exp (Kc

t ) =(exp (Ict ) + (1− δ) exp

(Kct−1

))Investment Euler equations:

(A.24)

exp (Qet ) = 1 +η

2

(exp (Iet )

exp(Iet−1

) − 1

)2

+

(exp (Iet )

exp(Iet−1

) − 1

)exp (Iet ) η

exp(Iet−1

)− exp

(Λet+1

)η

(exp

(Iet+1

)exp (Iet )

)2 (exp

(Iet+1

)exp (Iet )

− 1

)

(A.25)

exp (Qct) = 1 +η

2

(exp (Ict )

exp(Ict−1

) − 1

)2

+

(exp (Ict )

exp(Ict−1

) − 1

)exp (Ict ) η

exp(Ict−1

)− exp

(Λct+1

)η

(exp

(Ict+1

)exp (Ict )

)2 (exp

(Ict+1

)exp (Ict )

− 1

)

37

Aggregate budget constraint for EME households:

(A.26)

exp (Cet ) +Bet exp (RERt) = exp (Y et ) exp (P et )− exp(Ket−1

)αexp (He

t )1−α

exp (MCet ) exp (Aet ) α

+ exp (Qet ) exp (Iet )− exp (Iet )

1 +η

2

(exp (Iet )

exp(Iet−1

) − 1

)2

+ exp(Ket−1

)exp

(Qet−1

)exp

(Rk,et

)(1− θ) (1− Γt)

− exp(Ket−1

)exp (Qet ) δT + exp (RERt)

exp(R∗t−1

)exp (πct )

Bet−1

Goods market clearing conditions:

(A.27)exp(Ket−1

)αexp (Aet ) exp (He

t )1−α

= exp (Det ) exp (Y et )

(A.28)exp(Kct−1

)αexp (Act) exp (Hc

t )1−α

= exp (Dct ) exp (Y ct )

Aggregate demand for domestic and foreign goods:

(A.29)

exp (Y et ) = νe exp (P et )(−ηp)

exp (Cet ) + exp (Iet )

1 +η

2

(exp (Iet )

exp(Iet−1

) − 1

)2+ (1

− νc) 1− nn

(exp (P et )

exp (RERt)

)(−ηp)exp (Cct ) + exp (Ict )

1 +η

2

(exp (Ict )

exp(Ict−1

) − 1

)2

exp (Y ct ) =

exp (Cet ) + exp (Iet )

1 +η

2

(exp (Iet )

exp(Iet−1

) − 1

)2+

(1

− νe) n

1− n(exp (P ct ) exp (RERt))

(−ηp)

+

exp (Cct ) + exp (Ict )

1 +η

2

(exp (Ict )

exp(Ict−1

) − 1

)2 νc exp (P ct )

(−ηp)

(A.30)

Optimal lending contract between center country banks and EME banks:

(A.31)

exp(Λet+1

)(1− Γt+1) exp

(Rk,et+1

) (1− θ + θ exp

(αv,et+1

))+ exp (ϕt)

(exp

(Rk,et+1

)(Γt+1)

(exp (RERt)

exp (RERt+1)

)ld−

exp(Rbt)

exp(πct+1

)ldexp

(πet+1

)1−ld)

− γet κe exp (κet ) = 0

where ld = 1 under “liability dollarization” and 0, otherwise; where exp (αv,e) ≡ ∂exp (Jet )

∂exp (Net )

; where

(A.32)Γt =exp (RERt)

ldexp

(Rbt−1

)exp (πct )

ldexp (πet )

1−ldexp

(Rk,et

) exp(V et−1

)exp

(Qet−1

)exp

(Ket−1

)and where

(A.33)exp (ϕt) =exp (αv,et ) (1− γet )

exp(Rbt)

exp(πct+1)

ldexp(πe

t+1)1−ld

38

Envelope condition:

exp (Ket ) exp (Qet ) exp

(Rk,et+1

) (exp

(Λet+1

) (1− θ + θ exp

(αv,et+1

))−(

exp (RERt)

exp (RERt+1)

)ldexp (ϕt)

)= 0

(A.34)

Incentive compatibility constraint for EME banks (using the fact that exp (αv,et ) exp (Net ) = exp (Jet )):

(A.35)exp (αv,et ) exp (Net )− exp (Ke

t ) exp (Qet ) κe exp (κet ) = 0

Net worth:

(A.36)

exp (Net ) = exp

(Ket−1

)exp (Qet ) δT + θ

(exp

(Rk,et

)exp

(Qet−1

)exp

(Ket−1

)−

(exp (RERt)

ld exp(Rbt−1

)exp (πct )

ldexp (πet )

1−ld exp(V et−1

)))

Optimal choice of total assets by centre country banks:

(A.37)

exp(Λct+1

) (1− θ + θ exp

(αv,ct+1

)) exp(Rk,ct+1

)− exp (R∗

t )

exp(πct+1

)−

n1−n exp(V

et )

exp(RERt)1−ld

exp (Wt)

exp(Rk,ct+1

)−exp

(Rbt) ( exp(RERt)

exp(RERt+1)

)1−ld

exp(πct+1

)ldexp

(πet+1

)1−ld = κc exp (κct) γ

ct

where Wt denotes total assets of centre country banks.Centre country banks efficient investment in EME (asset pricing equation):

(A.38)exp(Λct+1

) (1− θ + θ exp

(αv,ct+1

)) exp(Rk,ct+1

)−exp

(Rbt) ( exp(RERt)

exp(RERt+1)

)1−ld

exp(πct+1

)ldexp

(πet+1

)1−ld = 0

Envelope condition for centre country optimal plan:

(A.39)(−exp (αv,ct )) (1− γct ) +exp (R∗

t ) exp(Λct+1

) (1− θ + θ exp

(αv,ct+1

))exp

(πct+1

) = 0

Incentive compatibility constraint:

(A.40)exp (αv,ct ) exp (N ct )− exp (Wt) κc exp (κct) = 0

Net worth of centre country banks:

exp (N ct ) = θ

(exp(Rk,ct )− exp(R∗t−1

)exp (πct )

)exp (Wt−1)

−n

1−n exp(V et−1

)exp (RERt)

1−ld

exp(Rk,ct )− exp(Rbt−1

) ( exp(RERt−1)exp(RERt)

)1−ld

exp (πct )ldexp (πet )

1−ld

+exp

(R∗t−1

)exp (πct )

exp(N ct−1

)+ exp

(Kct−1

)δT

(A.41)

39

Resource constraint for centre country banks:

(A.42)exp (Qet ) exp (Ket ) = exp (Ne

t ) + exp (RERt)ldexp (V et )

Definition of total assets of centre country banks:

(A.43)exp (Wt) =n

1−n exp (V et )

exp (RERt)1−ld + exp (Qct) exp (Kc

t )

Consumption Euler equation for centre country

(A.44)exp (Ret )exp

(Λet+1

)exp

(πet+1

) = 1

Leverage:

(A.45)exp (φet ) =exp (Qet ) exp (Ke

t )

exp (Net )

(A.46)exp (φct) =exp (Wt)

exp (N ct )

Spreads (financial efficiency wedges):

(A.47)exp (χet ) =exp

(Rk,et+1

)exp(Re

t )

exp(πet+1)

(A.48)exp (χct) =exp

(Rk,ct+1

)exp(R∗

t )

exp(πct+1)

40