ABSTRACT
Title of dissertation: ESSAYS IN INTERNATIONAL FINANCE
Christian Daude, Doctor of Philosophy, 2008
Dissertation directed by: Professor Enrique G. Mendoza
and Professor Carlos A. Vegh
Department of Economics
Access to private capital markets is the most salient difierence between emerg-
ing market economies and other developing countries. However, in contrast to de-
veloped economies, emerging markets have had a troubled relationship with capital
 ows. In particular, balance of payments and debt crises have been a recurrent prob-
lem. The three chapters of this dissertation contribute to the literature on emerging
markets and their relationship with capital markets.
Chapter 1 analyzes the efiects of volatility on sovereign default risk. Empir-
ically, the paper establishes a concave relationship between spreads and volatility.
While for low levels of volatility an increase in volatility is associated with an in-
crease in the sovereign risk premium, the risk premium increases at a decreasing
rate. This empirical relationship is robust to difierent estimation methods, sam-
ples and control variables. Furthermore, the relationship between volatility and risk
premia is non-monotonic: while at low levels of volatility an increase in volatility
implies an increase also in spreads, for su?ciently high levels of volatility this rela-
tionship turns negative. The chapter also presents a quantitative model of sovereign
debt with default risk consistent with this feature and other characteristics of EME
debt. The intuition for this result is the existence of a trade-ofi between prudential
behavior in order to avoid large consumption  uctuations under autarky and the
increased likelihood of a default, given default provides some short-run relief under
a very bad realization of shocks.
Chapter 2 addresses the determinants of the composition of cross-border in-
vestment positions. Using a novel database of bilateral capital stocks for all types
of investment - FDI, portfolio equity securities, debt securities as well as loans - for
a broad set of 77 countries, we show the importance of two key determinants of the
composition of cross-border asset positions: information frictions and the quality
of host country institutions. Overall, we flnd that in particular FDI, and to some
extent also loans, are substantially more sensitive to information frictions than in-
vestment in portfolio equity and debt securities. We also show that the share as
well as the size of FDI that a country receives are largely insensitive to corruption in
host countries, while portfolio investment is by far the most sensitive to the quality
of institutions.
Chapter 3 focuses on a related topic to chapter 2. Using bilateral FDI stocks
around the world, we explore the importance of a wide range of institutional vari-
ables as determinants of the location of FDI. While we flnd that better institutions
have overall a positive and economically signiflcant efiect on FDI, some institutional
aspects matter more than others do. Especially, the unpredictability of laws, reg-
ulations and policies, excessive regulatory burden, government instability and lack
of commitment play a major role in deterring FDI. For example, the efiect of a
one standard deviation improvement in the regulatory quality of the host country
increases FDI by a factor of around 2. These results are robust to difierent speci-
flcations, estimation methods and institutional variables. We also present evidence
on the signiflcance of institutions as a determinant of FDI over time.
ESSAYS IN INTERNATIONAL FINANCE
by
Christian Daude
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulflllment
of the requirements for the degree of
Doctor of Philosophy
2008
Advisory Committee:
Professor Enrique G. Mendoza, Co-Chair
Professor Carlos A. Vegh, Co-Chair
Professor John Shea
Assistant Professor Anton Korinek
Professor Carol Graham
c Copyright by
Christian Daude
2008
Dedication
To Catalina and Virginia.
ii
Acknowledgments
Many thanks to Enrique Mendoza, Carlos Vegh, John Shea, Anton Korinek
and Carol Graham for accepting to serve on my committee.
I am especially grateful to my advisors, Professors Mendoza and Vegh for giv-
ing me their valuable advice and their patience. I have learned a lot from Professor
Mendoza during the whole program. His research and approach to address economic
problems will have a long-lasting impact on the way I see the world. He always made
very detailed and constructive comments on my work. I am very thankful for his
generosity. My special thanks also to Professor Vegh for his kind support during
di?cult times as well as his very insightful comments and suggestions.
Many people made useful comments and contributions to the papers included
in this dissertation. I will always have fond memories of College Park due to the
good friends I have met during my years here. In particular, I would like to thank
my co-authors Ernesto Stein and Marcel Fratzscher for giving me the opportunity
to work with them and for many fruitful discussions. I have learned a lot from them.
I owe also special recognition to my collegues and friends Bora Durdu and Emine
Boz. I beneflted enormously from our lunch discussions and joint work. Professor
Guillermo Calvo had a great impact on my research. In particular, he helped me
sort out many \bad" ideas in the early stages. I am also grateful to Professors
Carmen Reinhart, John Rust and John Shea who gave very detailed comments and
insightful questions. I would also like to thank Vickie Fletcher for all her help.
I am also thankful to Enrique Alberola, Laura Alfaro, Paolo Angelini, Gian-
carlo Corsetti, Luca Dedola, Eduardo Fernandez-Arias, and two anonymous referees,
iii
as well as the participants of the workshop in international economics at the Uni-
versity of Maryland at College Park, the LACEA/LAMES Meetings in 2006 and
2007, the BIS/CGFS conference on \Reserach on global flnancial stability: The use
of BIS international flnancial statistics, the ECB Research Department, and the
2006 University of Navarra conference on emerging markets. I also appreciate the
generosity of Luis Catao, Michael Fidora Alejandro Izquierdo, Elias Papaioannou,
and Marcelo Oviedo who shared their data with me or helped to construct some of
the datasets used in this dissertation.
I would also like to thank my parents for their support during many years and
encouraging me to pursue my Ph.D. in economics, even if this meant that I would
be far away, again.
My deepest gratefulness is for my wife, Virginia. I could not imagine this
process without her. She has sacriflced many things for being my support and
companion. Even during very di?cult times for her, she always had the strength
and optimism to cheer me up when I needed it. She also patiently listed to my ideas
many times and made important comments that considerably improved my work. I
hope I can be such a great source of encouragement for her life and academic work
as she has been to me.
iv
Table of Contents
List of Tables vii
List of Figures ix
1 Sovereign Default Risk and Volatility 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Empirical Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 The Model Economy . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.3.1 The Household?s Problem . . . . . . . . . . . . . . . . . . . . 26
1.3.2 The Government?s Problem . . . . . . . . . . . . . . . . . . . 28
1.3.3 Creditors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.3.4 Some Intuition . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.3.5 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
1.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2 The Composition of Cross-Border Asset Positions (co-
authored with Marcel Fratzscher) 53
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.2 Related Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.3 Methodology, data and some stylized facts . . . . . . . . . . . . . . . 62
2.3.1 Methodology and hypotheses . . . . . . . . . . . . . . . . . . 62
2.3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.3.3 Composition of cross-border investment: some stylized facts . 69
2.4 The role of information frictions . . . . . . . . . . . . . . . . . . . . . 76
2.4.1 Benchmark results . . . . . . . . . . . . . . . . . . . . . . . . 76
2.4.2 Robustness: alternative proxies for information frictions . . . . 78
2.4.3 Robustness: Alternative model speciflcations and controls . . . 81
2.5 The role of institutions and flnancial market development . . . . . . . 89
2.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3 The Quality of Institutions and Foreign Direct Invest-
ment (co-authored with Ernesto Stein) 103
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.2 Data and Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . 109
3.2.1 FDI Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.2.2 Institutional Variables . . . . . . . . . . . . . . . . . . . . . . 109
3.2.3 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . 115
3.3 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
3.3.1 OLS Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . 120
v
3.3.2 Instrumental Variables Estimations . . . . . . . . . . . . . . . 124
3.4 Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.5 Time series evidence of the relevance of institutions . . . . . . . . . . 134
3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
A Appendix 139
A.1 Solution Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
A.2 Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
A.3 Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
Bibliography 144
vi
List of Tables
1.1 Spreads and Revenue Volatility - Cross-section Regressions . . . . . . 15
1.2 Spreads and Revenue Volatility - Robustness . . . . . . . . . . . . . . 18
1.3 Logit Regressions - Default Events . . . . . . . . . . . . . . . . . . . 22
1.4 Baseline Parameter Values . . . . . . . . . . . . . . . . . . . . . . . . 41
1.5 Sample and Simulated Moments . . . . . . . . . . . . . . . . . . . . . 43
1.6 Robustness under Alternative Parameterizations . . . . . . . . . . . . 49
2.1 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.2 Correlation Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.3 Information frictions: distance . . . . . . . . . . . . . . . . . . . . . . 77
2.4 Information frictions: distance versus telephone tra?c . . . . . . . . . 80
2.5 Information frictions: alternative information proxies . . . . . . . . . 82
2.6 Information frictions: Ratios as % of GDP and total capital stocks . . 83
2.7 Information frictions: Alternative Estimators . . . . . . . . . . . . . . 85
2.8 Information frictions: Developed countries versus emerging market
economiess . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
2.9 Robustness tests with trade, exchange rate volatility and global in-
terest rate shocks as controls . . . . . . . . . . . . . . . . . . . . . . . 88
2.10 Role of market development and transparency . . . . . . . . . . . . . 91
2.11 Role of investor protection and corruption . . . . . . . . . . . . . . . 93
2.12 Role of market development and transparency - 3SLS estimator . . . 96
2.13 Role of market development and transparency - 3SLS estimator . . . 98
3.1 Simple and Partial Correlations (controlling for GDP per capita) of
Kaufmann et al Variables . . . . . . . . . . . . . . . . . . . . . . . . 112
3.2 Cross Section OLS Estimates of Equation (3.1) . . . . . . . . . . . . 121
vii
3.3 IV Estimates of Equation (3.1) . . . . . . . . . . . . . . . . . . . . . 127
3.4 Robustness: Estimation Methods and Model Speciflcation . . . . . . 129
3.5 Robustness Alternative Institutional Variables . . . . . . . . . . . . . 133
3.6 Panel Data Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
viii
List of Figures
1.1 Annual Real Revenue Growth Venezuela and Norway . . . . . . . . . 3
1.2 Revenue Volatility and Level of Development . . . . . . . . . . . . . . 4
1.3 Revenue Volatility and Average Debt Holdings . . . . . . . . . . . . . 5
1.4 Revenue Volatility and Sovereign Bond Spreads . . . . . . . . . . . . 6
1.5 Revenue Volatility and Institutional Investors Rating . . . . . . . . . 7
1.6 Average Simulated Spreads and Volatility . . . . . . . . . . . . . . . 44
1.7 Average Net Asset Holdings, Debt Ceiling and Volatility . . . . . . . 45
1.8 Estimated VAR coe?cients . . . . . . . . . . . . . . . . . . . . . . . 46
1.9 Average Simulated Spreads for Difierent Relative Risk Aversion Co-
e?cients (?) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.1 Composition of Cross-border Investment Positions by GDP per capita
Quintiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.2 Shares in Cross-Border Investment and GDP per capita . . . . . . . . 73
2.3 Stylized facts: macro and exchange rate variables . . . . . . . . . . . 74
2.4 Stylized facts: market development and institutions . . . . . . . . . . 75
3.1 Partial Correlation between Average FDI In ows 1990-2000 as a Ra-
tio of GDP and Government E?ciency in 1996, controlling by GDP
per capita . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
ix
Chapter 1
Sovereign Default Risk and Volatility
1.1 Introduction
Two distinct features of emerging market economies (EME) are that they
are subject to large  uctuations compared to developed economies and also more
prone to sufier balance of payments crises as well as default on their debt. Mendoza
(1995), for example, provides compelling evidence that  uctuations in output as well
as the terms of trade are on average more than twice as large for EME compared
to developed economies. Therefore, not only endogenous business cycle  uctuations
but also relatively exogenous shocks are larger in EME.1 In line with this evidence, as
Mendoza and Oviedo (2006) show, government revenues are also much more volatile
in EME than in developed countries.
Since Ramey and Ramey?s (1995) empirical flnding of a negative impact of
volatility on economic growth, the literature on the interrelation between volatility
and economic growth has been growing rapidly.2 However, the link between volatil-
ity and sovereign debt issues - especially risk premia and default risk - has received
relatively little attention. Since the seminal paper on defaultable sovereign debt by
1For papers that quantify the importance of terms-of-trade shocks for developing countries see
Broda (2004), Kose (2002), and Mendoza (1995). The latter shows that terms-of-trade shocks
account for up to 50 percent of business cycle  uctuations in developing countries.
2See e.g. Mendoza (1997) for an early analytical contribution, as well as Calvo (2005) and
Aghion et al (2005) and references in these papers.
1
Eaton and Gersovitz (1981), the question of how volatility afiects spreads and debt
holdings has not been systematically addressed in the literature using a dynamic
stochastic general equilibrium model.
The contribution of this chapter to the existing literature is twofold. First, it
analyzes empirically the relationship between government revenue volatility and de-
fault risk (as well as risk premia). We flnd a concave relationship between sovereign
default risk and volatility; risk is increasing in the level of volatility, but at a decreas-
ing rate. In addition, evidence indicates a non-monotonic relationship; for low levels
of revenue volatility, a change in volatility increases the risk of default; however, for
su?ciently high levels of revenue volatility, default risk is actually decreasing in
volatility. We show that this relationship is robust to alternative measures of de-
fault risk and premia as well as econometric methods and speciflcations. Second,
we present a model of sovereign debt with potential repudiation that is consistent
with this previous flnding and other issues related to sovereign debt in EME.
The stylized facts that guide the empirical and quantitative analysis are the
following:
1. Revenues are much more volatile in emerging economies than in developed
economies, as shown in Figure 1.1 and Figure 1.2. This stylized fact has also
been addressed earlier by Gavin, Hausmann and Perotti (1996) for the case
of Latin America, as well as Mendoza and Oviedo (2006) for EME in gen-
2
Figure 1.1: Annual Real Revenue Growth Venezuela and Norway
-50%
-40%
-30%
-20%
-10%
0%
10%
20%
30%
40%
50%
1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Venezuela Norway
Source: World Economic Outlook database, IMF
eral. Figure 1.1 shows the annual growth rate of real government revenues
for Venezuela and Norway from 1981 to 2005. Both countries are large oil ex-
porters, with oil related revenues being a signiflcant share of the public sector?s
income. In particular, oil exports represented on average about 24 percent of
GDP in both countries during 1995 - 2005. In addition, the importance of oil
prices is re ected by the high co-movement between both series - with a corre-
lation coe?cient of 0.53. Also, on average real revenues have grown at a similar
pace (4 percent annually) in both countries from 1981 to 2005. However, the
difierences in the volatility of both series is striking. While the standard de-
viation of revenue growth is 7 percent in Norway, for the case of Venezuela
it is around 21 percent, three times larger.3 The fact that revenue volatility
3A similar difierence in magnitudes of volatility also holds for real GDP growth, with standard
deviations of 2 and 6 percent, respectively.
3
difiers signiflcantly according to the level of development is also conflrmed for
a larger sample of countries in Figure 1.2, which shows the correlation between
initial GDP per capita and the coe?cient of variation of the revenue to GDP
ratio from 1990 - 2004. The correlation between both variables is negative,
around -0.53, and signiflcant at conventional levels of confldence.
Figure 1.2: Revenue Volatility and Level of Development
ARG
AUS
AUT
BEL
BRA
CAN
CHL
CHN
COLCRI
DEU DNK
ECU
ESP FINFRA
GBR
HUN
IDN
IND
IRL
ISR ITA
JOR
JPN
KOR
MAR
MEX
MYS
NLD
NOR
NZLPAK
PAN
PER
PHL
PRT SWE
THA
URY USA
VEN
ZAF
0
5
10
15
20
Revenue/GDP Coefficient of Variation
0 10000 20000 30000
GDP per capita 1990
2. For the case of emerging markets, as also shown by Mendoza and Oviedo
(2006), debt holdings decrease with the level of revenue volatility. However,
there is no signiflcant relationship between volatility and debt holdings among
rich OECD countries, as can be seen in Figure Figure 1.3.
3. There is a non-monotonic relationship between risk premia, measured as the
spread of sovereign bond secondary-market yields over US treasuries, as well as
4
default risk, measured by the Institutional Investors Rating (IIR) (see Figures
1.4 and 1.5).4 As we document in section 1.2, this result is statistically signif-
icant and continues to hold when controlling for other potential determinants
of risk, estimation methods and excluding potential outliers.
Figure 1.3: Revenue Volatility and Average Debt Holdings
AUT
BEL
CAN
DEU
DNK
ESPFIN
FRA GBR
IRL
ITA
JPN
NLD
NOR
NZL
PRT
SWE
USA
20
40
60
80
100
120
Public Debt/GDP (%)
2 3 4 5 6 7
Revenue/GDP Coefficient of Variation
OECD Countries
ARG
BRA
CHL
CHN
COL
CRI
ECU
IDN
IND
ISR
JOR
KOR
MAR
MEX
MYS
PAKPAN
PER
PHL
THA
URY
ZAF
0
50
100
150
Public Debt/GDP (%)
0 5 10 15 20
Revenue/GDP Coefficient of Variation
Non?OECD Countries
This last stylized fact is \new" in the sense that the empirical literature has
not documented it so far, while theoretical and quantitative models of sovereign debt
have not addressed this issue neither. To my knowledge, the only paper that explores
the empirical relationship between volatility and sovereign default is Catao and
Kapur (2006). According to their estimates, while demand for debt is increasing in
volatility, debt ceilings - the level of debt beyond which a rational risk neutral lender
is not willing to extend further credit - decrease with macroeconomic volatility,
4For the case of the IIR, higher values represent less risk.
5
measured by the volatility of the output gap or the terms of trade. In addition, they
also estimate a logit model explaining sovereign default episodes and flnd that the
probability of default increases with macroeconomic volatility.5
Figure 1.4: Revenue Volatility and Sovereign Bond Spreads
ARG
AUS
AUTBEL
BRA
CAN
CHL CHN
COLCRI
DEUDNK
ECU
ESPFINFRA GBR
HUN
IND
IRL
ISR
ITA
JOR
JPN
KOR
MAR
MEX
MYS
NLDNOR
NZL
PAK
PAN
PER
PHL
POL
PRT
SWE
THA
UKR
URY
USA
VEN
ZAF
0
.02
.04
.06
.08
.1
log(1+Spread)
0 .05 .1 .15 .2
 Revenue/GDP Coefficient of Variation
Whole Sample
Excluding China and Korea
Excluding China
As seen in Figure 1.4, considering sovereign bond spreads, Korea and China
are in uential observations and when both are excluded, the relationship between
spreads and revenue volatility is concave, but not non-monotonic in the sample.
However, if we consider alternative risk measures - like the IIR in Figure 1.5 - the
non-monotonic relationship is still present in the data when these two observations
are excluded. Much of the efiort in the empirical part of this chapter is devoted
5Catao and Kapur provide also a simple two-period model in the spirit of Sachs and Cohen
(1982) that shows a positive relationship between volatility and risk premia under a uniform dis-
tribution. The present paper analyzes the relationship between volatility and sovereign debt issues
in a more standard inflnite-horizon small open economy model with a less restrictive assumption
on the distribution and persistence of shocks.
6
Figure 1.5: Revenue Volatility and Institutional Investors Rating
ARG
AUS
AUT
BEL
BRA
CAN
CHL
CHN
COLCRI
DEU
DNK
ECU
ESP
FIN
FRA GBR
IDN
IND
IRLITA
JOR
JPN
KOR
MAR
MEX
MYS
NLDNOR
NZL
PAK
PAN
PER
PHL
PRT
SWE
THAURY
USA
ZAF
0
20
40
60
80
100
Institutional Investors Rating
0 .05 .1 .15 .2
 Revenue/GDP Coefficient of Variation
Whole Sample
Excluding China and Korea
Excluding China
to show the robustness of the non-monotonic relationship between risk premia and
volatility in the data. Also, the model presented afterwards focuses especially on
the relationship between revenue volatility and default risk.
While there are several papers on sovereign debt that generate a negative cor-
relation between equilibrium debt holdings and volatility (stylized fact 2) based on
precautionary savings motives in incomplete asset market economies (see Mendoza
and Oviedo, 2006; Durdu et al, 2007), these papers assume that the government re-
pays its debt under all states of nature. Therefore, they do not generate an endoge-
nous risk premium and consequently do not analyze how changes in the volatility of
shocks afiect the incentive to default. In addition, default introduces the possibility
of making non-contingent debt an ex-post contingent instrument. Thus, it is also
7
interesting to analyze the relationship of debt holdings and volatility in a model that
allows for default to occur. There is a recent and growing literature of quantitative
models of sovereign debt with default risk inspired by Eaton and Gersovitz (1981)
(for two in uential contributions in this literature see Aguiar and Gopinath, 2006;
and Arellano, forthcoming). However, the quantitative implications for risk pre-
mia of changes in volatility have not been explored systematically in this literature.
Thus, a contribution of the present chapter is to analyze this issue in a similar set-up.
In their seminal paper on sovereign debt, Eaton and Gersovitz (1981) address
the efiects of volatility on debt in a non-stochastic model where the endowment
income of the economy  uctuates period-by-period between a high and a low level of
output. Inthisset-up, theyshowthatagreatergapbetweenbothoutputrealizations
increases the desired level of debt by the borrower and the credit ceiling, allowing
the borrower to hold higher levels of debt in equilibrium. The intuition for this result
is the following. Given that in the event of default borrowers are punished by being
excluded from the international credit market forever, a higher volatility implies a
larger welfare cost from losing the possibility to smooth income  uctuations using
debt. Thus, the argument goes, countries with higher volatility would be able to
commit to higher debt levels.
The non-stochastic nature of the model puts severe limitations on the results,
given that default never occurs in equilibrium. Therefore, the model is not able
to create an endogenous risk premium although it generates a credit ceiling - de-
flned as a level of debt beyond which creditor would not be willing to extend more
8
credit at any price, given that default would occur under all states of nature. In a
stochastic environment, more output variability implies also more uncertainty which
could lead to an increase in the probability of default and therefore push up risk
premia. This potentially important mechanism is not active in Eaton and Gerso-
vitz?s non-stochastic set-up. However, while Eaton and Gersovitz (1981) mention
this possibility, they suggest that their results would hold in a more general set-up.6
Although Eaton and Gersovitz (1981) also present econometric evidence that the
volatility of exports had a positive and signiflcant efiect on credit ceilings and debt
levels in the 1970?s, for a more recent time period, empirical evidence by Catao and
Kapur (2006) contradicts these results.
In general, the efiects of volatility on debt levels, default risk and risk premia
depend on the costs of default and sanctions that are imposed in the event of de-
fault. If the punishment in case of default is exclusion from the credit market, then
it might be that countries that face more volatility could commit to higher debt
levels, given that the cost of reverting to autarky would be more severe, whenever
the sovereign borrower is risk averse. This is the main argument made by Eaton
and Gersovitz (1981). However, a mean-preserving spread would also increase the
likelihood of having a very bad draw. If the borrower defaults under bad states of
nature, then higher volatility would increase the probability of default and credi-
tors would tend to charge a higher risk premium. This latter efiect is operating in
6Although they discuss brie y the potential negative efiect of volatility on debt ceilings in a
stochastic environment, they argue that only very high discount rates could cause this result in
their model. In addition, to address this item they must make very strong and limiting assumptions
on the model (e.g. that the current debt levels have to be zero in order to create lending).
9
Catao and Kapur?s (2006) set-up, while the two-period nature of their model makes
it impossible to analyze the efiects of exclusion from credit markets in the future.
Thus, potentially there are two efiects that go in opposite directions. This chapter
contributes to the literature by studying this issue quantitatively in a model where
both of these channels coexist. The model shows that at low levels of volatility the
increase in default risk dominates over the prudential reduction in borrowing, while
for su?ciently high levels of volatility this latter efiect tends to be relatively more
important. This implies a non-monotonic relationship between revenue volatility
and default risk, as observed in the data.
The remainder of the chapter is structured as follows. Section 1.2 analyzes the
empirical relationship between volatility and default risk as well as risk premia. In
section 1.3 I develop the model economy and discuss brie y the solution algorithm.
Section 1.4 presents the main quantitative results related to the impact of volatility
on risk premia and average debt holdings. I also present several robustness checks.
Section 1.5 concludes.
1.2 Empirical Evidence
This section presents empirical evidence of a non-monotonic relationship be-
tweenrevenuevolatilityand sovereign risk premia, as wellasmeasures ofdefault risk.
As discussed in the previous section, Figures 1.4 presents a non-monotonic relation-
ship between sovereign risk premia - measured by the average JPMorgan EMBIG
10
spreads over US treasury bills between 1998 and 2000 - and the volatility of govern-
ment revenues. Figure 1.5 shows a similar relationship for default risk measured by
the Institutional Investors Rating (IIR).7 This indicator has been used recently by
Reinhart, Rogofi and Savastano (2003) and Reinhart and Rogofi (2004) as a measure
of default risk. It is a rating on a scale from 0 to 100, where higher values represent
less risk. In order to reduce the in uence of potential outliers in the regressions
when using spreads, I transform the risk premium to log(1 + si=10000), where si
is the spread reported in basis points, so that the dependent variable is measured
approximately in percentage points. The econometric estimation corresponding to
the quadratic flt represented in the Figure 1.4 is shown in the flrst column of Table
1.1. The estimated coe?cients on revenue volatility and its quadratic term are pos-
itive and negative, respectively, and statistically signiflcant at conventional levels of
confldence. In particular, the estimates imply that for a coe?cient of variation of
government revenues greater than 0.121 the efiect of an increase in revenue volatil-
ity turns negative. A similar result using the IIR as dependent variable is shown
in column 2. In addition, revenue volatility and the squared term alone explain
more than 40 percent of the total cross-section variation in spreads in the sample.
One immediate concern from a visual inspection of Figure 1.4 is that the results
might be driven by Korea and China. Both countries present very high levels of
7The information of government revenues refers to annual series for the central government -
thus it excludes all subnational and government-enterprize revenues. The primary source for these
series are the IMF?s WEO and GFS databases. In some case of missing information the national
source, e.g. ministries of flnance and central banks, is used. The period used to compute the
coe?cient of variation is 1990 to 2004, expressed in percentage points. These data were kindly
provided by Marcelo Oviedo. They are also used by Mendoza and Oviedo (2006). For developed
countries I use EURO-GBI spreads vis-a-vis Germany for the years 1999 - 2000.
11
revenue volatility and very low levels of spreads. While these low spreads might be
explained by the very low levels of debt, possibly due to precautionary savings -
as e.g. Durdu, Mendoza and Terrones (2007) argue - it is important to check the
robustness of the correlation presented in columns 1 and 2. In order to do so, I
re-estimate the quadratic regression dropping both observations. As it can be seen
in column 3 and 4, the results remain signiflcant for the case of the IIR measure,
but the quadratic term is only marginally signiflcant considering the spreads. Alter-
natively, in columns 5 and 6, I estimate the following regression by non-linear least
squares (NLS): Risk = fi + revfl +". A coe?cient signiflcantly less that 1 (greater
than -1) implies a concave relationship between spreads (IIR) and volatility. As the
estimates show, for both dependent variables the coe?cient is signiflcant. In addi-
tion, the linearity hypothesis is rejected in both cases, such that the NLS regressions
indicate a signiflcant concave relationship.
The next four columns of Table 1.1 present estimations based on the inclusion
of additional control variables that have been found to be signiflcant in the literature
on the determinants of sovereign spreads and credit ratings.8 This relatively parsi-
monious model is able to explain a large fraction of the total cross-country variation
in spreads, with an R-squared of around 0.76, and an even better flt for the IIR
(R-squared of 0.84). I include the average in ation rate - deflned as the average of
log(1+in ation) from 1990 to 1999 - given that macroeconomic instability usually
tends to increase risk. This intuition is conflrmed by the estimates in column 7
8See Cantor and Paker (1996), as well as Reinhart (2002) on these issues. All explanatory
variables - except for debt levels - are taken from the World Bank?s WDI database.
12
and 8. In ation is highly signiflcant and positively (negatively) correlated with the
sovereign risk premium (IIR). The estimate implies that an increase in the annual
in ation rate from 2 percent to 12 percent would raise the real cost of borrowing
by around 1 percentage point. In addition, I include the average ratio of central
government expenditures to GDP for the period 1990 - 1999. The estimated coe?-
cient is not signiflcant.9 Another signiflcant variable is the initial GDP per capita in
PPP terms (in logs). This variable is in general included in the literature to proxy
a series of factors, e.g. the quality of institutions, that might be relevant to the
likelihood of default. The estimates show that GDP per capita has a signiflcant and
negative (positive) impact on sovereign risk premia (IIR). Next, given the relevance
that the theoretical and empirical literature on sovereign debt has assigned to direct
sanctions, I include trade openness - measured by the ratio of exports plus imports
to GDP - as a control.10 However, the coe?cient shows is not signiflcant. Regarding
the revenue volatility coe?cients, the estimates remain similar to the previous ones,
with tipping points, where the efiect of increases in volatility on spreads become
negative, are slightly below those estimated in columns 1 and 2. Overall, these es-
timations show that the non-monotonic relationship between spreads and revenue
volatility remain signiflcant when other determinants of risk are included in the re-
gressions.
In column 9, I include the central government?s gross debt to GDP ratio (ex-
9I also estimated alternative speciflcations with other macroeconomic variables such as the
central government budget deflcit or the current account deflcit, but they were also not signiflcant.
10See Bulow and Rogofi (1989) and Rose (2005) on this particular issue.
13
pressed in percentages) as an explanatory variable. In order to reduce endogeneity
problems I use the average debt ratios for the flrst flve years of the 1990?s. The esti-
mates show that countries with higher levels of debt pay a signiflcantly higher inter-
est rate. The estimates imply that a one-percentage point increase in the Debt/GDP
ratio increases the spread by 0.9 percentage points. The result regarding revenue
volatility remains robust. As can be seen in column 10 of Table 1.1, debt levels have
a very signiflcant impact on spreads if we restrict our sample to EME countries,
with the coe?cient almost doubling in size. While the other controls turn out to be
not signiflcant in this subsample, the non-monotonic relationship between revenue
volatility and spreads continues to be signiflcant. Again, tipping points are well
within the range of revenue volatility observed in the data. Thus, the result is not
driven by a systematic difierence between these two groups of countries.
So far, the evidence indicates a signiflcantly concave relationship between
sovereign risk and volatility in the sample. While in most speciflcations this re-
lationship is actually non-monotonic, the presence of two in uential observations
seems to be partially driving these results. In what follows, we use alternative
measures of volatility and default risk to further analyze the robustness of these
results.
Table 1.2 shows a series of robustness tests concerning the non-monotonic rela-
tionship between revenue volatility and default risk. First, I used the volatility of the
business cycle - computed as the standard deviation of the Hodrick-Prescott flltered
series of annual real GDP - as an alternative measure of volatility. As the flrst col-
umn shows, the non-monotonic correlation between volatility and spreads continues
14
Table
1.1:
Spreads
and
Rev
en
ue
Volatilit
y-
Cross-section
Regressions
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Estimation
Metho
d
OLS
OLS
OLS
OLS
NLS
NLS
OLS
OLS
OLS
OLS
Dep.
Variable
Spreads
(log)
IIR
Spreads
(log)
IIR
Spreads
(log)
IIR
Spreads
(log)
IIR
Spreads
(log)
Spreads
(log)
Sample
Excluding
China
Excluding
China
Excluding
China
Excluding
China
EME
and
Korea
and
Korea
and
Korea
and
Korea
only
Rev
enue
Volatilit
y
1.468***
-11.698***
0.926***
-10.827***
0.044***
-0.167***
0.625**
-3.983**
0.817***
0.628*
(0.315)
(2.115)
(0.319)
(2.932)
(0.007)
(0.021)
(0.301)
(1.569)
(0.265)
(0.312)
Rev
enue
Volatilit
ysq.
-6.101***
50.234***
-1.568
41.946**
-3.036*
25.847***
-3.727**
-3.320**
(1.907)
(11.202)
(1.797)
(18.732)
(1.550)
(8.094)
(1.643)
(1.540)
In ation
0.011***
-0.059***
0.010***
0.004
(0.002)
(0.012)
(0.002)
(0.003)
Gov.
Exp
enditure/GDP
0.000
0.002
0.000
0.000
(0.001)
(0.004)
(0.001)
(0.001)
GDP
per
capita
-0.007**
0.105***
-0.006*
-0.003
(0.003)
(0.017)
(0.003)
(0.003)
Openness
0.000
-0.001
0.000
0.000
(0.000)
(0.001)
(0.000)
(0.000)
Debt/GDP
0.009**
0.016***
(0.004)
(0.004)
Constan
t
-0.027***
1.101***
-0.014
1.082***
-0.844***
-1.018***
0.032
0.032
-0.011
-0.025
(0.009)
(0.067)
(0.009)
(0.079)
(0.021)
(0.114)
(0.025)
(0.175)
(0.033)
(0.026)
Observ
ations
44
40
42
38
42
38
44
40
44
21
R-squared
0.44
0.43
0.53
0.44
0.51
0.43
0.72
0.84
0.75
0.71
Tipping
poin
t
0.121
0.116
0.295
0.129
0.102
0.077
0.109
0.094
Max.
insample
0.180
0.180
0.152
0.141
0.180
0.180
0.180
0.180
Robust
standard
errors
in
paren
theses,
*,
**,***
signiflcan
tat
10,
5and
1p
ercen
t,
resp
ectiv
ely
.Dep
enden
tv
ariable
is
log(1+spread/10000).
GDP
per
capita
and
in ation
are
in
logs.
Columns
(5)
and
(6)
sho
w
the
estimate
of
fl
based
on
anon-linear
least
squares
estimation
of
y=
fi+
rev
fl
+
",
with
yeither
log
(1
+
spr
ead=
10000)
or
IIR
and
rev
is
rev
en
ue
volatilit
y.
15
to hold using this alternative indicator of volatility. It is important to point out that
for this indicator there are no clear outliers. One indication of this is the fact that
the median business cycle standard deviation in the sample is 0.047, very close to the
mean of 0.045. In addition, all countries lie within a two-standard deviation interval
from the mean. Column 2 explores an instrumental variable estimation using the
standard deviation of terms-of-trade growth as an instrument for the standard devi-
ation of the GDP business cycle. Correcting for potential endogeneity yields similar
results. Again, the quadratic term is negative and signiflcant. The tipping point
where the efiects of volatility turn negative is also considerably below the sample
maximum, so that it provides further evidences of the non-monotonic relationship
between default risk and volatility.
Next, as an additional robustness check, I consider a series of sovereign debt
ratings commonly used in the literature as proxies of default risk. Reinhart (2002)
shows that downgrades in sovereign credit ratings predict future defaults, in con-
trast to currency crises, which tend to take place before downgrades occur. Also,
Reinhart, Rogofi and Savastano (2003) show that sovereign credit ratings help sta-
tistically to separate defaulting from non-defaulting countries ex-ante. In particular,
I use ratings from Moody?s, Standard&Poors, and Fitch, for a similar period as the
spreads used above (1997 - 2000). The three rating agencies use a letter rating which
I recoded into 17 categories from 0 to 16. This procedure is the standard practice in
the literature (see e.g. Cantor and Packer, 1996; as well as Reinhart, 2002). Given
that the coding for all ratings is such that higher values represent a higher degree of
16
creditworthiness, the coe?cients of revenue volatility and its quadratic term should
now respectively be negative and positive if the relationship between default risk
and volatility established for spreads continues to hold, as for the case of the IIR
used in Table 1.1.
In columns 3 to 5, I present the estimations for these three ratings. Given
that OLS estimates do not exploit the fact ratings are actually ordinal variables
that take only 17 difierent ordered values instead of a continuous outcome, I esti-
mate an ordered probit model which accounts for this fact correctly. As can be seen
in Table 1.2, there is a signiflcant U-shape relationship between ratings and revenue
volatility for all three ratings. Furthermore, as column 6 shows, the same is true
if the sample includes only EME. In addition, columns 7 and 8 show that if the
ordered probit model is estimated considering the standard deviation of the busi-
ness cycle or the standard deviation of terms-of-trade growth, the non-monotonic
relationship between volatility and default risk remains signiflcant. Finally, the last
column shows that the logarithmic transformation for spreads does not have an im-
pact on the signiflcance of the non-monotonicity in the correlation between spreads
and revenue volatility. When estimated in levels, the non-monotonic relationship
continues to hold.
While the previous evidence is quite compelling, it considers only the cross-
section dimension and does not exploit any variation across time, basically due to
the fact that there are no su?ciently long time series of bond yields and consistent
revenue series for emerging markets. However, in order to analyze the robustness
17
Table
1.2:
Spreads
and
Rev
en
ue
Volatilit
y-
Robustness
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Estimation
Metho
d
OLS
IV
Ordered
probit
Ordered
probit
Ordered
probit
Ordered
probit
Ordered
probit
Ordered
probit
OLS
Dep.
Variable
Spread
(log)
Spread
(log)
Mo
ody?s
Fitc
h
S&P
S&P
S&P
S&P
Spread
(lev
el)
Sample
EME
only
Std
ofGDP
cycle
2.074**
5.094*
-93.030**
(0.773)
(2.682)
(42.714)
Std
ofGDP
cycle
sq.
-24.920***
-57.833**
831.534**
(8.727)
(28.190)
(398.925)
Rev
enue
Volatilit
y
-49.958**
-37.492*
-43.332**
-66.005***
8564.319**
(20.337)
(19.563)
(19.011)
(24.474)
(3193.199)
Rev
enue
Volatilit
ysq.
269.061**
211.355*
229.651**
381.228***
-42488.015**
(109.549)
(110.386)
(100.800)
(120.763)
(16366.125)
Std
TOT
gro
wth
-33.674**(13.997)
Std
TOT
gro
wth
sq.
116.039**(55.317)
In ation
0.012***
0.014***
-0.449***
-0.646***
-0.440***
-0.345**
-0.576***
-0.513***
84.160***
(0.003)
(0.002)
(0.151)
(0.152)
(0.141)
(0.174)
(0.176)
(0.141)
(21.641)
Gov.
Exp
enditure/GDP
-0.001
-0.001
0.047
-0.015
0.034
-0.159**
0.094**
0.071
1.3
(0.001)
(0.001)
(0.037)
(0.066)
(0.037)
(0.073)
(0.038)
(0.050)
(7.097)
GDP
per
capita
-0.003
-0.01
-0.021**
-0.385
-0.020*
0.642
-0.126
-0.836
2.703
(0.006)
(0.009)
(0.010)
(0.575)
(0.010)
(1.242)
(0.662)
(0.674)
(1.621)
Op
enness
-0.008***
-0.007***
0.768***
0.729***
0.798***
0.365
0.745***
0.613***
-77.461***
(0.002)
(0.002)
(0.177)
(0.196)
(0.183)
(0.296)
(0.187)
(0.169)
(26.797)
Constan
t
0.043
-0.026
431.299*
(0.026)
(0.062)
(254.079)
Observ
ations
38
38
42
37
42
20
48
48
44
R-squared
0.79
0.73
pseudo
R-squared
0.26
0.28
0.27
0.20
0.29
0.27
Tipping
poin
t
0.042
0.044
0.093
0.089
0.094
0.087
0.056
0.145
0.101
Max.
insample
0.076
0.076
0.180
0.180
0.180
0.180
0.090
0.200
0.180
White-corrected
robust
standard
errors
in
paren
theses,
*,
**,***
signiflcan
tat
10,
5and
1p
ercen
t,
resp
ectiv
ely
.GDP
per
capita
and
in ation
are
in
logs.
In
column
2,
the
standard
deviation
of
terms-of-trade
gro
wth
and
its
squared
term
are
used
as
instrumen
ts
for
the
standard
deviation
of
the
GDP
cycle
and
its
squared
term.
18
of the results presented in the cross-country analysis so far and to compare them
with the other study in the literature on this topic, we use similar econometric
techniques and the same data as Catao and Kapur (2006) to analyze the shape of
the relationship between volatility and default risk. These authors analyze the im-
pact of macroeconomic volatility on default probabilities applying an event study
approach to defaults and credit events in a panel of 26 EME and developing coun-
tries from 1970 to 2001. They estimate logit models to assess the impact of output
gap volatility or terms-of-trade volatility on the likelihood of observing a default
or rescheduling of sovereign debt. Volatility is measured by the standard deviation
of these variables using 10-year rolling windows previous to the year under con-
sideration.11 There are two aims of this exercise. First, given that the time-series
availability for spreads and revenue volatility is limited, in order to exploit the time
series dimension, one has to rely on ex-post episodes as measures of default risk
and alternative measures of volatility. Thus, this approach enables us to conduct
a further robustness check in terms of considering an alternative deflnition of the
dependent variables as well as additional measures of volatility. Second, this anal-
ysis allows us to check the robustness of our non-monotonicity flnding using the
same data and methodology as the only other study in the literature. Therefore, I
augment the econometric model of Catao and Kapur (2006) to include a quadratic
term of their volatility measures, and ask whether this term has a negative and
signiflcant impact on the probability of default, as the cross-section evidence above
11See their paper for more details and descriptive statistics on the dependent and independent
variables used below. All explanatory variables are lagged one period in order to reduce endogeneity
problems.
19
indicates. The logit estimates are presented in Table 1.3. All regressions include
the international interest rate, which enters positively and highly signiflcantly in
all speciflcations. This re ects the common wisdom that defaults are more likely
during periods of tight international liquidity. In addition, exports as a fraction
of GDP (which is included to capture the potential cost of trade sanctions in the
event of default), shows the expected negative sign in most speciflcations although
it is only signiflcant in two of them. With respect to debt indicators, debt as a
fraction of exports, as well as debt service as a fraction of exports, perform better
than debt-to-GDP ratios.
Regarding the variables of interest, the linear speciflcation in column 1 shows
that the volatility of the terms of trade has a positive efiect on the probability of
default. In column 2, I add the quadratic term of terms-of-trade volatility, which
comes in highly signiflcant with the expected negative sign. In addition, the flt
of the regression measured by the pseudo-R-squared improves from 0.15 to 0.18.
Hence, this evidence is consistent with the cross section regressions presented in
Table 1. Adding the debt to exports ratio or estimating the model using random
efiects (column 3) yield similar results. While the implied tipping points at which
the efiects of terms-of-trade volatility on the probability of default become negative
is well below the in-sample maximum, there could be concerns that the results are
driven by some extreme observations. In column 4, I re-estimate the model exclud-
ing all observation with a 10-year rolling standard deviation above 50%. Again,
the non-monotonic shape is signiflcant in this subsample with an estimated tipping
point at around values of 23%. Similar results are obtained when estimating con-
20
sidering the standard deviation of the residuals of a growth forecasting regression.
These are the residuals from regressing real GDP growth on two lags and a seg-
mented time trend (with a break in 1974) as in Ramey and Ramey (1995). Finally,
the results are robust to including the debt service to exports ratio and the devia-
tion of the real exchange rate from its Hodrick-Prescott trend as additional controls.
Summing up, the evidence presented in this section shows that there is a
non-monotonic relationship between default risk/risk premia and revenue volatility.
For low levels of volatility, an increase in volatility is associated with an increase
in spreads and the perceived default risk. However, for su?ciently large levels
of volatility, this relationship reverts. This empirical relationship is found using
difierent measures of default risk, as well as alternative econometric methods and
measures of volatility.
In terms of the discussion presented in the introduction, this empirical fact
can be interpreted as a trade-ofi between precautionary savings motives and the
increased risk of default due to a higher variance of the relevant shocks. According
to Eaton and Gersovitz (1981), a higher volatility of the relevant income process
increases the cost of exclusion from credit markets. Therefore, default becomes less
attractive for countries that face higher volatility. However, if this is the only chan-
nel through which volatility afiects default incentives, the risk premium should be
a decreasing function of volatility. This clearly is not observed in the data. Alter-
natively, as argued by Catao and Kapur (2006), volatility increases potentially the
likelihood of receiving a very bad draw and therefore increases the fraction of the
21
Table
1.3:
Logit
Regressions
-Default
Ev
en
ts
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Std
TOT
gro
wth
?0.5
Random
Efiects
US
Real
Interest
Rate
0.421***
0.419***
0.417***
0.363***
0.437***
0.432***
0.482***
0.495***
0.463***
(0.128)
(0.123)
(0.121)
(0.092)
(0.104)
(0.132)
(0.150)
(0.170)
(0.148)
Exp
orts/GDP
-0.031*
-0.025
-0.009
0.001
-0.01
-0.008
-0.016*
-0.012
-0.003
(0.018)
(0.017)
(0.009)
(0.008)
(0.013)
(0.009)
(0.010)
(0.010)
(0.006)
Debt/GDP
0.011
0.009
(0.008)
(0.008)
Std
TOT
gro
wth
3.078***
9.837***
10.706***
32.561***
10.775***
9.228***
(0.873)
(2.848)
(3.033)
(11.806)
(3.891)
(3.049)
Std
TOT
gro
wth
sq.
-9.036***
-9.707***
-69.905**
-9.574*
-7.899**
(2.983)
(3.263)
(31.126)
(5.242)
(3.533)
Debt/Exp
orts
0.003**
0.003***
0.003*
0.000
0.003***
0.000
0.003*
(0.001)
(0.001)
(0.001)
(0.002)
(0.001)
(0.002)
(0.001)
Std
Gro
wth
Forecast
Residuals
227.32***
245.63***
280.45***
(83.35)
(86.30)
(88.21)
Std
Gro
wth
Forecast
Residuals
sq.
-9932.45*
-12031.27*
-16950.85**
(5866.97)
(6399.67)
(0.793)
Debt
Service/Exp
orts
2.927***
3.256***
3.033***
(0.751)
(0.824)
(0.852)
Real
Exc
hange
Rate
Gap
10.283***(2.737)
Constan
t
-4.677***
-5.426***
-6.111***
-7.776***
-6.250***
-6.919***
-5.505***
-6.788***
-17.798***
(0.790)
(0.864)
(0.989)
(1.507)
(0.924)
(1.217)
(0.815)
(1.196)
(3.318)
Observ
ations
584
584
584
564
584
583
543
543
543
pseudo
R-squared
0.16
0.18
0.18
0.19
0.26
0.17
0.27
0.34
Tipping
poin
t
0.544
0.551
0.233
0.562
0.563
0.011
0.010
0.008
Max
insample
0.958
0.958
0.498
0.958
0.958
0.018
0.018
0.018
White-corrected
robust
standard
errors
in
paren
theses,
*,
**,***
signiflcan
tat
10,
5and
1p
ercen
t,
resp
ectiv
ely
.Column
4presen
ts
random
efiects
logit
estimations.
22
state space where default might be optimal. Clearly, in this case, default should be
monotonically increasing in the variance of shocks. This is also not the whole story,
according to the empirical evidence presented in this section. In the next section, I
present a model that tries to disentangle these efiects in a set-up where both forces
are potentially relevant.
Before flnishing this section, it is useful to discuss some of the evidence re-
garding debt holdings and their relationship with volatility. In this case, although
there is a clearly negative correlation, as shown in Figure 1.3 and also documented
by Mendoza and Oviedo (2006), this empirical evidence is a little bit harder to in-
terpret. Figure 1.3 plots the average debt holdings and revenue volatility for two
distinct samples, rich OECD and Non-OECD countries.12 Clearly, the cross-section
evidence shows a negative correlation between volatility and debt for EME. For the
case of revenue volatility the correlation coe?cient is -0.40, which is signiflcant at
conventional levels. However, in the case of developed countries there is no signifl-
cant correlation between these variables.
This distinct pattern across EME and developed countries can be interpreted
in two ways. First, following Durdu, Mendoza and Terrones (2007), a higher volatil-
ity may induce agents to accumulate less debt due to precautionary saving motives
in the face of incomplete asset markets. This model can explain the data if EME
12Average debt holdings are computed over the period 1990 - 2005. The data are taken from
JaimovichandPanizza(2006). Thisdatabasehasbeenconstructedespeciallytoallowcrosscountry
comparisons and has been used as a primary data source for the 2007 IDB Report on Economic
and Social Progress in Latin America on sovereign debt. The classiflcation of countries is based on
those countries that were members of the OECD in the 1970?s. Therefore, e.g. Mexico is classifled
as Non-OECD, given that it joined the OECD in 1994.
23
face a lot of market incompleteness, while developed countries have access to more
sophisticated forms of flnance and state-contingent assets. Second, in models with
limited commitment on the borrower?s side, debt holdings re ect the minimum be-
tween desired debt levels and potential debt ceilings. In economies that always have
access to capital markets, the government would use debt to smooth revenue shocks
over the business cycle. Given that there would be no commitment problems, the
government would be expected to repay always. Combined with the previous ar-
gument, precautionary motives would also be less severe in this case, given that
permanent access to credit markets would allow the government to borrow more
even during severe recessions. In addition, developed market economies might have
a signiflcantly large menu of contingent flnancing options, which would also tend
to reduce the need for self-insurance, reducing precautionary savings motives and
consequently the link between volatility and precautionary savings. Thus, debt lev-
els primarily re ect debt demand for the OECD countries, and there are no strong
reasons to expect a systematic correlation between volatility and debt levels.13 In
contrast, for the case of EME, the signiflcantly negative correlation could re ect
the fact that volatile economies are more likely to be credit constrained because
they have a higher probability of default. This would be in line with the flnding by
Catao and Kapur (2006) that debt ceilings are negatively correlated with volatility,
while demand for debt increases with volatility. On the other hand, countries with
very high revenue volatility (as could be the case of China, e.g.) might want to
13This refers to eliminating obvious short-run changes in debt levels over the business cycle.
Countries that undergo a large shock are naturally expected to hold more debt than those that
are close to their long-run equilibrium.
24
avoid a debt crisis by holding a more balanced position of net foreign assets. This
precautionary motive might induce countries to demand less debt when they are at
very high levels of volatility. Both of these features are distinct for EME because
they are linked to potential loss of market access due to default risk.
While we do not explore these issues empirically further here, this section
yields the following main conclusions. First, default risk - and as a consequence
interest rates and risk premia - show a positive correlation with revenue volatility.
However, there is a non-monotonic relationship between these variables such that
for su?ciently high levels of volatility, the probability of default - and therefore the
sovereign spread also - actually decrease. Second, debt holdings are decreasing with
volatility in EME, while there is no signiflcant correlation between debt holding and
volatility for developed economies. Finally, it should be mentioned that Reinhart,
Rogofi and Savastano (2003) provide a further stylized fact related to those in es-
tablished above: EME tend to default on lower debt to GDP ratios than developed
countries.
1.3 The Model Economy
This section presents a simple model to analyze the impact of revenue volatil-
ity on risk premia and the probability of default, as well as equilibrium debt levels.
In order to generate endogenous default, I follow the existing literature on sovereign
debt with incomplete markets and limited commitment, such as Arellano (forth-
coming) and Aguiar and Gopinath (2006). Although these models have problems
25
in matching some moments of debt dynamics in emerging markets (especially debt
levels and/or the magnitude of spreads), they do ofier a framework to analyze the
relationship between volatility and default risk. Given that the empirical evidence
presented in section 1.2 shows a non-monotonic relationship between default risk
and volatility, a theory that wants to match this empirical fact has to be capable of
generating a risk premium and default (endogenously) in equilibrium. This is the
main reason to opt for this framework.
In contrast to the standard in the literature, I use a similar set-up to Alesina
and Tabellini (2006) and Mendoza and Oviedo (2006) in which the government?s
objective function difiers from that of the representative household?s. The gov-
ernment?s objective is to smooth its expenditure, which reports no utility to the
households. This implies that the competitive equilibrium in this economy will not
reproduce the social optimum. In addition, I assume that the government might
discount the future at a higher rate, re ecting the fact that political turnover and
instability might induce the government to have a more myopic behavior.
1.3.1 The Household?s Problem
The representative household maximizes the expected discounted utility value
of consumption and leisure given by:
E0
1X
t=0
flt
h
Ct ? L1+?t1+?
i1??
1?? ; (1.1)
where fl is the subjective discount factor, and ? is the coe?cient of relative
26
risk aversion. The relationship between consumption (Ct) and labor (Lt) is modeled
using the formulation by Greenwood et al (1988) which is standard in the real busi-
ness cycle literature, given that it has the property that the elasticity of substitution
between consumption and labor is independent of the level of consumption, which
makes labor supply independent of consumption decisions. The elasticity of labor
supply is given by the inverse of ?.
Households also carry out production in the economy according to a linear
production function given by:
Qt = AtLt; (1.2)
where total factor productivity (At) is a random variable.
The household?s budget constraint is given by:
Ct = Qt(1??t); (1.3)
where ?t is an output tax rate set by the government. In particular, I follow
Mendoza and Oviedo (2006) and assume that this efiective tax rate is a combination
of a constant tax rate over time and states and a stochastic revenue shock, such that:
?t = ?ezt: (1.4)
27
The parameter ? is the average revenue as a fraction of GDP, while the zt
shock process is assumed to capture tax as well as non-tax revenue shocks. Given
that in many developing countries non-tax revenues - linked to commodity export
income e.g. - represent an important fraction of total revenues, this shock might
also capture  uctuations in the terms-of-trade.14 The representative household?s
problem is static in nature. The flrst order condition of maximizing (1) subject to
(2) - (4), yields a labor supply given by:
Lt =
?
At(1??t)
?1
?: (1.5)
1.3.2 The Government?s Problem
I assume that the government?s objective is to maximize the expected dis-
counted utility of government expenditure given by:
E0
1X
t=0
(fl?)t g
1??
t
1??; (1.6)
where ? represents the probability that the current government will stay in power.
This parameter is included in order to account for a higher impatience of the gov-
ernment due to political uncertainty. For simplicity, I assume that ? is constant
over time and across states.15
14See Gavin and Perotti (1997) for evidence of the importance of non-tax revenues in Latin
America.
15There is some evidence (e.g. Inter-American Development Bank, 2006) that the probability of
the government being replaced or losing power increases after a default episode. Thus, ? could be
state-dependent.
28
Following the literature on quantitative models of sovereign debt with strategic
default (see e.g. Arellano, forthcoming; as well as Aguiar and Gopinath, 2006), I
assume that international asset markets are incomplete such that the government
can only issue a one-period bond. The outstanding stock of net foreign assets is
denoted by b, so that a negative value of b represents the level of outstanding debt.
In addition, I assume that the government cannot commit to repay its debt, so that it
will only repay its debt if the expected discounted utility value of doing so is greater
than the value of defaulting on its obligations. In particular, the value function
of the government?s maximization problem is given by V(b;z;A) = maxfV D;V Rg,
where V D is the value of default and V R the value of repayment. The latter is given
by:
V R(b;z;A) = max
n
u(g)+fl?
Z
V(b0;z0;A0)dF(z0;A0jz;A)
o
(1.7)
s.t. g +q(b0;z;A)b0 = ?ezQ+b; (1.8)
lim
T!1
(fl?)TbT+1 ? 0; (1.9)
where q is the bond price, u(:) is the CRRA utility function, F(:) is the joint
cumulative distribution of the exogenous state variable, which is assumed to follow
a Markov chain, and the second condition is a standard non-Ponzi game condition.
29
Alternatively, the government defaults on its debt. When it does so, the
government is excluded for an uncertain period of time from international credit
markets, so that the government has to rely exclusively on national tax revenue
to flnance expenditure. In addition, I assume that default involves disorder and
potential deadweight losses due to sanctions imposed by creditors, so that revenues
follow a difierent stochastic process under default, denoted by h(Q). Also, if the
country is in a state of default, there is an exogenous re-entry probability fi, so
that the country on average is excluded only 1fi years from credit markets. If the
government reains access to international capital markets after a default, it does so
with zero debt. The value of default under these assumptions - which are standard
in the literature - is given by:
V D(z;A) = u?h(Q)?+fl?
(Z ?
fiV(0;z0;A0)+(1?fi)V D(z0;A0)
?
dF(z0;A0jz;A)
)
(1.10)
1.3.3 Creditors
As it is standard in the literature, international creditors are modeled as risk-
neutral agents. They have access to an international capital market in which they
can trade a risk-free bond at the interest rate r? which for simplicity is assumed to
be constant over time and states of nature. In addition, there is perfect competition
among creditors, such that their expected proflts are zero. Under these conditions
30
it is straightforward to show that the following condition has to hold:
1+rt = 1+r
?
1??(bt+1;zt;At); (1.11)
where r is the interest rate of the bond, given by rt = 1qt ?1, and ?(bt+1;zt;At) is
the endogenous default probability, deflned as:
?(bt+1;zt;At) = E
h
D(bt+1;zt+1;At+1)jbt;zt;At
i
; (1.12)
where D is an indicator function deflned as:
D(bt;zt;At) =
8
>>>>
>><
>>>
>>>
:
1 if V D(zt;At) > V R(bt;zt;At)
0 otherwise
(1.13)
This equation just states that the expected returns of investing in the risk free
asset and making a risky loan to the agent have be be equal. It also shows that
the interest rate is a function of the level of net foreign assets (bt+1), as well as the
current state of revenues given by (zt;At). A higher expected probability of default
clearly implies a higher risk premium. Also, given that the default probability is
decreasing in bt+1, rt will be a non-increasing function of bt+1. Clearly, the risk
premium or spread is analogously deflned as:16
s(bt+1;zt;At) = ?(bt+1;zt;At)1??(b
t+1;zt;At)
(1.14)
16The spread s is computed using the multiplicative formula: (1+r?)(1+s) = 1+r.
31
1.3.4 Some Intuition
In order to provide some intuition of the forces present in the model, this
subsection presents a brief discussion on how the incentives to borrow and default
might change with volatility.
Assume that government revenues can be characterized by the following pro-
cess: Tt = Tezte? 22 , with zt being an i.i.d. shock, such that zt ? N(0; 2), which
implies the expected value of Tt always equals T. This allows us to analyze the
efiects of a mean-preserving spread. In addition, assume that fi = 0, such that there
is permanent exclusion from credit markets after a default, but there is no further
cost of default, i.e. h(Qt) = Tt.17 Under these assumptions, the value of default will
be given by:
V D = E0
1X
t=0
fl T
1??
t
1?? =
(Te?12? 2)1??
(1?fl)(1??) (1.15)
In addition, let us assume that under repayment, the government is able to
complete smooth its expenditure, such that each period the government pays the
amount of R = ?rb of interest payments.18 Under these assumptions, the value of
repayment will be given by:
17Aguiar and Gopinath (2006) use a very similar set-up to analyze sustainable debt levels in
their model.
18We are assuming b < 0 to make the discussion interesting. Observe also that even if it were
feasible, complete expenditure smoothing will not be optimal in our calibrated exercises presented
in the subsequent sections due to the relative impatience we assume.
32
V R = (T ?R)
1??
(1?fl)(1??) (1.16)
The government will default on its debt, only if the V D > V R, i.e.:
R
T > 1?e
1
2? 
2 (1.17)
This equation has a series on interesting implications. If the variance of rev-
enues tends to zero, the right-hand side of the equation converges to zero, which
means that the government would default whenever it has to make any positive
repayment or if it holds a positive amount of debt. The rationale for this is that the
utility cost of the exclusion punishment declines when volatility declines and there-
fore the government would be more tempted to default on its obligations. Consider-
ing the opposite case, when the variance of revenues tends to inflnity, the right-hand
side of the equation converges to one. This implies that the government would not
default on any debt level below the level at which repayment would compromise
all available resources.19 Thus, as also suggested by Eaton and Gersovitz (1981),
for given levels of debt and interest rates, the probability of default is a decreasing
function of the volatility of shocks in this set-up. In addition, the equation also
shows that the speed at which the default probability decreases is a positive func-
tion of the coe?cient of relative risk aversion (?). Observe that if the government
is not risk averse, it would default on any positive level of debt.20 If we assume
19Of course, other optimality considerations and prudential behavior would probably prevent
the government to borrow up to this level of debt in the flrst place.
20Clearly, in our model without risk aversion the only reason to borrow for the government is it
relative impatience that induces it to prefer current expenditure to future expenditure.
33
that there is an additional cost of default that reduces the revenues in the state of
default by a fraction ?, such that h(Qt) = (1??)Tt, it is straightforward to show
that the right-hand side of the equation becomes: 1?(1??)e12? 2. This shows that
the level of debt that the government can support without defaulting increases with
the severity of the punishment (?).21
However, several of the strong assumptions made so far are unlikely to hold.
In particular, the government might not be able to smooth consumption completely
under repayment. In this case, the value of repayment will be afiected negatively
by volatility and therefore default could act as partial insurance mechanism. As
Eaton and Gersovitz (1981) also argue, under uncertainty volatility has an ambigu-
ous efiect. On the one hand, it makes it more attractive to honor the current debt,
given the possibility that tomorrow?s revenue income will be low and therefore ex-
penditures have to be cut. On the other hand, if the economy is hit by a larger
shock today, default becomes more attractive. Therefore, the relationship between
volatility and default risk will depend on which efiect dominates. It is interesting
to point out that this trade-ofi will also depend on the degree of impatience of the
government. If the future is discounted at a higher rate - due to a low probability
of staying in power, for example - the current gain of not having to repay the debt
will become more attractive than the inter-temporal future gains from repaying the
debt and conserve market access.
A second efiect that could potentially afiect the relationship between volatility
21This point is also made by Aguiar and Gopinath (2006).
34
and default risk is related to the impact of an increase in volatility on the demand
for assets in incomplete market economies under uncertainty. In particular, given
that I assume a CRRA utility function, it is well-known that this type of preferences
generate prudential behavior. Thus, self-insurance via precautionary savings could
reduce the demand for debt and therefore it could lower the risk of default. Further-
more, given that the utility function exhibits a positive third derivative (u000 > 0)
a \natural" debt limit as proposed by Aiyagari (1994) could arise. Let us assume
for a moment that there is no commitment problem. In this case, the government
solves the the following problem:22
maxE0
1X
t=0
(fl?)tu(gt) (1.18)
s.t. gt +qtbt+1 = Tt +bt; (1.19)
Given that for utility functions that exhibit prudence, the following condition
holds: limg!0 u0(g) = 1, gt ? 0 has to hold. Marginal utility becomes very large
as expenditure tends to zero, so that the government would always avoid getting
close to very low levels of expenditure. Combining this condition with the budget
constraint, the fact that qt = 11+r?, and solving it forward, implies that:
bt+1 ??Tminr? ; (1.20)
22In addition, it is assumed that a non-Ponzi game condition also holds.
35
where Tmin is the lowest possible realization of the revenue process. The right-hand
side of this equation is the natural debt limit. Basically, the equation states that in
order to avoid very low levels of expenditure, the government would always choose
a debt level that would allow to service it under the worst possible realization of
future revenues which is to receive the lowest possible draw forever.
Now suppose that Tt follows the process specifled above: Tt = Tezte? 22 , with
zt ? N(0; 2) and i.i.d. If zt is approximated using a discrete symmetric grid
fzmin;:::;zmaxg, with zmin = ?zmax < 0, this will translate into a grid for revenues
given by fTmin;:::;Tmaxg. An increase in the variance of zt will reduce the value
of Tmin. In the limit, as  2 ! 1, Tmin will converge to 0. Thus, the government
would never hold a negative level of net assets.
Although this previous debt limit was derived for the case where default does
not occur, it is straightforward to show that a similar argument can be extended
to the case without commitment. As Eaton and Gersovitz (1981) show, in the
sovereign lending problem outlined at the beginning of this subsection there exists
a flnite net foreign asset ceiling, bt+1 > ?1 beyond which lender would not extent
further credit, because the sovereign borrower would default with probability 1.23
The proof of this proposition follows from the fact that V R is increasing in the level
of net foreign assets (b), i.e. it is a decreasing function of the outstanding debt level,
while V D does not depend on the debt level. Therefore, the probability of default ?
is a monotonically increasing function of debt. Under these conditions, the sovereign
23The subindex is used to highlight the fact that this ceiling will depend on the state of the
economy.
36
faces two possible constraints: the natural debt limit and the credit ceiling. Thus,
the following has to hold:
b0 ? max
h
? Tminr? ;b0(b;T)
i
(1.21)
Independently from the efiects of volatility on the credit ceiling, in the limit
the natural debt limit will tend to zero, while b0(b;T) ? 0. Therefore, for extremely
high levels of volatility, the demand for debt will converge to 0. Given that as debt
converges to 0, the risk premium has to fall, in the limit the risk premium converges
to 0. However, at intermediate levels of volatility, the efiect of volatility on the credit
ceiling might be the relevant restriction, and therefore we could observe higher risk
premia and tighter ceilings. The next subsection will explore these issues further
using numerical methods.
1.3.5 Calibration
Tax shocks and the TFP process are modeled jointly as a VAR(1) process,
similar to Mendoza and Oviedo (2006). I use the TFP series for Argentina from
1960 to 2003 from Fern?andez-Arias, Manuelli and Blyde (2006) and the ratio of
central government revenues to GDP from the GFS database and the Ministry
of Economics and Production.24 Both series were de-trended using the Hodrick-
Prescott fllter with a smoothing parameter of 6.25, as recommended by Uhlig and
Ravn (2002). Given that revenue data are available only since 1970, the sample
24For 1989, revenue data are missing. In this case I interpolated linearly between the surrounding
observations.
37
period is 1970 - 2003. Let the vector xt be given by de-trended TFP (tfpt) and the
de-trended revenue to GDP ratio (revt), such that xt = (tfpt;revt)0. I estimate the
following VAR(1): xt = ?xt?1 + "t, where ? is a coe?cient matrix of dimension 2
? 2 and "t is a white noise error vector with variance-covariance matrix ?. The
resulting estimates and variance-covariance matrix of residuals is given by (standard
errors are in parenthesis):
^? =
0
BB
BB
BB
BB
BB
BB
BB
@
0:8583 1:4052
(0:1271) (0:8274)
?0:0309 0:6927
(0:0251) (0:1634)
1
CC
CC
CC
CC
CC
CC
CC
A
; ^? =
0
BB
@
0:002664 0:000266
0:000266 0:000104
1
CC
A (1.22)
As I solve the model using discrete optimization methods, the VAR process
has to be translated into a discrete approximation. Therefore, I use the quadrature
procedure by Tauchen and Hussey (1991) to approximate the continuous VAR by
a discrete Markov chain. In particular, I use 25 pairs of realizations of TFP and
revenue shocks (5 difierent realizations for each particular shock). Given that the
ofi-diagonal elements of ^? are marginally signiflcant, I set them equal to zero. The
average revenue to GDP ratio ? is set equal to 16.6%, which is the sample mean
over the period considered above.
Several of the remaining parameters are drawn from the existing literature on
38
economic  uctuations in EME. As Mendoza (1991), I set the coe?cient of relative
risk aversion ? equal to 2 and the labor elasticity parameter ? equal to 0.455,
which are both standard in the literature. For the probability of redemption fi, I
use the historical evidence presented by Tomz and Wright (2007), who estimate an
average exclusion duration of around ten years. Thus, fi is set equal to 0.10. This
parameter is somewhat smaller than alternative values used in the literature (e.g.
Aguiar and Gopinath, 2006) based on Gelos et al (2004) estimate of an average
exclusion duration of 2.5 years after a default in the 1990?s. However, Gelos et
al (2004) focus on a very short time period for such a relatively low probability
event as a default. Also, the latest default by Argentina indicates that exclusion
from international capital markets might be signiflcantly longer; since defaulting at
the end of 2001, Argentina has not been yet able to re-access international capital
markets, as of 2007. Thus, I prefer the estimate of Tomz and Wright (2007), which
is also consistent with the flndings of Arraiz (2006) of long periods of exclusion from
credit markets.
The risk-free real interest rate r? is set to 2.5% per annum, which is the
average ex-post real interest rate on 10-year US Treasury Bonds for the period
1997 - 2006. The discount rate fl is set equal to 1=(1 + r?), i.e. 0.9756, while the
probability of staying in power ? is calibrated using the information from Alesina et
al (1996) regarding the unconditional frequency of a major change in the executive
in developing countries over the period 1950 - 1982 for 108 countries. This includes
all \irregular" changes as well as \regular" changes. The resulting value for ? is
0.735. Thus, the resulting efiective discount factor of the government is around
39
0.72. While this value is low, it is relatively high compared to many models in
the literature; Aguiar and Gopinath (2006) use a discount rate of 0.8 calibrated to
quarterly data which would imply a rate of around 0.41 for annual data. Similarly,
Cuadra, Hatchondo and Sapriza (2007) model default with two difierent types of
politicians that difier in their degree of impatience with quarterly discount rates are
0.9 and 0.6, respectively, which again result in lower annual discount factors than
the one used in this paper. Arellano (forthcoming) uses a rate of 0.953 for quarterly
data, resulting in an annual discount rate of 0.82.25
Finally, with respect to the evolution of revenues under default, I follow Arel-
lano (forthcoming) and model the function h(:) as:
h(Q) =
8
>><
>>:
(1??)E(?ezQ) if ?ezQ ? E(?ezQ)
(1??)?ezQ else
(1.23)
The empirical motivation of this way to model revenues under default comes
from the fact during the latest default episode in Argentina, output has remained
below trend for almost 4 years while revenues have followed a similar path. From a
mechanical point of view, the asymmetry between the revenue process under default
and repayment increases the probability of default, given that default reduces the
volatility of shocks. The parameter ? is calibrated in order to match a probability
of default of 3% per annum, which is also targeted by Arellano (forthcoming). The
25From a quantitative point of view, most of these models need low discount factors to create
enough default episodes. This is also true in other models of lack of commitment, like Alvarez and
Jermann (2001), who analyze the asset pricing implications of an endogenous incomplete markets
model using a relatively low discount factor to make incentive compatibility constraints tighter
and match equity premia for the US.
40
Table 1.4: Baseline Parameter Values
Parameter Value
fl 0.9761
r? 0.025
fi 0.10
? 0.455
? 2
? 0.166
? 0.735
? 0.031
resulting value for ? is 0.031, which is equal to the parameter chosen by Arellano
(forthcoming) and close to the relative output loss of 2% per year estimated by
Chuhan and Sturzenegger (2005). All parameters are presented in Table 1.4.
1.4 Results
This section presents the main results from simulating the model outlined in
section 1.3. The solution algorithm is presented in the appendix. First, I compare
some moments of the ergodic distribution to the sample moments for Argentina.
The flrst column of Table 1.5 shows the sample moments for Argentina, using an-
nual data from 1980 to 2001, except for the spreads, which come from JPMorgan?s
EMBI spreads for sovereign debt and which are only available since 1993. The cor-
relations and standard deviations refer to HP-flltered series. Finally, the debt series
I consider is the net external debt of the central government, from which I subtract
net international reserves, so that the debt aggregate is more similar to the one used
in the model, as there are no debt holdings at the national level nor by the private
sector in the model, so that the variable b refers to net foreign assets.
41
As shown in Table 1.5, the model matches the sign of the empirical correlation
between output and spreads, as well as the trade balance, although the absolute
values are too small. The flnding that model interest rates are higher when out-
put is below trend and that capital  ows are pro-cyclical are standard results in
this literature (see Arellano, forthcoming; and Aguiar and Gopinath, 2006). These
results are driven by the fact that in the model the incentive to default is higher
during bad times. Therefore, when output and revenues are low, spreads will be
higher, creating a disincentive to borrow. In contrast, when the model is solved
assuming no commitment problems, such that the sovereign always repays the debt
(column 3), capital  ows are counter-cyclical, given that the government borrows
from abroad during recessions to smooth expenditure and repays during booms,
which is the standard result in small open economy models (see Mendoza, 1991). In
addition, the baseline model generates an average debt-to-revenue level of around
23%. While this is around a third of the observed average debt-to-revenue level,
it should be taken into account that the model allows only for one-year bonds. In
the case of Argentina, e.g. short-term debt represented less than 1/4 of its total ex-
ternal obligations in the sample period. In addition, observe if the sovereign could
fully commit to repay in all states of nature, debt holdings would be much higher,
precisely more than 21-times the average revenue, which is completely at odds with
the empirical evidence. In addition, it is interesting to point out that the model is
able to create a relatively large spread.
Next, in order to analyze how volatility afiects spreads in the model, I simulate
42
Table 1.5: Sample and Simulated Moments
Moments Argentina Model Full Commitment Model
corr(output;spread) -0.71 -0.09 0.00
corr(output;tradebalance) -0.88 -0.17 0.57
E(b=revenue) -0.97 -0.23 -21.81
E(spread) (bps) 783 418 0.00
 (spread) 0.09 0.08 0.00
Default Probability 3.00% 3.01% 0.00%
the model for difierent levels of macroeconomic volatility. In order to do so, I
hold flxed the coe?cient matrix (?) of the joint TFP and revenue shock process
remains the same, but multiply the variance and covariance matrix (?) by a factor
that ranges from 0.25 to 3, which is close to the variation in the data used in
section 1.2. The resulting average spread levels for difierent levels of volatility are
presentedinFigure1.6. Asthisgraphshows, themodelgeneratesthenon-monotonic
relationship between the spread and volatility observed in the data. The intuition
for this result is the following. At relatively low levels of volatility, an increase in
volatility primarily raises the probability of default, making default more likely for
every level of debt. In addition, the incentives provided by the threat of exclusion at
low levels of volatility are relatively low, and therefore the reduction in the demand
for debt due to precautionary motives is relatively low. However, for su?ciently high
levels of debt, the exclusion from capital markets in the event of default becomes
more costly given the large  uctuations in the provision of public goods that would
take place under autarky. Consequently, the government would borrow less, which
makes it less likely that it would end up with a risky debt level on which the
incentives to default are high.
43
Figure 1.6: Average Simulated Spreads and Volatility
0 0.5 1 1.5 2 2.5 30
100
200
300
400
500
600
Variance relative to baseline
Average Spread (bps)
 
 
Thispreviousintuitioncanbecorroboratedbytherelationshipbetweenvolatil-
ity and average asset holdings and credit ceilings, presented in Figure 1.7. This flgure
shows the average level of asset holding as a fraction of revenues for difierent levels
of volatility. In addition, it also reports the ceiling - deflned as the level of assets
beyond which creditors are not willing to extend any further funds, given that the
sovereign would default under all states of nature. Clearly, this ceiling is difierent
for each point in the state space. The one reported in the graph refers to the av-
erage level of revenues in the model which does not change wit the increase in the
variance of shocks. As Figure 1.7 shows, both average debt levels and the credit ceil-
ing decrease initially with volatility, i.e. for higher levels of volatility the sovereign
holds less debt and creditors are also willing to extend less credit. While the flrst
fact can be explained by the standard precautionary savings result in incomplete
44
asset market models (as in Mendoza and Oviedo, 2006), the lower supply of credit
is explained by the fact that an increase in volatility makes default a more likely
outcome. However, for su?ciently high levels of volatility, average asset holdings
continue to decrease, while the credit ceiling becomes relatively less sensitive to in-
creases in volatility, i.e. the slope of the credit-ceiling curve is smaller for higher
levels of volatility. Thus, at su?cient high levels of volatility, precautionary savings
tend to dominate the dynamics of debt, making default less likely and therefore
dampening the efiects of volatility on credit ceilings.
Figure 1.7: Average Net Asset Holdings, Debt Ceiling and Volatility
0 0.5 1 1.5 2 2.5 3?0.32
?0.3
?0.28
?0.26
?0.24
?0.22
?0.2
Variance relative to baseline
 
 
Average asset holdings (E[b])
ceiling evaluated at average revenues
In Figure 1.8, I present the estimated coe?cients of the VAR process for dif-
ferent levels of volatility, to check whether any of the results are driven by approxi-
mation errors in the Tauchen and Hussey (1991) discretization procedure. As it can
45
be seen, the estimated autoregressive coe?cients do not vary signiflcantly with the
relative variance, such that the discretization process seems to be a relatively good
approximation even for high levels of volatility.
Figure 1.8: Estimated VAR coe?cients
This section so far has shown that the empirical non-monotonic relationship
between spreads and volatility can be explained by a model with incomplete asset
markets and default risk based on the trade-ofi between precautionary motives and
increased default risk due to higher volatility. Next, I present some more intuition
for this result based on a sensitivity analysis.
A key parameter related to the intuition for the non-monotonic relationship
between spreads and volatility presented in the previous section is the coe?cient of
46
relative risk aversion. As discussed in subsection 1.3.4, lower levels of this parameter
should imply that default is more likely for a given level of net foreign assets, given
that the sovereign would be less concerned about precautionary motives. Therefore,
spreads should be higher for lower levels of risk aversion. In addition, give that
prudential behavior is less relevant, the precautionary-savings efiect should kick-
in at higher levels of volatility. Figure 1.9 shows the relationship for the baseline
parametrization with a coe?cient of relative risk aversion (?) equal to 2, as well
as 1 (log utility) and 3. As it can be seen in the graph, the previous reasoning
is conflrmed by the simulations. Higher levels of risk aversion are associated with
lower spreads for all levels of volatility. Furthermore, the tipping point happens at
lower levels of volatility the coe?cient of relative risk aversion is higher, which also
conflrms the intuition that precautionary motives become more relevant.
The next series of sensitivity analyses relate to changes in the efiective discount
factor via a higher probability of remaining in power for the sovereign (? = 0:9
instead of 0.735), changes in the cost of default parameter (? = 0:045 instead of
0.031), and a lower average period of exclusion (fi = 0:4 versus 0.1). This higher
value of fi corresponds to an average exclusion period of 2.5 years after a default.
Results arereported inTable1.6. With respect tospreads, qualitativelyall results go
in the expected direction. A higher level of political stability induces the sovereign to
be less impatient and therefore worry more about the future utility costs of default.
This allows the sovereign to borrow at a substantially lower cost. For example,
at the level of volatility estimated for Argentina, a higher value of ? allows the
47
Figure 1.9: Average Simulated Spreads for Difierent Relative Risk Aversion Coe?-
cients (?)
0 0.5 1 1.5 2 2.5 3?100
0
100
200
300
400
500
600
700
Variance relative to baseline
Average Spread (bps)
 
 
delta = 2
delta = 1
delta = 3
government to pay on average the risk-free rate, while it contracts signiflcantly less
debt (10 percentage points less of a fraction of revenues than under the baseline
parametrization). It is interesting to point out that for ? = 0:9 the risk premium is
very small and declines very fast with the level of volatility, which shows that the
prudential motives dominate at all levels of volatility considered. This is conflrmed
by the fact that for ? = 0:9 the average level of debt also declines faster with the
level of volatility, in contrast with the other cases presented in Table 1.6 where there
is a signiflcant risk of default and average debt levels decline at a slower pace. Thus,
a higher political stability induces more patience on behalf of the sovereign, reducing
the demand for debt, the probability of default and the risk premium on sovereign
debt.
48
Table
1.6:
Robustness
under
Alternativ
eP
arameterizations
Variance
relativ
eto
baseline
Av
erage
Spreads
(basis
poin
ts)
Av
erage
Asset
Holdings/Rev
en
ues
Baseline
?=
0:9
?=
0:045
fi
=
0:4
Baseline
?=
0:9
?=
0:045
fi
=
0:4
0.25
40
6
24
738
-0.25
-0.24
-0.34
-0.12
0.50
213
6
123
746
-0.24
-0.20
-0.31
-0.15
0.75
268
5
170
779
-0.24
-0.17
-0.29
-0.17
1.00
418
0
238
752
-0.23
-0.13
-0.28
-0.18
1.25
477
0
362
1160
-0.23
-0.11
-0.26
-0.19
1.50
510
0
408
1445
-0.23
-0.09
-0.26
-0.22
1.75
534
0
495
2106
-0.23
-0.07
-0.26
-0.22
2.00
530
0
504
2940
-0.22
-0.05
-0.25
-0.24
2.25
543
0
496
3264
-0.22
-0.05
-0.24
-0.24
2.50
515
0
481
3240
-0.22
-0.03
-0.23
-0.25
2.75
498
0
484
4080
-0.21
-0.03
-0.22
-0.27
3.00
478
0
456
4753
-0.21
-0.02
-0.21
-0.28
49
With respect to the efiects of changing the cost of default parameter ?, the
simulations show that the risk premium is smaller at all levels of volatility, which
is reasonable, given that if default is more costly the sovereign will try harder not
to default for a given level of debt. Average debt holdings are higher than under
the baseline case and decrease also with the level of volatility. These results conflrm
the common wisdom that higher cost of default can be beneflcial for the borrower,
because it reduces the cost of credit and increases the supply of credit.
Finally, I analyze also the case of a higher probability of regaining access to
capital markets after a default (fi). In this case, the spreads are extremely high
compared to those observed in the data. It should also be pointed out that for
the range of volatility presented in Table 1.6 the risk premium is monotonically
increasing. However, for a level of volatility 5.5 times larger than the baseline, a
decline in spreads similar to the other cases is observed. This result is explained
by the fact that a very high probability of being pardoned after a default reduces
greatly any prudential behavior on the part of the sovereign. The simulated average
debt levels under this parametrization are also consistent with this less prudent
behavior. The average levels of debt holdings increase with volatility, contrary to all
other parameterizations in which debt declines with volatility. Overall, these results
again show the importance of the trade-ofis between precautionary motives and a
higher default risk in the presence of an increase in volatility.
50
1.5 Conclusions
This chapter analyzes the efiects of volatility on sovereign default risk. Em-
pirically, the papers establishes a \new" empirical fact, namely a non-monotonic
relationship between spreads and macroeconomic volatility. While for low levels of
volatility an increase in volatility is associated with an increase in the sovereign risk
premium, for su?ciently high levels of volatility this relationship turns negative.
This empirical relationship is robust to difierent estimation methods, samples and
control variables. The paper also provides a quantitative model of default risk con-
sistent with this feature and other characteristics of EME debt. In the model, there
is a trade-ofi between a higher default probability due to the increase in the variance
of shocks and an increase in precautionary savings motives by the sovereign due to
a higher degree of volatility in public expenditures if the sovereign cannot access
capital markets for some time after a default episode. At low levels of volatility, the
flrst efiect dominates, given that the welfare loss under autarky is relatively small
in comparison to the present gain associated with debt. However, for su?ciently
high levels of volatility, the cost of potential exclusion becomes more relevant and
dominates the trade-ofi.
The analysis presented in the paper also yields some insights regarding the
difierent behavior of EME in recent times regarding the accumulation of net foreign
assets, especially in the form of reserves. Extremely volatile economies would tend
to accumulate less debt - or hold more net assets - in order to self-insure against
adverse shocks, while economies with lower levels of volatility would hold more debt.
51
This result has been standard in the incomplete asset market literature since Aiya-
gari (1994) and has also been applied by Durdu, Mendoza and Terrones (2007) to
EME. However, this paper is the flrst to explain the mechanics of why EME coun-
tries with very high levels of volatility tend to have lower spreads than countries
with lower levels of volatility. For example, the quantitative model as well as the
empirical evidence imply that countries with very high revenue volatility like Korea
and China would demand less debt and pay a lower risk premium than Argentina
or Brazil. While a reduction in volatility in general is associated with an increase
in social welfare, the simulations of the model presented in the paper show that for
certain levels of volatility a reduction in volatility might increase the incentives to
default on sovereign debt, which in general implies a reduction in the possibility to
smooth consumption. The paper shows that higher political stability that induces
more patience on behalf of the sovereign, as well as higher default costs, are associ-
ated with lower spreads and higher sustainable debt levels. Thus, a reform agenda
that targets these problems in addition to volatility seems promising in reducing
vulnerabilities. However, in order to reduce the pro-cyclicality of capital  ows and
increase international risk-sharing, additional mechanisms that allow for more state-
contingent instruments and reduce the frictions that generate lack of commitment
by the sovereign are needed.
52
Chapter 2
The Composition of Cross-Border Asset Positions
(co-authored with Marcel Fratzscher)
2.1 Introduction
The debate in the literature on trade in flnancial assets makes the important
point that the type of foreign flnancing of cross-border investment may not pursue
a random pattern, but follows some systematic pattern regarding the composition
of capital  ows. One key focus has been on the role of information frictions, with
some important theoretical contributions arguing that portfolio investment should
be more sensitive to information frictions than FDI or bank loans due to a lack of
ownership control of the former (Razin, Sadka and Yuen 1998). A second important
strand of the literature has concentrated on the role of institutions in in uencing the
composition of cross-border investment (Albuquerque 2003; Wei 2000a), with the
empirical work still being inconclusive on which types of capital are most afiected
by the institutional environment.
The main contribution of this chapter is to test empirically for the existence of
such patterns in private cross-border asset position and to identify its determinants
in a bilateral country-pair setting. We concentrate on two determinants that have
53
been central in the literature on trade in flnancial assets: the role of information
frictions, and the role of institutions as drivers of cross-border investment. We
build on several seminal studies. In particular, Portes, Rey and Oh (2001) show
that information frictions for a number of countries indeed exert a larger efiect on
portfolio equity and corporate debt than on government bond  ows with the United
States. The present chapter is complementary to this as well as other studies, but
innovates in a number of ways. First, using a novel dataset on bilateral holdings,
the present paper is the flrst that includes all types of capital, i.e. also FDI and
other investment/loans, and thus allows for a systematic comparison of all types of
investment in the capital account. This is an important difierence because especially
FDI and loans are the dominant types of investment received by many if not most
emerging markets and developing countries.1
Second, the empirical analysis covers 77 countries and thus is much broader in
scope by addressing the issue of cross-border investment also from an emerging mar-
ket (EME) perspective. This allows us to investigate and indeed empirically conflrm
that the efiect of information on cross-border investment exhibits a sizeable asym-
metry across countries, exerting a larger in uence on EMEs. Third, our empirical
methodology is distinct from most of the literature by building on the trade litera-
ture on the border efiect (Anderson and van Wincoop 2003; Cheng and Wall 2005)
which stresses the importance of including source and host country flxed efiects and
shows that the exclusion of such flxed efiects may generate a sizeable estimation bias.
1For example, in our sample the average share of FDI in total foreign investment is 46% for
developing countries but only 22% for developed countries. Moreover, the share of combined FDI
and loans accounts for 76% of total foreign inward investment for EMEs. We discuss these issues
in detail in section 2.3.
54
Our empirical results show that information frictions have a substantial efiect
on the pecking order as we flnd that FDI and loans are the most sensitive and FPI
equity and FPI debt securities the least sensitive types of investment to information
frictions. For instance, the distance among country pairs has a 1.5 to 2 times larger
impact on FDI stocks than on equity securities and debt securities. Similarly, we
flnd loans to be as sensitive as FDI to information asymmetries, thus conflrming
and being in line with the literature on the capital structure of flrms which has
emphasized the special role of loans and its sensitivity to information (Myers 1984;
Bolton and Freixas 2000). We use various proxies for information frictions - dis-
tance, the volume of bilateral telephone tra?c, bilateral trade in newspapers and
periodicals, and the stock of immigrants from the source country in the host - show-
ing the robustness of this result to alternative speciflcations. While these empirical
flndings are new, we also conflrm some of the existing flndings, in particular that
equity portfolio investment are not more sensitive to information frictions than debt
securities (Portes, Rey and Oh 2001). Using our difierent econometric approach also
reveals that the efiects of information frictions tend to be larger than some found
in the literature, though a precise comparison is impossible due to difierent country
samples across studies.
Regarding the second determinant - the impact of institutions on the composi-
tion of cross-border investment - we make two key points. First, while many papers
in the literature have focused on the efiects of institutions on one or two particular
types of capital  ows, our analysis is the flrst to test for difierences across all major
55
components of the capital account. Our results show that portfolio investment is
much more sensitive than FDI or loans to a broad set of institutional indicators,
such as the degree of information disclosure in local credit market regulations, as
well as accounting standards in the host country. Portfolio investment also reacts
much more strongly to the risk of expropriation and repudiation costs, conflrming
the hypothesis put forward by Albuquerque (2003) who argues that portfolio invest-
ment is easier to expropriate than other types of investment. Other hypotheses of
the literature are, however, not conflrmed by our analysis. For instance, portfolio
investments in particular, but also loans, decrease substantially with the degree of
corruption. By contrast, the stock of FDI is found to be less sensitive to corrup-
tion, which is consistent with some flndings in the literature (see Daude and Stein,
2007) but contrary to others (e.g. Wei, 2000a). Overall, portfolio investment, and
in particular equity securities, appear to be the most sensitive type of investment
to institutional factors. Our results prove robust to various alternative proxies of
institutions and country samples.
An additional point of the chapter is that we also study the impact of flnancial
market development on the pecking order of cross-border investment positions. We
flnd that portfolio investment is substantially more sensitive to the degree of market
openness and development than FDI or loans. For instance, capital account liber-
alization and flnancial development change the composition of flnancial liabilities of
a country by raising the share of portfolio investment substantially. Moreover, we
flnd that the volume of FDI and loans is relatively insensitive to market develop-
56
ments as, for instance, capital account liberalization does not have a statistically
signiflcant efiect on the volume or stock of FDI or loans. This is in line with the
evidence for capital  ows of previous studies that use a difierent empirical strategy
(see e.g. Montiel and Reinhart, 1999, Magud and Reinhart, 2005).
Our flndings have a number of policy implications. We underline the role of
bilateral information frictions as a barrier to cross-border investment, in particu-
lar for FDI and loans. Importantly, the chapter emphasizes that FDI should not
necessarily be seen as an unconditional blessing for host countries. We present ev-
idence that the share of inward FDI and also foreign loans is highest for countries
with weak institutions and poorly developed or badly functioning capital markets.
Therefore, although FDI may have beneflcial efiects on the economy, a composition
of foreign investment that is heavily tilted towards FDI is likely to be a signal of
some fundamental weaknesses of the host country economy, thus providing support
for the argument of Hausmann and Fernandez-Arias (2000) and Albuquerque (2003).
The remainder of the chapter is organized in the following way. The next
section provides a brief overview of the literature on the determinants of capital
 ows and the pecking order of cross-border investment. Section 2.3 then outlines
the empirical methodology and presents the data, together with a number of stylized
facts on cross-border investment. The empirical results are discussed in sections 2.4
and 2.5, including various robustness and sensitivity tests. Section 2.6 concludes
and ofiers a short discussion of policy implications.
57
2.2 Related Literature
Information frictions have been at the core of the debate on international cap-
ital  ows.2 Razin, Sadka and Yuen (1998) present a model that extends the pecking
order argument from the corporate flnance literature by Myres and Majlauf (1984)
and Myres (1984) to international capital  ows to analyze issues of capital taxation.
In particular, they assume that FDI circumvents the informational problems com-
pletely, while portfolio debt and equity are subject to informational asymmetries
where domestic investors observe the real productivity of the flrm, while foreign
investors do not. Therefore, FDI is the preferred form of flnancing in the pres-
ence of information frictions, followed by portfolio debt and then equity. Neumann
(2003) presents a version of lending with moral hazard model by Gertler and Rogofi
(1990) that focuses on the difierences between international debt and equity flnanc-
ing. In contrast to Razin, Sadka and Yuen (1998), she assumes that ownership,
even in the form portfolio equity, conveys some control and therefore information
on the investment. Assuming that monitoring costs are decreasing in ownership,
the implied pecking order is that FDI and equity are less costly ways of flnancing
domestic investment than instruments that do not convey some degree of ownership
and therefore information, like loans or debt.
Goldstein and Razin (2006) present a model that explains difierences in volatil-
ity of FDI versus FPI through information asymmetries. Again the key assumption
2Portes and Rey (2005) and Portes, Rey and Oh (2001) provide references and a discussion
of the flnance literature related to information frictions. Also, see Harris and Raviv (1991) for
an earlier survey on the empirical corporate flnance literature on information frictions and asset
markets.
58
is that FDI implies ownership control of the flrm and therefore more information
than FPI. In addition, FDI is subject to a flxed cost in contrast to FPI. They as-
sume that foreign investors are subject to privately observed liquidity shocks which
drive down the price of selling the asset before maturity due to a standard \lemons"
problem. Thus, there is a trade-ofi between e?ciency and liquidity for foreign in-
vestors. Under these conditions, they show that in equilibrium, if production costs
are higher in developed countries, developed countries will receive more FPI that
developing countries, given that it would be less profltable to pay the flxed cost
associated to FDI. Finally, Mody, Razin and Sadka (2003) present a similar model
that predicts also that more countries with good corporate governance attract more
FPI. While several of these theoretical models assume difierent sensitiveness to in-
formation frictions across the difierent components of the capital account, it has
not been tested systematically. Our paper tries to flll this gap in the literature.
Despite limited empirical evidence, the perceived wisdom is that certain types of
capital in ows are more beneflcial for receiving countries than others. In particu-
lar, foreign direct investment (FDI) is generally seen as a \good" type of capital
because it may promote growth in host countries by encouraging a transfer of tech-
nology and knowledge and by opening market access abroad (e.g. Aitken, Hanson
and Harrison, 1997; Borensztein, De Gregorio and Lee, 1998).3 On the other hand,
3For papers that flnd a positive and difierential impact of FDI on domestic investment and
economicgrowthcomparedtoportfolioinvestments, seeBosworthandCollins(1999), Razin(2004),
and Mody and Murshid (2005). However, the literature is not conclusive on the impact of FDI
on growth or the channels through which it acts. Alfaro et al (2004) flnd that FDI has a positive
impact on economic growth provided that the domestic flnancial sector is su?ciently developed.
Alternatively, Borensztein, De Gregorio and Lee (1998) flnd a positive impact of FDI in interaction
with human capital. For some evidence of the efiects of capital  ow composition on currency crises
see Frankel and Rose (1996).
59
portfolio investment  ows are considered to be more volatile, may exacerbate the
magnitude of business cycles and also induce or at least worsen flnancial crises (e.g.
Claessens, Dooley and Warner, 1995; Chuhan, Claessens and Mamingi, 1998; Sarno
and Taylor, 1999).
Other papers have challenged the view of considering FDI necessarily as \good
cholesterol" (e.g. Hausmann and Fernandez-Arias, 2000, Albuquerque, 2003). These
papers show that actually the richest and least volatile economies, and countries
with good institutions and well functioning markets, receive more foreign portfolio
investment (FPI) and relatively less FDI from abroad as a fraction of total capital
in ows.
Finally, the existence and functioning of markets is potentially an important
determinant of foreign investment, and which is closely linked to the efiects of in-
formation asymmetries. If markets are absent or are functioning poorly, flrms may
have no other choice than to use FDI to carry out an investment project (Hausmann
and Fernandez-Arias, 2000). In this sense, FDI may function as a substitute for a
functioning market mechanism. Thus, portfolio investment or bank loans may be
preferred options for flrms in an environment in which markets function well. In a
broader sense, the quality of economic and political institutions is an analogy to the
functioning of markets. In a country where property rights are poorly enforced and
the risk of expropriation is high, flrms may prefer FDI as it is harder to expropriate
due to its information intensity and its inalienability (Albuquerque, 2003). More-
over, difierent types of investment may react difierently to factors such as the degree
of corruption, the functioning of the legal system and transparency (e.g. Wei, 2000b;
60
Faria and Mauro, 2004; Alfaro, Kalemli-Ozkan and Volosovych, 2005; Papaioannou,
2005; Gelos and Wei 2005). While several of these papers look at the efiects on total
capital  ows, a speciflc type of  ows, or the difierence between portfolio and FDI,
we contribute to the literature by analyzing the efiect of institutional variables on
all major concepts of the capital account. 4 Moreover, other important difierences
of the present paper with the existing literature are the focus on bilateral capital
stocks as well as the methodological approach, which allows us to control for infor-
mation asymmetries as well as for both source and host country factors. Finally, we
also study the impact of flnancial market development on the composition of the
capital account.
We view this research also as a complement to the literature on the cyclical
determinants and characteristics of capital  ows to emerging markets. Much of
this literature has been focused on the drivers (\push" and \pull" factors) of  ows.
Several studies including Fernandez-Arias (1996), Calvo, Leiderman and Reinhart
(1993), and Montiel and Reinhart (1999) focus on the relevance of \push" and \pull"
factors regarding capital in ows towards Latin America or emerging economies dur-
ing the 1990?s. Using difierent methodologies, they tend to emphasize on external
factors, like changes in the international (U.S.) interest rate as a key determinant
of capital  ows.5 The focus of the present chapter is on the determinants of the size
4While Alfaro et al. (2005) also test the efiects of institutions on the capital account, their focus
is on aggregate capital  ows (deflned as the sum of FDI and portfolio investment  ows). Therefore,
they do not include bank loans nor do they test or comment on difierences among the difierent
types of investment. As we will show below, we flnd this distinction to be important as difierent
types of capital react fundamentally difierently to information frictions as well as institutions.
5See also Kaminsky, Reinhart and Vegh (2004) for evidence on the pro-cyclicality of capital
61
and composition of country portfolio positions rather than cyclical changes in  ows.
2.3 Methodology, data and some stylized facts
This section gives an outline of the methodology and the main hypotheses
for the empirical analysis (section 2.3.1). The subsequent presentation of our data
(section 2.3.2) is then followed by a discussion of some key stylized facts of the
pecking order of cross-border investment positions derived from our data (section
2.3.3).
2.3.1 Methodology and hypotheses
The empirical analysis consists of two parts. In the flrst part, we address the
role of information frictions as a determinant of the pecking order of cross-border
investment. The efiects of information frictions are likely to be difierent across
country pairs, i.e. one particular source country i may face a difierent degree of in-
formation costs and asymmetries vis--vis host country j than other source countries.
For this purpose, we use a pseudo-flxed efiects model of bilateral capital stocks held
by residents of source country i in host country j:
log(1+ykij) = fii +fij +flkXij +"ij (2.1)
with ykij is the holdings in US dollars of asset type k - where k = FDI, portfolio
equity, portfolio debt securities, or loans - of residents of source country i in host
 ows and macroeconomic policies in emerging markets.
62
country j; Xij is a matrix that includes a proxy of bilateral information frictions
and additional controls; and fii and fij are source country and host country flxed
efiects.
Given that in our flrst step we want to identify consistently the efiect of infor-
mation frictions - a pair-efiect variable - we also need to control for all other relevant
factors that afiect the volume of bilateral investment from a particular source coun-
try by including source and host country dummies as well as other bilateral controls
that are likely to afiect the level of bilateral investment.6 In the second step, we
then try to explain the country flxed efiects in order to understand which factors
make host countries attractive places for investment.7 The vector of coe?cients of
interest to us in this flrst step is flk, i.e. we want to test whether difierent types
of asset holdings have a difierent degree of sensitivity to various proxies of informa-
tion frictions in Xij. Note that we are interested in two separate hypotheses, one
relating to the volume efiect of information frictions (H1) and the second one to
the composition efiect (H2), i.e. that one type of flnancial asset holdings (k1) reacts
difierently to information frictions than other types of assets (k2):
Volume efiect hypothesis: H1 : flk = 0
Composition efiect hypothesis: H2 : flk1 = flk2
Our empirical analysis is cross-sectional, hence the explanatory power of the
6The inclusion of these country flxed efiects has also been recommended by Anderson and van
Wincoop (2003) in empirical trade models to control for multilateral resistance. In the case of
investment positions, the problem of omitted and unobserved variables at the source or host level
might also be more serious, given the lack of an overall accepted theory of bilateral investment
positions that could be used as a benchmark for the empirical exercise.
7See Cheng and Wall (2005) for the relevance of such a two-step approach for trade. Lane
and Milesi-Ferretti (2004) use a very similar approach to ours for the case of bilateral portfolio
positions.
63
model comes purely from the cross-section, which is sensible given the focus on
capital stocks and the fact that the independent variables on information frictions
and institutions are mostly changing little over time.
Note also that we estimate the model using ykij as the stocks in US dollars
of asset type k. More precisely, we take the log value of the value in million US
dollars and add one in order to be able to keep observations that are zero.8 As there
are several observations with a value of zero, it may raise the problem of censoring
at zero. Although we use a Tobit estimator and a two-step Heckman procedure to
show that the results are largely robust to this speciflcation, our preferred estimation
technique is via seemingly unrelated regressions (SUR). This means that we estimate
the four equations for each type of capital k simultaneously. The advantage of the
SUR estimator is that it improves the e?ciency of the estimates by allowing for
cross-correlations of the residuals of the four equations. Moreover, it allows us to
directly test our pecking order hypothesis H2 in the model.
Note that we do not \normalize" the dependent variable by dividing by host
country GDP for H1 on the volume efiect or by dividing by total asset liabilities
of host country j for H2 on the pecking order efiect, as is frequently done in the
literature. The reason is that each of these \normalizations" imposes restrictions
on the parameters of the model that may not hold. Although such a normalization
is possible, our preferred speciflcation is the one using the log of the levels of cross-
border investment, given that it allows for more  exibility and it allows use to test
8However, in our flnal sample the number of zeros is relatively small. Out of the flnal 1116
observations, FDI values are all strictly positive, FPI portfolio has 187 zero observations, FPI debt
125, and Loans 84, respectively. Our results do not change if we drop these observations.
64
the volume and composition hypothesis in the same equation.9
More generally, although it may seem appealing to exclude the flxed efiects
in order to explicitly allow for including vectors of source country-speciflc variables
Xi and of host country-speciflc variables Xj, this would imply excluding important
unobserved components of relevant flxed efiects and is likely to bias the estima-
tors of interest flk. We show below that the estimates of flk indeed mostly change
substantially when excluding the flxed efiects.
In the second part of the analysis, our aim is to understand the factors that
explain the host country flxed efiects. More precisely, we want to understand the
role of markets and institutions in host countries as determinants of the composition
of cross-border flnancial positions. As these factors are symmetric, i.e. investors in
all source countries face the same conditions in a particular host country, we use the
flxed efiects obtained from the gravity model (2.1) to test for the role of host country
institutions and market conditions Xj on the pecking order and volume efiects:
fikj = ?k +?kXj +?kj; (2.2)
where ?kj is an error term. Analogously to model (2.1), this speciflcation allows us
to formulate and test the two hypotheses with regard to the volume efiect (H3) and
the pecking order efiect (H4) of flnancial market development and institutions:
Volume efiect hypothesis: H3 : ?k = 0
Composition efiect hypothesis: H4 : ?k1 = ?k2
9It should be pointed out that the country dummies capture the size efiects of the source and
host in an accurate way.
65
Our preferred estimator is again the SUR, and the same caveats and discussion
apply to this second stage as to the estimation of model (2.1).
2.3.2 Data
As the focus of the paper is on the pecking order of cross-border investment,
our data is on stocks of various types of foreign investment, rather than capital  ows
per se. We use three difierent data sources to construct a comprehensive database
that covers all four categories of the flnancial account - or what is still often referred
to as the capital account; two terms which we use interchangeably throughout the
paper - i.e. for FDI, for portfolio investment - distinguishing also between equity
and debt securities - and for loans.
For FDI, we use the UNCTAD database on bilateral FDI stocks. A database
that is often employed in studies on FDI is the one provided by the OECD. However,
the UNCATD database is more comprehensive as it includes both industrialized
countries and developing countries. The UNCTAD data has annual entries in US
dollars for around 90 reporting countries vis--vis most countries in the world from
1980 to 2003. Unfortunately, there are many missing entries, so that we do not have
bilateral stocks for all country pairs. Moreover, country pairs are excluded from the
analysis if there are no entries for the past ten years.
For portfolio investment, we use the Consolidated Portfolio Investment Survey
(CPIS) by the IMF. It provides bilateral assets of portfolio equity and portfolio
debt securities for 68 reporting countries.10 We use the average flgures for equity
10In fact, the efiective number of reporting countries ends up being 67, because Pakistan reports
66
securities and for debt securities for 2001, 2002 and 2003. The CPIS also provides a
breakdown between short-term and long-term debt securities. We conducted several
tests but did not flnd systematic difierences with this distinction, and thus ignore
this dimension in the remainder of the paper.
For loans, we use the International Locational Banking Statistics (ILB) data
provided by the Bank of International Settlement?s (BIS). The database comprises
aggregate assets as well as aggregate liabilities of banks in 32 reporting countries
vis--vis banking and non-banking institutions in more than 100 partner countries,
capturing exclusively private claims. The reported assets and liabilities capture
mostly loans and deposits, but may also include other transactions that fall under
portfolio or direct investment (see BIS 2003). To minimize this overlap, we use inter-
bank claims, i.e. the data for assets and liabilities of banks in reporting countries
vis--vis banks in partner countries. Although the number of reporting countries is
smallest for this database, the fact that it includes data not only for assets but also
for liabilities allows us to obtain a proxy also for asset holdings of non-reporting
countries vis--vis reporting countries.
There are several caveats that are present for the various data sources. A flrst
potential caveat is that the data stems from difierent sources, thus raising the issue of
how comparable they are, though the deflnitions used are the same across sources.
Moreover, one potentially important issue is that the data collection is generally
based on the residence principle. This may imply that countries may report their
asset holdings vis--vis their direct counterpart country but not vis--vis the country
only missing data.
67
where the flnancial asset is ultimately invested. This of course would give enormous
importance to flnancial centers as a lot of capital is channeled through these, but do
not re ect the true bilateral holdings of flnancial assets. Hence we exclude flnancial
centers from our analysis.
Moreover, note that our empirical analysis is purely cross-sectional for two
reasons: due to the fact that capital stocks obviously change little from one year
to the next and also due to data availability. Due to the potential importance
of valuation changes and other special factors afiecting the size of capital stocks
in individual years, our cross-section is the average size of capital stocks over the
flve-year period of 1999-2003.
It is important to emphasize that we include only those country pairs for
which all four types of asset holdings are available. This reduces the sample size
to 77 countries. The appendix shows the countries which are included. It reveals
that the sample includes 22 rich, industrialized countries and 55 mainly emerging
markets, but also some poorer developing countries. The country sample for the
EMEs is roughly balanced across regions with 12 in Africa/Middle East, 13 in Cen-
tral and Eastern Europe, 13 in Asia and 17 in Latin America. The exclusion of
many of the poorer developing countries is required by the fact that they do not
have stock markets and/or bond markets. Thus the results on the composition of
cross-border investment are not driven by the absence of stock and bond markets
in less developed countries. Further tests focusing only on industrialized countries
and only for emerging market economies (EMEs) are conducted below and show the
robustness of the flndings to difierent country samples.
68
Figure 2.1: Composition of Cross-border Investment Positions by GDP per capita
Quintiles
Note: GDP per capita is measured as the average PPP GDP per capita over the period
1999-2003. The x-axis shows the flrst to flfth quintile of countries, ranging from those with the
lowest to those with the highest GDP per capita. Sources: IMF CPIS and IFS; UNCTAD; BIS;
authors? calculations.
2.3.3 Composition of cross-border investment: some stylized facts
Figure 2.1 shows for a broad set of developed and emerging market economies
(EMEs) that the poorest countries have the highest shares and the richest the lowest
shares of FDI in total capital stocks.
This stylized fact - as well as several others discussed in detail in the paper -
makes the important point that the type of foreign flnancing of cross-border invest-
ment does not pursue a random pattern, but follows a certain \pecking order".
Table 2.1 presents some summary statistics for the difierent types of flnancial
liabilities, i.e. the table shows the total stocks of difierent types of capital held
69
by foreigners in the host countries implied by the data described in the previous
subsection. There are clear, systematic difierences in the composition and volume
of capital stocks across countries. First, developing countries receive on average a
higher share of FDI and loans than developed countries. For example, the average
share of FDI in total foreign capital for developing countries is 44% while in the
case of the developed countries FDI amounts only to 22%. In contrast, the share
of portfolio equity and portfolio debt holdings is signiflcantly higher for developed
countries. Second, in terms of the volume of investments, developed countries receive
signiflcantly higher volumes of all types of capital. Developed countries receive
on average - as a ratio of their GDP - around 2.5 times more FPI portfolio, 6.6
times more FPI debt, 2 times more loans, and 1.3 times more FDI than developing
countries.
Table 2.2 shows the correlation coe?cients and the signiflcance of investment
shares with regard to selected indicators of income, market development and insti-
tutions. First, there is a large negative correlation of -0.38 between the share of FDI
instocks and per capita income of a country. Loans are also negatively correlated,
though the correlation coe?cient is not statistically signiflcant. The same flnding
applies to domestic flnancial market development - as proxied by the degree of cap-
ital account liberalization and by the ratio of credit to the private sector as percent
of GDP: the more developed flnancial markets are, the lower the shares of FDI and
loans a country receives. Figure 2.2 illustrates in more detail the relationships be-
tween these difierent types of capital and per capita GDP. Moreover, countries with
a higher risk of expropriation (indicated by a lower value in the flgure) receive a
70
Table 2.1: Summary Statistics
Variable Observations Mean Std. Dev. Min Max
EMEs / Developing Countries
FDI share 55 0.46 0.22 0.1 0.9
Loans share 55 0.3 0.18 0 0.7
FPI debt share 55 0.14 0.11 0 0.4
FPI portfolio share 55 0.1 0.11 0 0.5
FDI/GDP 55 0.42 0.48 0 2.7
Loans/GDP 55 0.34 0.91 0 6.8
FPI debt/GDP 55 0.13 0.14 0 0.5
FPI equity/GDP 55 0.11 0.22 0 1.4
Developed Countries
FDI share 22 0.22 0.1 0.05 0.38
Loans share 22 0.26 0.09 0.13 0.49
FPI debt share 22 0.35 0.16 0.03 0.66
FPI equity share 22 0.17 0.11 0.03 0.44
FDI/GDP 22 0.56 0.59 0.03 2.44
Loans/GDP 22 0.65 0.51 0.11 1.65
FPI debt/GDP 22 0.86 0.86 0.05 3.17
FPI equity/GDP 22 0.4 0.43 0.05 1.45
Total
FDI share 77 0.394 0.23 0.05 0.92
Loans share 77 0.292 0.16 0.05 0.74
FPI debt share 77 0.198 0.16 0 0.66
FPI portfolio share 77 0.117 0.11 0 0.51
FDI/GDP 77 0.462 0.51 0.03 2.66
Loans/GDP 77 0.424 0.82 0.01 6.79
FPI debt/GDP 77 0.339 0.57 0 3.12
FPI equity/GDP 77 0.19 0.32 0 1.43
Sources: IMF CPIS and IFS; UNCTAD; BIS; authors? calculations.
71
Table 2.2: Correlation Matrix
FDIshare Loansshare FPIdebtshare FPIequity share
FDIshare 1
Loansshare -0.514 1
FPIdebtshare -0.427 -0.338 1
FPIequity share -0.281 -0.296 -0.057 1
GDPpercapita(log) -0.405 -0.17 0.475 0.319
PrivateCredit/GDP -0.357 -0.106 0.246 0.471
KAOpenness -0.137 -0.152 0.208 0.202
Property Rights 0.347 0.147 -0.342 -0.449
GDPpercapitagrowth 0.03 -0.144 -0.097 0.287
Note: Signiflcant correlations at the 95% level are shown in bold.
signiflcantly higher share of FDI and loans.
By contrast, both equity security and debt security holdings are strongly posi-
tivelycorrelatedwithGDPper capita. Moreover, countriesthathave a largeshareof
portfolioequity anddebtstocks alsohave moredeveloped domesticflnancialmarkets
and better institutions. Moreover, when considering the correlation of the shares of
difierent types of assets with the average growth rate of GDP per capita over 1980
- 2003, the correlations show that there is a positive and signiflcant correlation only
for portfolio investment.
Figures 2.3 and 2.4 illustrate these points in more detail by showing the dis-
tributions of the shares of individual types of capital by quintiles of the variable of
interest, e.g. starting with the quintile of countries with the lowest GDP per capita
on the left and leading up to those with the highest GDP per capita (top left panel
of Figure 2.3). The top right panel of Figure 2.3 shows that countries that had
the highest volatility in GDP growth rates - as measured as the standard deviation
of annual real GDP growth rates over the period 1980-2003 - also experienced the
72
Figure 2.2: Shares in Cross-Border Investment and GDP per capita
Sources: IMF CPIS and IFS; UNCTAD; BIS; authors? calculations.
73
Figure 2.3: Stylized facts: macro and exchange rate variables
Notes: GDP growth volatility is the standard deviation of annual real GDP growth rates over the
period 1980-2003. Exchange rate volatility is deflned as the standard deviation of the monthly
nominal exchange rate changes vis--vis the US dollar over the period 1980-2003. The x-axis shows
the flrst to flfth quintiles of countries. Sources: IMF CPIS and IFS; UNCTAD; BIS; authors?
calculations.
highest degree of output volatility.
Figure 2.4 shows corresponding charts for market development and various
institutional indicators. For instance, countries with the least developed domestic
flnancial markets - as proxied by credit to the private sector to GDP - have the
highest share of the inward investment from abroad in the form of FDI and loans,
which both fall as domestic flnancial development improves. Moreover, the bottom
left panel of Figure 2.4 indicates that countries with higher corruption receive rela-
tively more FDI and loans, and substantially less portfolio investment. Finally, also
countries with a worse protection of property rights - as indicate by a rise in the
74
Figure 2.4: Stylized facts: market development and institutions
Notes: A higher value of the expropriation risk indicator means a lower degree of risk, and a larger
indicator for property rights indicates a worse protection of property rights. The x-axis shows
the flrst to flfth quintiles of countries. Sources: IMF CPIS and IFS; UNCTAD; BIS; authors?
calculations.
indicator shown - have a larger share of FDI and loans and relatively fewer equity
and debt securities.
Overall, these stylized facts provide some flrst, descriptive evidence that there
is indeed a pecking order in cross-border investment, as the various types of foreign
capital stocks are strongly correlated with indicators of market development and
institutions. A detailed analysis of the causality underlying these relationships is
provided in the subsequent sections.
75
2.4 The role of information frictions
We now turn to our econometric results. We start with the analysis of the role
of information frictions (section 2.4), before presenting the flndings with regard to
the role of markets and institutions (section 2.5).
2.4.1 Benchmark results
What is the role of information frictions in explaining the pecking order of
cross-border investment positions? A flrst important issue is how to measure infor-
mation frictions. We start by following the common practice in the literature both
on trade in goods and on trade in flnancial assets and proxy information frictions
through the log geographic distance between country pairs. We then proceed by
using various alternative measures for information.
Table 2.3 shows the results of our benchmark model (2.1), which includes in
addition to distance a set of standard gravity variables, such as dummy variables
on whether or not the two countries have a common language, have a common legal
origin, colonial links, and whether they have a trade agreement or a joint invest-
ment treaty to facilitate cross-border investment. The results are compelling both
with regard to our hypothesis H2 about the composition of cross-border investment
positions as well as with regard to the volume efiects hypothesis H1.
FDI and loans are substantially more sensible to changes in distance than
portfolio equity and portfolio debt investment. The difierences in the efiects are
sizeable as the coe?cients for FDI and loans are both around -1.2 as compared
76
Table 2.3: Information frictions: distance
FDI FPIequity FPIdebt Loans Signiflcanceforpeckingorder:FDI Equity Debt
vs. vs. vs.equity debtloansdebtloansloans
distance-1.180*** -0.676*** -0.808*** -1.231*** 0.000.00 0.520.07 0.00 0.00(0.068) (0.057) (0.063) (0.068)
commonlanguage0.433*** 0.324** 0.111 0.247 0.540.11 0.320.22 0.67 0.46(0.160) (0.135) (0.149) (0.161)
commonlegalorigin 0.713*** 0.568*** 0.395*** 0.438*** 0.240.02 0.040.15 0.31 0.74(0.112) (0.094) (0.104) (0.113)
coloniallinks 0.924*** 0.333 * 0.198 0.321 0.010.01 0.020.56 0.96 0.62(0.216) (0.182) (0.200) (0.217)
tradeagreement -0.167 -0.336** 0.617*** 0.230 0.380.00 0.060.00 0.00 0.05(0.175) (0.147) (0.163) (0.176)
investmenttreaty 0.260** 0.027 0.094 0.429*** 0.060.24 0.200.58 0.00 0.01(0.113) (0.095) (0.105) (0.113)
#observations 1116 1116 1116 1116R-squared0.828 0.907 0.881 0.847
The right-hand side of the table shows the p-values for the equality tests across distance coe?cients.
***,**,*showstatisticalsigniflcanceofthecoe?cientsatthe99%, 95%and90%levels, respectively.
to point estimates of -0.67 and -0.80 for portfolio equity and debt. Also, these
difierencesarehighlystatisticallysigniflcantas shownintheright-hand-sidecolumns
of the table.11
It is interesting to point out that the size of the estimated coe?cients for
distance is in line with the empirical literature on trade in assets, e.g. Portes and
Rey (2005) report a coe?cient of -0.89. In addition, the efiect of distance on asset
trade is greater than its efiect on trade in goods, which according to Leamer and
Levinsohn (1995) is mostly around -0.6. In the case of goods, Grossman (1998)
shows that for sensible values of transportation costs, the distance elasticity should
be around -0.03.12 Thus, he concludes that information costs must be behind the
11Note that while the information variables have a large efiect on FDI than on portfolio in-
vestment (our pecking order hypothesis), the goodness-of-flt of the model for FDI is generally
somewhat lower than that or portfolio investment equity and debt. This flnding comes from the
lower explanatory power of the flxed efiects in the models for FDI, which can be seen by estimating
the models including only the flxed efiects.
12For a recent survey on the importance of trade costs see Anderson and van Wincoop (2004).
77
empirical result that the efiect is around 20 times larger. For trade in assets it
therefore seems that the case for distance re ecting information rather than trade
costs is even more compelling.
We explore this information hypothesis in more detail below. The point es-
timates for the variables on what is often referred to as \familiarity" efiects are
sensible as they have the correct sign and are mostly statistically signiflcant. Like
for the distance variable, FDI reacts much more strongly to these familiarity efiects
than this is the case for portfolio equity and debt investment. For instance, when
both countries speak the same language FDI stocks in host countries are 54 percent
higher and portfolio equity investment 38 percent larger, whereas portfolio debt
investment and loans are not statistically signiflcantly difierent.13
2.4.2 Robustness: alternative proxies for information frictions
How robust are these flndings to difierent proxies for information frictions?
Clearly, it may seem odd to proxy information frictions for trade in flnancial assets
through geographic distance as one would expect that geography should have lit-
tle to do with flnancial transactions. However, the literature on capital  ows has
repeatedly found distance to be highly signiflcant, see e.g. Portes and Rey (2005)
for equity  ows. Nevertheless, it is useful to employ alternative and ideally more
direct proxies for information frictions. We use three proxies: the amount of tele-
phone tra?c between two countries, the trade in newspapers, and bilateral stock of
13Note that the coe?cients for the dummy variables are not strictly elasticities. The calculation
of the elasticity, for instance for the former variable can be done by using: exp(0.43) - 1 = 0.537.
78
immigrants of the source country living in the host country and vice versa.
The intuition for the use of these variables as proxies for the degree of informa-
tion frictions is straightforward. The volume of telephone call tra?c was proposed
flrst by Portes and Rey (2005) and has been used in the most recent empirical
literature.14 Telephone tra?c is a proxy of the amount of information that  ows
between both countries and it is assumed that a larger volume of information  ows
- controlling additionally for the size of both economies - implies less informational
frictions. A similar rationale has been put forward to use trade in newspapers and
periodicals by Nicita and Olarreaga (2000) to study information spillovers in goods
markets. They report a high correlation of trade in newspapers with telephone traf-
flc (a simple correlation of 0.77), but prefer their measure due to a greater data
availability. Finally, Gould (1994) analyzes the impact of the stock of immigrants
in the U.S. on trade between the U.S. and the immigrants? country of origin. The
intuition is that immigrants have better information on the markets and institutions
in their home country which would lower transaction costs.
Table 2.4 shows the results when adding telephone tra?c to the benchmark
model. One important result is that when adding telephone tra?c it is not only
highly signiflcant, but distance becomes insigniflcant for FDI and portfolio equity
and debt investment. Distance retains its signiflcance for loans, albeit with a much
smaller coe?cient of -0.34 as compared to -1.23 in the benchmark model of Table
2.3. It is important to point out that this result is not driven by multicollinearity
14See Portes, Rey and Oh (2001) for the case of equity  ows; Loungani, Mody and Razin (2002),
as well as Di Giovanni (2005) for FDI; and Mody, Razin and Sadka (2003) for FDI and equity.
79
Table 2.4: Information frictions: distance versus telephone tra?c
FDI FPIequity FPIdebt Loans Signiflcanceforpeckingorder:FDI Equity Debt
vs. vs. vs.equity debtloansdebtloansloans
distance-0.072 -0.091 -0.071 -0.341** 0.901.00 0.100.90 0.12 0.09(0.130) (0.112) (0.131) (0.134)
telephonetra?c 0.721*** 0.447*** 0.399*** 0.595*** 0.000.00 0.230.63 0.15 0.06(0.083) (0.072) (0.084) (0.086)
commonlanguage-0.016 0.130 0.126 -0.144 0.490.56 0.570.98 0.22 0.22(0.181) (0.157) (0.184) (0.187)
commonlegalorigin 0.505*** 0.448*** 0.327** 0.402*** 0.700.30 0.520.42 0.77 0.63(0.126) (0.109) (0.128) (0.130)
coloniallinks 0.353 -0.055 -0.177 -0.357 0.130.10 0.010.66 0.30 0.53(0.233) (0.201) (0.236) (0.240)
tradeagreement -0.106 -0.299 * 0.845*** 0.304 0.370.00 0.070.00 0.01 0.02(0.184) (0.159) (0.186) (0.190)
investmenttreaty 0.078 0.314** 0.313 * 0.591*** 0.210.29 0.011.00 0.17 0.16(0.162) (0.140) (0.164) (0.167)
#observations 595 595 595 595R-squared0.873 0.928 0.884 0.850
The right-hand side of the table shows the p-values for the equality tests across coe?cients for
information friction variables. ***,**,* show statistical signiflcance of the coe?cients at the 99%,
95% and 90% levels, respectively.
problems between telephone tra?c and distance, given that the simple correlation
between both variables in our sample is just -0.13. In addition, although the sample
is reduced due the availability restrictions on telephone tra?c, if we re-estimate the
regression from Table 2.3 for this sub-sample, the distance coe?cients are negative,
signiflcant, and not difierent from the estimates for the whole sample. Therefore,
distance seems to be a proxy for overall information frictions in asset trade. When
comparing the pecking order efiect of information frictions, telephone tra?c is again
signiflcantly larger for FDI and also loans than for equity and debt.
Table 2.5 gives the estimates for the other two alternative information proxies
as well as for a model that instead includes the flrst principal component of the
three proxies. We include the principal component of all three alternative proxies
80
because it may help alleviate measurement errors related to each individual vari-
able.15 The results conflrm that FDI and loans are more sensitive to information
frictions. However, distance remains signiflcant in most of these speciflcations, and
with the same order as before as information generally has the largest efiects on FDI
and loans and the smallest impact on portfolio equity and debt.
2.4.3 Robustness: Alternative model speciflcations and controls
Finally, we conduct a battery of sensitivity tests by using alternative economet-
ric speciflcation and by adding various controls to the empirical speciflcation of the
model. A flrst test is to ask whether the results are robust to taking ratios, of GDP
or of total capital stocks, as dependent variables, which is a commonly done in the
literature, despite the controversial underlying assumptions behind such a speciflca-
tion, as discussed in section 2.3.1. Table 2.6 shows the estimates for the benchmark
model where the dependent variable is measured as a percentage of source and host
country GDP and as a percentage of total capital  ows from source country i to
host country j. The results indicate that although the coe?cients are very difierent,
our overall results with regard to the pecking order still hold: FDI and loans are in
both speciflcations signiflcantly and substantially larger than portfolio equity and
portfolio debt investment.
As the next step, we investigate the robustness of the results to using alterna-
tive econometric estimators. Table 2.7 provides the results for a Tobit estimator and
15About 81 percent of the total variation in the three alternative proxies is explained by their
flrst principal component. The factor loadings are high for all three variables, so that they seem
to be well represented by the flrst factor.
81
Table
2.5:
Information
frictions:
alternativ
einformation
pro
xies
FDI
FPI
equit
y
FPI
debt
Loans
(1)
(2)
(3)
(1)
(2)
(3)
(1)
(2)
(3)
(1)
(2)
(3)
distance
-1.021
***
-0.736
***
-0.225
-0.602
***
-0.521
***
-0.258
**
-0.722
***
-0.345
***
-0.073
-1.062
***
-0.855
***
-0.438
***
(0.073)
E,D
(0.115)
E,D
(0.148)
L(0.062)
F,L
(0.085)
F,D,L
(0.111)
(0.069)
F,L
(0.095)
F,E,L
(0.124)
(0.074)
E,D
(0.122)
E,D
(0.163)
F
trade
innewspap
ers
0.064
***
0.030
***
0.035
***
0.069
***
(0.012)
E,D
(0.010)
F,L
(0.011)
F,L
(0.012)
E,D
stoc
kof
foreigners
0.180
***
0.105
***
0.107
**
0.127
**
(0.050)
(0.037)
(0.041)
(0.053)
principal
comp
onen
t
0.498
***
0.298
***
0.209
***
0.406
***
(0.087)
E,D
(0.065)
F
(0.073)
F,L
(0.096)
D
common
language
0.364
**
0.244
0.066
0.292
**
0.146
0.128
0.073
-0.069
0.097
0.173
-0.089
0.082
(0.159)
(0.207)
(0.229)
(0.135)
(0.153)
(0.172)
(0.148)
(0.172)
(0.191)
(0.159)
(0.220)
(0.252)
common
legal
origin
0.665
***
0.767
***
0.759
***
0.545
***
0.451
***
0.418
***
0.368
***
0.654
***
0.712
***
0.386
***
1.012
***
0.967
***
(0.111)
D,L
(0.166)
E(0.186)
E(0.094)
(0.123)
F,L
(0.139)
F,D,L
(0.104)
F(0.138)
L(0.155)
E(0.112)
F(0.176)
E,D
(0.204)
E
colonial
links
0.778
***
-0.274
-0.380
0.265
-0.279
-0.358
0.118
-0.071
-0.324
0.166
-0.577
-0.597
(0.215)
E,D,L
(0.357)
(0.370)
(0.182)
F(0.264)
(0.278)
(0.201)
F(0.296)
(0.309)
(0.216)
F(0.378)
(0.407)
trade
agreemen
t
-0.162
0.241
0.171
D
-0.333
**
-0.037
0.022
0.620
***
0.917
***
0.693
***
0.235
0.438
**
0.446
**
(0.173)
D,L
(0.197)
D
(0.205)
(0.147)
D,L
(0.146)
D,L
(0.154)
D,L
(0.162)
F,E,L
(0.163)
F,E,L
(0.172)
F,E
(0.174)
F,E,D
(0.209)
E,D
(0.226)
E
investmen
ttreat
y
0.218
*
-0.026
-0.435
0.007
0.334
**
0.318
0.071
0.617
***
0.698
***
0.385
***
0.483
**
0.766
**
(0.112)
E(0.225)
D,L
(0.279)
E,D,L
(0.095)
F,L
(0.167)
(0.209)
F(0.105)
L(0.187)
F(0.233)
F(0.112)
E,D
(0.239)
F(0.307)
F
#observ
ations
1116
474
332
1116
474
332
1116
474
332
1116
474
332
R-squared
0.832
0.864
0.876
0.908
0.928
0.938
0.882
0.904
0.913
0.851
0.842
0.839
The
sup
erscripted
letters
F
,E
,D
and
L
indicate
the
test
of
the
hyp
othesis
H2
,that
the
co
e?cien
tis
difieren
tto
that
of
FDI,
equit
y
portfolio,
debt
securities,
and
loans,
resp
ectiv
ely
.***,**,*
sho
w
statistical
signiflcance
of
the
co
e?cien
ts
at
the
99%,
95%
and
90%
lev
els,
resp
ectiv
ely
.
82
Table 2.6: Information frictions: Ratios as % of GDP and total capital stocks
FDI FPIequity FPIdebt Loans FDI FPIequity FPIdebt Loans
Ratioas%ofGDP Ratioas%oftotalcapitalstocks
distance-0.005***-0.001***-0.002***-0.005***-0.282*** 0.086 0.051 -0.339***(0.000)E,D(0.000)F,D,L(0.000)F,E,L(0.000)E,D(0.065)E,D(0.062)F,L(0.062)F,L(0.050)E,D
commonlanguage0.004*** 0.001 0.000 -0.003 * 0.097 0.012 -0.238 * -0.286**(0.001)E,D,L(0.000)F,L(0.000)F,L(0.001)F,E,D(0.150)L (0.142) (0.143) (0.115)F
commonlegalorigin0.000 0.001 ** 0.002*** 0.001 0.246** 0.212** 0.021 0.109(0.000)D (0.000)D (0.000)F,E(0.000) (0.104) (0.098) (0.099) (0.080)
coloniallinks0.003 ** 0.002*** 0.003*** 0.005*** 0.457** -0.100 0.009 -0.102(0.001) (0.000)L (0.000) (0.001)E (0.204)E,L(0.193)F (0.194) (0.157)F
tradeagreement 0.001 0.000 0.005***-0.001 -0.219 -0.335** 0.482***-0.117(0.001)D (0.000)D (0.000)F,E,L(0.001)D (0.159)D (0.151)D (0.151)F,E,L(0.122)D
investmenttreaty -0.003***-0.001** -0.001 -0.003***-0.058 -0.002 -0.081 0.293***(0.000)E,D(0.000)F,L(0.000)F,L(0.000)E,D(0.121)L (0.115)L (0.115)L (0.093)F,E,D
#observations1027 1027 1027 1027 842 842 842 842R-squared0.323 0.499 0.549 0.369 0.985 0.932 0.937 0.756
The superscripted letters F,E,D and L indicate the test of the hypothesis H2, that the coe?cient
is difierent to that of FDI, equity portfolio, debt securities, and loans, respectively. ***,**,* show
statistical signiflcance of the coe?cients at the 99%, 95% and 90% levels, respectively.
for an OLS estimator without source and host country flxed efiects. The estimates
of the Tobit model are in line with those obtained from our OLS benchmark. Recall
that the tobit model is a non-linear estimator that uses a mixture of a continuous
distribution over the non-censored observations and a discrete distribution for the
censored ones. The point estimates shown in the table are the marginal efiects
evaluated at the mean of the independent variables. Hence the size of the marginal
efiects is not so meaningful.
There are some interesting difierences between the models with and without
flxed efiects. The model without flxed efiects is estimated by including nominal
GDP (in US dollar) and population of both the source country and of the host
country instead of the flxed efiects. There are two important points to note from
the results. First, almost all point estimates for the proxies of information frictions
83
are substantially difierent from those of the benchmark flxed-efiects model. This
lends support to our point made above that it is important to estimate the model
by including flxed efiects as otherwise the point estimates are biased due to omitted
variables. Nevertheless, even without the flxed efiects our pecking order hypothesis
is conflrmed. Second, note that the hypothesis that the point estimates of the GDP
variables are equal to one is rejected in almost all equations. This is a notewor-
thy fact because it stresses that a \normalization" of the model, i.e. including the
dependent variables as ratios of GDP imposes incorrect restrictions on the param-
eters of the model.16 Next, we test for the presence of asymmetries in the efiects
of information flctions across samples. In particular, it is possible that some types
of countries are much more sensitive to information than others; for instance those
where information is already very scarce. Table 2.8 shows the results when estimat-
ing the benchmark model (2.1) separately for when only industrialized countries and
when only EMEs are the host countries. Overall, the results conflrm that FDI and
loans are most sensitive to information frictions.
Moreover, some interesting difierences across country groups emerge. In par-
ticular, capital stocks are much more sensitive to information and familiarity efiects
when the host country is an emerging market economy. The elasticity for FDI, for
instance, is -1.54 for EMEs but only -0.89 for industrialized countries. Investment
in EMEs also appears to be more sensitive to the common language and the colonial
links. Taken together, these flndings conflrm our hypothesis on the pecking order,
16We also tested for the importance of censoring, due to a few of the observations in our sample
being zero, by using Heckman?s (1979) two-step procedure. While the results are not shown for
brevity reasons, the point estimates are very similar, underscoring the there is no signiflcant bias
stemming from a censoring problem in our data.
84
Table 2.7: Information frictions: Alternative Estimators
FDI FPIequity FPIdebt Loans FDI FPIequity FPIdebt Loans
Withoutflxedefiects Tobitestimator
GDP-sourcecountry 1.985*** 2.167*** 1.821*** 1.100***(0.065)E,D,L(0.062)F,D,L(0.069)F,E,L(0.074)F,E,D
GDP-hostcountry 0.854*** 1.647*** 1.386*** 1.137***(0.045)E,D,L(0.043)F,D,L(0.048)F,E,L(0.051)F,E,D
Population-sourcecty -1.108***-1.543***-1.100***-0.106(0.071)E,L(0.067)F,D,L(0.075)E,L(0.080)F,E,D
Population-hostcty -0.093 * -0.650***-0.619***-0.326***(0.051)E,D,L(0.048)F,L(0.054)F,L(0.058)F,E,D
distance-0.462***-0.181***-0.460***-0.717***-1.072***-0.988***-0.954***-1.445***(0.064)E,L(0.061)F,D,L(0.067)E,L(0.073)F,E,D(0.051) (0.056) (0.057) (0.054)
commonlanguage0.949*** 1.263*** 0.309 0.307 0.641*** 0.556*** 0.425*** 0.183(0.179)E,D,L(0.169)F,D,L(0.188)F,E(0.202)F,E(0.134) (0.125) (0.127) (0.114)
commonlegalorigin0.940*** 0.603*** 0.574*** 0.805*** 0.704*** 0.547*** 0.380*** 0.271***(0.146)E,D(0.138)F (0.154)F (0.165) (0.092) (0.090) (0.091) (0.080)
coloniallinks1.181*** 0.729*** 0.889*** 1.370*** 1.096*** 0.917*** 0.456** 0.898***(0.282) (0.267)D (0.297)E (0.319) (0.167) (0.168) (0.176) (0.146)
tradeagreement 0.486*** 0.723*** 1.576*** 0.837*** 0.592*** 0.453*** 0.929*** 0.011(0.181)E (0.172)F,D(0.191)E,L(0.205)D (0.146) (0.146) (0.153) (0.152)
investmenttreaty 0.310 ** -0.153 0.048 0.504***-0.129 -0.039 0.002 0.579***(0.125)E (0.119)F,L(0.132)L (0.142)E,D(0.086) (0.087) (0.088) (0.074)
#observations1030 1030 1030 1030 1116 1116 1116 1116(Pseudo)R-squared0.651 0.757 0.679 0.589 0.357 0.486 0.418 0.369
The superscripted letters F,E,D and L indicate the test of the hypothesis H2, that the coe?cient
is difierent to that of FDI, equity portfolio, debt securities, and loans, respectively. ***,**,* show
statistical signiflcance of the coe?cients at the 99%, 95% and 90% levels, respectively.
but also underline the presence of important asymmetries in the efiect of informa-
tion frictions.
One set of explanations that we have not analyzed so far is risk sharing or risk
diversiflcation as a driver of cross-border investment. The motivation for the type
and direction of cross-border capital  ows may not only be information frictions
and institutions but also the attempt to diversify idiosyncratic, home-country risk.
Obstfeld and Rogofi (2000), Lane and Milesi-Ferretti (2004) and Aviat and Cour-
dacier (2007) argue that a source country that receives a high share of its imports
from a particular host country will want to acquire more capital in this speciflc
host country in order to ensure itself against terms of trade shocks to this country.
Extending this argument to risk diversiflcation, it may be optimal for investors to
85
Table 2.8: Information frictions: Developed countries versus emerging market
economiess
FDI FPIequity FPIdebt Loans FDI FPIequity FPIdebt Loans
DevelopedCountries EmergingMarketEconomies
distance-0.893***-0.693***-0.513***-1.047***-1.543***-0.589***-1.019***-1.595***(0.097)E,D(0.076)F,D,L(0.076)F,E,L(0.096)E,D(0.106)E,D(0.095)F,D,L(0.097)F,E,L(0.100)E,D
commonlanguage0.097 0.153 0.175 -0.091 0.942*** 0.444 * 0.336 0.975***(0.203) (0.158) (0.159) (0.200) (0.260)E,D(0.232)F,L(0.238)F,L(0.244)E,D
commonlegalorigin0.975*** 0.655*** 0.393*** 0.853*** 0.550*** 0.543*** 0.411** -0.104(0.144)E,D(0.112)F,D(0.113)F,D,L(0.142)D (0.177)L (0.158)L (0.162)L (0.167)F,E,D
coloniallinks0.681** 0.326 0.397 * -0.218 0.998*** 0.083 0.387 0.851***(0.294)L (0.229) (0.231)L (0.291)F,D(0.330)E (0.294)F,L(0.302) (0.310)E
tradeagreement 0.206 -0.183 1.099*** 0.441 ** 0.212 1.226 * 0.808 0.663(0.218)D(0.170)D,L(0.171)F,E,L(0.216)E,D(0.748) (0.667) (0.684) (0.702)
investmenttreaty 0.150 0.079 0.374 ** 0.879*** 0.238* 0.016 -0.065 -0.006(0.224)L (0.175)L (0.176)L (0.222)F,E,D(0.140)D(0.125) (0.128)F (0.132)
#observations573 573 573 573 543 543 543 543R-squared0.872 0.928 0.917 0.848 0.780 0.857 0.842 0.854
The superscripted letters F,E,D and L indicate the test of the hypothesis H2, that the coe?cient
is difierent to that of FDI, equity portfolio, debt securities, and loans, respectively. ***,**,* show
statistical signiflcance of the coe?cients at the 99%, 95% and 90% levels, respectively.
invest relatively more in those countries with the lowest or even a negative degree
of output correlation to its own.
We therefore add to our benchmark model imports of source country i from
host country j (see left panel of Table 2.9) to investigate whether the flndings for
information frictions change when controlling for proxies of risk sharing. The table
shows that trade is indeed positively correlated with all four types of capital invest-
ment. As an alternative control, we include bilateral real exchange rate volatility,
measured over the period 1990-2003, as a regressor (middle panel of Table 2.9) in
order to test whether uncertainty and risk afiects cross border investment. It is,
however, found to be signiflcant only for investment in debt securities, and to a
lesser degree for FDI.17 We also attempt to control for the efiect of global factors
on cross-border investment. The intuition is that two countries that exhibit a very
17We also tested for the interaction efiect of exchange rate volatility and information, but did
not flnd any additional efiect of this interaction in the empirical model.
86
difierent responsiveness to global shocks should also have less bilateral investment.
We use daily US short-term interest rate changes as our proxy for global shocks, and
take the difierence in the reaction of short-term interest rates between the source
country and host country as our measure of the difierent response to global shocks.18
The right panel of Table 2.9 shows that the difierence in the response to such global
shocks indeed reduces bilateral portfolio investment and loans, though not FDI.
It is important to stress that trade, exchange rate volatility and possibly are
likely to be to some extent endogenous to cross-border investment and one would
need to flnd suitable instruments if one wanted to investigate the link between risk
sharing and capital  ows. However, the important point to note for the objective of
this paper is that information frictions as proxied by distance (or other information
proxies when substituted for distance) retain their signiflcance and the pecking or-
der of FDI and loans to be the most sensitive to information frictions and portfolio
investment the least sensitive is conflrmed.19
Overall, the flrst key result that we take from this section is that there is a
clear pecking order with regard to information frictions. FDI and loans are sub-
stantially more sensitive to information frictions than portfolio investment. The
difierences are large and statistically signiflcant. These flndings are also robust to
several alternative proxies for information frictions, in particular when using tele-
18Short-term interest rates for most countries are 3-month money market rates, if available. The
estimation is based on daily interest rate changes over the period 1990 to 2004.
19As a flnal check, we flnd that the results are robust to using alternative country samples, i.e.
our pecking order hypothesis in that FDI and loans are most sensitive to information frictions
is conflrmed for both emerging market countries as well as industrialized countries. Results are
available upon request.
87
Table
2.9:
Robustness
tests
with
trade,
exc
hange
rate
volatilit
yand
global
interest
rate
sho
cks
as
con
trols
FDI
FPI
equit
y
FPI
debt
Loans
FDI
FPI
equit
y
FPI
debt
Loans
FDI
FPI
equit
y
FPI
debt
Loans
With
control
for
trade
With
control
for
exchange
rate
volatilit
y
With
control
for
global
interest
rate
shoc
ks
distance
-0.692
***
-0.362
***
-0.689
***
-0.773
***
-1.232
***
-0.655
***
-0.694
***
-1.246
***
-0.995
***
-0.605
***
-0.563
***
-1.063
***
(0.088)
E(0.074)
F,D,L
(0.083)
E(0.088)
E(0.073)
E,D
(0.062)
F,L
(0.068)
F,L
(0.074)
E,D
(0.079)
E,D
(0.063)
F,L
(0.076)
F,L
(0.083)
E,D
trade
-imp
orts
0.384
***
0.247
***
0.119
***
0.382
***
(0.047)
E,D
(0.040)
F,D,L
(0.044)
F,E,L
(0.047)
E,D
Exchange
rate
volatilit
y
10.875
*
-4.388
-23.816
***
3.136
*
(5.968)
D
(5.028)
D
(5.502)
F,E,L
(6.004)
D
Global
interest
rate
shoc
k
0.191
-0.389
***
-0.345
**
-0.322
**
(0.150)
E,D,L
(0.119)
F(0.142)
F(0.158)
F
common
language
0.384
**
0.334
**
0.037
0.176
0.454
***
0.316
**
0.066
3.136
*
0.368
**
0.257
*
0.050
0.025
(0.161)
D
(0.136)
D
(0.152)
F,E
(0.162)
(0.160)
(5.028)
D
(5.502)
E(6.004)
D
(0.175)
D,L
(0.139)
(0.166)
F(0.184)
F
common
legal
origin
0.603
***
0.500
***
0.390
***
0.330
***
0.703
***
0.572
***
0.417
***
0.435
***
0.628
***
0.538
***
0.443
***
0.631
***
(0.113)
L(0.095)
(0.106)
(0.113)
F(0.112)
D;L
(0.094)
(0.103)
F(0.113)
F(0.124)
D
(0.098)
(0.118)
F(0.184)
colonial
links
0.811
***
0.208
0.135
0.209
0.922
***
0.334
*
0.200
0.321
0.905
***
0.380
*
0.221
0.304
(0.216)
E,D,L
(0.183)
F(0.203)
F(0.218)
F(0.215)
E(0.181)
(0.199)
(0.217)
(0.252)
E,D,L
(0.199)
F(0.239)
F(0.265)
F
trade
agreemen
t
-0.075
-0.283
*
0.632
***
0.267
-0.068
-0.376
**
0.399
**
0.258
0.080
-0.170
0.877
***
0.463
**
(0.173)
D
(0.147)
D,L
(0.163)
F,E,L
(0.175)
E,D
(0.183)
D,L
(0.154)
D,L
(0.169)
F,E
(0.184)
F,E
(0.184)
D,L
(0.145)
D,L
(0.174)
F,E
(0.193)
F,E
investmen
ttreat
y
0.042
-0.038
-0.010
0.297
**
0.222
*
0.042
0.177
*
0.419
***
0.458
***
0.335
***
0.330
**
0.532
***
(0.118)
L(0.100)
D
(0.111)
E,L
(0.119)
F,E,D
(0.115)
L(0.096)
L(0.106)
L(0.115)
F,E,D
(0.144)
(0.113)
(0.137)
(0.152)
#observ
ations
1027
1027
1027
1027
782
782
782
782
1027
1027
1027
1027
R-squared
0.841
0.911
0.883
0.851
0.845
0.920
0.895
0.849
0.830
0.912
0.876
0.823
The
sup
erscripted
letters
F
,E
,D
and
L
indicate
the
test
of
the
hyp
othesis
H2
,that
the
co
e?cien
tis
difieren
tto
that
of
FDI,
equit
y
portfolio,
debt
securities,
and
loans,
resp
ectiv
ely
.***,**,*
sho
w
statistical
signiflcance
of
the
co
e?cien
ts
at
the
99%,
95%
and
90%
lev
els,
resp
ectiv
ely
.
88
phone tra?c. Moreover, various robustness tests conflrm the speciflcation of the
model and underline the robustness of the results on the pecking order hypothesis
to alternative speciflcations and difierent econometric estimators.20 Thus, the re-
sults indicate that FDI and loans are more sensitive to information frictions - or
more information-intensive - than portfolio investment, equity and debt. A possible
explanation for this fact is that FDI and loans in general require frequent interaction
and a deeper knowledge of the markets where they operate. Also, especially for the
case of FDI, once an asset has been acquired, direct ownership makes the asset less
liquid given the potential lemon problem in case of a re-sale as Goldstein and Razin
(2006) point out. Thus, FDI becomes partially irreversible or costlier to liquidate,
and therefore more sensitive to information in the flrst place.21
2.5 The role of institutions and flnancial market development
We now turn to the role of flnancial markets and institutions. The central
focus is on the question of whether we can identify a pecking order of cross-border
capital positions with regard to the degree of development and openness of markets
and the quality of institutions in the host country. For this purpose, we extract
the host country flxed efiects from model (2.1) and then estimate model (2.2), i.e.
we attempt to explain the host country flxed efiects through market conditions and
20We have also conducted further robustness checks, especially splitting the sample between
industrialized and emerging economies and the results hold for both groupings of countries. They
can be found in an earlier working paper version (Daude and Fratzscher, 2006).
21Although we cannot distinguish between greenfleld investments and mergers and acquisitions
in our data, this informational friction is linked to ownership control and thus applies to both
types of FDI.
89
institutions. Note that given the speciflcation of model (2.1) where the dependent
variable is measured in value terms, we need to control for size efiects in model
(2.2). We do so by including host country GDP in each of the speciflcations below,
though we omit showing the point estimates for this variable for brevity reasons.
All variables used are described in more detail in the appendix.
We start with the role of market development and openness. We use three
difierent proxies. First, we employ a capital account openness dummy. This dummy
takes the value of one if the country had fully liberalized its capital account by the
mid-1990s, and is zero otherwise. Data for this variable comes from the IMF?s
Annual Report of Exchange Arrangements and Exchange Restrictions (AREAER).
The flnding is remarkably strong as portfolio equity and portfolio debt investment
react strongly to capital account openness, whereas the coe?cients for FDI and
loans are positive but only marginally statistically signiflcant (see Table 2.10). The
magnitude of the efiects is large: a country that is open receives about 80% more
equity capital and 80% more debt investment compared to an economy with a closed
capital account.
Second, we investigate the efiect of the development of the domestic flnancial
sector on the pecking order. We include credit to the private sector as a proxy
for flnancial development. Table 2.10 shows that the elasticities are by the far
the largest for equity investment, which is about twice as large as that for debt
securities and FDI. These difierences are statistically signiflcant, while in the case
of FDI investment appears to not react to changes in the degree of flnancial market
90
Table
2.10:
Role
of
mark
et
dev
elopmen
tand
transparency
FDI
FPI
equit
y
FPI
debt
Loans
FDI
FPI
equit
y
FPI
debt
Loans
Mark
et
openness
and
dev
elopmen
t:
Transparency:
capital
accoun
t
0.262
0.825
**
0.803
**
0.387
qualit
yof
0.137
*
0.389
***
0.191
0.134
openness
(0.232)
(0.354)
(0.372)
(0.288)
disclosure
(0.074)
E
(0.111)
F,D,L
(0.121)
E
(0.093)
E
#observ
ations
69
69
69
69
#observ
ations
65
65
65
65
R-squared
0.7556
0.8184
0.6801
0.7019
R-squared
0.7449
0.8355
0.6968
0.7079
flnancial
0.462
2.270
***
1.396
***
1.344
***
accoun
ting
0.019
0.067
***
0.033
**
0.024
**
dev
elopmen
t
(0.321)
E,D,L
(0.424)
F,D,L
(0.469)
F,E
(0.366)
standards
(0.011)
E
(0.014)
F,D,L
(0.015)
E
(0.012)
E
#observ
ations
64
64
64
64
#observ
ations
37
37
37
37
R-squared
0.7311
0.8693
0.7537
0.7575
R-squared
0.5543
0.8117
0.6446
0.6261
sto
ck
mark
et
0.435
***
1.104
***
0.560
**
0.743
***
prop
ert
y
-0.139
-0.847
***
-0.904
***
-0.570
***
capitalisation
(0.126)
E,L
(0.131)
F,D,L
(0.219)
E
(0.133)
F,E
righ
ts
(0.118)
E,D,L
(0.158)
F,L
(0.156)
F,L
(0.133)
F,E,D
#observ
ations
46
46
46
46
#observ
ations
63
63
63
63
R-squared
0.6589
0.888
0.6045
0.7349
R-squared
0.741
0.868
0.7992
0.7684
The
sup
erscripted
letters
F
,E
,D
and
L
indicate
the
test
of
the
hyp
othesis
H4
,that
the
co
e?cien
tis
difieren
tto
that
of
FDI,
equit
yp
ortfolio,
debt
securities,
and
loans,
resp
ectiv
ely
.***,**,*
sho
w
statistical
signiflcance
of
the
co
e?cien
ts
at
the
99%,
95%
and
90%
lev
els,
resp
ectiv
ely
.
91
development in the host country.
Third, we analyze the role of the development of the local stock market, and
proxy this through stock market capitalization. The bottom panel of Table 2.10
indicates again that equity investment is most strongly related to changes in market
capitalization but nevertheless also cross-border investment in debt securities, loans
and FDI react, though to a lesser extent.
As a next step, we analyze the role of institutions for the pecking order of cross-
border capital positions. As discussed in section 2.2, there have been a number of
studies arguing that difierent types of capital should react difierently to various
institutional features. For instance, Albuquerque?s (2003) model implies that FDI
is harder to expropriate as the information required for and obtained by FDI is
inalienable. Various other studies have focused on individual types of capital  ows
and how they are linked to other institutional elements such as e.g. corruption,
transparency and political risk etc. (e.g. Wei 2000a, Papaioannou 2005, Gelos and
Wei 2005).
We test the efiect of various institutional features. While it is hard to deter-
mine which institutional factors to focus on, we are guided in our choice of insti-
tutional variables by the mostly theoretical literature discussed in section 2.2. The
sources for these variables are manifold, partly stemming from the work by La Porta
et al. (1998), Djankov et al. (2002) and partly from the databases by the World
Bank Doing Business and by the International Country Risk Guide (ICRG).
Tables 2.10 and 2.11 show the flndings for three sets of institutional variables.
92
Table
2.11:
Role
of
inv
estor
protection
and
corruption
FDI
FPI
equit
y
FPI
debt
Loans
FDI
FPI
equit
y
FPI
debt
Loans
Inv
estor
protection:
Corruption:
expropriation
0.054
0.913
***
0.952
***
0.492
***
TI
-0.048
-0.407
***
-0.372
***
-0.198
***
risk
(0.125)
E,D,L
(0.166)
F,L
(0.175)
F,L
(0.143)
F,E,D
(0.049)
E,D,L
(0.059)
F,L
(0.061)
F,L
(0.052)
F,E,D
#observ
ations
66
66
66
66
#observ
ations
61
61
61
61
R-squared
0.7497
0.8644
0.7562
0.7326
R-squared
0.7508
0.8835
0.7896
0.7618
repudiation
0.078
0.781
***
0.701
***
0.445
***
WDR
-0.146
-0.583
***
-0.434
***
-0.335
***
costs
(0.093)
E,D,L
(0.115)
F,L
(0.132)
F,L
(0.103)
F,E,D
(0.099)
E,D,L
(0.137)
F,L
(0.143)
F
(0.107)
F,E
#observ
ations
66
66
66
66
#observ
ations
56
56
56
56
R-squared
0.7516
0.8832
0.7528
0.7542
R-squared
0.733
0.8372
0.7109
0.7382
days
of
-0.103
-0.626
***
-0.573
**
-0.277
German
surv
ey
-0.025
-0.254
***
-0.242
***
-0.113
**
enforcemen
ts
(0.147)
E,D
(0.222)
F,L
(0.229)
F
(0.182)
E
(0.036)
E,D,L
(0.048)
F,L
(0.048)
F,L
(0.044)
F,E,D
#observ
ations
65
65
65
65
#observ
ations
57
57
57
57
R-squared
0.7335
0.8258
0.7127
0.7091
R-squared
0.712
0.8477
0.763
0.7051
The
sup
erscripted
letters
F
,E
,D
and
L
indicate
the
test
of
the
hyp
othesis
H4
,that
the
co
e?cien
tis
difieren
tto
that
of
FDI,
equit
y
portfolio,
debt
securities,
and
loans,
resp
ectiv
ely
.***,**,*
sho
w
statistical
signiflcance
of
the
co
e?cien
ts
at
the
99%,
95%
and
90%
lev
els,
resp
ectiv
ely
.
93
First, we look at the role of transparency. For this, we employ both a measure on
the quality of information disclosure and on the quality of the accounting standards
required by law in the host country - with higher values indicating a better quality.
For both measures, portfolio equity investment reacts the strongest to changes in
these transparency measures, while in the case of accounting standards the coe?-
cient for debt securities and loans are also signiflcant at a 10% level. FDI and loans
are the least responsive. In fact, the elasticity of equity investment is about three
times larger than that for FDI and for loans.
Second, we analyze the role of investor protection (last regression in Table 2.10
and Table 2.11). In particular, a lower risk of expropriation - indicated by a higher
value of the variable in the table - has a highly signiflcant impact mainly on portfolio
investment. By contrast, the elasticity of loans is only about one half of that of
portfolio investment, while FDI does not react at all to difierences in expropriation
risk. This flnding thus provides strong support for the hypothesis formulated by
Albuquerque (2003) and is line with the stylized facts presented above in section
2.3.
Moreover, Table 2.10 shows that an improvement in the quality of property
rights - indicated by a decline in the variable in the table - has a signiflcant and the
largest impact on portfolio equity and debt investment, a lower efiect on loans, but
no efiect on FDI. An almost identical picture emerges for repudiation costs and for
the quality of enforcement of laws and regulations - which is measured in the days it
takes to enforce a particular ruling, so that a higher number for the latter indicates
a worse system of enforcement. Overall, all three measures therefore indicate that
94
investor protection has the largest efiect on portfolio investment but does not appear
to have any signiflcant efiect on FDI stocks.
Third, we analyze the importance of corruption for the pecking order. We use
three alternative proxies for corruption; a flrst one from Transparency International,
a second one from the World Development Report of the World Bank and the third
one from a survey of German manufacturing flrms. All three indicators have been
used previously by Wei (2000b). In all cases, a higher value indicates a higher degree
of corruption. Overall, the same flnding emerges for all three of the proxies: cor-
ruption has the strongest negative efiect on portfolio investment and some, though
smaller efiect on loans. Corruption does not appear to have any signiflcant efiect on
FDI. This flnding is in line with Daude and Stein (2007) who do not flnd a robust
relation between difierent corruption indicators and FDI in contrast to other institu-
tional indicators.22 We conduct various sensitivity tests to check for the robustness
of these flndings. For instance, we flnd very similar results when controlling also for
GDP per capita in model (2.2). The stylized facts of section 2.3 underline that there
is a high correlation between per capita GDP and the pecking order of cross-border
capital positions. However, the fact that the results hold also when controlling for
GDP per capita stresses that market development and institutions have a large and
signiflcant efiect on the pecking order independent of the level of development of a
country.
As a further important sensitivity test, we use an IV estimator to take into
account the possibility that institutional arrangements and market development
22See also chapter 3 of the present dissertation.
95
Table
2.12:
Role
of
mark
et
dev
elopmen
tand
transparency
-3SLS
estimator
FDI
FPI
equit
y
FPI
debt
Loans
FDI
FPI
equit
y
FPI
debt
Loans
Mark
et
openness
and
dev
elopmen
t:
Transparency:
capital
accoun
t
0.828
*
1.796
**
2.953
***
1.112
*
qualit
yof
-0.035
0.548
***
0.117
0.147
openness
(0.460)
E,D
(0.717)
F,D
(0.850)
F,E,L
(0.572)
D
disclosure
(0.136)
E
(0.199)
F,D,L
(0.215)
E
(0.165)
E
#observ
ations
65
65
65
65
#observ
ations
65
65
65
65
R-squared
0.7162
0.8015
0.5682
0.6843
R-squared
0.7238
0.8303
0.695
0.7079
flnancial
0.898
*
2.376
***
1.491
*
2.207
***
accoun
ting
0.019
0.069
***
0.039
*
0.022
dev
elopmen
t
(0.538)
E,L
(0.700)
F
(0.775)
(0.630)
F
standards
(0.015)
E
(0.018)
F,D,L
(0.019)
E
(0.015)
E
#observ
ations
64
64
64
64
#observ
ations
37
37
37
37
R-squared
0.7233
0.8692
0.7535
0.7364
R-squared
0.5543
0.8117
0.6434
0.6256
sto
ck
mark
et
0.520
***
1.036
***
0.595
*
1.027
***
prop
ert
y
0.008
-1.017
***
-1.015
***
-0.690
***
capitalisation
(0.199)
E,L
(0.205)
F
(0.341)
(0.218)
F
righ
ts
(0.154)
E,D,L
(0.204)
F,L
(0.201)
F,L
(0.171)
F,E,D
#observ
ations
45
45
45
45
#observ
ations
63
63
63
63
R-squared
0.6374
0.8793
0.6017
0.697
R-squared
0.7347
0.8656
0.7976
0.7654
The
sup
erscripted
letters
F
,E
,D
and
L
indicate
the
test
of
the
hyp
othesis
H4
,that
the
co
e?cien
tis
difieren
tto
that
of
FDI,
equit
yp
ortfolio,
debt
securities,
and
loans,
resp
ectiv
ely
.***,**,*
sho
w
statistical
signiflcance
of
the
co
e?cien
ts
at
the
99%,
95%
and
90%
lev
els,
resp
ectiv
ely
.
96
may be. We estimate the system using a three-stage least square estimator (3SLS),
which in essence implies instrumenting the institutional variables. An additional
advantage of this approach is that we also address potential measurement errors in
the institutional variables with our estimation technique.
We draw our instruments from the literature on law and flnance and the lit-
erature on institutions and economic development. Speciflcally, we use legal origin
dummies and dummies for religion which have been found to be important determi-
nants of flnancial markets development and regulations (see La Porta et al. 1997,
1998). In the case of institutions, we use the mortality of settler from Acemoglu et
al (2001). Our approach therefore also draws on the work by Alfaro et al. (2005).
The results for the 3SLS estimates are given in Tables 2.12 and 2.13. Overall,
the key point is that the results are highly robust to those without instrumenting
the institutions. All the results described above are qualitatively identical when
using 3SLS, underlining that portfolio investment is substantially more sensitive to
institutions and market development than FDI, and to some extent also than loans.
It is also reassuring to observe that with the IV estimates the efiects on FDI turn
signiflcant, but still signiflcantly smaller than for portfolio investment. Moreover,
the fact that the size of the coe?cients and their signiflcance increase somewhat
also helps to stress the robustness of the results.
In summary, we flnd that market development and institutions are strongly
related to the pecking order of cross-border investment. The key flnding of this
section is that portfolio investment, in particular in equity securities, is the type of
97
Table
2.13:
Role
of
mark
et
dev
elopmen
tand
transparency
-3SLS
estimator
FDI
FPI
equit
y
FPI
debt
Loans
FDI
FPI
equit
y
FPI
debt
Loans
Inv
estor
protection:
Corruption:
expropriation
-0.407
1.397
***
1.430
***
0.273
TI
0.108
-0.371
***
-0.346
***
-0.145
**
risk
(0.259)
E,D,L
(0.328)
F,L
(0.352)
F,L
(0.277)
F,E,D
(0.068)
E,D,L
(0.078)
F,L
(0.080)
F,L
(0.069)
F,E,D
#observ
ations
63
63
63
63
#observ
ations
61
61
61
61
R-squared
0.6678
0.8488
0.7279
0.7228
R-squared
0.7096
0.8828
0.7889
0.7579
repudiation
-0.128
0.957
***
0.802
***
0.434
**
WDR
0.133
-0.825
***
-0.755
***
-0.450
***
costs
(0.160)
E,D,L
(0.197)
F,L
(0.223)
F,L
(0.172)
F,E,D
(0.151)
E,D,L
(0.200)
F,L
(0.213)
F,L
(0.155)
F,E,D
#observ
ations
63
63
63
63
#observ
ations
56
56
56
56
R-squared
0.709
0.8758
0.7506
0.7551
R-squared
0.6955
0.8281
0.6849
0.7329
days
of
0.184
-0.954
***
-0.602
-0.628
**
German
surv
ey
0.106
-0.272
***
-0.204
***
-0.050
enforcemen
ts
(0.243)
E,D,L
(0.363)
F
(0.368)
F
(0.300)
F
(0.064)
E,D,L
(0.076)
F,L
(0.077)
F,L
(0.072)
F,E,D
#observ
ations
65
65
65
65
#observ
ations
57
57
57
57
R-squared
0.718
0.82
0.7126
0.6925
R-squared
0.647
0.8473
0.7603
0.6945
The
sup
erscripted
letters
F
,E
,D
and
L
indicate
the
test
of
the
hyp
othesis
H4
,that
the
co
e?cien
tis
difieren
tto
that
of
FDI,
equit
y
portfolio,
debt
securities,
and
loans,
resp
ectiv
ely
.***,**,*
sho
w
statistical
signiflcance
of
the
co
e?cien
ts
at
the
99%,
95%
and
90%
lev
els,
resp
ectiv
ely
.
98
capital that is the most sensitive to difierences in market development/openness and
the quality of host country institutions. A second key result is that FDI appears to
bethetypeofcapitalthatismostimmunetothequalityofdomesticinstitutions. We
flnd that FDI is least sensitive in all institutional categories, including with regard to
transparency, investor protection, to the degree of corruption and to expropriation
risk.
2.6 Conclusions
Is there a pecking order of cross-border investment in that countries become
flnancially integrated primarily through one type of investment rather than others?
The perceived wisdom in much of the debate on flnancial integration and trade in
flnancial assets is that FDI constitutes a type of investment that is desirable from
a host country perspective because it brings about a transfer of know-how, creates
access to foreign markets and reduces the risks of flnancial distress. However, the
facts of cross-border capital positions also show that countries that are richer, have
higher growth and better institutions receive a higher share of their foreign invest-
ment in the form of portfolio investment and a much lower share through FDI and
loans.
The objective of this chapter has been to analyze whether there is a natural
pecking order in cross-border investment. We focus on the role of two key deter-
minants for the trade in flnancial assets that have been central in this literature in
99
recent years: the importance of information frictions, and the role of institutions.
Recent theoretical contributions to this literature emphasize the importance of dif-
ferences in the ownership structure of difierent forms of investment. In particular,
FDI has stronger ownership implications and thus tends to be more information
sensitive than portfolio equity or debt investment. A second strand of the literature
has focused on the implications of this theory for the role of institutions. One line
of reasoning is that due to the larger information sensitivity of FDI, it is also harder
to expropriate and thus it may be more immune to difierences in the quality of
institutions and market development.
Our contribution is to test these hypotheses empirically for a broad set of
countries. To our knowledge, this is the flrst study that provides a comprehensive
comparison of all four types of cross-border investment - distinguishing between
FDI, portfolio equity securities, debt securities as well as loans. We develop and use
a unique, combined data source of the capital stocks, rather than capital  ows, for
77 countries.
The empirical results are compelling and conflrm our hypotheses on the com-
position of cross-border investment. First, information frictions across countries are
an important determinant of the pecking order of cross-border capital positions. In
line with the theory on the capital structure of the flrm, we flnd that FDI, and
to some extent loans, are the most sensitive types of capital to information fric-
tions, whereas portfolio investment is much less responsive. The magnitude of these
pecking order efiects is large: FDI and loans are about 1.5 to 2 times more sen-
100
sitive to information frictions than equity and portfolio investment. This flnding
is robust to several sensitivity tests, including the use of alternative proxies for in-
formation frictions; various speciflcations of the econometric model; controlling for
other determinants, such as risk diversiflcation; and across country samples, both
for industrialized and for emerging market economies.
The second key result of the analysis is that the degree of market develop-
ment and the quality of host country institutions are important determinants of
the pecking order of cross-border investment. We flnd that portfolio investment is
substantially more sensitive than FDI and loans to both market development - such
as the openness of the capital account and the development of the domestic flnancial
sector - and to domestic institutional features. We use three proxies for the quality
of institutions - the degree of transparency, investor protection and corruption - and
show that this result is robust across all these difierent elements of host country
institutions. These results conflrm some hypotheses formulated in the literature but
contradict others. For instance, in line with the argument by Albuquerque (2003),
we flnd that FDI does not react to difierences to the risk of expropriation, whereas
portfolio equity and debt investment is highly sensitive to this risk. Similarly, we do
not flnd that corruption has a more detrimental efiect on FDI, as hypothesized in
the literature, but that the magnitude of FDI is not sensitive to corruption, whereas
portfolio investment is. This implies that in fact corruption tilts the composition
of foreign investment signiflcantly towards FDI, and to a lesser extent towards loans.
101
Ourflndingshaveanumberofimportantpolicyimplications. Inparticular, the
empirical results indicate that a large share of foreign investment that takes the form
of FDI - despite the various beneflts FDI may ultimately entail - may not necessarily
be a blessing, but may in fact also be a signal of some underlying weaknesses - either
intermsofweakinstitutionsorintermsofthepoorfunctioningorunderdevelopment
of domestic flnancial markets - of the host country. By contrast, a large share of
foreign investment that comes through portfolio equity or debt securities is likely,
at least in part, to signal well-functioning domestic flnancial markets and the trust
of foreign investors in domestic institutions.
102
Chapter 3
The Quality of Institutions and Foreign Direct Investment
(co-authored with Ernesto Stein)
3.1 Introduction
One of the most notorious features of the trend towards globalization in re-
cent times has been the increased importance of foreign direct investment (FDI)
around the world. Over the last couple of decades, worldwide FDI  ows have grown
by a factor of almost 10. To put this evolution in perspective, trade  ows around
the world, by comparison, only doubled during a similar period. In this context,
a deeper understanding of the determinants of the location of multinational enter-
prises is becoming more and more relevant for the design of successful policies to
attract investors.
While the existing literature has focused mainly on the efiects of corruption
or political risk on FDI, we contribute to the literature by testing a broader set of
institutional variables that may afiect the decision of foreign investors to undertake
investment projects in a particular country.1 This also allows us to assess what
1Among the papers that focus on the impact of political risk on US investment abroad are Fathi-
Sedeh and Saflzadeh (1989, 1994), Loree and Guisinger (1995), and Schneider and Frey (1985),
among others. For a more recent study see Sethi et al (2003). Schollhammer and Nigh (1987)
focus on the impact of international con icts German FDI. In general, the main message from this
literature is that the evidence for political instability as a signiflcant determinant is weak.
103
dimensions of the quality of governance institutions afiect foreign investors? location
decisions more. In addition, while most papers in the literature analyze the efiects
of host country institutions on investment from a particular source, we use a large
sample of bilateral investment data. The use of bilateral data allows us also to esti-
mate the impact of institutional variables embedded in an empirical model backed
by FDI theory rather than ad-hoc formulations as most studies in the literature.
The impact of institutions on investment, either domestic or foreign can be
related to two difierent channels. First,\bad" institutions might act as a tax by
increasing the cost of doing business. Second, imperfect enforcement of contracts
might also increase uncertainty regarding future returns and thus have a negative
impact on the level of investment.2 Thus, for example corruption may deter invest-
ment by increasing the cost of doing business, as investors need to bribe o?cials
in order to obtain licenses and permits. In addition, corruption may increase un-
certainty, which may deter investment as well. According to Shleifer and Vishny
(1993), the secrecy of corruption is what makes it much more distortionary than
taxes.
In the empirical literature of FDI location decisions an often-cited paper re-
lated to ours is Wheeler and Mody (1992). They flnd that a composite measure
2Although this seems to be a natural argument, Dixit and Pindyck (1993) show that impor-
tant restrictive assumptions are required to create a negative efiect of uncertainty on the level
investment. See Stasavage (2002) for an empirical application that analyzes the efiects on do-
mestic investment of checks and balances, as a mechanism to reduce time-inconsistency in capital
taxation.
104
of risk factors, which includes institutional variables such as the extent of bureau-
cratic red tape, political instability, corruption and the quality of the legal system,
does not afiect the location of US foreign a?liates. However, their index aggregates
these variables together with others such as attitudes towards the private sector,
living environment, inequality, risk of terrorism, etc, making it impossible to assess
the role of individual variables. In particular, the question of whether any of the
institutional aspects have a signiflcant impact on FDI is left unanswered. Mauro
(1995) represents the flrst systematic empirical study on a related topic. He shows
that corruption has a negative impact on the ratio of total and private investment
to GDP and therefore causes harm to economic growth.3Wei (2000), using data on
bilateral FDI stocks from OECD countries, flnds that corruption has an economi-
cally signiflcant and negative impact on FDI. His results imply that an increase in
the level of corruption from Singapore to that of Mexico is equivalent to increasing
the tax rate on multinationals by more than twenty percentage points. In addition,
Wei (1997) flnds that uncertainty regarding corruption has also important negative
efiects on FDI location.4
However, investment decisions may depend on difierent dimensions of public
institutions in addition to corruption, like the regulatory framework, the predictabil-
ity of economic policy, the protection of property right, or the e?ciency of law en-
forcement. In this chapter, we provide evidence on the impact of these difierent
3See also Henisz (2000) for an analysis of the impact of formal political institutions on economic
growth, rather than outcome variables like those used by Mauro (1995).
4See Smarzynska and Wei (2000) for a flrm-level study in transition economies of the impact of
corruption on FDI.
105
dimensions of governance institutions on the location of FDI. In addition, we use
difierent types of institutional variables based on experts? reports, surveys and a
combination of both in order to ensure the robustness of our results.
A difierent literature that is related to the present chapter includes Albu-
querque (2003), Aizmann and Spiegel (2002), Hausmann and Fernndez-Arias (2000),
as well as Mody et al (2003). This literature focuses on the efiects of institutions
on the composition of capital  ows. Albuquerque?s paper develops an imperfect en-
forcement model, where FDI has a risk-sharing advantage over other capital  ows,
because it contains more intangible assets that are inalienable and make FDI there-
fore less attractive to expropriation. The optimal contract implies that share of FDI
in total capital  ows is higher for flnancially constrained countries. In a set of cross-
country regressions with the average FDI shares in gross private capital  ows as
dependent variable and controlling by GDP per capita and trade openness, he flnds
that the ICRG variable of Law and Order has a negative but not signiflcant efiect.
However, once credit ratings are included in the regression, the institutional quality
has a positive and signiflcant efiect on the FDI share. Mody et al present a model
where multinational flrms have an advantage over domestic flrms in the screening
process of projects with a noisy signal concerning their real level of profltability. In
this context, the value of this advantage is decreasing in the host country?s degree
of corporate transparency. Thus, their model predicts that the proportion of FDI in
comparison to portfolio investment is lower in countries where institutions are more
transparent. They present empirical evidence in favor of this prediction, using an
106
index of creditors? rights from La Porta et al (2000) in a gravity model to explain
the ratio of FDI  ows to trade. Aizmann and Spiegel present an e?ciency wage
model where ex-post monitoring costs and enforcement of labor contracts are lower
for domestic flrms than for multinationals, but the later are more productive. In
this situation, multinationals will be more sensitive to changes in the enforcement
cost (quality of institutions) and pay higher wages than domestic flrms do. They
flnd that the share of FDI to gross flxed investment, as well as the ratio of FDI to
private domestic investment, is negatively and signiflcantly correlated with the level
of corruption, such that FDI seems to be more sensitive than domestic investment
to the institutional quality. Finally, Hausmann and Fernndez-Arias study the efiects
of institutional variables on the composition of capital in ows, using six difierent
institutional variables compiled by Kaufmann et al (1999a), as well as indices of
creditor and shareholder rights from La Porta et al (1998).5 The authors flnd that
better institutions lead to a reduction of the share of in ows represented by FDI.
They conclude that, in comparison to FDI, other forms of capital are more sensitive
to the quality of institutions. When they look at the efiects of their institutional
variables on FDI as a share of GDP, only a small subset of the institutional vari-
ables - regulatory quality, government efiectiveness and shareholder rights - remain
signiflcant after including some controls. Their summary index of institutions, the
flrst principal component of the six institutional variables of Kaufmann et al, does
not have signiflcant efiects on the ratio of FDI to GDP.
5The institutional variables from Kaufmann et al (1999a) are regulatory quality, voice and ac-
countability, government efiectiveness, political stability and lack of violence, control of corruption
and rule of law. We will describe these in more detail below, as we will use them here as well.
107
Unlike these studies, our focus is on FDI per se, rather than on the composi-
tion of capital in ows. As in Wei (1997, 2000), we use bilateral data on FDI stocks,
but we use a wider range of institutional indicators. The use of bilateral data allows
us to use a much richer set of control variables.
Another contribution of our analysis is that we avoid the shortcoming of the
existing empirical literature, especially the studies that analyze the efiects of the
some institutional dimensions on FDI as Wheeler and Mody (1992), Hausmann and
Fernndez-Arias (2000), Wei (1997, 2000), and Mody et al (2003), that they rely on
ad-hoc empirical speciflcations. In this sense, we test the signiflcance of the quality
of institutions on FDI in an empirical model that follows recent developments in
the theory of multinational enterprise location (see Markusen, 1997 and 2001) more
closely. Carr, et al (2001) and Blonigen et al (2002) have used very similar econo-
metrics speciflcations recently.
The rest of the chapter is organized as follows: In Section 2, we present the
data, and discuss our empirical strategy. Section 3 presents our main results on the
institutional quality as a determinant of the location of FDI, while in Section 4 we
perform some sensitivity analysis and robustness checks. In Section 5 we extend to
the efiects of institutions over time using a panel data analysis. Finally, in Section
6 we present our main conclusions.
108
3.2 Data and Empirical Strategy
3.2.1 FDI Data
We use bilateral outward FDI stocks from the UNCTAD FDI database. The
dataset covers FDI from 34 source countries, most of them developed, to 152 host
countries.6 By using outward stocks, we ensure that difierences across countries in
the deflnition and measurement of FDI do not alter the relative allocation of FDI
for each of the source countries. The data are available from 1982 to 2002. For the
cross-section analysis, we use the 2002 information.7
3.2.2 Institutional Variables
In order to assess the role of institutions as a determinant of the location of
FDI, we primarily use a set of institutional variables developed by Kaufmann et al
(1999a). These indicators are constructed based on information gathered through a
wide variety of cross-country surveys as well as polls of experts. The authors use a
model of unobserved components, which enables them to achieve levels of coverage
of approximately 160 countries for each of their indicators.8They construct six dif-
ferent indicators, each representing a difierent dimension of governance: Voice and
6Thus the number of annual observations is 34 x (152 - 1) = 5134. However, data availability
in our regressions reduces signiflcantly the efiective number of observations. In addition, most of
the observations present no investment at all (around 75 percent are 0). An important part of our
robustness checks deals with this issue.
7This dataset has become a primary source for empirical studies. For example, Daude and
Fratzscher (2007) - chapter 2 - use the same database to study the impact of information frictions
on the composition of cross-border investments; Daude et al (2007) study the impact of regional
integration agreements on FDI using UNCTAD FDI data.
8For more technical details see Kaufmann et al (1999b). The database is available at
http://www.worldbank.org/wbi/governance/govdata/index.html
109
Accountability, Political Stability and Lack of Violence, Government Efiectiveness,
Regulatory Quality, Rule of Law, and Control of Corruption. This clustering of
institutional indicators into difierent dimensions allows us to study whether some
dimensions of governance matter for FDI location, while others do not. The indi-
cators are recoded such that they all have mean zero and unit standard deviation.
In all cases, larger values indicate better institutions. We re-standardize these vari-
ables to have mean zero and standard deviation of one in our own sample, in order
to simplify the interpretation of the coe?cients, as well as the comparison of their
relative importance. In order to reduce simultaneity problems, we consider data for
1996, the earliest date these indicators are available.9 Thus, in the cross-section
analysis our institutional variables precede the stock of FDI by 6 years.
Voice and Accountability, as well as Political Stability and Lack of Violence
aggregate those aspects related to the way authorities are selected and replaced.
The flrst variable focuses on difierent indicators related to the political process, civil
rights, and institutions that facilitate citizens? control of government actions, such
as media independence. The second variable combines indicators that measure the
risk of a destabilization or removal from power of the government in a violent or
unconstitutional way.
The indicators clustered in Government Efiectiveness and in Regulatory Qual-
9The correlation for all indicators between their value in 1996 and the value in 2002 is above
0.95 except Political Stability, which has a correlation of 0.85. Thus, as it is common knowledge
these institutional aspects tend to change slowly over time and identiflcation will mainly come
from the cross-section variation in the data.
110
ity are related to the ability of the government to formulate and implement policies.
The flrst variable aggregates indicators on the quality of bureaucracy, the compe-
tence of civil servants, the quality of public service provision and the credibility of
the government?s commitment to its policies. The second brings together indicators
related to the content of policies, like the existence of market-unfriendly regulations
such as price controls and other forms of excessive regulation.
The last two variables, Rule of Law and Control of Corruption, consider as-
pects related to the respect, on the part of both citizens and the government, for
the institutions that resolve their con icts and govern their interactions. The flrst
one includes variables that measure the perceptions on the efiectiveness and pre-
dictability of the judiciary, as well as enforceability of contracts, while the second
aggregates difierent indicators of corruption.
In Table 3.1, we present the simple correlations between the six variables, and
the partial correlation between them controlling for GDP per capita. There is a
remarkably signiflcant correlation between the variables, even when controlling for
GDP per capita. For example, in our sample, the simple correlation between Rule
of Law and Control of Corruption is 0.93 and it remains at 0.76 once we control for
GDP per capita.
As Mauro (1995) points out, there may be good reasons to expect this posi-
tive correlation between most variables. For example, Krueger (1993) argues that
corruption may induce a less e?cient bureaucracy since o?cials may introduce re-
111
Table 3.1: Simple and Partial Correlations (controlling for GDP per capita) of
Kaufmann et al Variables
VARIABLE (1) (2) (3) (4) (5) (6)
(1) Voice and Accountability 1.000
1.000
(2) Political Instability 0.718 1.000
0.435 1.000
(3) Control of Corruption 0.781 0.766 1.000
0.475 0.457 1.000
(4) Regulatory Quality 0.691 0.683 0.768 1.000
0.424 0.417 0.534 1.000
(5) Government Efiectiveness 0.736 0.785 0.963 0.782 1.000
0.372 0.513 0.880 0.569 1.000
(6) Rule of Law 0.700 0.851 0.928 0.727 0.929 1.000
0.270 0.677 0.760 0.437 0.774 1.000
(7) Log GDP per capita 0.714 0.701 0.843 0.653 0.825 0.833
Partial correlations controlling for GDP per capita are in italic.
quirements and additional obstacles in order to receive bribes. However, from an
econometric point of view this correlation can induce serious problems of multi-
collinearity and might limit the extent to which the relevance of each institutional
dimension can be identifled. The most standard solution is to group those variables
that capture similar dimensions. In this sense, in several regressions we will use
the average of Voice and Accountability and Political Stability and Lack of Violence
as Political Stability and Freedom, while we group Rule of Law, Control of Corrup-
tion, Government Efiectiveness and Regulatory Quality as Government E?ciency.
This grouping may also reduce measurement problems of the individual components.
In Figure 3.1 we plot the partial correlation between the ratio of average FDI
in ows to GDP in the 1990?s and Government E?ciency - after controlling for GDP
per capita - for the countries that will be considered in our subsequent regression
112
Figure 3.1: Partial Correlation between Average FDI In ows 1990-2000 as a Ratio
of GDP and Government E?ciency in 1996, controlling by GDP per capita
ARG
AUS
AUT
BGR
BRA CAN CHE
CHL
CHN
COL
CRICZE
DEU
DNK
DZA
EGYESP FINFRA
GBR
GRC
HUN
IDN
IND
IRL
IRN ISL
ISR
ITAJPN
KOR
MAR
MEX
MYS
NLD
NOR
NZL
PAN
PHL
POL
PRTROM
SGP
SVK SVN
SWE
THA
TURUKR USA
VEN
ZAF
?.02
0
.02
.04
.06
.08
FDI/GDP controlling for GDP per capita
?1 ?.5 0 .5 1
Government Effectiveness controlling for GDP per capita
analysis. There is a positive and signiflcant partial correlation between this institu-
tional variable and the rate of FDI to GDP.10
In our robustness analysis and for the panel regressions, we use variables from
the International Country Risk Guide (ICRG) compiled by the PRS Group. Un-
like those of Kaufmann et al (1999a), these indicators rely exclusively on polls of
experts. The main advantage of this dataset is that they are available for a consid-
erable time span, allowing us also to test the relevance of institutions in attracting
FDI exploiting the time variation. This also enables us to control for potential un-
observed heterogeneity that could bias our cross-section estimates. The variables
10The partial correlation between government e?ciency and the ratio of FDI to GDP is 0.36,
which is signiflcant at a 99 percent confldence level.
113
we consider are a subset of the ones available from the ICRG database that refer to
political risk.11 Speciflcally, we use the following indicators: Risk of Expropriation,
Government Stability, Democratic Accountability, Law and Order and Corruption.
While the flrst two variables are coded on a 0 to 10 scale, the other three are coded
between 0 and 6. In order to facilitate comparability we standardize all variables
in our sample to mean zero and unit variance. In all cases, higher rankings imply
better institutions.
Finally, a third source of institutional data we use comes from the World
Business Environment Survey (WBES) from the Worldbank.12 In particular, we
consider the average by country of the answers to the following questions in the
survey: i) Quality of the courts (1 very good to 6 very bad); ii) Quality of central
government (1 very good to 6 very bad), iii) Corruption is a general constraint to
do business (1 no obstacle to 4 major obstacle, iv) Change in law and regulations
are predictable (1 completely predictable to 6 completely unpredictable). Thus, for
the WBES, higher values imply bad institutions.
11This dataset has been used extensively to analyze the impact of institutions on economic
performance (see e.g. Knack and Keefer, 1995).
12This survey was conducted between1999 and 2000, collecting information regarding constraints
to business activities that flrms face, including institutional and governance aspects. The sample
covers 80 countries and approximately 100 enterprises in each country. The advantage of this kind
of surveys is that they report in a more accurate way the perception of entrepreneurs about the
difierent risks. However, the main purpose of this survey is to ensure a representative measure at
a country level, such that their comparability across countries may be lower than in the case of
indicators based on subjective perceptions of experts.
114
3.2.3 Empirical Strategy
Most of the empirical studies of FDI location are based on some variation of
the so-called "gravity model", which is a standard speciflcation in empirical models
of bilateral trade.13 In its simplest formulation, it states that bilateral trade  ows
(in our case bilateral FDI stocks) depend positively on the product of the GDPs
of both economies and negatively on the distance between them. Typical variables
added to the simplest gravity speciflcation in the trade literature include GDP per
capita, as well as dummies indicating whether the two countries share a common
border, a common language, past colonial links, etc.
While in the trade literature the gravity model has good theoretical founda-
tions, the use of this model for the case of FDI is somewhat ad-hoc. Although we
will look at the results of the standard gravity model for reasons of robustness, we
base our empirical evaluation of the efiects of the difierent institutional variables on
FDI on an empirical model recently developed by Carr et al (2001) which in turn
follows closely a theoretical model of location of multinational activity developed by
Markusen (1997, 2001). The model incorporates horizontal and vertical motives for
FDI. The type of FDI that is observed between two countries is determined endoge-
nously in a general equilibrium framework considering a two-country, two-factor,
two-good world. The types of flrms that can arise in this context are: horizontal
13For a discussion of the empirical application and theoretical foundations of the gravity equation
in trade theory see Frankel (1997). Papers that have used the gravity model to study the location
of FDI include Wei (1997, 2000), Mody et al (2003), Stein and Daude (2002) and Daude et al
(2003).
115
flrms with plants in both countries and headquarters in one, vertical flrms that have
a single production facility in one country and headquarters in the other country,
and national flrms that maintain headquarters and the production plant in only one
country and may serve the other market through trade. One good (A) is produced
in a competitive industry with constant returns to scale using unskilled labor, while
the other good (B) is produced under imperfect competition with increasing returns
to scale at the flrm level due to R&D, and management services. The model includes
similar assumptions as earlier models of vertical FDI as Helpman (1984), Helpman,
and Krugman (1985) such as the possibility of fragmenting the production and the
location of the headquarters and the operational plant. Also, plant scale economies
are assumed for this sector. Finally, a key assumption of the model is the factor
intensity in the difierent production facilities. In this sense, headquarters activities
are the most skilled-labor intense, followed by the flrm that produces good B and
has headquarters in the same location. Moreover, a production plant in sector B
is supposed to be less skilled-labor intensive than the former, but more intensive
than one in sector A. It is clear that while difierences in factor endowments tend to
favor vertical FDI, flrm level economies of scale would favor horizontal FDI, given
the existence of trade costs.
The type and volume of FDI between two countries depends on the size of each
economy, difierences in the size between the host and the source country, relative
factor endowments, trade costs and investment costs. When countries difier in size,
but not in factor endowments, there is an inverted U-shaped relation, indicating
116
that horizontal FDI is highest between countries that are of the same size. In this
sense, the empirical speciflcation should include the squared difierence in size in
order to account for this relationship. Additionally, vertical FDI takes place if the
difierence in the size of the economies is signiflcant and the small country is skilled
labor intensive, so that the production facility tends to be installed abroad. Notice
that since headquarters location decisions are based on factor endowments and plant
location on the basis of the factor endowments and the market size, an interaction
term between both variables should be included in the empirical speciflcation of the
model.
As in the pure horizontal model (see Horstmann and Markusen, 1987, 1992),
trade costs in the host country encourage horizontal FDI, while investment restric-
tions in the host country - captured in our institutional variable - and trade costs
in the source country restrict FDI activity of vertical nature. However, since trade
costs favor horizontal FDI but not vertical FDI, and horizontal FDI increases if fac-
tor endowments are similar, Carr et al (2001) include an interaction between trade
costs and the squared endowment difierences.
To the benchmark model of Carr et al (2001), we add our measure of institu-
tional quality, such that the empirical speciflcation is as follows:
117
ln(FDIij) = fl0 +fl1SUMGDPij +fl2SQDIFGDPij +fl3ADIFGDPij
?ADIFSKILLij +fl4ADIFSKILLij +fl5ln(Distanceij)+fl6TARIFFj
+fl7TARIFFj ?SQDIFSKILLij +fl8Institutionsj +?i +"ij: (3.1)
The deflnitions of the variables are as follows. FDI is the outward stock of FDI
from country i (source) in country j (host) from the UNCTAD database. SUMGDP
is the sum of the logs of the host country and the source country GDPs, in current
dollars from the WDI database in 2000. The variable SQDIFGDP is the squared
difierence in the GDPs of the host and the source country, while ADIFGDP is the
absolute difierence between them. Similarly, ADIFSKILL is the absolute difierence
between the countries? endowments of skilled labor and SQDIFSKILL is the corre-
sponding squared difierence. We use the average percentage of the labor force with
secondary education from the WDI database over 1990 and 2000 as our variable
of skilled labor endowment. DISTANCE is the great circle distance between the
countries? capitals.14 Trade costs in the host are measured by the average tarifi level
between 1990 and 2000.15 Finally, source country dummies (?i) are included in
order to capture the efiects of possible systematic difierences in the FDI accounting
methodology of reporting countries, as well as other relevant source country char-
acteristics.
14only exceptions are the U.S. and China, where we consider Chicago and Shanghai respectively.
15Tarifi data are from the Worldbank.
118
The log speciflcation is used because it has typically shown the best adjust-
ment to the data in the empirical literature. A problem that arises when using the
log of FDI as a dependent variable, however, is how to deal with the observations
with zero values. Our dataset includes more than two-third of observations with
zero FDI stocks16, which would be dropped by taking logs. The problem of zero
values of the dependent variable is typical in gravity equations for trade, and it has
been dealt with in difierent ways.
Some authors (see for example Rose, 2000) simply exclude the observations in
which the dependent variable takes a value of zero. A problem with this approach
is that those observations may convey important information for the problem under
consideration. Giventheimportanceofzeroobservationsinoursample, thisstrategy
could lead to a serious estimation bias. One alternative we explore is that used by
Eichengreen and Irwin (1995) who use a simple transformation to deal with the
zeros problem: work with log (1 + trade), instead of the log of trade. This has the
advantage that the coe?cients can be interpreted as elasticities when the values of
trade tend to be large, since in this case log (1 + trade) is approximately equal to
log (trade). However, if zero FDI is a consequence of the existence of flxed costs,
it would be inappropriate to deal with this problem in a linear way as Eichengreen
and Irwin (1995) do for the case of trade. Therefore, an alternative method is
to use Tobit estimations instead of OLS. Santos and Tenreyro (2006) propose an
alternative estimation method based on Poisson regressions in order to address the
16More speciflcally, in 2002 there are 3970 zero observations out of 5134.
119
potential bias in gravity models in the presence of heteroscedasticity. Our approach
here is to flrst present the estimates for the sub-sample that excludes the zeroes and
then to show that the results are robust to the use of several of these alternative
estimation techniques and solutions to the zero-FDI problem.
3.3 Empirical Results
3.3.1 OLS Estimates
In the flrst column of Table 3.2, we present our estimate of equation (3.1)
without including any institutional indicator. A flrst interesting point is that the
model explains a high proportion - approximately 71 percentage points - of the total
variation in FDI stocks.17The signiflcant variables are the sum of GDPs, the squared
difierence of GDPs, distance, and the absolute difierence in factor endowments.
Distance has a negative efiect on bilateral FDI, while economic mass measured by
the sum of GDPs has a signiflcantly positive impact. In addition, large difierences
of scale between the source and host country discourage FDI, as well as difierences
in factor endowments. These results are consistent with those obtained by Blonigen
et al (2002).
In columns 2 to 7 we introduce the Kaufmann et al (1999a) variables into
equation (3.1). The estimates show that the impact of institutions depends on the
17While the source country dummies are jointly signiflcant, they do not drive this result, since
an estimation of equation (1) without the source dummies explains approximately 61 percent of
the variation in the dependent variable, while a regression with only source country dummies as
explanatory variables explains 31 percent of the total variation. While not presented in the tables,
the source dummies are jointly signiflcant at conventional levels.
120
Table 3.2: Cross Section OLS Estimates of Equation (3.1)
INDEP.VARIABLE (1) (2) (3) (4) (5) (6) (7) (8) (9)
SUMGDP 0.632 0.629 0.635 0.586 0.577 0.622 0.61 0.705 0.605(4.90)***(4.54)***(4.88)***(4.30)***(4.73)***(4.52)***(4.28)***(5.92)***(4.50)***
SQDIFGDP -0.049 -0.049 -0.056 -0.054 -0.045 -0.057 -0.049 -0.048 -0.054(2.33)**(2.37)**(2.62)**(2.47)**(2.19)**(2.45)**(2.29)**(2.62)**(2.36)**
ADIFLAB -0.027 -0.027 -0.027 -0.025 -0.027 -0.026 -0.026 -0.031 -0.025(2.96)***(2.90)***(2.90)***(2.70)***(2.83)***(2.79)***(2.84)***(3.35)***(2.85)***
ADIFGDPxADIFLAB 0.003 0.003 0.003 0.003 0.004 0.003 0.003 0.003 0.003(1.05) (1.12) (1.11) (1.26) (1.45) (1.18) (1.19) (1.33) (1.23)
TARIFF -0.022 -0.021 -0.007 0.004 0.018 -0.009 -0.013 0.024 -0.003(1.56) (1.21) (0.40) (0.23) (0.95) (0.54) (0.86) (1.57) (0.19)
TARIFFxSQDIFLAB 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001(0.95) (0.90) (0.63) (0.14) (0.12) (0.58) (0.72) (0.85) (0.31)
DISTANCE(logs) -0.763 -0.761 -0.728 -0.756 -0.794 -0.756 -0.761 -0.813 -0.805(8.32)***(8.27)***(7.65)***(8.40)***(8.53)***(8.38)***(8.32)***(8.58)***(8.14)***
VoiceandAccountability 0.022 -0.317(0.10) (1.30)
PoliticalStability 0.251 0.67(1.25) (2.17)**
GovernmentEfiectiveness 0.355 1.474(1.93)* (2.74)***
RegulatoryQuality 0.702 1.259(3.38)*** (2.78)***
ControlofCorruption 0.185 0.18(1.00) (0.46)
RuleofLaw 0.129 -2.445-0.7 (3.51)***
PoliticalStabilityandFreedom -0.379(1.23)
GovernmentE?ciency 0.635(2.05)**
Observations 714 714 711 711 714 710 714 710 710R-squared 0.71 0.71 0.72 0.72 0.72 0.72 0.71 0.75 0.72
OLS estimates. The dependent variableis the log of the bilateralstock of FDI in 2002. Institutional
variables are from 1996. All regressions include source country dummies, not reported. Absolute
robust White-corrected t-statistics in parentheses. *, **, *** signiflcant at 10, 5 and 1 percent,
respectively.
121
speciflc dimension considered. While Voice and Accountability, Political Stability,
Rule of Law, and Control of Corruption have no signiflcant impact on FDI, Regula-
tory Quality and Government Efiectiveness have a positive and signiflcant impact
on the volume of FDI. The largest efiect seems to be associated to Regulatory Qual-
ity where a one standard deviation increase in this dimension of governance would
increase FDI stocks by a factor of 2.18 While initially the magnitude of this efiect
might seem very large, it should be kept in mind that a one standard deviation
improvement in the regulatory quality of government implies a substantial change,
e.g. from the level of Thailand to that of Canada or Germany. A slightly lower
efiect corresponds to Government Efiectiveness, where a one standard deviation
improvement would imply an increase in FDI stocks by a factor of 1.4. This flrst
evidence indicates that several dimensions of government institutions clustered in
Government E?ciency seem to be especially relevant in explaining the location of
FDI.
There are however at least two possible problems with the preceding regres-
sions. On the one hand, if various institutional dimensions determine simultane-
ously the location of FDI, by including them one by one as before there might be
an omitted variable bias, especially severe given the correlation among the difierent
indicators reported in the previous section. On the other hand, as mentioned pre-
viously, the difierent variables might be subject to measurement errors. In order to
address these problems, we run two additional regressions, one including the six re-
18exp(0.702)=2.018.
122
gressors simultaneously, and the other clustering all variables into Political Stability
and Freedom or Government E?ciency as previously mentioned.
In column 8, we present the results from including all six variables together.
Given the high correlation among them, there might be important multicollinearity
problems, as indicated by the negative sign of Voice and Accountability or Rule
of law. Taking into account this caveat it is still interesting to point out that
the variables Regulatory Quality and Government Efiectiveness seem to the most
relevantgovernancedimensions. InthelastcolumnofTable3.2weestimateequation
(3.1) incorporating the clusters Political Stability and Freedom and Government
E?ciency. A one standard deviation improvement in Government E?ciency - e.g.
from Slovenia to Sweden or Argentina to Chile - would increase FDI by a factor
of 1.89. Let us consider Argentina and Chile to assess the economic signiflcance
and plausibility of the estimates. In 2002, the FDI stock is 7.3 percentage points
of GDP for Argentina, while in the case of Chile it is 25.8. An improvement in
Argentina?s institutional quality to the level of Chile would therefore lead to an FDI
stock to GDP ratio of approximately 13.8 percent, still signiflcantly below that of
Chile. Political Stability and Freedom has no signiflcant efiect on FDI.19
19The fact that Political Stability and Freedom is not signiflcant in our regression means that
it has no direct efiect on FDI. This does not exclude the possibility that it might still have an
important indirect efiect, for example via the accumulation of human capital.
123
3.3.2 Instrumental Variables Estimations
Although we use institutional variables for 1996 and FDI stocks for 2002 to
reduce simultaneity problem, these problems might subsist. Thus, the previous es-
timates could potentially be biased due to endogeneity.20 It might be reasonable
to consider the possibility that the quality of institutions might be endogenous for
two reasons. First, once foreign investors are located in a country, they might be-
come a constituency that demands better institutions. Therefore, there could be a
feedback efiect on the quality of institutions. Second, there is a potential subjectiv-
ity bias, where experts report a better score on the quality of institutions because
they observe a high level of FDI, which generates the same econometric problems.
In order to address this issue, we re-estimated the regressions in Table 3.2 using
instrumental variables. We use two distinct sets of instruments for the two difier-
ent sets of institutional variables. First, to instrument Voice and Accountability as
well as Political Stability, we use an index of ethnolinguistic fragmentation (ELF)21
from Easterly and Levine (1997) and the average number of homicides per 100,000
inhabitants during the 1990?s. Both variables have been used extensively in the
literature to analyze political violence and social risk. The simple correlation coe?-
cients of Voice and Accountability and Political Stability with ELF are - 0.35 and -
0.21, respectively. For the case of homicides these correlations are - 0.34 and - 0.55,
20We do not consider institutional variables that are strongly related to macroeconomic factors
(e.g. "investment climate") in order to reduce the possibility that experts opinions might be caused
by the observed volume of FDI. In addition, the bilateral nature of the FDI data reduces partially
the potential severity of endogeneity.
21Mauro (1995) to instrument corruption has also used this variable. It measures the probability
with in a country that two randomly selected persons are members of difierent ethnic groups. See
Easterly and Levine (1997) and Mauro (1995) for more details.
124
respectively.22 For the second group of variables clustered in Government E?ciency
we use the fraction of population that speaks English and the fraction of the popula-
tion that speaks a Western European language from Hall and Jones (1999). Hall and
Jones (1999) use these variables to instrument institutions in cross-country growth
regressions. It seems natural to assume that the extent to which this constitutes
the mother tongue of a country is positively correlated with the degree of in uence
of Western Europe. La Porta et al (1999) flnd that the origin of the legal code is
an important and signiflcant determinant of a series of government institutions and
economic outcomes. In addition, Chong and Zanforlin (2000) flnd that countries
with law tradition based on the French Civil code display signiflcantly lower levels
of bureaucratic development, lower levels of credibility of the government and higher
levels of corruption, while countries with English Common Law show a higher level
of institutional quality. Thus, we consider a set of dummy variables for Common
Law, French Law, German Law and Scandinavian Law as instruments.
In Table 3.3, we present the results for the IV estimations. In columns 1 - 6,
we flrst present the regressions including one variable at the time. The flrst inter-
esting result is that in terms of signiflcance, the results are analogous to the OLS
estimations. Thus, only Government Efiectiveness and Regulatory Quality have a
signiflcant efiect on FDI stocks, while the remaining variables are not signiflcant.
Similarly, for the clustered institutional variables, we flnd that Government E?-
ciency has a positive and signiflcant impact on FDI, while for Political Stability and
22These correlations are signiflcant at conventional levels of confldence.
125
Freedom the estimate is negative.23 Taking a look at the size of the coe?cients, the
IV estimates look rather large. A one standard deviation increase in Government
E?ciency would increase FDI by a factor of 10!24However, this huge increase in
the estimated coe?cient is partially due to the change in the sample due to limited
data availability of some instruments. In column 8, we present the OLS estimates
for the same reduced sample. Observe that the coe?cient of Government E?ciency
changes from 0.635 (see Table 3.2, column 9) to 1.541 just due to the change in
the sample. Thus, if we assume that the change in the coe?cient would be similar
for the IV estimations if we could estimate the model for the whole sample, the
efiect of a one standard deviation improvement in Government E?ciency would be
an increase of bilateral FDI by a factor of \only" 2.58.25 This order of magnitude
implies that, e.g., if Kenya had the same level of institutional quality than South
Africa it would almost receive the same amount of FDI as a fraction of GDP.26
Overall, the results so far show that the quality of institutions has a signifl-
cant and economically important impact on the location of FDI. In addition, not
all dimensions of the institutional framework have the same direct importance for
foreign investors? investment decisions. We flnd that the regulatory framework and
the efiectiveness of the government in \getting things done" are the most sensitive
aspects to foreign investors. Thus, variables that refer to the predictability and
23This last result should not be interpreted that in order to attract FDI a country should reduce
its civil liberties. It only shows that, once we take into account other institutional aspects, there
is no direct positive efiect of political stability on FDI. However, it might still be the case that
without political stability it is di?cult to maintain a predictable regulatory framework.
24exp(2.303) = 10.004.
25exp(0.635*2.303/1.541) = 2.583.
26South Africa has a one standard deviation better institutional quality than Kenya, while the
FDI stock to GDP ratios are 17 percent versus 5.1 percent.
126
Table 3.3: IV Estimates of Equation (3.1)
INDEP.VARIABLE (1) (2) (3) (4) (5) (6) (7) (8)
SUMGDP 0.904 0.652 0.633 0.68 0.687 0.666 0.721 0.693(3.77)***(4.94)***(6.00)***(8.05)***(5.94)***(5.42)***(6.53)***(6.58)***
SQDIFGDP -0.026 -0.01 -0.01 -0.001 -0.016 -0.017 -0.003 -0.004(0.94) (0.45) (0.47) (0.03) (0.76) (0.75) (0.11) (0.19)
ADIFLAB -0.048 -0.029 -0.01 -0.011 -0.013 -0.015 -0.014 -0.018(2.85)***(3.32)***(0.87) (1.03) (0.97) (1.27) (1.07) (1.91)*
ADIFGDPxADIFLAB 0.003 0.002 0.001 0.003 0.001 0.001 0.004 0.003(0.60) (0.63) (0.40) (1.05) (0.31) (0.40) (0.94) (0.92)
TARIFF -0.124 -0.027 0.038 0.085 0.013 0.013 0.011 0.008(1.75)* (0.90) (1.07) (1.84)* (0.39) (0.35) (0.31) (0.32)
TARIFFxSQDIFLAB 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001(2.38)**(1.54) (0.63) (0.84) (0.07) (0.03) (0.21) (0.47)
DISTANCE(logs) -0.903 -0.679 -0.639 -0.714 -0.675 -0.647 -0.823 -0.765(3.74)***(6.42)***(6.89)***(7.10)***(7.14)***(6.17)***(6.73)***(6.89)***
VoiceandAccountability -2.042 - - - - - - -(1.23)
PoliticalStability - 0.143 - - - - - -(0.31)
GovernmentEfiectiveness - - 0.886 - - - - -(2.10)**
RegulatoryQuality - - - 2.098 - - - -(2.96)***
ControlofCorruption - - - - 0.546 - - -(1.41)
RuleofLaw - - - - - 0.572 - -(1.26)
PoliticalStabilityandFreedom - - - - - - -1.835 -1.059(2.46)**(2.21)**
GovernmentE?ciency - - - - - - 2.303 1.541(2.95)***(3.37)***
Observations 514 514 566 566 566 566 500 500R-squaredofflrststage 0.51 0.66 0.79 0.75 0.8 0.79 0.79;0.81 0.82
F-testoninstruments 3.35 19.33 5.93 3.68 10.84 8.18 9.79;8.70 -[0.046] [0.000] [0.000] [0.006] [0.000] [0.000][0.00;0.00]
Notes: Instrumental variable estimations, except column 8 which is OLS. Instruments for the
flrst two variables are ethnolinguistic fragmentation and homicides, while for the other variables
the instruments are the English speaking fraction of the population, European language speaking
fraction, and legal code dummies. The dependent variable is the log of the bilateral stock of FDI
in 2002. Institutional variables are from 1996. All regressions include source country dummies,
not reported. Absolute robust White-corrected t-statistics in parentheses. P-values in brackets. *,
**, *** signiflcant at 10, 5 and 1 percent, respectively.
127
stability of policies are especially important to establish a foreign investor friendly
environment. The results show no evidence of a direct efiect of civil liberties and
political violence. This holds for OLS as well as 2SLS estimates. In the next section,
we analyze the robustness of these results. There is little evidence of any signiflcant
impact of political instability and violence on FDI. This result is also in line with
the empirical literature on this issue (see e.g. Sethi et al, 2003), which has often
failed to flnd a signiflcant impact of political violence on FDI.
3.4 Robustness
The flrst issue we address in our robustness tests is whether our results are
sensitive to the solution used to deal with the observations with zero FDI. In the
flrst two columns of Table 3.4, we present estimates of equation (3.1) considering
adding the minimum observed FDI stock to the log of the bilateral FDI stock.27 In
the flrst column, we restrict the sample to exclude all zero FDI observations, while
in column 2 we include these observations. Focusing on the institutional variables,
column 1 shows that the transformation of the dependent variable does not alter
signiflcantly the results, given that the estimates are virtually identical to those
in column 9 of Table 3.2. By contrast, the inclusion of the zero FDI observations
alters signiflcantly the estimated impact of Government E?ciency on FDI, while
Political Stability and Freedom remaininsigniflcant. Inparticular, thepointestimate
indicates that the estimated impact of a one standard deviation improvement in
27The results do not change if we consider adding the unity instead of the minimum. While not
reported here, they are available upon request.
128
Table 3.4: Robustness: Estimation Methods and Model Speciflcation(1) (2) (3) (4) (5) (6) (7) (8)
EstimationMethod OLS 2SLS OLS POISSONTOBIT OLS OLS OLSDep.Variable LnStock+minLnStock+minStockinlevelsStockinlevelsLnStock+minLnStock LnStock+minLnFlow
SUMGDP 1.200 1.165 0.008 0.877 0.110 - - -(13.59)***(6.97)***(2.15)**(20.07)***(0.71)
SQDIFGDP -0.079 -0.069 0.001 0.006 -0.086 - - -(4.83)***(2.64)** (1.31) (0.49) (5.16)***
ADIFLAB -0.012 -0.008 48.85 -0.038 -0.027 - - -(0.87) (0.45) (1.28) (1.86)* (1.60)
ADIFGDPxADIFLAB -0.001 -0.001 0.001 0.004 0.006 - - -(0.21) (0.21) (3.50)***(0.85) (1.30)
TARIFF 0.001 -0.001 35.466 0.028 0.022 - - -(0.01) (0.03) (0.71) (1.11) (0.91)
TARIFFxSQDIFLAB 0.001 0.001 0.005 0.001 0.001 - - -(0.69) (0.64) (0.17) (0.25) (2.12)**
DISTANCE(log) -1.313 -1.381 -0.519 -0.572 -0.235 -0.817 -1.341 -0.817(10.04)***(6.11)***(3.08)***(6.66)***(0.78) (7.20)***(7.85)***(7.56)***
LPGDP - - - - - 0.716 1.033 0.76(7.19)***(10.13)***(9.75)***
LPGDPPC - - - - - -0.374 -0.141 -0.347(1.34) (0.47) (1.36)
COMLANG - - - - - 0.125 0.513 0.262(0.51) (2.22)**(0.98)
COL - - - - - 1.157 0.049 0.662(4.14)***-0.1 (2.42)**
ADJACENCY - - - - - 1.065 1.745 0.769(3.37)***(3.30)***(2.89)***
COMCOL - - - - - 1.489 - -(2.25)**
PoliticalStabilityandFreedom-0.484 -1.838 -938.339 -0.724 0.263 -0.068 0.391 -0.312(1.19) (1.15) (1.74)* (2.09)** (0.52) (0.26) (1.15) (1.48)
GovernmentEfiectiveness 1.291 2.572 2 1.518 1.043 0.752 0.89 0.763(3.80)***(1.71)* (2.71)***(3.82)***(1.98)*(2.47)**(2.31)**(2.20)**
Observations 1375 1375 2173 2173 2173 863 4007 925R-squared 0.84 0.84 0.31 - 0.75 0.72 0.68 0.71
Notes: Notes: Instruments used in column (2) are ethnolinguistic fragmentation and homicides
for Political Stability and Freedom, and English speaking fraction of the population, European
language speaking fraction, and legal code dummies for Government Efiectiveness. In column (3)
the explanatory variables are also in levels. All regressions include source country dummies, not
reported. Absolute robust White-corrected t-statistics in parentheses. P-values in brackets. *, **,
*** signiflcant at 10, 5 and 1 percent, respectively.
Government E?ciency raises FDI stocks by a factor of 3.6. While this magnitude
is large, it is in line with the estimates discussed in the previous section. The fact
that the point estimate is greater when we include the zero FDI observations shows
that the sample selection bias goes in the expected direction. If the likelihood of
observing a zero FDI stock is higher for host countries with \bad" institutions,
excluding these observations would bias the estimates downwards towards zero.
In column 3, we explore a difierent alternative to deal with the zero FDI obser-
vations by estimating the model in levels.28 The estimates indicate that on average
28The right hand variables, like SUMGDP and distance, are also expressed in levels rather than
129
an improvement of one standard deviation in Government E?ciency increases FDI
stocks on average by 2482 million dollars.29
In column 4, we present the results of estimating the regression using the Pois-
son regression approach recently proposed by Santos and Tenreyro (2006) which
corrects for the potential bias of the log-linearlized model under heteroscedasticity.
Clearly, the results regarding your main variables of interest remain unchanged. As
we mentioned above, an additional alternative approach could be the estimation
of a TOBIT model of equation (3.1). In column 5 of Table 3.4, we present the
estimates considering this alternative estimation method. The main results remain;
Government E?ciency has a positive and signiflcant efiect on FDI, while Political
Stability and Freedom has again no signiflcant efiect. While the estimated coe?-
cient is slighter higher, the implied impact on FDI is in line with the OLS estimated
presented above.
In the next column, we present a standard gravity model in order to ex-
plore the sensitiveness of our results to the speciflcation of our baseline regression.
We include the product of GDPs (LPGDP) and GDP per capita of the host and
source countries - LPGDPPC - (in logs), distance (in logs), a common language
dummy (COMLANG), a dummy if both countries were colonized by the same coun-
try (COMCOL), a dummy that equals unity if the source country was the colonizer
logs.
29While this estimate might seem very large, it is important to remember that the standard
deviation of the estimation is high. For example, the 95 percent interval goes from 658 to 4305
million dollars.
130
of the host country (COL), an adjacency dummy (ADJ), and a dummy for common
membership in a Free Trade Area (SAMEFTA). The estimates show that overall
the gravity equation is successful in explaining the variation in FDI across countries
with an R-squared of 0.72. Regarding our variables of interest, the estimates show
that a one standard deviation improvement in the degree of e?ciency of the gov-
ernment would increase FDI by a factor of 2.1, this point estimate is very close to
the one presented in Table 3.2. As before, the estimated efiect of Political Stability
and Freedom on FDI is not signiflcant.
In column 7, we present the gravity model estimate considering the alterna-
tive dependent variable that adds the minimum observed FDI. Again, the results
are very similar to our previous estimates. A one standard deviation improvement
in Government E?ciency increases FDI by a factor of 2.4, while Political Stability
and Freedom does not have any signiflcant direct efiect.
Next we consider FDI  ows instead of stocks. While our preferred dependent
variable is stocks, the estimates for our variables of interest remain at similar levels
of signiflcance and magnitude. As before, Political Stability and Freedom does not
have any impact on FDI, a one standard deviation change in Government E?ciency
change FDI  ows by a factor of 2.1.30
30The similarity in the size of the coe?cient is quite logical, given the high correlation between
 ows and stocks. The simple correlation coe?cient is 0.86. This high correlation is because on
average the FDI stock re ects recent large FDI  ows.
131
In Table 3.5, we consider a set of alternative measures of institutional out-
comes in order to test the robustness of our results. In the flrst three columns, we
consider the ICRG variables. The only variable that is systematically signiflcant is
Government Stability, with a one-standard deviation improvement in Government
Stability implying an increase of between 38 and 46 percentage points. All other
variables do have no signiflcant impact on FDI.31 In next two columns, we consider
the WBES variables. In this case, only the predictability of laws and regulations has
consistently a signiflcant impact on FDI. A deterioration of a one standard deviation
in this dimension of the institutional quality of a country decreases FDI between 54
and 94 percentage points.32
Overall, the results from this section show that our results are robust to dif-
ferent estimation methods, deflnitions of the dependent variable, and speciflcations.
Furthermore, some institutional dimensions have a greater impact on FDI than
others. Especially, institutions that create predictable regulatory and legal frame-
works, as well as policy stability are the most important. This result is consistent
with those of Stasavage (2002) who flnds that formal institutions that produce poli-
cies that are more predictable and stable have a positive and signiflcant impact on
domestic investment.
31This result does not seem to be driven by multicollinearity, given that regressions including
on one variable at the time produce the same results. Results are available upon request.
32It should be kept in mind that in this case higher values of the institutional variables imply
worse institutions.
132
Table 3.5: Robustness Alternative Institutional Variables(1) (2) (3) (4) (5)
Dep.Variable LnStock LnStock+minLnStock+minLnStock+minLnStock+minModel CMM CMM Gravity CMM Gravity
ExpropriationRisk -0.034 -0.155 0.229 - --0.18 -0.37 -0.71
GovernmentStability 0.374 0.377 0.278 - -(2.40)** (1.82)* (1.92)*
DemocraticAccountability -0.101 0.22 0.393 - --0.43 -0.59 -1.57
Corruption -0.047 0.07 0.087 - --0.2 -0.24 -0.4
LawandOrder -0.018 0.147 0.178 - --0.12 -0.44 -0.53
QualityoftheCourts - - - -0.044 0.043-0.11 -0.16
QualityoftheGovernment - - - 0.419 0.157-1.63 -0.86
GovernmentCorruption - - - -0.219 -0.061-0.45 -0.14
LegalandRegulatoryPredictability - - - -0.664 -0.431(2.59)** (1.92)*
Observations 596 1635 2728 1182 1995R-squared 0.74 0.8 0.8 0.77 0.72
Note: Columns with CMM include the Carr et al (2001) controls, while Gravity stands for the
gravity model. Estimation results for controls are similar to previous tables and not reported here.
All regressions include source country dummies, not reported. Absolute robust White-corrected
host country clustered t-statistics in parentheses. *, **, *** signiflcant at 10, 5 and 1 percent,
respectively.
133
3.5 Time series evidence of the relevance of institutions
In this section, we extend our analysis to assess the impact of institutions over
time. This panel data analysis is also an alternative approach to the IV regressions
presented in Section 3 to deal with potential endogeneity problems. Thus, it repre-
sents an additional important robustness check. In this section, we use the ICRG
component Government Stability that turned out to be consistently the most signif-
icant in our cross section analysis. We use flve-year periods for our panel, such that
we are left with 4 periods: 82-86, 87-91, 92-96, and 97-02. The dependent variable
is the FDI stock at the end of the period, while for controls we use period averages,
except for institutions where we use the value at the beginning of the period in order
to reduce simultaneity problems. All regressions include period dummies to account
for common shocks.
In the flrst column of Table 3.6, we present the pooled OLS estimates. The
impact of institutions is positive and signiflcant. The coe?cient implies that a
one-standard deviation improvement in Government Stability increases FDI by 17
percentage points. While this estimate is smaller than the cross-section estimate, it
remains economically important. Next, we estimated the model using random efiects
and flxed efiects estimators. Again, the coe?cient of our institutional variable is
signiflcant and positive. Moreover, the magnitude is only slightly higher than in the
case of pooled OLS, with an estimated impact of between 22 and 26 percent of a
one-standard deviation change.
134
Table 3.6: Panel Data Estimates(1) (2) (3) (4) (5) (6) (7)
EstimationMethod PooledOLSRandomEfiectsFixedEfiectsPoissonRegressionPraisWinstenGMM GEE
LagDependentVariable - - - - - 0.566 -(8.63)***
LPGDP 1.172 0.872 0.175 0.863 0.895 0.316 0.936(34.31)***(19.80)***(2.15)** (25.91)*** (20.48)***(2.48)**(19.73)***
LPGDPPC -0.034 0.185 0.894 0.613 0.138 -0.29 0.168-0.37 (2.08)** (6.23)*** (3.10)*** -1.47 -1.15 -1.63
DISTANCE(log) -1.025 -0.903 - -0.325 -0.931 - -1.009(9.95)***(7.50)*** (2.70)*** (7.74)*** (7.40)***
COMLANG 1.678 1.377 - 0.558 1.285 - 1.828(12.98)***(6.23)*** (2.76)*** (5.97)*** (8.36)***
COMCOL 0.482 0.341 - 2.299 0.196 - 0.21-1.45 -0.67 (5.62)*** -0.39 -0.35
COLONY -0.683 -0.086 - 0.784 -0.025 - -0.62(2.54)** -0.21 (4.45)*** -0.06 -1.64
ADJANCENCY 2.502 2.647 - 0.173 2.726 - 2.141(6.10)***(5.27)*** -0.6 (5.46)*** (3.73)***
SAMEFTA 0.175 0.88 1.052 0.276 0.684 -0.26 0.286-0.89 (5.77)*** (5.73)*** -1.56 (4.28)***-0.82 (1.65)*
GovernmentStability 0.153 0.2 0.232 0.508 0.209 0.121 0.208(2.37)** (5.80)*** (6.58)*** (2.91)*** (5.81)***(2.12)**(5.27)***
Observations 9760 9760 9760 9127 9760 5761 8484R-squaredoverall 0.61 0.6 0.26 - 0.46 - -
R-squaredwithin - 0.07 0.08 - - - -R-squaredbetween - 0.59 0.28 - - - -
Numberofpairs - 3496 3496 - - 2054 2439AR(1)coe?cient - - - - 0.875 - -
Sargantest - - - - - 5.76 -[0.060)
Note: The dependent variable in all regressions is the log of the FDI stock plus the minimum
positive value expect column 4 where the dependent variable is the FDI stock in levels. All
regressions include source country dummies, not reported. Absolute robust White-corrected t-
statistics in parentheses. P-values in brackets. *, **, *** signiflcant at 10, 5 and 1 percent,
respectively.
135
The results for the Poisson estimation, proposed by Santos and Tenreyro
(2006), are presented in column 4. Again, while the point estimate is slightly higher
than for the other regressions the efiect of institutions on FDI is statistically sig-
niflcant and in line with our cross section results in terms of economic signiflcance.33
In column 5, we use the Prais - Winsten estimator that corrects for flrst order
autocorrelation in the residuals, which could potentially be a problem. As it can
be seen, while we flnd evidence of a signiflcantly autocorrelated error term, our
result regarding the institutional variable is not sensitive to this issue. The point
estimate for the impact of institutions over time almost identical to the random and
flxed efiects estimates. Given the evidence of a signiflcant autocorrelation in the
residuals, an alternative is to formulate an explicitly dynamic model by including a
lag of the dependent variable in our model. In this case, it is well known that OLS
estimates tend to be inconsistent. Thus, we proceed to estimate the equation using
the Arellano-Bond GMM estimator. Results are reported in column 6. Again, the
coe?cient of Government Stability is positive and signiflcant, although somewhat
smaller. However, it should be kept in mind that the efiect of the explanatory
variables is no longer straightforward to compute if the lagged dependent variable
is included. Given the size of the coe?cient on the lagged dependent variable,
the long-run coe?cient of the institutional variable is actually very near to the
previous estimates. A last robustness check we perform is to estimate the model
33Fixed efiects estimates using the Poisson approach yield a smaller but signiflcant estimate of
around 0.04. Results are not reported due to space considerations but are available upon request.
136
using panel-corrected standard errors GEE, assuming an AR(1) process. The results
are reported in column 7. Again, the estimate for our coe?cient of interest remains
signiflcant. Furthermore, the point estimate is very close to those reported in the
previous levels. Overall, this section shows that the panel data evidence also shows
a signiflcant and important impact of institutions on FDI.
3.6 Conclusions
In this chapter, we have shown the relevance of the institutional quality as
a factor of attraction of FDI. We flnd that the quality of institutions has positive
efiects on FDI. The impact of institutional variables is statistically signiflcant, and
economically very important. For example, a one standard deviation change in the
regulatory quality of the host country?s government changes FDI by a factor of 2.
Additionally, not all institutional dimensions have the same importance for the de-
cision of where to invest. We flnd that unpredictable policies, excessive regulatory
burden, and lack of commitment on the part of the government seem to play a ma-
jor role in deterring FDI. These results are robust to the use of a wide variety of
institutional variables, collected from difierent sources, and using difierent method-
ologies. Furthermore, they are also robust to difierent speciflcations, and difierent
estimation techniques. In addition, we have also contributed panel data evidence
that conflrm our results from the cross section.
Thus, countries that would increase foreign investment would be able to do so
137
by increasing their institutional framework, especially by establishing a predictable
framework for economic policies and enforcement. In particular, the results pre-
sented In addition, this development strategy would also have positive spillovers to
other economic activities that are key to economic growth and development. The
results of our paper are clearly in line with the empirical growth literature that has
stressed the importance of institutions for economic growth (e.g., see Acemoglu et
al, 2001 and Hall and Jones, 1999). In particular, our paper highlights one channel
through which institutions might afiect growth: by increasing FDI. In addition, rais-
ing the institutional quality would have also a positive efiect on domestic investment
as Mauro (1995) and Stasavage (2002) have shown.
138
Appendix A
Appendix
A.1 Solution Algorithm
The following algorithm is used to solve for the equilibrium of the model
outlined in section 1.3. It follows Arellano (forthcoming) very closely.
1. Start with a guess for ? and a equally spaced grid of 300 points.
2. Initiate the process formulating a guess for asset prices q0(b0;z;A). I start
considering the risk free rate, such that q0(b0;z;A) = 11+r?.
3. For this intial guess of bond prices, solve the model by value function itera-
tion until convergence and compute policy functions and the implied default
decisions, as well as asset prices q1(b0;z;A).
4. Check whether the resulting asset prices matrices q1 and the initial guess q0
are su?ciently close, by computing jjq0 ? q1jj < " = 10?6. If this condition
holds, continue to the next step, else set q0 = q1 and start at step 3 again.
5. Compute business cycle statistics for 100 samples of 100 years. Compare the
average default frequency with the 3% default frequency in the data. If the
model matches this frequency, stop. Else, adjust ? and start at step 2 again.
139
A.2 Sample
? Latin America: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica,
Ecuador, El Salvador, Guatemala, Honduras, Jamaica, Mexico, Paraguay,
Peru, Trinidad and Tobago, Uruguay, Venezuela
? Asia: Bangladesh, China, Hong Kong, India, Indonesia, Kazakhstan, Korea,
Malaysia, Pakistan, Philippines, Singapore, Sri Lanka, Vietnam
? Eastern Europe: Bulgaria, Croatia, Cyprus, Czech Republic, Estonia, Hun-
gary, Latvia, Lithuania, Poland, Romania, Russia, Slovenia, Turkey
? Africa/Middle East: Cote d?Ivoire, Egypt, Ghana, Israel, Kenya, Morocco,
Namibia, Nigeria, South Africa, Tanzania, Tunisia, Zambia
? Europe: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ice-
land, Ireland, Italy, Netherlands, Norway, Portugal, Spain, Sweden, Switzer-
land, United Kingdom
? Other: Australia, Canada, Japan, New Zealand, United States
A.3 Variable
? Bilateral FDI stocks FDI asset holdings of source country i in host country j
in million US dollar; UNCTAD
? Bilateral portfolio equity and portfolio debt stocks average 2001-2003 holdings
of source country i in host country j in million US dollar; Coordinated Portfolio
140
Investment Survey (CPIS), IMF
? Bilateral loans aggregate assets and aggregate liabilities of banks in report-
ing countries vis--vis banking and non-banking institutions in host countries;
International Locational Banking Statistics (ILB), BIS
? Distance log bilateral great circle distance in miles between economic centers
of source country and host country; Andy Roses website
? Telephone tra?c volume of telephone call tra?c between source and host
country; ITU Directions of Trade
? Trade in newspapers and periodicals Exports from country i to country j plus
exports from j to i in million US dollar; UN Comtrade database Exports of
item 8922 SITC Rev.2
? Bilateral stock of foreigner sum of foreigners born in country i currently living
in country j and vice-versa; OECD Database on Foreign-born and Expatriates
? Common language dummy equal to one if both countries speak the same
language and zero otherwise; Andy Roses website and CIA World Factbook
? Common legal origin dummy equal to one if both countries have legal system
with same origin and zero otherwise; La Porta et al (1998)
? Colonial links dummy equal to one if both countries have been linked through
colonization; Andy Roses website and CIA World Factbook
141
? Trade agreement dummy equal to one if both countries have a bilateral trade
agreement or are part of a common agreement and zero otherwise; Andy Roses
website
? Investment treaty dummy equal to one if both countries have a bilateral
investment treaty and zero otherwise; UNCTAD
? Bilateral trade the imports of goods and services of host country from and
source country in US dollar million; IFS, IMF
? Capital account openness dummy equal to one if the host country had fully
liberalized its capital account by 1996 and zero otherwise; Annual Report of
Exchange Arrangements and Exchange Restrictions, IMF
? Financial development credit to the private sector in USD million; IFS, IMF
? Stock market capitalization average stock market capitalization in USD mil-
lion over the period 1999-2003; Datastream and national sources
? Quality of information disclosure index that goes from 0 to 7 with higher
values indicating that regulation requires more disclosure of information (see
source for more details); World Bank Doing Business Database
? Accounting standards rating of companies in seven difierent categories in
1990. The index goes from 0 to 100, with higher values representing better
standards; La Porta et al (1998)
142
? Property rights index that goes from 0 to 5, with higher values representing
bad protection of property rights; Heritage Foundation
? Expropriation risk index goes from 0 to 10, with high values representing low
risk; ICRG PRS
? Repudiation risk index goes from 0 to 10, with high values representing low
risk; ICRG PRS
? Days of enforcement the time of dispute resolutionin calendar dayscounted
from the moment the plaintifi flles the lawsuit in court until settlement or
payment; World Bank Doing Business Database
? TI corruption value of index goes from 0 to 10, with higher values indicating
higher levels of corruption; Transparency International (Wei, 2000b)
? WDR corruption index goes from 1 to 8, with higher values indicating higher
levels of corruption; World Bank (Wei, 2000b)
? German exporters corruption index survey based index that goes from 0 to
10. Higher values represent higher levels of corruption; Wei (2000b)
143
Bibliography
[1] Acemoglu, Daron, SimonJohnson, James Robinson, and YunyongThaicharoen,
2003. \Institutional Causes, Macroeconomic Symptoms: Volatility, Crises and
Growth", Journal of Monetary Economics, Vol. 50, 49 - 123.
[2] Acemoglu, Daron, Simon Johnson, and James Robinson, 2001. \The Colonial
Origins of Comparative Development: An Empirical Investigation", American
Economic Review, Vol. 91(5), 1369-1401.
[3] Aghion, Philippe, George-Marios Angeletos, Abhijit Banerjee, and Kalina
Manova, 2005. \Volatility and Growth: Credit Constraints and Productivity-
Enhancing Investment", NBER Working Papers #11349.
[4] Aguiar, Mark and Gita Gopinath, 2006. \Defaultable debt, interest rates and
the current account", Journal of International Economics, Vol. 69(1), 64 - 83.
[5] Aiyagari, S. Rao, 1994. \Uninsured Idiosyncratic Risk and Aggregate Saving",
Quarterly Journal of Economics, Vol. 109(3), 17 - 31.
[6] Aizmann, Joshua and Mark Spiegel, 2002. \Institutional E?ciency, Monitoring
Costs, and the Share of FDI", NBER Working Paper # 9324.
[7] Aitken, Brian, Gordon H. Hanson, and Ann E. Harrison, 1997. \Spillovers,
Foreign Investment and Export Behavior", Journal of International Economics,
Vol. 43(1-2), 103 - 132.
[8] Albuquerque, Rui, 2003. \The Composition of International Capital Flows:
Risk Sharing Through Foreign Direct Investment", Journal of International
Economics, Vol. 61, 2:353-383.
[9] Alesina, Alberto and Guido Tabellini, 2006. \Why is flscal policy often pro-
cyclical?", mimeo, Harvard University.
[10] Alesina, Alberto, Sule Ozler, Nouriel Roubini, and Phillip Swagel, 1996. \Po-
litical Instability and Economic Growth", Journal of Economic Growth, Vol.
1(2), 189 - 211.
[11] Alfaro, LauraSebnem Kalemli-Ozcan, and Vadym Volosovych, 2005. \Why
Doesn?t Capital Flow from Rich to Poor Countries? An Empirical Investi-
gation", Review of Economics and Statistics (forthcoming).
144
[12] Alvarez, Fernando and Urban J. Jermann, 2001. \Quantitative Asset Pricing
Implications of Endogenous Solvency Constraints", Review of Financial Stud-
ies, Winter 2001 Vol. 14(4), 1117 - 1151.
[13] Anderson, James E. and Eric van Wincoop, 2003. \Gravity with Gravitas: A
Solution to the Border Puzzle", American Economic Review Vol. 93, 170 - 192.
[14] Anderson, James E. and Eric van Wincoop, 2004. \Trade Costs", Journal of
Economic Literature Vol. 42, 691 - 751.
[15] Arraiz, Irani, 2006. \Default, Settlement, and Repayment History: A Unifled
Model of Sovereign Debt", mimeo, University of Maryland at College Park.
[16] Arrellano, Cristina, 2004. \Default Risk and Income Fluctuations in Emerging
Economies", American Economic Review, forthcoming.
[17] Arrellano, Cristina and Enrique G. Mendoza, 2003. \Credit Frictions and ?Sud-
den Stops? in Small Open Economies: An Equilibrium Business Cycle Frame-
work for Emerging Markets Crises" in Dynamic Macroeconomic Analysis: The-
ory and Policy in General Equilibrium, S. Altug, J. Chadha and C. Nolan (eds.),
Cambridge University Press.
[18] Aviat, Antonin and Nicolas Coeurdacier, 2007. \The Geography of Trade in
Goods and Assets", Journal of International Economics, Vol. 71, 22 - 51.
[19] Bank of International Settlements, 2003. Guide to International Banking Statis-
tics. BIS Papers No. 16, April.
[20] Barro, Robert J., 1991. \Economic Growth in a Cross-section of Countries",
Quarterly Journal of Economics, Vol. 106, 2(May): 407 - 443.
[21] Blonigen, Bruce A., Ronald B. Davies and Keith Head, 2002. \Estimating the
Knowledge-Capital Model of the Multinational Enterprise: Comment", NBER
Working Paper # 8929.
[22] Bolton, Patrick and Xavier Freixas, 2000. \Equity Bonds and Bank Debt: Cap-
ital Structure and Financial Market Equilibrium Under Asymmetric Informa-
tion", Journal of Political Economy, Vol. 108(2), 324 - 351.
[23] Borensztein, Eduardo, De Gregorio, Jose, and Jong-Wha Lee, 1998. \How Does
Foreign Direct Investment Afiect Economic Growth?", Journal of International
Economics, Vol. 45 (1), 115 - 135.
145
[24] Bosworth, Barry and Susan Collins, 1999. \Capital Flows to Developing Coun-
tries: Implications for Saving and Investment", Brooking Papers on Economic
Activity, Vol. 1, 143 - 180.
[25] Bulow, Jeremy and Kenneth S. Rogofi, 1989. A Constant Recontracting Model
of Sovereign Debt. Journal of Political Economy, Vol. 97 (1), 155 - 178.
[26] Broda, Christian, 2004, \Terms of Trade and Exchange Rate Regimes in De-
veloping Countries", Journal of International Economics, Vol. 63, 31 - 58.
[27] Calvo, Guillermo, 1998. \Capital Flows and Capital-Market Crises: The Simple
Economics of Sudden Stops", Journal of Applied Economics, Vol. 1, 35 - 54.
[28] Calvo, Guillermo, 2005. \Volatility as an innovation deterrent: Adam Smith in
Stormy Weather", mimeo, Inter-American Development Bank.
[29] Calvo, Guillermo, Alejandro Izquierdo and Rudy Loo-Kung, 2006. \Relative
price variablility under Sudden Stops: The relevance of balance sheet efiects",
Journal of International Economics, Vol. 69, 231 - 254.
[30] Cantor, Richard and Frank Packer, 1996. \Determinants an Impact of Sovereign
Credit Ratings\, Federal Reserve of New York Economic Policy Review (Octo-
ber): 1 - 15.
[31] Carr, David L., James R. Markusen and Keith E. Maskus, 2001. \Estimat-
ing the Knowledge-Capital Model of the Multinational Enterprise", American
Economic Review, Vol. 87, 4 (September): 520 - 544.
[32] Catao, Luis and Sandeep Kapur, 2006. \Volatility and the Debt Intolerance
Paradox", IMF Stafi Papers, Vol. 52(2), 195 - 218.
[33] Cavallo, Eduardo A. and Andres Velasco, 2006. \Quid pro Quo: National Insti-
tutions and Sudden Stops in International Capital Movements", Working Paper
#581, Research Department, Inter-American Development Bank.
[34] Chaterjee, Satyajit, Dean Corbae, Nakajimo Makoto, Jose Victor Rios-Rull,
2002. A Quantitative Theory of Unsecured Consumer Credit with Risk of De-
fault, Working Paper.
[35] Cheng, I-Hui, Howard J. Wall, 2005. \Controlling for heterogeneity in gravity
models of trade and integration", Federal Reserve Bank of St. Louis Review,
Vol. 87(1), 49 - 63.
146
[36] Choe, Hyuk, Bong-Chan Kho, and Rene Stulz, 2004. \Do Domestic Investors
Have An Edge? The Trading Experience of Foreign Investors in Korea", Review
of Financial Studies, Vol. 18(3), 795 - 829.
[37] Chong, Alberto and Luisa Zanforlin, 2000. \Law Tradition and the Quality of
Institutions: Some Empirical Evidence", Journal of International Development,
Vol. 12, 1057 - 1068.
[38] Chuhan, Punam, Stijn Claessens, and Nlandu Mamingi, 1998. \Equity and
Bond Flow to Latin America and Asia: The Role of Global and Country Fac-
tors", Journal of Development Economics, Vol. 55(2), 441 - 465.
[39] Chuhan, Punam and Federico Sturzenegger, 2005. \Defaults Episodes in the
1980s and 1990s: What have we learned?", in Aizenman, Joshua and Brian
Pinto (eds.), Managing Economic Volatility and Crises, Cambridge University
Press, 471-519.
[40] Claessens, Stijn, Michael Dooley, and Andrew Warner, 1995. \Portfolio Capital
Flows: Hot or Cool?", World Bank Economic Review, Vol. 9 (1), 153 - 179.
[41] Dahlquist, Magnus and Goran Robertsson, 2001. \Direct foreign ownership, in-
stitutional investors, and flrm characteristics", Journal of Financial Economics,
Vol. 59 (3), 413 - 440.
[42] Daude, Christian and Marcel Fratzscher, 2006. \The pecking order of cross-
border investment", European Central Bank Working Paper Series No. 590 /
February.
[43] Daude, Christian, Eduardo Levy Yeyati, and Ernesto Stein, 2003. \Regional
Integration and the Location of FDI", Inter-American Development Bank, Re-
search Department Working Paper #492, Washington D.C.
[44] Daude, Christian and Ernesto Stein, 2007. \The Quality of Institutions and
Foreign Direct Investment", Economics & Politics, Vol. 19 (3), 317 - 344.
[45] Di Giovanni, Julian, 2005. \What drives capital  ows? The case of cross-border
M&A activity and flnancial deepening", Journal of International Economics
Vol. 65 (1), 127 - 149.
[46] Dixit, Avinash and Robert Pindyck, 1993. Investment under Uncertainty.
Princeton University Press: Princeton, New Jersey.
147
[47] Djankov, Simon, Edward Glaeser, Rafael La Porta, Florencio Lopez-de-Silanes,
and Andrei Shleifer, 2002. \The Regulation of Entry", Quarterly Journal of
Economics, Vol. 117 (1), 1 - 37.
[48] Durdu, Ceyhun Bora, Enrique Mendoza, and Marco Terrones, 2007. \Precau-
tionary Demand For Foreign Assets in Sudden Stop Economies: An Assessment
of the New Merchantilism", IMF Working Paper WP/07/146.
[49] Dvorak, Tomas 2005. \Do Domestic Investors Have an Information Advantage?
Evidence from Indonesia", Journal of Finance, Vol. 60 (2), 817 - 839.
[50] Eaton, Jonathan and Mark Gersovitz, 1981. \Debt with Potential Repudiation:
Theoretical and Empirical Analysis", Review of Economic Studies, Vol. 48 (2),
289 - 309.
[51] Easterly, William and Ross Levine, 1997. \Africa?s Growth Tragedy: Policies
and Ethnic Divisions", Quarterly Journal of Economics, Vol. 112 (4), 1203 -
1250.
[52] Eichengreen, Barry and Douglas Irwin, 1995. \Trade Blocks, Currency Blocs
and the Reorientation of Trade in the 1930s", Journal of International Eco-
nomics Vol. 38, 1-2 (February): 1 - 24.
[53] Faria, Andre and Paolo Mauro, 2004. \Institutions and the External Capital
Structure of Countries" IMF Working Paper 04/236, December 2004.
[54] Fatehi-Sedeh, Kamal and Hossein M. Saflzadeh, 1989. \The association be-
tween political instability and  ow of foreign direct investment", Management
International Review, Vol. 29 (4), 4 - 13.
[55] Faethi-Sedeh, Kamal and Hossein M. Saflzadeh, 1994. \The efiect of sociopo-
litical instability on the  ow of difierent types of foreign direct investment",
Journal of Business Research, Vol. 31 (1), 65 - 73.
[56] Fernandez-Arias, Eduardo, 1996. \The New Wave of Private Capital In ows:
Push or Pull?" Journal of Development Economics, Vol. 48, 389-418.
[57] Fernandez-Arias, Eduardo, Rodolfo Manuelli and Juan Blyde (Editors), 2006.
Sources of Growth in Latin America. What is Missing?, Inter-American Devel-
opment Bank.
[58] Frankel, Jefirey A. and Andrew Rose, 1996. \Currency crashes in emerging
markets: An empirical treatment", Journal of International Economics, Vol.
41, 351 - 366.
148
[59] Frankel, Jefirey A. with Ernesto Stein, and Shang-Jin Wei, 1997. Regional Trad-
ing Blocs in the World Economic System, Institute for International Economics,
Washington, DC
[60] Gelos, Gaston, Ratna Sahay, and Guido Sandleris, 2004. \Sovereign Borrowing
by Developing Countries: What Determines Market Access?", IMF Working
Paper WP/04/221.
[61] Gelos, Gaston and Shang-Jin Wei, 2005. \Transparency and International In-
vestors? Behavior", Journal of Finance (forthcoming).
[62] Gertler, MarkandKennethS.Rogofi, 1990.\North-SouthLendingandEndoge-
nous Domestic Capital Market Ine?ciencies", Journal of Monetary Economics,
Vol. 26, 245 - 266.
[63] Goldstein, Itay and Assaf Razin, 2006. \An information-based trade ofi be-
tween foreign direct investment and foreign portfolio investment", Journal of
International Economics, Vol. 70, 271 - 295.
[64] Gould, David M., 1994. \Immigrant Links to the Home Country: Empirical Im-
plications for U.S. Bilateral Trade Flows", Review of Economics and Statistics,
Vol. 76 (2), 302 - 316.
[65] Greene, William H., 1980. \On the Asymptotic Bias of the Ordinary Least
Squares Estimator of the Tobit Model", Econometrica, Vol. 48, 505 - 514.
[66] Hall, Robert E. and Charles I. Jones, 1999. \Why Do Some Countries Pro-
duce So Much More Output Per Worker Than Others?", Quarterly Journal of
Economics, Vol. 114 (1), 83 - 116.
[67] Hausmann, Ricardo and Eduardo Fernandez-Arias, 2000. \Foreign Direct In-
vestment: Good Cholesterol?", Research Department Working Paper # 417,
Inter-American Development Bank, Washington D.C.
[68] Harris, Milton and Artur Raviv, 1991. \The Theory of Capital Structure",
Journal of Finance, Vol. 46 (1), 297 - 355.
[69] Heckman, James, 1979. \Sample Selection Bias as a Speciflcation Error",
Econometrica, Vol. 42, 153 - 168.
[70] Helpman, Elhanan, 1984. \A Simple Theory of Trade with Multinational Cor-
porations", Journal of Political Economy, Vol. 92, 451 - 471.
149
[71] Helpman, Elhanan and Paul Krugman, 1985. Market Structure and Interna-
tional Trade, Cambridge, United States: MIT Press.
[72] Henisz, Witold J., 2000. \The institutional environment for economic growth",
Economics & Politics, Vol. 12, 1 - 31.
[73] Horstmann, Ignatius J. and James R. Markusen, 1987. \Strategic Investments
and the Development of Multinationals.", International Economic Review, Vol.
28 (1), 109 - 121.
[74] Horstmann, Ignatius J. and James R. Markusen, 1992. \Endogenous Market
Structures in International Trade", Journal of International Economics, Vol.
32 (1-2), 109 - 129.
[75] Jaimovich, Danny and Ugo Panizza, 2006. \Public Debt Around the World:
A New Dataset for Central Government Debt", Research Department Working
Paper WP-561, Inter-American Development Bank.
[76] Kaminsky, Graciela L., Carmen M. Reinhart and Carlos A. Vegh, 2004. \When
It Rains, It Pours: Procyclical Capital Flows and Macroeconomic Policies,"
NBER Macroeconomics Annual.
[77] Kaufmann, Daniel, Aart Kraay, and Pablo Zoido-Lobatn, 1999a. \Governance
Matters", Policy Research Working Paper # 2196, The World Bank, Washing-
ton, DC.
[78] Kaufmann, Daniel, Aart Kraay, and Pablo Zoido-Lobatn, 1999b. \Aggregating
Governance Indicators" Policy Research Working Paper # 2195, The World
Bank, Washington, DC.
[79] Klein, Michael W., Joe Peek, and Eric P. Rosengren, 2002. \Troubled Banks,
Impaired Foreign Direct Investment: The Role of Relative Access to Credit",
American Economic Review, Vol. 92 (3), 664 - 682.
[80] Knack, Stephen and Philip Keefer, 1995. \Institutions and Economic Perfor-
mance: Cross-Country Tests Using Alternative Institutional Measures", Eco-
nomics & Politics, Vol. 7 (3), 207 - 228.
[81] Kose, M. Ayahn, 2002. \Explaining business cycles in small open economies
?How much do world prices matter?"?, Journal of International Economics,
Vol. 56, 299 - 327.
150
[82] Krueger, Anne O., 1993. \Virtuous and Vicious Circles in Economic Develop-
ment", American Economic Review (Papers and Proceedings), Vol. 83 (3), 351
- 356.
[83] Lane, Philip, and Gian Maria Milesi-Ferretti, 2004. \International Investment
Patterns", Institute for International Integration Studies Discussion Paper No.
24.
[84] La Porta, Rafael, Florencio Lpez-de-Silanes, and Andrei Shleifer, 1997. \Legal
Determinants of External Finance", Journal of Finance, Vol. 52 (3), 1131 -
1150.
[85] La Porta, Rafael, Florencio Lpez-de-Silanes, Andrei Shleifer and Robert W.
Vishny, 1998. "Law and Finance". Journal of Political Economy (December),
Vol. 106, 6: 1113-55.
[86] La Porta, Rafael, Florencio Lpez-de-Silanes, Andrei Shleifer and Robert W.
Vishny, 1999. \The Quality of Government", Journal of Law, Economics and
Organization, Vol. 15 (1), 222 - 279.
[87] La Porta, Rafael, Florencio Lpez-de-Silanes, Andrei Shleifer and Robert W.
Vishny, 2000. \Investor Protection and Corporate Governance", Journal of Fi-
nancial Economics, Vol. 58 (1-2), 3 - 27.
[88] Leamer, Edward E. and James Levinsohn, 1995. \International Trade Theory:
The Evidence", in Gene M. Grossman and Kenneth S. Rogofi, (Eds.), The
Handbook of International Economics: Vol. III, North Holland, Amsterdam ;
New York, 1339 - 1394.
[89] Loree, David W. and Stephen E. Guisinger, 1995. \Policy and Non-Policy De-
terminants of U.S. Equity Foreign Direct Investment", Journal of International
Business Studies, Vol. 26 (2), 281 - 299.
[90] Lucas, Robert E., 1990. \Why Doesn?t Capital Flow From Rich to Poor Coun-
tries?", American Economic Review (Papers and Proceedings), Vol. 80(2), 92 -
96.
[91] Magud, Nicolas, and Carmen M. Reinhart, 2005. \Capital Controls: An Eval-
uation", in Sebastian Edwards (ed.), International Capital Flows, Chicago:
Chicago University Press for the NBER.
[92] Markusen, James R., 1997. \Trade versus Investment Liberalization", NBER
Working Paper # 6231.
151
[93] Markusen, James R., 2001. Multinationals and the Theory of International
Trade, Cambridge: MIT Press.
[94] Martin, Philippe, and Helene Rey, 2004. \Financial super-markets: size matters
for asset trade", Journal of International Economics Vol. 64 (2), 335 - 361.
[95] Mauro, Paolo, 1995. \Corruption and Growth", Quarterly Journal of Eco-
nomics, Vol. 110 (3), 681 - 712.
[96] Mendoza, Enrique, 1991. \Real Business Cycles in a Small Open Economy",
American Economic Review, Vol. 81(4), 797 - 818.
[97] Mendoza, Enrique, 1995. \The Terms of Trade, the Real Exchange Rate, and
Economic Fluctuations", International Economic Review, Vol. 36, 101 - 137.
[98] Mendoza, Enrique, 1997. \Terms-of-trade uncertainty and economic growth",
Journal of Development Economics, Vol. 54, 323 - 356.
[99] Mendoza, Enrique, 2002. \Credit, Prices and Crashes: Business Cycles with
Sudden Stops" in S. Edwards and J. Frankel (eds.) Preventing Currency Crises
in Emerging Markets, NBER, University of Chicago Press.
[100] Mendoza, Enrique and P. Marcelo Oviedo, 2006. \Fiscal Policy and Mar-
coeconomic Uncertainty in Developing Countries: The Tale of the Tormented
Insurer", NBER Working Paper #12586.
[101] Mody, Ashoka, Assaf Razin and Efraim Sadka, 2003. \The Role of Informa-
tion in Driving FDI Flows: Host-Country Transparency and Source-Country
Specialization", NBER Working Paper # 9962.
[102] Mody, Ashoka and Antu P. Murshid, 2005. \Growing up with capital  ows",
Journal of International Economics, Vol. 65, 249 - 266.
[103] Montiel, Peter and Carmen M. Reinhart, 1999. \Do capital controls and
macroeconomic policies in uence the volume and composition of capital  ows?
Evidence from the 1990s", Journal of International Money and Finance, Vol.
18, 619 - 635.
[104] Myers, Stewart C., 1984. \The Capital Structure Puzzle", Journal of Finance,
Vol. 39 (3), 575 - 592.
152
[105] Myers, Stewart C. and Nicholas S. Majlauf, 1984. \Corporate flnancing and in-
vestment decisions when flrms have information investors do not have" Journal
of Financial Economics Vol. 5, 147 - 175.
[106] Neumann, Rebecca, 2003. \International capital  ows under asymmetric in-
formation and costly monitoring: implications of debt and equity flnancing",
Canadian Journal of Economics, Vol. 36(3), 674 - 700.
[107] Nicita, Alessandro and Marcelo Olarreaga, 2000. \Exports and information
spillovers", Policy Research Working Paper #2474, The World Bank, Wash-
ington DC.
[108] Obstfeld, Maurice, and Kenneth S. Rogofi, 2000. \The Six Major Puzzles in
International Economics: Is there a Common Cause?", in NBER Macroeco-
nomics Annual 2000, University of Chicago Press, Chicago.
[109] Papaioannou, Elias, 2005. \What drives international bank  ows? Politics,
institutions and other determinants", ECB Working Paper No 417.
[110] Portes, Richard, Helene Rey, Yonghyup Oh, 2001. \Information and Capi-
tal Flows: The Determinants of Transactions in Financial Assets", European
Economic Review, Vol. 45 (4-6), 783-96.
[111] Portes, Richard and Helene Rey, 2005. \The determinants of cross-border
equity  ows", Journal of International Economics Vol. 65 (2), 269 - 296.
[112] Ramey, Garey and Ramey, Valerie, 1995. \Cross-country Evidence on the Link
between Volatility and Growth", American Economic Review, Vol. 85, 1138 -
1151.
[113] Razin, Assaf, 2004. \The Contribution of FDI Flows to Domestic Investment
in Capacity, and Vice Versa", in Ito, Takatoshi, and Andrew K. Rose (eds.),
Growth and Productivity in East Asia, NBER-East Asia Seminar on Economics,
Vol. 13, 149 - 167.
[114] Razin, Assaf, Yona Rubinstein, and Efraim Sadka, 2005. \Fixed Costs and Bi-
lateralFDIFlows: Con ictingEfiectsofCountrySpeciflcProductivityShocks",
unpublished mimeo, Tel Aviv University.
[115] Razin, Assaf, Efraim Sadka, and Chi-Wa Yuen, 1998. \A pecking order of cap-
ital  ows and international tax principles", Journal of International Economics
Vol. 44, 45 - 68.
153
[116] Reinhart, Carmen, 2002. \Default, Currency Crises, and Sovereign Credit
Ratings", World Bank Economic Review, Vol. 17(3), 151 - 170.
[117] Reinhart, Carmen and Kenneth S. Rogofi, 2004. \Serial Default and the \Para-
dox" of Rich to Poor Capital Flows", American Economic Review, Papers and
Proceedings.
[118] Reinhart, Carmen, Kenneth S. Rogofi and Miguel A. Savastano, 2003. \Debt
Intolerance", Brookings Papers on Economic Activity, Vol. 1, 1-74.
[119] Rose, Andrew K., 2000. \One market, one money: the efiect of common cur-
rencies on trade", Economic Policy, Vol. 14 (30), 7 - 46.
[120] Rose, Andrew K., 2005. \One Reason Countries Pay Their Debts: Renegotia-
tion and International Trade", Journal of Development Economics, Vol. 77(1),
189 - 206.
[121] Sachs, Jefirey D. and Daniel Cohen, 1982. \LDC borrowing with default risk",
NBER Working Paper #925.
[122] Sachs, Jefirey D. and Andrew M. Warner, 1995. \Economic Reform and the
Process of Global Integration", Brookings Papers on Economic Activity, Vol. 1,
1 - 118.
[123] Sarno, Lucio and Mark P. Taylor, 1999. \Hot Money, Accounting Labels and
the Permanence of Capital Flows to Developing Countries", Journal of Devel-
opment Economics, Vol. 59 (2), 337 - 364.
[124] Schneider, Friedrich G. and Bruno Frey, 1985. \Economic and Political De-
terminants of Foreign Direct Investment", World Development Vol. 13, 161 -
175.
[125] Santos, J.M.C. and Silvana Tenreyro, 2006. \The Log of Gravity", Review of
Economics and Statistics, Vol. 88 (4), 641 - 658.
[126] Schollhammer, Hans and Douglas Nigh, 1987. \Foreign Direct Investment,
Political Con ict and Co-operation: The Asymmetric Response Hypothesis",
Managerial and Decision Economics, Vol. 8 (4), 307 - 312.
[127] Sethi, D., S.E Guisinger, S.E. Phelan, and D.M. Berg, 2003. \Trends in for-
eign direct investment  ows: A theoretical and empirical analysis", Journal of
International Business Studies, Vol. 34, 315 - 326.
154
[128] Shleifer, Andrei and RobertW.Vishny, 1993. \Corruption", Quarterly Journal
of Economics, Vol. 109, 599 - 617.
[129] Smarzynska, Beata and Shang-Jin Wei, 2000. \Corruption and Composition
of Foreign Direct Investment: Firm-Level Evidence", NBER Working Papers
# 7969, National Bureau of Economic Research, Cambridge, MA.
[130] Stasavage, David, 2002. \Private Investment and Political Institutions", Eco-
nomics and Politics, Vol. 14, 41 - 63.
[131] Stein, Ernesto and Christian Daude, 2002. \Institutions, Integration and the
Location of FDI", in New Horizonts of Foreign Direct Investment, OECD
Global Forum on International Investment, OECD, Paris.
[132] Tomz, Michael and Mark Wright, 2007. \Do Countries Default in Bad
Times?", Journal of the European Economic Association, Vol. 2(2-3), May,
352 - 360.
[133] Tauchen, George and Robert Hussey, 1991. \Quadrature-Based Methods for
Obtaining Approximate Solutions to Nonlinear Asset Pricing Models," Econo-
metrica, Vol. 59(2), 371 - 396.
[134] Uhlig, Harald and Morten Ravn, 2002. \On Adjusting the HP-Filter for the
Frequency of Observations", Review of Economics and Statistics, Vol. 84(2),
May, 371 - 376.
[135] Wei, Shang-Jin, 1997. \Why is Corruption so Much More Taxing than Tax?
Arbitrariness Kills", NBER Working Paper # 6255, National Bureau of Eco-
nomic Research, Cambridge, MA.
[136] Wei, Shang-Jin, 2000a. \Local Corruption and Global Capital Flows", Brook-
ings Papers on Economic Activity, Vol. 2, 303 - 354.
[137] Wei, Shang-Jin, 2000b. \How Taxing is Corruption on International In-
vestors", Review of Economics and Statistics, Vol. 82 (1), 1-11.
[138] Wheeler, D. and Ashoka Mody, 1992. \International Investment Location De-
cisions", Journal of International Economics, Vol 33, 57 - 76.
[139] Yue, Vivian Z., 2006. \Sovereign Default and Debt Renegotiation", unpub-
lished mimeo, New York University.
155