Accounting undergraduate Honors theses: Essays on international trade and finance

The dissertation consists of three papers exploring the macroeconomic implications of heterogeneity of countries in financial development, economic interconnectedness via trade and financial linkages.

University of Arkansas, Fayetteville

University of Arkansas, Fayetteville

Follow this and additional works at:http://scholarworks.uark.edu/etd

Part of theFinance Commons,International Economics Commons, and theMacroeconomicsCommons

This Dissertation is brought to you for free and open access by ScholarWorks@UARK It has been accepted for inclusion in Theses and Dissertations by

an authorized administrator of ScholarWorks@UARK For more information, please contact scholar@uark.edu, ccmiddle@uark.edu

Recommended Citation

Adarov, Amat, "Essays on International Trade and Finance" (2012) Theses and Dissertations 307.

http://scholarworks.uark.edu/etd/307

ESSAYS ON INTERNATIONAL TRADE AND FINANCE

ESSAYS ON INTERNATIONAL TRADE AND FINANCE

A dissertation submitted in partial fulfillment

of the requirements for the degree of Doctor of Philosophy in Economics

By

Amat B Adarov Altai State Technical University Bachelor of Arts in Economics, 2003

University of Arkansas Master of Arts in Economics, 2007

May 2012 University of Arkansas

ABSTRACT

The dissertation consists of three papers exploring the macroeconomic implications of

heterogeneity of countries in financial development, economic interconnectedness via trade and

financial linkages

Chapter 1 examines whether countries which are more centrally located in the global

trade network have more synchronized stock markets Global trade data is used to construct a

novel measure of random walk betweenness centrality (RWBC), measuring the extent to which a

country lies on random pathways in-between other countries and is therefore likely to be a

conduit in the transmission of a shock across global markets Based on a panel dataset of 58

countries over the period 1990–2000, the study finds that higher centrality of a country in the

world trade network is indeed associated with greater stock market synchronicity, ceteris paribus

Chapter 2 uses aggregate macroeconomic experiences of 118 countries over the period

1994–2008 to establish benchmark relationships between macroeconomic fundamentals and

levels of financial development of the banking sector, equity markets, and private bond markets

The analysis quantifies the extent to which de-facto financial development of emerging market

economies (EMEs) deviates from the levels predicted by their macroeconomic stance While

financial markets in Latin American EMEs are found to be well aligned with their

macroeconomic fundamentals, Asian EMEs exceed their reference levels, and European EMEs

are found to be systematically financially underdeveloped No support is found for the argument

that these misalignments are caused by heterogeneity in institutional development

Finally, chapter 3 studies the properties and evolution of the product space—a network of

relatedness between products We use bilateral trade data for 187 countries to construct the

product space and export specialization of individual countries over the period 1965—2000 The

study shows that the product space changed significantly during the 20th century and represents

a highly uneven core-periphery structure The highly interconnected core consists of three

industries—chemicals, industrial machinery, and crude materials, each forming around 20% of

all linkages Product synergies that these “commanding heights” industries yield are strategically

important for industrialization policies Regression analysis confirms that specialization in these

industries is associated with higher real income levels

This dissertation is approved for recommendation to the Graduate Council

DISSERTATION DUPLICATION RELEASE

I hereby authorize the University of Arkansas Libraries to duplicate this dissertation

when needed for research and/or scholarship

Agreed

Amat B Adarov

Refused

Amat B Adarov

ACKNOWLEDGEMENTS

I am grateful to my dissertation co-chairs Javier Reyes and Raja Kali for their invaluable

guidance and support Special thanks are due to Gary Ferrier for insightful comments that helped

improve the quality of the dissertation I am in debt to all my friends and colleagues at the

Department of Economics for making the years in the doctorate program so pleasant

My gratitude goes to all my friends around the world whose support was crucial to me

Finally, and most importantly, I am extremely grateful to my parents, Boris and

Aleksandra, and my sister, Aradyana, for their continuous love and encouragement

DEDICATION

To my parents, Adarov Boris and Adarova Aleksandra

TABLE OF CONTENTS

Introduction ……… ……… 1

Chapter 1: Stock Market Synchronicity and the Global Trade Network:

a Random-Walk Approach ……….……… 4

Chapter 2: Macroeconomic and Institutional Determinants of Financial

Development: Implications for Emerging Markets ……… 43

Chapter 3: International Trade and Export Specialization Dynamics:

a Network Perspective ……… 83 Conclusion ……… ……….…… …… 127

INTRODUCTION

The dissertation consists of three papers exploring the macroeconomic implications of

heterogeneity of countries in financial development, economic interconnectedness via trade and

financial linkages

In Chapter 1, recent advancements in network theory are used along with conventional

panel data techniques to analyze cross-border financial synchronicity and propagation of

financial shocks The central question is whether countries that are better integrated into the

world economy and more centrally located in the global trade network have more synchronous

financial markets The paper uses a novel measure of random walk betweenness centrality

(RWBC) to gauge the extent to which a country lies on random pathways in-between other

countries in the global trade network and is therefore likely to be a conduit in the cross-border

transmission of a shock resulting in higher stock market synchronicity Based on a panel dataset

of 58 countries over the period 1990-2000 the analysis demonstrates that higher centrality of a

country in the world economy is indeed associated with higher synchronicity, ceteris paribus

The analysis also reveals that the global trade network has a well-defined core-periphery

structure, where the highly interconnected core, comprising China, France, Germany, Italy,

Japan, and the UK, is characterized by significantly lower synchronicity The study has

important policy implications as it demonstrates the importance of network centrality in the

world economy for understanding financial synchronicity and global shock propagation This

contrasts sharply with conventional measures of economic integration, e.g trade openness,

which do not reflect the risks associated with economic partners Therefore, the network-based

approach may serve as an important tool for monitoring systemic risks and resilience of the

world economy, as well as risk exposures of individual countries to financial shocks

Chapter 2 examines an often-sounded claim that financial markets of emerging market

economies (EMEs) are weak, and this is primarily the result of institutional impediments Based

on aggregate experience of 118 countries excluding EMEs over the period of 1994-2008 I find

that financial development measured by financial market size can be consistently related to a

country's general stage of economic development (proxied by real per capita income), economic

openness, inflation, and inflation volatility Then, I use a full sample of countries and original

specification augmented by fixed effects for major EME groups to test if there is any residual

effect pertinent to EMEs that is unexplained by macroeconomic fundamentals Notably, while

financial markets in Latin American EMEs are found to be well-aligned with their

macroeconomic parameters, European EMEs are underperforming and Asian EMEs are

overperforming relative to their expected levels However, the quality of institutions does not

contribute much to explaining these misalignments

Finally, in Chapter 3, bilateral trade data on 1006 product categories (SITC4

classification) for 187 countries over the period 1965-2000 is used to construct the product

space—a network of relatedness between nodes-products, where the weight of a link between

individual products is proportional to the probability that they are produced and exported

together Along these lines, export specialization of a country is a subnetwork of the product

space formed by products in which it enjoys revealed comparative advantage (RCA) Network

properties and evolution of the product space and export specialization patterns of individual

countries are then examined in order to understand their implications for economic development

The paper demonstrates that the product space changed significantly during the twentieth century

and evolved into a highly uneven core-periphery structure Specifically, the highly

interconnected core consists of only three industries—chemicals, industrial machinery, and crude

materials—each forming around 20 percent of all product linkages Product synergies that these

commanding heights industries yield are strategically important for industrialization policies

Regression analysis confirms that specialization in the commanding heights industries is

associated with higher real income levels, controlling for other relevant factors

nations Recently however, as the housing sector in the United States slowed sharply and turmoil erupted in many financial markets, a different theme has come to the foreground: “decoupling.” This refers to the apparent divergence in economic performance among different regions of the

world economy In the context of these opposing discussions it seems reasonable to ask, to what

extent does integration into the global economy influence synchronized movements in markets

around the world? Are there meaningful differences between groups of countries in this

relationship?

In this paper, we aim to cast some light on these issues by focusing on a narrow version

of the questions above Specifically, how does integration into the global economy affect

synchronicity in financial markets? Our approach involves two methodological novelties First,

we construct a network of economic connectedness among nations by using the NBER–United

Nations World Trade Database associated with Feenstra et al (2005) We view individual

countries in the world trade network as nodes connected by bilateral trade linkages that are

weighted by trade volume Our assumption here is that the global trade network is a meaningful

proxy for economic connectedness among nations, and that alternative proxies of the global

economic network are likely to be closely related to the trade network, e.g a network of bilateral

capital flows is linked to trade flows via the balance of payments As trade linkages are relatively

stable over time, highly correlated with other cross-country linkages1, and we are primarily interested in stock market synchronicity over a relatively long time horizon (1990–2000), this

seems particularly suitable

Second, we use a novel approach to computing country-level connectedness that is

agnostic about the way in which each country receives and transmits shocks Specifically, based

on the notion of random walk betweenness centrality introduced in Newman (2005), for each

country in our sample we compute its random walk betweenness centrality in the world trade

network (RWBC) In brief, random walk betweenness centrality of node i is equal to the number

of times that a random walk starting at s and ending at t passes through i along the way, averaged

over all possible combinations of s and t in the network In computing RWBC, we assume that

the probability that a financial shock follows a particular link along its propagation path in the

trade network is proportional to the intensity of bilateral trade flow the link represents

Hence, in the context of the world trade network, RWBC summarizes the connectedness

of a country in the world economy and its ultimate exposure to a financial shock that can

originate anywhere in the system and spread through the network in a manner that is not

necessarily optimal An attraction of this measure is that it is agnostic about which path a shock actually takes between any particular “epicenter” country and a “target” country with regard to the transmission of shocks Since the propagation mechanism of international economic shocks

is not well understood, with different hypotheses vying for attention in the literature, this

approach seems especially useful as it does not favor one transmission mechanism over another

We then examine whether a position of a country in the world trade network contributes

to explaining financial synchronicity Financial synchronicity is measured as comovement

intensity between stock market indices of individual countries and the index of a benchmark

economy (in our case, the USA and the Dow Jones Industrial Average) that is inspired by the

work of Morck, Yeung and Yu (2000) If stock prices are based mainly on the capitalization of

country-specific information we expect a low degree of synchronicity, while greater degree of

interdependence will be reflected in higher synchronicity, ceteris paribus

Our basic hypothesis is then formulated as follows Other things equal, a country that has

a high measure of random walk betweenness centrality lies on more random pathways

in-between countries and is therefore more likely to be affected by an external shock, regardless of

the exact transmission mechanism, than a country with lower RWBC This will be reflected in a

higher level of stock market synchronicity of a high-RWBC country than a low-RWBC country

Our empirical analysis supports this hypothesis Greater connectedness of an economy in

the global trade network as measured by RWBC is associated with higher stock market

synchronicity, after controlling for other relevant characteristics However, we find that a group

of nations that are highly central in the global trade network (we refer to them as the “core” of the network) are characterized by uniformly lower financial synchronicity than others The high-

RWBC core is comprised of the UK, Germany, France, Italy, China, and Japan

In terms of the literature, only a few studies have attempted to take into consideration

multilateral linkages in the global economy to explain stock market correlations Forbes and

Rigobon (2002) find that trade linkages are important factors for stock market dynamics and therefore a country’s susceptibility to financial crisis Reinhart and Kaminsky (2008) analyze the three emerging markets that experienced financial crises in the late 1990s: Brazil, Russia, and

Thailand, and suggest that financial turbulence in these countries spreads globally only when the

shock reaches world financial centers and remains local otherwise A recent paper by Kali and

Reyes (2010) explicitly uses a network approach to international economic integration to study

financial crisis episodes and associated contagion

A separate strand of the literature, pioneered by Imbs (2004, 2006), focuses on business

cycle synchronization and uses simultaneous equations systems to disentangle the complex

interactions between trade, finance, specialization, and business cycle synchronization The

overall effect of trade on business cycle synchronization is confirmed to be strong and a sizable

portion is found to work through intra-industry trade

Our approach here is differentiated from the prior literature along several dimensions

Most importantly, we apply a network approach to understand stock market synchronicity Using

the network of global trade linkages enables us to use a completely multilateral approach to the

propagation of financial shocks Our measure of network position, RWBC, is novel to the

literature and well-suited to the application Second, unlike most studies we address stock market

synchronicity over the long run rather than focusing on financial turmoil periods alone This is an

important distinction because financial crisis years are likely to be characterized by downward

financial trends in stock markets resulting in a bias towards higher synchronicity values Our

empirical analysis is based on a panel dataset spanning the 1990–2000 period that includes both

tranquil periods and periods of economic crises Third, we assess stock market synchronicity of a

wide range of countries from diverse regions of the world and different in terms of economic

development to ensure results are not driven by individual country properties2

The rest of the paper is organized as follows Section 2 discusses our empirical strategy

and data Regression results are discussed in Section 3 Section 4 presents concluding remarks

1.2 Research framework and data

The question the paper focuses on is whether more interconnected economies have more

synchronous stock market dynamics In order to test this hypothesis, in this section we first

develop a measure of stock market synchronicity and a measure of interconnectedness among

individual economies – RWBC We also recognize that other macroeconomic factors can

potentially affect the degree of financial synchronicity and consider control variables deemed to

be relevant in the related literature Then, the benchmark econometric specification is described

1.2.1 Stock market synchronicity

Computation of a stock market synchronicity measure that is comparable across countries

requires selection of a common benchmark country and associated stock market index to which

all other countries are compared We use the United States of America as the benchmark country

and the Dow Jones Industrial Average (the DJIA) as the benchmark index3 Based on our

2

We would be remiss not to mention a rich strand of work in finance that uses cointegration methods to demonstrate international stock market interdependence However, this literature does not concern itself with the channels of transmission and is therefore orthogonal to our focus Noteworthy papers are Awokuse, Bessler and Chopra (2009) and Arshanapalli and Doukas (1993)

3

We use the DJIA index as it is the most widely recognized of the stock market indices We realize it is often criticized, e.g for being a price-weighted measure, which affects its accuracy as

methodology, the US is the most integrated country in the world economy as identified by

centrality in the global trade network, and, in general, it is hard to find a better representative

financial center by any standard

Stock market index data are obtained from the Bloomberg stock market database, where

each data point represents a daily stock market index closing value Our dataset comprises 58

countries between the 1990–2000 period For each country in the sample we identify a

representative stock market index and employ several techniques to compute its synchronicity

with respect to the benchmark index, the DJIA Our main analysis is based on two synchronicity

measures, denoted further as Synch (FREQ) and Synch (R-SQ), that are inspired by the synchronicity

measures of Morck, Yeung and Yu (2000) We also use a third viable measure denoted as

baseline synchronicity variables are developed

Calculation of Synch (FREQ) involves two steps First, we compute the frequency of stock

market index comovements in year t for country i as a simple fraction:

t

t t

Days

s Comovement Frequency

where Comovements i,t is the number of days in year t in which stock market index of country i

moves in the same direction as the DJIA, and Days i,t is the total number of days for which both

stock markets were operating in year t

Equation (1) provides an intuitive assessment of stock market comovements with the

benchmark US equity market at daily frequency for a given year For instance, in the case of

Brazil and its representative stock market index, the Bovespa Index, comprised of the most liquid

stocks traded on the Sao Paulo Stock Exchange, the frequency value in year 2000 is 0.6929,

implying the Bovespa Index moved in the same direction as the DJIA 69.29% of days Figure 1

lists the frequency of stock market comovements for all countries assessed in our study

[Insert Figure 1 here]

However, the computed frequency variable is confined in the interval of [0,1] and

therefore cannot be used in our regression analysis directly In order to map frequency values to

the real number set we apply the standard statistical technique of logistic transformation as

t

Frequency

Frequency Ln

Synch

,

, )

( ,

Hence, our first stock market synchronicity measure, Synch (FREQ) , is merely a logistic

transformation of stock market comovements frequency Although it is a simple measure, we

believe it is adequate for our purposes and is robust to most issues associated with alternative

measures as we keep track only of the direction of stock market index dynamics Synch (FREQ)

values for the year 2000 are presented in Figure 25

[Insert Figure 2 here]

Another measure of stock market synchronicity used in our formal analysis, Synch (R-SQ),

is based on a goodness-of-fit approach rather than a fraction of comovement days:

R t

R

R Ln Synch

, 2 , 2 )

( ,

1

(3)

where R 2 i,t is the coefficient of determination from the linear regression

t t

log-differenced form) on those of the DJIA Again, as the coefficient of determination is

bounded in the [0,1] interval, we apply logistic transformation to map the variable to the real

number set

We construct a panel dataset for synchronicity based on daily closing stock market values

with appropriate adjustments made for time zone differences in the operation of corresponding

stock exchanges For instance, a shock affecting the New York Stock Exchange on December 1

is reflected in the US in the reported December 1 daily closing prices, while the London Stock

Exchange would reflect the effects of this shock, if any, in the reported December 2 daily closing

prices because of the time zone difference Therefore, in calculations of synchronicity measures

5

For brevity we present diagrams for the Synch (FREQ) variable only, as all three stock market

synchronicity measures that we develop in the study are highly correlated (see Table 1) and

diagrams for Synch (R-SQ) and Synch (CORR) look virtually identical to those of Synch (FREQ)

we ensure the data is within the same frame of reference by lagging stock market index values by

one day for stock exchanges located to the east of New York (hence, all countries not located in

North America or South America), e.g local index values reported in non-American economies

on December 2 would correspond to the DJIA values reported in the US on December 1 In

addition, for robustness we drop synchronicity values which are based on less than 100 days of

equity market data reported per year since we believe these to be unreliable

We acknowledge the fact that our synchronicity measures are imperfect since there can

be scenarios which may or may not be captured by them explicitly For example, if capital

market-relevant news or events are announced after a stock market is closed for the day, the effects would be reflected in the next day’s price, but there is no way to control for this unless we had an explicit “news” indicator Recent studies that have looked at the effects of macroeconomic news (economic information) on stock market prices, e.g Albuquerque and

Vega (2009), Karolyi and Stulz (1996), McQueen and Roley (1993), Wongswan (2006), examine

the mechanisms of price discovery and spillovers on interdependent asset markets after public

economic news releases Notably, the literature suggests that the effect of macroeconomic

announcements on stock market comovements manifests itself in high-frequency stock market

data (daily or intra-day return dynamics), and in this case constitutes an important source of

international stock market comovements At low frequencies the effect is minimal

In fact, our synchronicity variables take advantage of high-frequency (daily) stock market

data, and can be viewed as statistics summarizing the daily dynamics of relevant macroeconomic

fundamentals and the effect of economic news However, our analysis differs from this literature

as we focus on systematic factors that explain cross-country stock market synchronicity over a

long time horizon

It could also be the case that a shock may affect different markets with a time lag for

simple reasons, like holidays, or more complicated ones, such as markets not fully realizing the

extent of the effects due to such factors as regulations by domestic or international agencies

(central banks, rating agencies, etc.) Despite these potential shortcomings of the synchronicity

measures, we believe that, given large sample size, the impact of anomalous adjustments due to

the reasons mentioned here is minimal

Summary statistics and pairwise correlations between the three alternative stock market

synchronicity measures Synch (FREQ) , Synch (R-SQ) and Synch (CORR) are reported in Table 1 As

correlations between the three synchronicity measures are fairly high, we expect regression

results to be similar regardless of the particular choice of the synchronicity variable

[Insert Table 1 here]

1.2.2 Random walk betweenness centrality (RWBC)

Our hypothesis of interest calls for the application of a measure that is a proxy for the

degree of economic interconnectedness among countries We believe a measure of economic

integration that is based on network approach is superior to conventional measures of integration

relying on trade or investment intensity and not taking into account asymmetries in the world

economy that are captured by network modeling Connectedness of a country in the global

economic network seems to be especially important for understanding how financial shocks

spread across countries Therefore, we derive a measure of random walk betweenness centrality

of a country in the global trade network, denoted further as random walk betweenness centrality

or RWBC

In the networks literature centrality is a measure that summarizes the position of a given

node in the network based on the value of its relations, and relations of the nodes it is connected

to There are different measures of centrality (binary and weighted) that are based on closeness

or betweenness We use random walk betweenness centrality, first suggested by Newman (2005)

and later expanded by Fisher and Vega-Redondo (2006) Insightful discussion of the measure

and its technical properties is provided in Newman (2005)

In particular, he describes RWBC of a given node i as the number of times that a random

walk starting at node s and ending at node t passes through i along the way, averaged over a large

number of trials of the random walk for all possible source-target pairs of s and t RWBC is most

appropriate for a network in which a signal spreads randomly and the actual paths along which it

travels are not necessarily optimal Newman (2005) derives a sequence of equations to obtain

RWBC values for node i by manipulating a diagonal matrix of node degrees D and an adjacency

matrix A:

RWBCi =

,

where s,t; ; n is the number of nodes in the network, and matrix T is obtained by (1) inverting the matrix D – A with any single row and a corresponding column removed, and (2) adding back a row and a column of zeros to

the position where they were removed in step 1

Fisher and Vega-Redondo (2006) introduce important generalizations to matrix

methodology for obtaining RWBC for weighted directed networks, in particular, the

Moore-Penrose pseudo-inverse, which is more suitable for networks based on real economic data

Following Newman (2005) and Fisher and Vega-Redondo (2006) methodology, Fagiolo, Reyes,

and Schiavo (2008) use international trade flows data provided in Gleiditsch (2002) to build time

series of weighted directed networks and compute RWBC values for 159 countries over the

period of 1981 to 20006

[Insert Figure 3 here]

Based on the original bilateral trade data from the NBER-United Nations World Trade

Database associated with Feenstra et al (2005), we construct a panel dataset of RWBC for the 58

countries assessed in our study over the period 1990–2000 Figure 3 shows RWBC data for our

sample for the year 2000 As can be seen, RWBC is not directly related to the level of economic

development, as, for instance, the RWBC range is comprised of both high-income and

low-income countries At the same time, inequality in RWBC between individual economies is vast,

e.g RWBC value for Malta is 18 times smaller than RWBC value for Germany The benchmark

country of our analysis, the US, has RWBC value of 0.6297 in the year 2000, which is by far

higher than RWBC of any other country This speaks in favor of the argument that the USA is a

good benchmark economy to relate other countries to For comparison, the second-highest

RWBC economy in the sample is Germany with the value of only 0.2743

Notably, the pairwise correlation coefficient between RWBC and trade openness,

measured as trade to GDP ratio, that is also used in our study is -0.24, implying the two variables

reflect rather different information about the degree of integration in the world economy Hence,

6

The literature applying network perspectives to the study of international economic integration has expanded substantially in recent years and the principal framework for this approach has defined linkages in terms of international trade flows (in many cases weighted by GDP) Recent studies in this area include Bhattacharya et al (2007), Saramaki et al (2007), Fagiolo, Reyes and Schiavo (2008), Kali and Reyes (2007), Serrano et al (2007)

we would like to examine whether the proposed network measure explains financial

synchronicity better than conventional measures of economic integration

1.2.3 Controlling for macroeconomic factors

We also consider other macroeconomic variables that can affect stock market

synchronicity, including stock market capitalization to GDP ratio, trade to GDP ratio, exchange

rate regime, and a dummy variable for financial centers

Stock market capitalization to GDP ratio Higher synchronicity could be a result of

equity market size effects, with higher volumes of publicly traded stocks leading to greater

significance of financial shocks Hence, we control for the level of equity market development

by including stock market capitalization (the total market value of stocks listed on the stock

exchange), as a share of GDP The data for stock market capitalization is obtained from the

Database on Financial Development and Structure, associated with Beck et al (2000), recent

revision introduced in Beck and Demirgüç-Kunt (2009)

Trade to GDP ratio, calculated as the sum of exports and imports as a share of GDP, is

commonly used in the related literature to control for the degree of economic openness A more

open economy, characterized by higher trade to GDP ratios, is expected to be more susceptible to

global economic shocks and have more synchronized financial markets, ceteris paribus In

addition, given that trade openness is a classical measure of economic integration, we would like

to explicitly examine whether the network-based measure of economic integration that we

suggest in the study is superior in explaining financial synchronicity International trade data as a percentage of GDP comes from the World Bank’s World Development Indicators

Exchange rate regime Exchange rate regime is claimed to be an important

macroeconomic characteristic that could influence synchronicity as a part of the financial shock

may be mediated by exchange rates However, tracking exchange rate regime of individual

countries over time is problematic Although de-jure most countries in our sample claim to have

floating exchange rates, in practice in many cases their governments use some sort of peg or

persistently intervene in exchange markets, so de-facto the real exchange rate regime is either

fixed or intermediate, whereas only few countries have truly floating exchange rates To address

this issue, we use the de-facto exchange rate regime database constructed in Levy-Yeyati and

Sturzenegger (2005), that categorizes exchange rate regimes in individual countries on a yearly

basis with a 3-way classification system: fixed, intermediate and flexible We include a fixed

exchange rate regime and an intermediate exchange regime dummy variable in our regression

model to account for de-facto government control of exchange rates

Financial center Possible effects of large financial centers on stock market synchronicity

are accounted for with the financial center dummy variable Some empirical studies (e.g

Reinhart and Kaminsky, 2008) suggest that countries hosting financial centers may be important

in further propagation of a shock The dummy variable takes the value of unity for countries

hosting a major financial center (Japan, Germany and the UK) and zero otherwise

1.2.4 Benchmark model specification

To analyze the effects of RWBC on stock market synchronicity we estimate several

versions of a regression equation of the form:

t t

where subscripts i and t denote country and year, respectively, Synch i,t is the stock market

synchronicity variable (Synch (FREQ) or Synch (R-SQ) developed earlier in section 2.1), RWBC i,t is

random walk betweenness centrality, as defined in section 2.2, Г is the vector of control

variables, that includes, depending on specification, stock market capitalization to GDP ratio,

trade to GDP ratio, financial center dummy variable, fixed exchange rate regime and

intermediate exchange rate regime dummy variables Table 2 reports descriptive statistics for the

explanatory variables

We would like to explicitly track the impact of RWBC Specifically, significant and

positive β would speak in favor of our hypothesis that economic connectedness as measured by

centrality in the trade network matters for financial synchronicity

[Insert Table 2 here]

1.3 Empirical results

Our empirical analysis is based on a panel dataset of 58 countries spanning the period

1990-2000 The choice of countries was made contingent upon data availability for the key

variables - RWBC and stock market synchronicity The complete list of countries assessed in our

study and corresponding stock market indices can be found in Table 7 As our sample covers all

geographic regions and major income level groups, we believe it is a relatively good

representation of the world economy

1.3.1 Baseline results: panel data regression analysis

Several versions of the benchmark equation (4) are estimated with varying combinations

of control variables to ensure robustness of results to inclusion of additional explanatory

variables The panel data regression model is estimated with the random effects model7, which is more appropriate for describing cross-country variation over time Whereas the model does not

suffer from multicollinearity issues8, we use robust standard errors to account for heteroskedasticity that could arise for different reasons, for example periods of financial turmoil

Before expanding on the regression results, we perform a brief exploratory analysis in

which we simply plot stock market synchronicity against RWBC for our sample of countries The

scatter plots for the period 1990–2000 and for the year 2000 with the fitted linear regression line

for Synch (FREQ) are presented in Figure 4 (diagrams for Synch (R-SQ) look virtually identical and

therefore are not reported)

[Insert Figure 4 here]

As can be seen from Figure 4, there is a clear positive relationship between RWBC and

countries naturally splits into the high-RWBC “core” cluster comprised of the UK, Germany,

France, Italy, China and Japan and the low-RWBC “periphery” cluster, comprising the rest of the

sample The observed positive relationship between RWBC and Synch (FREQ) is significantly stronger for the “periphery” cluster than for the “core” economies It should be noted that in the networks literature (Kali and Reyes, 2007) the core-periphery structure of the world trade

network has been also identified Therefore, it is logical to consider not only the connectedness

of individual economies in the world trade network, but also its clustering properties, and interpret our main hypothesis of interest along with the “decoupling” hypothesis For robustness,

in the formal analysis we first treat the sample in the context of a non-clustered network and then consider clustering effects by controlling for the densely connected “core” and the less interconnected “periphery” in Section 3.2

Along these lines, the following econometric analysis provides a more rigorous

assessment of whether the positive effect of RWBC on synchronicity persists after controls for

other economic characteristics are included Estimation results are presented in Table 3 for

the full sample of countries, not controlling for the core-periphery structure of the world

economy) indicate that in both cases RWBC has a positive and statistically significant effect on

stock market synchronicity In particular, for Synch (FREQ) the coefficient of RWBC is greater than

0.80 across specifications and it is significant at the 5% level, while for the Synch (R-SQ) the

coefficient of RWBC is greater than 8.1 and statistically significant at the 1% level

Magnitude-wise, using the estimated coefficient for RWBC in column (1) of Table 3, one standard deviation

of RWBC (which is equal to 0.0721 for the 1990–2000 period) translates into a change in

0.3091).9 In the case of Synch (R-SQ) one standard deviation of RWBC translates to about 25% of a standard deviation of Synch (R-SQ)

[Insert Tables 3 and 4 here]

9

The change in stock market synchronicity is computed by multiplying the estimated coefficient

for RWBC in column 1 of Table 3 by the standard deviation of RWBC (0.8539*0.0721 = 0.0616)

With regard to the control variables, as expected, stock market capitalization to GDP ratio

is positive and statistically significant in all specifications, and therefore constitutes another

major factor explaining stock market synchronicity Its magnitude effects are relevant as well

For instance, for regression results with Synch (FREQ), using the estimated coefficient for stock

market capitalization to GDP ratio in column (2) of Table 3, a one standard deviation change in

stock market capitalization (equal to 0.4943 for the 1990–2000 period) translates into a change in

robust and intuitive, since stock market capitalization to GDP ratio describes the general level of

equity market development of a country Higher stock market capitalization implies higher

market value of stocks comprising the underlying stock market index and leads to higher

synchronicity with the benchmark index Publicly traded equity is the most susceptible asset of a

firm that quickly adjusts for relevant news Therefore, it constitutes a major source of market

value volatility of individual firms which translates to stock market indices, with the magnitudes

of this effect on the entire economy contingent upon the stock market capitalization level

Hence, our results suggest that, while better developed financial markets are more prone

to synchronous dynamics, centrality of a country in the global economic network is another

critical factor that explains financial synchronicity On the contrary, after we control for these

two factors, other explanatory variables - trade to GDP ratio, exchange rate regime and financial

center dummy variable - do not seem to enhance the model much Nevertheless, they still lead to

certain important conclusions Notably, economic integration measured by trade openness enters

insignificantly in most specifications and is only weakly significant for several regressions

involving Synch (R-SQ) as the dependent variable (Table 4) At the same time, RWBC remains

significant with little changes to its magnitude, even after trade openness is included This

implies, besides robustness of RWBC, that economic integration measures that do account for the

network structure of the world economy, are superior to conventional integration measures, and

financial synchronicity can be better understood with the application of network analysis

Contrary to expectations, exchange rate regime, to the extent captured by the de-facto

classification that we use, has little effects on mitigating financial synchronicity The financial

center dummy variable also enters all specifications insignificantly, controlling for RWBC and

stock market capitalization This may suggest that the “financial center” effect that is recurrently

discussed in the literature in the context of financial contagion and general financial

synchronicity may arise largely due to the large size of equity markets in these economies,

namely, Germany, the UK and Japan, and their highly central position in the global economic

network The capitalization effect and the network centrality effect manifest in the same way not only for the countries hosting a financial center, but also for other “core” economies that we identified, including France, China and Italy The core-periphery argument is developed in detail

next

1.3.2 Addressing the core-periphery structure of the world trade network

As discussed before, the international trade network has a clearly defined core-periphery

structure that cannot be ignored in empirical analysis In order to explore this clustering effect we

estimate the regression model with adjustments made for the core economies (the UK, Germany,

France, Italy, China and Japan) We use two viable methods to address this As a first approach,

we estimate the equations with additional controls for the core economies and report results in

columns (5) and (6) of Tables 3 and 4 As can be inferred from the scatter plots and fitted linear

regression lines of Figure 4, the cluster of the core economies differs from the rest of the sample

in terms of both its fixed effect (different constant term) and the slope coefficient of the RWBC

variable Therefore, we include the Core economies dummy variable and an interaction term

RWBC*Core economies in the benchmark specifications to properly address the core-periphery

argument As a second approach, we estimate the regression model with a subsample excluding

the core economies These results are reported in column (7) of Tables 3 and 4

As expected, the results for the two benchmark synchronicity measures Synch (FREQ)

(Table 3) and Synch (R-SQ) (Table 4) are similar and support our hypothesis of interest In

particular, results reported in columns (5) and (6) confirm the significance of RWBC for both the

core and the periphery cluster For both synchronicity measures the slope coefficient of RWBC

remains statistically significant at the 1% probability level However, the marginal effect of

RWBC is considerably different for the core and the periphery clusters: for the core economies

the constant term is larger and the magnitude of the slope coefficient is significantly smaller

relative to the periphery economies The Wald chi-squared test confirms that both the slope

coefficient of RWBC and the constant term are statistically significantly different for the core and

the non-core subsamples10

The economic effect of RWBC on stock market synchronicity in terms of standard

deviations is also notable In the case of Synch (FREQ) , using the estimated coefficient for RWBC in

column (7) of Table 3, a one standard deviation change in RWBC translates to a change in stock

10

We perform a conventional Wald chi-squared test to examine whether the slope coefficient of

RWBC and the constant term are statistically significantly different from each other for the core

and the periphery economies In particular, we use the panel data specifications (5) and (6) in

Tables 3 and 4 to test whether the coefficients of the interaction term RWBC*Core economies and the Core economies dummy variable are jointly and independently statistically significant from zero For instance, for Synch (FREQ) with the full array of control variables (specification (6)

of Table 3) the Wald test yields χ2

statistic of 18.14 with the associated P-value of 0.0001

market synchronicity of 0.1148, or 36% of one standard deviation in synchronicity11 Similarly,

for Synch (R-SQ) , a one standard deviation change in RWBC translates to a change in synchronicity

of 0.7862, or 30% of one standard deviation in Synch (R-SQ)

Importantly, results for both synchronicity measures suggest the core economies are

characterized by different RWBC-stock market synchronicity relationship than the non-core

cluster In particular, the interaction term between RWBC and the Core economies dummy

variable enters negatively and is significant at the 1% level for Synch (FREQ) and at the 5% level

for Synch (R-SQ) In the case of Synch (FREQ) , the Core economies dummy variable is also positive

(the slope coefficient is about 0.42) and significant at the 5% level

An alternative approach that involves regressions with only the subsample excluding the

core economies yields similar results for both synchronicity variables, suggesting that RWBC is a

highly significant determinant of stock market comovements Superior estimation results relative

to the benchmark model, where we do not control for clustering, support the “decoupling” hypothesis and suggest that the structure of the global trade network does have significant

implications for vulnerability of individual stock markets These results remain robust to the

inclusion of additional control variables that we used for robustness checks—GDP, GDP per

capita, dummy variables for high-GDP and low-GDP countries

Hence, summarizing the discussion of the core-periphery structure of the world trade

network and its implications for financial shock propagation, several key features can be

explicitly recognized as a result of our analysis First, along with stock market capitalization, the

position of a country in the global trade network, in our case measured by RWBC, is an important

factor determining its stock market synchronicity, with higher values of RWBC associated with

11

In this case, we use the subsample excluding the core economies to compute the effects of an

increase in RWBC by one standard deviation on stock market synchronicity

higher stock market synchronicity Second, confirming the “decoupling” hypothesis, the world trade network has a clearly identifiable core-periphery structure Based on centrality of a country

in the world trade network, the UK, Germany, France, Italy, China and Japan form a cluster of

high-RWBC countries that remains consistent throughout the period 1990–2000, while the rest

of the sample forms a cluster of low-RWBC economies Third, the core-periphery structure of

the world economy has critical implications for global stock market synchronicity patterns

Specifically, for the non-core economies the slope coefficient of RWBC is substantially higher

than for the core economies across specifications, suggesting their greater susceptibility to global

financial shocks At the same time, the core economies, although highly central in the global

trade network, have uniformly lower levels of stock market synchronicity and lower sensitivity

to RWBC than the non-core economies

1.3.3 Robustness checks

Cross-section estimation for the year 2000 As a robustness check, we replicate the

analysis using a cross-section approach instead of a panel data approach to check whether results

hold for financially tranquil periods The results for the most recent year of our analysis, year

2000, are presented in Table 5 As expected, the effects of RWBC on synchronicity are positive

and statistically significant, while their magnitude differs from the panel-data case Specifically,

for Synch (FREQ), regression results for the full sample controlling for the core economies

(specification (1) of Table 5) suggest that a one standard deviation change in RWBC (0.0656 in

the year 2000) translates to a change in Synch (FREQ) of 0.1893, which corresponds to 70% of a

standard deviation of Synch (FREQ) in 2000 (equals 0.2694) In the case of Synch (R-SQ)

(specification (4) of Table 5), a one standard deviation change in RWBC in 2000 translates to a

change in Synch (R-SQ) of 1.69, which corresponds to 78% of a standard deviation of Synch (R-SQ)

[Insert Table 5 here]

Alternative measure of stock market synchronicity We check whether results hold with

an alternative stock market synchronicity measure, which is also based on daily stock market

index data with proper adjustments for time zone differences, but rather involves correlations

between stock market indices:

t

Corr

Corr Ln Synch

,

, )

( ,

1

1 (5)

where Corr i,t is the correlation coefficient between daily index values (in log-differenced form)

of the Dow Jones Industrial Average and country i’s stock market index in year t The correlation

coefficient is bounded in the [-1,1] interval, hence logistic transformation is applied in this case

also We replicate the same set of estimations as performed with the benchmark synchronicity

measures with Synch (CORR) and report results in Table 6

[Insert Table 6 here]

As the three alternative measures of synchronicity are fairly highly correlated, it is not

surprising that regression analysis utilizing Synch (CORR) yields similar outcomes Results confirm

statistical significance of RWBC (RWBC is significant at the 1% level of significance across all

specifications), supporting the original hypothesis that a country’s position in the global trade network is one of the key factors of stock market comovements, as well as provide additional

evidence in favor of the core-periphery argument Regarding the economic effect of RWBC, in

the case of panel data estimation controlling for the core economies (column (1) of Table 6) a

one standard deviation change in RWBC translates to a change in Synch (CORR) of 0.45, which

corresponds to 98% of a standard deviation of Synch (CORR) For the year 2000 data, a one

standard deviation change in RWBC in 2000 translates to a change in Synch (CORR) of 0.29, which

corresponds to 83% of a standard deviation of Synch (CORR)

Controlling for bilateral trade with the US Finally, it could be argued that the level of

bilateral trade with the US should enter our specification as an additional control variable12 This

is based on the argument that, since the US is the benchmark economy in our synchronicity

analysis, higher levels of bilateral trade should lead to higher stock market synchronicity as

shocks affecting the US economy are likely to affect its major trading partners, and vice versa

The RWBC measure that we develop captures to a certain extent the relevance of bilateral

trade flows with the US, as it is based on the trade volume-weighted network with links with

larger trade flows having a higher likelihood of being chosen as the paths through which shocks

travel Nevertheless, for further robustness we test this possibility by estimating the original

specifications augmented with the level of bilateral trade with the US as a share of GDP variable

The results (not reported here for brevity) suggest that this variable is not significant, whereas

RWBC remains statistically significant with minimal changes to its magnitude

12

The impact of bilateral trade intensity on stock market comovements have been studied, e.g in Tavares (2009), Wälti (2010)

1.4 Conclusion

In this paper we apply a network approach to analyze stock market synchronicity

between nations We find that assessing connectedness of an economy from a network

perspective yields significant insights into understanding financial synchronicity Moreover,

network measures of economic integration appear to be superior to traditional measures, which

ignore the network properties of the world economy, e.g international trade volume as a share of

GDP Specifically, our analysis suggests that random walk betweenness centrality of a country in

the world trade network and stock market capitalization levels are the two significant and robust

factors explaining stock market synchronicity Notably, we also find evidence of nonlinearity in

the relationship between RWBC and financial synchronicity, and identify a group of nations,

namely, the UK, Germany, France, Italy, China, and Japan, forming the highly interconnected

“core” of the global trade network, characterized by uniformly lower synchronicity and its

sensitivity to RWBC than the rest of our sample

References

Abeysinghe, T and K Forbes (2005) Trade Linkages and Output-Multipliers: A Structural VAR

Approach with a Focus on Asia Review of International Economics, 13, 356-375

Albuquerque, R and C Vega, C (2009) Economic News and International Stock Market

Co-movement Review of Finance, 13 (3), 401-465

Ammer, J and J Mei (1996) Measuring International Economic Linkages with Stock Market

Data Journal of Finance, 51, 1743-1763

Arshanapalli, B and J Doukas (1993) International stock market linkages: Evidence from the

pre- and post-October 1987 period Journal of Banking and Finance, Elsevier, Vol.17(1),

193-208

Awokuse T.O., Bessler, D.A and A Chopra (2009) Structural change and international stock

market interdependence: Evidence from Asian emerging markets Economic Modelling, Vol 26,

Issue 3, May 2009, 549-559

Bae, K., Karolyi A.G and R.M Stulz (2003) A New Approach to Measuring Financial

Contagion Review of Financial Studies, 16(2003), 717-63

Beck T., Demirgüç-Kunt A and R Levine (2000) A New Database on Financial Development and Structure World Bank Economic Review 14, 597-605

Beck T and A Demirgüç-Kunt Financial Institutions and Markets Across Countries and over Time: Data and Analysis World Bank Policy Research Working Paper No 4943, May 2009

Bhattacharya, N., Black E., Christensen T., and R Mergenthaler (2007) Who trades on pro

forma earnings information The Accounting Review, 82, 581-619

Boyer, D., Kumagai, T and K Yuan (2003) How Do Crises Spread? Evidence from Accessible

and Inaccessible Stock Indices The Journal of Finance, Vol LXI, N 2, April 2006

Connolly, R and A Wang (2003) International Equity Market Co-movements: Economic

Fundamentals or Contagion Pacific Basin Finance Journal, 11, 23–44

Connolly R and A Wang (1998) Economic News and Stock Market Linkages - The Evidence from the U.S., U.K., and Japan Proceedings of the 2nd Joint Central Bank Research Conference

on Risk Management and Systemic Risk 1 211-240

Fagiolo, G., Reyes, J and S Schiavo (2008) On the Topological Properties of the World Trade

Web: A Weighted Network Analysis Physica A, 387: 3868–3873

Fagiolo, G., Reyes, J and S Schiavo (2009) The World-Trade Web: Topological Properties,

Dynamics, and Evolution Physical Review E, Vol 79, No 3 (2009), 036115

Feenstra, R (1998) Integration of Trade and Disintegration of Production in the Global

Economy Journal of Economic Perspectives, 12(4): 31-50

Feenstra, R.C., Lipsey R.E., Deng, H., Ma, A.C., and H Mo (2005) World Trade Flows:

1962-2000 NBER Working Paper No W11040

Fisher, E and F Vega-Redondo (2006) The Linchpins of a Modern Economy Working paper California Poly

Fisman, R and I Love (2004) Financial Development and Growth in the Short- and Long-Run Working Paper, Columbia University

Forbes, K and R Rigobon (2001) Measuring Contagion: Conceptual and Empirical Issues In

Stijn Claessens and Kristin Forbes (Eds.), International Financial Contagion Boston: Kluwer

Academic Publishers, 43-66

Forbes, K and R Rigobon (2002) No contagion, only interdependence: Measuring stock market

comovements Journal of Finance 57, 2223-2262

Gleditsch, K (2002) Expanded Trade and GDP data Journal of Conflict Resolution, 46: 712–

Kalemli-Ozcan, S., Sorensen B., and O Yosha (2003a) Risk-Sharing and Industrial

Specialization: Regional and International Evidence American Economic Review, 93(1)

Kalemli-Ozcan, S., Sorensen B., and O Yosha (2003b) Economic Integration, Industrial

Specialization and the Asymmetry of Macroeconomic Fluctuations Journal of International

Economics, 55: 107-137

Kali, R and J.A Reyes (2007) The Architecture of Globalization: A Network Approach to

International Economic Integration Journal of International Business Studies, 38 595-620

Kali, R and J.A Reyes (2010) Financial Contagion on the International Trade Network

Economic Inquiry Volume 48, Issue 4, pages 1072–1101, October 2010

Kaminsky, G.L., Reinhart C and C Végh (2003) The Unholy Trinity of Financial Contagion

Journal of Economic Perspectives, Vol 17, Issue 4, Fall 2003, 51-74

Kaminsky, G.L and C.M Reinhart (2000) On Crises, Contagion, and Confusion Journal of

International Economics June, 51:1, pp 145-168.