14. Januar 2010

Does Export ConcentrationCause Volatility?Christian Busch

14. Januar 2010 2

Overview

� Countries with undiversified export structure are plausibly morevulnerable to external shocks.

� But difficult to evaluate “causal” influence of export concentration on external volatility (exchange rate, terms of trade, export growth)

� Use of a new instrument that is based entirely on geographic characteristics, and thus plausibly uncorrelated with other determinants, including institutions, income, and policies.

� Causal effect found for terms of trade and export growth, but not for exchange rates.

14. Januar 2010 3

Introduction

� Large variation in volatility of macroeconomic aggregates (output, exports, exchange rates, terms of trade etc.) across countries

� Volatility decreases with level of development

� Lack of access to financial markets to mitigate external risks� External shocks can be amplified by weak institutional environment

(Rodrik, 1999, JEG)

� Instability in economic performance critical for many developing countries

� Uncertainty lowers private investment and leads to slower growth� Increases inequality and poverty

� Instability can result in crises

14. Januar 2010 4

Determinants of Volatility

� Policy Mismanagement (Fatas and Mihov, 2003, QJE)

� Financial Institutions (Easterly et al. 2001, WB)

� Political Institutions (Acemoglu et al., 2003, JME)

� Export Concentration and Geography (Malik and Temple, 2009, JDE)

14. Januar 2010 5

Determinants of Volatility: Institutional Constraints and Export Concentration

ALB

DZA

AGO

ATGARGARM

AUSAUT

AZE

BGD

BRB

BLR

BEL

BLZ BENBOL

BWA

BRA

BGR

BFA

BDI

KHMCMR

CAN

CPV

CAFCHL

CHN

COL

COM

COG

CRIHRV

CUB

CYP

CZE

DNK

DMA

DOM ECUEGY

SLVEST

ETHFJI

FIN

FRA

GAB

GMB

GEO

DEU

GHA

GRCGRD

GTM

GIN

GNB

GUY

HTI

HND

HKG

HUN ISL

INDIDN

IRN

IRL

ISR

ITA

CIVJAM

JPN

JOR

KAZ

KEN

KOR

KGZ

LVA

LBN

LSO

LTU

LUX MAC

MKD

MDG

MWI

MYS

MLI

MLT

MRT

MUS

MEX

MDAMAR

MOZ

NAM

NPL

NLD

NZL

NICNER

NGA

NORPAK

PAN

PNG

PRY

PER

PHL

POL

PRT

ROM

RUS

RWA

KNA

LCA

VCT

STP

SEN

SYCSLE

SGP

SVK

SVN

ZAF

ESP

LKASWZ

SWECHE

SYR TJK

TZA

THA

TGOTTO

TUNTUR

UGAUKR

GBRUSA

URYVEN

VNM

YEM

ZMB

ZWE

0.0

5.1

.15

.2V

olat

ility

of G

DP

per

cap

ita g

row

th r

ate

1 2 3 4Export concentration

ALB

DZA

AGO

ATG

ARG

ARM

AUS

AUT

AZE

BGD

BRB

BLR

BEL

BLZBEN

BOL

BWA BRA

BGR

BFA

BDI

KHMCMR

CAN

CPV

CAF

TCD

CHL

CHN

COL

COM

ZAR

COG

CRI

HRV

CUB

CYP CZE

DNK

DMA

DOMECU

EGY

SLV EST

ETH

FJI

FIN

FRA

GAB

GMB

GEO

DEU

GHA

GRC

GRD

GTM

GIN

GNB

GUY

HTI

HND

HKGHUN

ISL

IND

IDN

IRN

IRL

ISRITA

CIV

JAM

JPN

JOR

KAZ

KEN

KOR

KGZ

LVA

LBN

LSOLTU

LUX

MKD

MDG

MWI

MYS

MLI

MLT

MRT

MUS

MEX

MDA

MAR

MOZ

NAM

NPL

NLD

NZL

NICNER

NGA

NORPAK

PAN

PNG

PRY

PER

PHL

POL

PRT

ROM

RUS

RWA

KNA

LCA

VCT

STP

SEN

SYC

SLE

SGP

SVK

SVN

ZAF

ESPLKA

SWZSWECHE

SYR

TJK

TZA

THA

TGO

TTO

TUN

TUR

UGA

UKR

GBRUSA

URY

UZB

VEN

VNM

YEM

ZMB

ZWE

0.0

5.1

.15

Vo

latil

ity o

f G

DP

per

ca

pita

gro

wth

rat

e

0 .2 .4 .6 .8Political constraints

14. Januar 2010 6

Export Concentration versus Income per Capita

ALB

DZA

AGO

ATG

ARG

ARM

AUS

AUT

AZEBHS

BHRBGD

BRB

BLR

BEL

BLZ

BEN

BMU

BTN

BOLBWA

BRABGR

BFA

BDI

KHMCMR

CAN

CPV

CAF

CHL

CHN

COL

COMCOG

CRI

HRV

CUB

CYP

CZE

DNK

DJIDMA

DOM

ECU

EGY

SLV

ERI

EST

ETH

FJI

FIN

FRA

GAB

GMB

GEO

DEU

GHA

GRC

GRD

GTM

GIN

GNB

GUYHTI

HND

HKG

HUN

ISL

IND

IDN

IRN

IRL

ISR

ITA

CIV

JAM

JPN

JOR

KAZ

KEN

KOR

KWT

KGZ

LVALBN

LSO

LTU

LUX

MAC

MKD

MDG

MWI

MYS

MLI

MLT

MRT

MUS

MEX

MDA

MNG

MAR

MOZ

NAM

NPL

NLD

NZL

NIC

NER

NGA

NOR

OMN

PAK

PAN

PNGPRY

PER

PHL

POL

PRT

QAT

ROM

RUS

RWA

KNALCAVCT

STP

SAU

SEN

SYCSLE

SGP

SVKSVNZAF

ESP

LKA

SDN

SWZ

SWECHE

SYR

TJK

TZA

THA

TGOTTO

TUN

TUR

TKM

UGA

UKR

GBR USA

URY

VEN

VNMYEM

ZMB

ZWE

12

34

Exp

ort C

onc

entr

atio

n

6 7 8 9 10GDP per Capita

ALB

DZA

AGO

ATG

ARG

ARM

AUS

AUT

AZE

BGD

BRB

BLR

BEL

BLZBEN

BOL

BWABRA

BGR

BFA

BDI

KHM CMR

CAN

CPV

CAF

TCD

CHL

CHN

COL

COM

ZAR

COG

CRI

HRV

CUB

CYPCZE

DNK

DMA

DOMECU

EGY

SLV EST

ETH

FJI

FIN

FRA

GAB

GMB

GEO

DEU

GHA

GRC

GRD

GTM

GIN

GNB

GUY

HTI

HND

HKGHUN

ISL

IND

IDN

IRN

IRL

ISRITA

CIV

JAM

JPN

JOR

KAZ

KEN

KOR

KGZ

LVA

LBN

LSOLTU

LUX

MAC

MKD

MDG

MWI

MYS

MLI

MLT

MRT

MUS

MEX

MDA

MAR

MOZ

NAM

NPL

NLD

NZL

NICNER

NGA

NORPAK

PAN

PNG

PRY

PER

PHL

POL

PRT

PRI

ROM

RUS

RWA

KNA

LCA

VCT

STP

SEN

SYC

SLE

SGP

SVK

SVN

ZAF

ESPLKA

SWZSWECHE

SYR

TJK

TZA

THA

TGO

TTO

TUN

TUR

UGA

UKR

GBRUSA

URY

UZB

VEN

VNM

YEM

ZMB

ZWE

0.0

5.1

.15

Vo

latil

ity o

f G

DP

per

ca

pita

gro

wth

rat

e

6 7 8 9 10Income per capita

14. Januar 2010 7

Causality and Model Uncertainty

� Instrumental Variables

� Acemoglu et al. (2003, JME): Institutions versus policies

� Bayesian Methods

� Malik and Temple (2009, JDE): Geography and export concentration

� Panel Data

� Yang (2008, JM): Influence of democracy depends on ethnic homogeneity

� Here: New instrument for export concentration based on geographic characteristics

14. Januar 2010 8

Approach

� Use geographically determined component of export concentration as an instrument for actually observed export concentration to study influence on external volatility

� Remoteness as a natural barrier to trade (Malik and Temple, 2009)

� Transport costs affect sectors differently (imported intermediate goods)

� countries may be locked into concentrated export structures

� Export concentration determines how external shocks are absorbed

� Terms of trade: Undiversified countries are hit more severely by price shocks

� Exports: Diversification dampens exposure to demand shocks

� Exchange rates: currency demand fluctuates with exports

14. Januar 2010 9

Geography-based Instrument for Export Concentration

� Appropriately weighted average of countries’ remoteness (based on Frankel and Romer, 1999)

� Include information about distance, population, area, landlocked, border

� Do not use common geographic variables

� Distance from equator and tropical area (Sachs, 2003)

� Soil quality (Sokoloff and Engerman, 2000)� Coastal area (Acemoglu et al., 2005)

� Control for trade � Remoteness reduces trade; Increased specialization induces more

volatility (di Giovanni and Levchenko, RES, forthcoming)

14. Januar 2010 10

New Zealand: Influence of Distance on Export Concentration

Argentina

Australia

AustriaBahrain

BangladeshBarbadosBelgium

Bolivia

Brazil

Bulgaria

Canada

ChileChina

Colombia

Cook Islands

Costa Rica Cyprus

Denmark

EcuadorEgypt

Fiji

Finland

FranceGermany West

GhanaGreece

Guatemala

Hong Kong

HungaryIceland

India

Indonesia

IranIraq

IrelandIsrael

Italy

Jamaica

Japan

Jordan

Kenya

Kiribati

Korea South

Kuwait

Malaysia

MaltaMauritiusMexico

Myanmar (Burma)

Nepal

Netherlands

New Caledonia

Nigeria

Niue

Norway

OmanPakistan

Panama

Papua New Guinea

Peru

Philippines

PolandPortugalQatar

Romania

Saudi Arabia

Seychelles

Singapore

Solomon Islands

South Africa

Spain

Sri Lanka

SwedenSwitzerland

Syria

Tanzania

Thailand

Trinidad and TobagoTurkey

Tuvalu

USSR

United Arab Emirates

United Kingdom

United States

Uruguay

Venezuela

Vietnam

Zimbabwe

1.5

22

.53

3.5

4ex

port

con

cent

ratio

n

7 7.5 8 8.5 9 9.5log distance

14. Januar 2010 11

Belgium: Influence of Trading Partner‘s Size on Export Concentration (Controlling for Distance)

Seychelles

Bahamas

Malta

Djibouti

Iceland

SurinameBarbados

Qatar

Bahrain

Comoros

Fiji

Cyprus

GuyanaGambia

OmanGabon

Guinea-Bissau

Trinidad and Tobago

Mauritania

Jordan

Kuwait

Mauritius

United Arab Emirates

Congo Republic

Liberia

Panama

Mongolia

Costa Rica

Ireland

Nicaragua

JamaicaTogo

Uruguay

Central African Republic

Singapore

Paraguay

Sierra Leone

Honduras

Israel

Norway

El Salvador

New Zealand

Chad

Benin

TunisiaDominican Republic

Denmark

Finland

Bolivia

Guinea

MaliSwitzerland

SyriaYemenZambia

Guatemala

Burundi

Haiti

Austria

Senegal

Ecuador

Rwanda

Niger

Greece

Zimbabwe

Malawi

Sweden

Saudi Arabia

AngolaNetherlandsBulgariaCameroon

Portugal

Hong Kong

Iraq

Ivory Coast

Burkina Faso

Hungary

AlgeriaGhana

Chile

Madagascar

Morocco

Venezuela

Sri Lanka

Uganda

Peru

Malaysia

Sudan

Nepal

Mozambique

Kenya

Australia

Yugoslavia

Romania

Colombia

Tanzania

Spain

Egypt

ArgentinaSouth Africa

Congo Democratic Republic

Canada

Iran

Poland

Myanmar (Burma)

Korea SouthFrance

Italy

Ethiopia

Turkey

United Kingdom

Philippines

Mexico

PakistanThailand

Bangladesh

Nigeria

BrazilIndonesia

Japan

United States

India

China

-1-.

50

.51

e( e

xpor

t co

ncen

tra

tion

| X )

-4 -2 0 2 4 6e( lpop2 | X )

coef = -.08958648, se = .01800051, t = -4.98

14. Januar 2010 12

Method

� Estimate bilateral export concentration equation

� Include only purely geographic determinants

� Aggregate fitted values to estimate a geographic component of countries’ overall export concentration

� Problem: bilateral export concentration cannot be easily aggregated across countries (as is the case for trade)

� Solution: use an additively decomposable inequality measure

14. Januar 2010 13

Generalized Entropy Inequality Measure…

� x value of trade in product h between country i and country j� n number of products traded

� µ mean value of trade flows across products

� α inequality weighting parameter (set to 0.5)

14. Januar 2010 14

Concentration Measures Compared

0.2

.4.6

.81

Her

finda

hl c

onc

entr

atio

n

1 2 3 4Generalized entropy concentration

14. Januar 2010 15

… allows Decomposition by Export Country

� Where the first term is the concentration between all bilateral trade relationships for country i,

� and the second term is the concentration within each bilateral trade relationship

� Thus, we need to estimate Xi,j, Ni,j, and GEi,j

14. Januar 2010 16

Data

� Export concentration: � Concentration across products (4-digit SITC)� Comtrade dataset; averaged over 1980-2000

� Volatility (standard deviation; WDI)� Terms of trade� Real effective exchange rate� Export growth

� Gravity variables from Frankel and Romer (1999)

� Trade openness (EX+IM / GDP; WDI)

� Institutional environment (Aggregate Governance Indicators)

14. Januar 2010 17

How does Geography Influence Concentration? Dependent Variable

ln Export ln Numberall > 50 value of sectors original

valuesComtrade

values

ln distance 0.213*** 0.212*** -0.583*** -0.243*** -0.85*** -1.225***(0.007) (0.008) (0.033) (0.011) (0.04) (0.036)

ln population (country i) -0.262*** -0.246*** 0.663*** 0.270*** -0.24*** -0.411***(0.003) (0.005) (0.020) (0.006) (0.03) (0.030)

ln area (country i) 0.100*** 0.094*** -0.164*** -0.094*** -0.12*** -0.012 (0.003) (0.003) (0.016) (0.005) (0.02) (0.023)

ln population (country j) -0.042*** -0.025*** 0.718*** 0.085*** 0.61*** 1.400***(0.004) (0.005) (0.019) (0.007) (0.03) (0.028)

ln area (country j) 0.010*** -0.005 -0.184*** -0.011** -0.19*** -0.307***(0.003) (0.004) (0.016) (0.005) (0.02) (0.023)

Landlocked dummy 0.107*** 0.033** -0.905*** -0.148*** -0.36*** -1.149***(0.011) (0.014) (0.060) (0.021) (0.08) (0.103)

Common border dummy-0.231*** -0.075** 0.426*** 0.123***(0.039) (0.038) (0.156) (0.045)

Constant 4.800*** 4.275*** 7.103*** 3.120*** -6.38*** -7.725***(0.068) (0.086) (0.351) (0.120) (0.42) (0.469)

R2 0.358 0.304 0.319 0.226 0.360 0.455

Observations 12'503 6'773 6'773 6'773 3'220 6'284

Bilateral trade flows

Frankel/Romer

Bilateral export concentration Bilateral exports

Number of sectors

14. Januar 2010 18

Actual versus Constructed Export Concentration

DZA

AGO

ARG

AUS

AUT

BHS

BHRBGD

BRB

BEL

BLZ

BENBTN

BOL

BRABGR

BFA

BDI

CMR

CAN

CAF

CHL

CHN

COL

COMCOG

CRI

CYP

DNK

DJIDMA

DOM

ECU

EGY

SLV

ETH

FJI

FIN

FRA

GAB

GMBGHA

GRC

GRD

GTM

GIN

GNB

GUYHTI

HND

HKG

HUN

ISL

IND

IDN

IRN

IRL

ISR

ITA

CIV

JAM

JPN

JORKEN

KOR

KWT

LBR

MDG

MWI

MYS

MLI

MLT

MRT

MUS

MEX

MNG

MAR

MOZ

MMRNPL

NLD

NZL

NIC

NER

NGA

NOR

OMN

PAK

PAN

PNGPRY

PER

PHL

POL

PRT

QAT

ROM

RWA

LCAVCT

SAU

SEN

SYCSLE

SGP

SLBSOM

ZAF

ESP

LKA

SDN

SUR

SWECHE

SYR

TZA

THA

TGOTTO

TUN

TUR

UGAARE

GBRUSA

URY

VEN

YEM

ZMB

ZWE

12

34

Act

ual E

xpor

t Con

cent

ratio

n

2 2.5 3 3.5 4Constructed Export Concentration

14. Januar 2010 19

External Volatility and Export Concentration

Dependent Variable

OLS IV OLS IV OLS IV(1) (2) (3) (4) (5) (6)

Export 0.068*** 0.043*** 0.054*** 0.006 0.076*** 0.054***Concentration (0.008) (0.013) (0.012) (0.023) (0.010) (0.019)

Trade -0.040*** -0.038*** -0.066*** -0.056*** -0.040*** -0.035***Openness (0.009) (0.009) (0.019) (0.019) (0.012) (0.013)

Constant 0.087** 0.143*** 0.248*** 0.326*** 0.121** 0.161***(0.039) (0.055) (0.071) (0.098) (0.047) (0.058)

R2 0.404 0.353 0.323 0.139 0.324 0.296 N 110 110 73 73 127 127

Excluded Instrument Predicted Predicted PredictedConcentration Concentration Concentration

Levels of statistical significance indicated by: *** p<0.01; ** p<0.05; * p<0.1

Numbers reported in parentheses are heteroscedasticity-robust standard errors.

Regressions of Volatility

Terms of Trade Real Effective Exchange Rate Export Growth

14. Januar 2010 20

Importance of Controlling for Trade (e.g. ToT)-1

.5-1

-.5

0.5

1R

esid

uals

from

Re

gre

ssin

g E

xp. C

onc.

on

Pre

d. T

rad

e

-1 -.5 0 .5 1Resdiuals from Regressing Predicted Export Concentration on Predicted Trade

-2-1

01

2R

esid

uals

from

Re

gre

ssin

g T

rade

on

Pre

d. E

xp. C

on

c.

-2 0 2 4 6Resdiuals from Regressing Predicted Trade on Predicted Export Concentration

-1-.

50

.51

1.5

Res

idu

als

fro

m R

egr

ess

ing

Exp

. Con

c. o

n P

red.

Exp

.Con

.

-2 0 2 4 6Resdiuals from Regressing Predicted Trade on Predicted Export Concentration

-3-2

-10

12

Res

idua

ls fr

om R

egre

ssin

gTra

de o

n P

red

. Tra

de

-1 -.5 0 .5 1Resdiuals from Regressing Predicted Export Concentration on Predicted Trade

14. Januar 2010 21

Export Concentration and InstitutionsRegressions of Volatility in Terms of Trade

Method OLS OLS IV IV IV2SLS First Stage First Stage

Exp. Conc. Governance(1) (2) (3) (4) (5)

Export 0.046*** 0.047*** 0.042* Concentration (0.010) (0.014) (0.025)

Trade -0.017** -0.011 -0.011 0.008 1.130 Openness (0.008) (0.008) (0.007) (0.107) (0.928)

Aggregate -0.005*** -0.004** -0.003 Governance (0.001) (0.002) (0.004)

Predicted Exp. 0.505** 2.540

Concentration (0.230) (2.067)

Settler 0.323*** -2.394***

Mortality (0.044) (0.458)

Constant 0.057* 0.024 0.042 -0.302 -1.394 (0.034) (0.036) (0.067) (0.964) (6.893)

R2 0.468 0.384 0.374 0.512 0.431 N 104 58 58 58 58

Excluded Instrument Predicted Predicted PredictedConcentration, Concentration, Concentration,Settler Mortality Settler Mortality Settler Mortality

Levels of statistical signif icance indicated by: *** p<0.01; ** p<0.05; * p<0.1

Numbers reported in parentheses are heteroscedasticity-robust standard errors.

14. Januar 2010 22

Robustness Checks

� Overidentification tests using concentration across countries (predicted from weighted bilateral distance) as additional IV

� Control for export shares of various natural resources (oil, agriculture, various raw materials)

� Control for the size of the manufacturing sector: inconclusive results

14. Januar 2010 23

Conclusions

� Bilateral export concentration is well described by a gravity equation

� Inequality decomposition methods allow to aggregate bilateral export concentration to an overall measure of export concentration

� Predicted export concentration contains valuable information about actual concentration

� Using predicted export concentration as an IV, export concentration exhibits a causal influence on volatility in terms of trade and growth in exports, but not so on volatility in exchange rates

� The instrumental variable is not suited to elucidate the separate roles of institutions and export concentration in determining aggregate volatility