The International Diversification Puzzle: Home Bias in Countries’ Investment Portfolios [1 ed.] 9783896447166, 9783896737168

Citation preview

The International Diversification Puzzle: Home Bias in Countries’ Investment Portfolios

Studienreihe der Stiftung Kreditwirtschaft an der Universität Hohenheim Herausgeber: Prof. Dr. Hans-Peter Burghof

Band 52

Helena Kleinert

The International Diversification Puzzle Home Bias in Countries’ Investment Portfolios

Verlag Wissenschaft & Praxis

Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.dnb.de abrufbar.

D100 ISBN 978-3-89673-716-8 © Verlag Wissenschaft & Praxis Dr. Brauner GmbH 2016 D-75447 Sternenfels, Nußbaumweg 6 Tel. +49 7045 930093 Fax +49 7045 930094 [email protected] www.verlagwp.de

Alle Rechte vorbehalten Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Druck und Bindung: Esser printSolutions GmbH, Bretten

5

Preface Numerous theoretical and empirical studies document the substantial benefits through international portfolio diversification. In spite of this established state of knowledge, many analyses dealing with the actual behavior of investors find strong evidence for large unutilized benefits from a better diversification of investment portfolios. Thus, with regard to geographical portfolio allocation we find that investors largely prefer domestic assets in their investment portfolios over international positions. This overweighting of domestic assets is commonly known as home bias, and it is recently discussed in the light of the Single Supervisory Mechanism of the European Central Bank. Topical financial research focuses on the structure of the home bias and thus gives better insights into the question which factors determine the decision of international portfolio allocation. Applying this approach on an aggregated country-level analysis provides an improved understanding on country-specific determinates of the home bias. The work of Mrs. Kleinert contributes to this academic literature by combining traditional approaches with cultural dimensions in order to explain the home bias puzzle in international portfolio allocation. In addition, it puts the anomaly in a macroeconomic context and analyzes the preference for the domestic market in international portfolio allocation with regard to consumption risk. Overall, the study presents evidence that the growing fraction of foreign positions in international investment portfolios leads to increased income smoothing and thus improves international consumption risk sharing among economies. This volume contributes to the study series of the Stiftung Kreditwirtschaft pursuing the goal to give interested expert readers insights on the latest academic research in banking and finance. Hohenheim, December 2015 Prof. Dr. Hans-Peter Burghof

6

Table of Contents List of Figures........................................................................................................... 8 List of Tables ............................................................................................................ 9 Abbreviations ......................................................................................................... 10 Country Abbreviations ............................................................................................ 12 Variables and Indices .............................................................................................. 13 Introduction ............................................................................................................ 17 A.

Home Bias in International Investment Portfolios – A Literature Review ....... 21 1. Introduction ................................................................................................................21 2. International Diversification and Home Bias in Portfolio Allocation.........22 2.1 Gains from International Diversification ....................................... 22 2.2 Empirical Evidence of the Home Bias ........................................... 25 2.3 Measuring Home Bias in international portfolio allocation ........... 27 3. Calculation optimal portfolio weights .................................................................29 3.1 Classical Mean-Variance Portfolio Model ..................................... 29 3.2 International Capital Asset Pricing Model ..................................... 30 3.3 Bayesian Portfolio Weights ........................................................... 31 3.4 Gravity Model Approach............................................................... 33 3.5 Discussion of the optimization frameworks................................... 33 4. Potential Explanations for the Home Bias .........................................................35 4.1 Institutional Perspective ................................................................ 35 4.2 Behavioral Perspective .................................................................. 43 4.3 Discussion of the explanation attempts.......................................... 47 5. Conclusion..................................................................................................................48

TABLE OF CONTENTS

7

B.

Cultural Influences on Domestic and Foreign Bias in International Asset Allocation ....................................................................................................... 51 1. Introduction ................................................................................................................51 2. Data and Placement in Literature .........................................................................53 2.1 Data .............................................................................................. 53 2.2 Placement in Literature ................................................................. 54 3. Calculation of the dependent variables ...............................................................56 4. Explanatory Variables and Regression Framework.........................................59 4.1 Explanatory Variables ................................................................... 59 4.2 Regression Framework.................................................................. 67 5. Empirical Results......................................................................................................71 5.1 Results for the Domestic Bias ....................................................... 71 5.2 Results for the Foreign Bias .......................................................... 76 5.3 Robustness Test ............................................................................ 81 6. Conclusion..................................................................................................................96

C.

Is increasing Financial Integration related to improved International Risk Sharing? .......................................................................................................... 97 1. Introduction ................................................................................................................97 2. Data ..............................................................................................................................99 3. Theoretical Background and Placement and Literature ................................101 3.1 International Portfolio Holdings and the Equity Home Bias ........ 101 3.2 International Risk Sharing ........................................................... 104 4. Patterns of Risk Sharing and International Asset Positions.........................110 4.1 Channels of Risk Sharing ............................................................ 110 4.2 The Increase in International Risk Sharing and the growth in Gross International Asset Positions ............................. 113 5. Conclusion................................................................................................................118

Summary .............................................................................................................. 121 References ............................................................................................................ 124

8

List of Figures Figure 1: Average GNI per capita (Atlas Method) between 2001 and 2011 ......... 84 Figure 2: Ratio of net exports of financial services to GDP in % ......................... 87 Figure 3: Evolution of the Home Bias in international equity portfolios 1988 – 2010 ............................................................ 103 Figure 4: International Consumption Risk Sharing from 1988 to 2010 .............. 108 Figure 5: Evolution of the Foreign Equity Holdings to GDP.............................. 116

9

List of Tables Table 1: Overview of selected studies on the empirical evidence of the home bias ................................................................................... 26 Table 2: International portfolio investment allocations ...................................... 55 Table 3: Domestic Bias of Domestic Investors and Average Foreign Bias of Foreign Investors ............................................................................ 58 Table 4: Descriptive Statistics for the explanatory variables from 2001 to 2011................................................................................ 69 Table 5: Explaining the Domestic Bias in countries' equity portfolios ............... 74 Table 6: Explaining the Foreign Bias in countries' equity portfolios .................. 79 Table 7: Regression Results – Additional Variables........................................... 82 Table 8: Explaining the Domestic and Foreign Bias – Income Analysis ............ 85 Table 9: Explaining Domestic and Foreign Bias – Financial Centers ................. 88 Table 10: Explaining Domestic and Foreign Bias – the European Monetary Union.................................................................................. 89 Table 11: Explaining Domestic and Foreign Bias – Excluding the U.K. and the U.S................................................................................. 90 Table 12: Geographical Classification of country sample ................................... 92 Table 13: Explaining Domestic and Foreign Bias– Geographic Analysis............ 93 Table 14: Explaining Domestic and Foreign Bias – Excluding the Financial Crisis ............................................................................. 95 Table 15: Home Bias in international equity portfolios in 1990, 2000, and 2010 ................................................................... 102 Table 16: Consumption Risk Sharing Pooled Regression.................................. 109 Table 17: Channels of Risk Sharing .................................................................. 112 Table 18: Income Channel as a function of the Equity Home Bias.................... 114 Table 19: Income Channel as a function of Foreign Equity to GDP .................. 117

10

Abbreviations ADR

American Depositary Receipt

BPM5

5th Edition of the Balance of Payment by the International Monetary Fund

CAPM

Capital Asset Pricing Model

CPI

Consumer Price Index

CPIS

Coordinated Portfolio Investment Survey

CRRA

Constant Relative Risk Aversion

ECB

European Central Bank

EMU

European Monetary Union

EU

European Union

FSF

Financial Stability Forum

G7

Group of the seven wealthiest countries: Canada, France, Germany, Italy, Japan, U.K., and U.S.

GDP

Gross National Product

GNI

Gross National Income

GNS

Gross National Savings

IBM

International Business Machines Corporation

IFCI

Industrial Finance Corporation of India

IMF

International Monetary Fund

km

Kilometer

M&A

Mergers and acquisition

MNC

Multinational company

OECD

Organization for Economic Co-operation and Development

OFC

Offshore financial centers

PHI

Permanent Income Hypothesis

PPP

Purchasing Power Parity

ABBREVIATIONS

S&P

Standard & Poor’s

SDR

Special drawing rights

SOE

Small but financially open economies

SSM

Single Supervisory Mechanism

w/o

Without

11

12

Country Abbreviations AUS

Australia

JAP

Japan

AUT

Austria

KOR

South Korea

BEL

Belgium

MAL

Malaysia

BRA

Brazil

MEX

Mexico

CAN

Canada

NDL

Netherlands

CHL

Chile

NOR

Norway

DEN

Denmark

PHI

Philippines

FIN

Finland

POL

Poland

FRA

France

POR

Portugal

GER

Germany

SIN

Singapore

GRE

Greece

SOU

South Africa

HGK

Hong-Kong

SPA

Spain

HUN

Hungary

SWE

Sweden

IDA

India

SWI

Switzerland

IDO

Indonesia

THA

Thailand

IRE

Ireland

TUR

Turkey

ISR

Israel

U.K.

United Kingdom

ITA

Italy

U.S.

United States

13

Variables and Indices α

ACT

regression coefficient indicating the intercept

β

risk shared through the income channel

CCONTROL

per capita consumption in country

β

actual portfolio allocation of country

C

unshared consumption risk

β

risk shared through the consumption channel

C

world consumption

CD

capital control

db

culture

CORR CULT d

db

DCREDIT DIST

DLANG Ε

cultural distance between home and target country correlation between the returns in the holder and the target country domestic bias in country . dividends of a risky project in the gravity model approach domestic bias in country ’s investment portfolio domestic credit provided by the banking sector geographic distance in kilometer between the capitals in the holder and the target country. dummy variable for the common language between holder and target country. It takes the value of 1 if they share a common language and 0 otherwise.

$

dollar

EHB

disturbance term

ECD

index of the equity home bias in the portfolio allocation of country economic development

14 EFREEDOM

VARIABLES AND INDICES

EQINDEX

index of the economic freedom

FL

foreign equity holdings of country

GDPCAP

parameter for the relative risk aversion

fb

growth rate of the S&P Equity Index

FAM

total domestic equity stocks held by foreign investors

FE

foreign bias in country ’s investment portfolio

γ

familiarity

GDPG HB h

IDV

gross domestic product per capita real gross domestic product growth rate index of the home bias in the portfolio allocation of country risk-averse immobile agent in country model approach

INC

amount of individualism

κ&

strength of investor protection

INTERNET IPROT κ'

κ( κ λ

in the gravity

income number of internet users per 100 people idiosyncratic output growth including the equity home bias year-by-year amount of income smoothing including the equity home bias impact of the equity home bias on income smoothing average co-movement of the idiosyncratic NFI growth with idiosyncratic GDP growth

μ

share of foreign assets in the wealth portfolio

η&

market capitalization of country

MAS

MCap

expected returns country’s masculinity idiosyncratic output growth including foreign equity holdings to GDP

VARIABLES AND INDICES

η'

η( N

NFI

OPT

year-by-year amount of income smoothing including foreign equity holdings to GDP impact of the foreign equity holdings to GDP on income smoothing number of assets available net factor income from abroad, i.e. the net claims on flows of foreign output

OTHERVARIABLES other variables PD r0 r1

r2

r3 r4

optimal portfolio allocation of country country’s power distance real return of the domestic market portfolio risk free rate rate of return of foreign assets owned by foreign investors rate of return of foreign assets owned by domestic investors

R(

return of the world market portfolio

σ

country’s rating score

adj. R(

RATING ∑

S2

coefficient of determination adjusted coefficient of determination degree of mistrust in the international CAPM; Variable used in the Bayesian Approach for calculating of optimal portfolio weights variance-covariance matrix

S3

stock of foreign assets owned by foreign investors

SMD

market capitalization of listed companies

S

SIZE

stock of foreign assets owned by domestic investors exogenously determined states of nature in the gravity model approach stock market development

15

16

VARIABLES AND INDICES

T

total number of projects in the world in the gravity model approach

θ

ϑ

elasticity of substitution

ω

total value of stocks traded to GDP

UAI

country specific effects in the risk sharing analysis

VALUE ω∗

portfolio weight

x

risky project of agents ℎA in the gravity model approach

x

ϱ

Y

?

country’s uncertainty avoidance

optimal portfolio weight

risky project of agent ℎA

units of a freely traded good in the gravity model approach output

INTRODUCTION

17

Introduction Increasing opportunities for investors to diversify their portfolios internationally is one of the effects of the globalization in financial markets (Chan et al. 2005, p. 1495). Despite a wide range of theoretical and empirical approaches that documented substantial benefits through international portfolio diversification (among others Grubel, 1968 and Solnik, 1974a), many studies show that investors do not take advantage of these potential gains. On the contrary, they show that investors largely prefer domestic assets in their investment portfolios over international positions. This overweighting is commonly known as home bias and represents one of the major puzzles in financial economics (Lewis, 1999). „Several decades of international financial liberalization have shown that simply declaring the global capital market an open field does not suffice to achieve full integration.” (Flandreau, 2006, p. 634). Economists from the nineteenth-century were already familiar with this phenomenon and called it “the disinclination of capital to migrate” (Flandreau, 2006, p. 634). Putting this home bias puzzle into historical context Bayoumi (1990) shows that correlations between domestic savings and investments across countries were lower in the late nineteenth century than nowadays, indicating less integration today. But why does home bias in international portfolio allocation still persist? What drives this puzzle? Recently, home bias is discussed in light of the Single Supervisory Mechanism (SSM) of the European Central Bank (ECB) by the Deutsche Bundesbank. “The direct link between public finances and bank balance sheets prompted calls for a banking union in the euro-area and posed the danger of a ‘home bias’ among national supervisors” (Deutsche Bundesbank, Monthly Report 07/2013, p. 14). The report states that the cross-border supervisions of banking groups should be effective and transparent. Furthermore, it should be able to identify risks to the financial system already in an early stage and counter national preferences of supervisors, i.e. to be more lenient with banks just because they are from the supervisors’ home countries. In order to achieve further integration among national banking sectors and to complete the monetary union in the European Union (EU), the European Commission initiated the project of a banking union, which is not faced to the home bias phenomena.

18

INTRODUCTION

New approaches in finance literature focused on the structure of the home bias and thus gave better insights, which factors determine the decision of international portfolio allocation. They added significant value to research on the anomaly. Putting this approach for equity holdings on an aggregated country-level and being therefore able to analyze determinants of the home bias with country-specific variables, is one of the major contributions of my thesis.1 Previous studies on a country-level predominantly employ only traditional approaches to explain the overweighting of national positions and/or the over- and underweighting of foreign markets in national portfolios. But the question, whether an investor puts less of his wealth in a foreign country, because it is culturally very distant to his home country, is largely neglected by academic literature on an aggregated country-level. Furthermore, this study contributes to academic literature by combining traditional approaches with cultural dimensions in order to explain the home bias puzzle in international portfolio allocation. This allows me to combine economic and stock market variables with dimensions, which capture the individualism of an economy and the cultural distance between home and target country. Going a step further and analyzing the discussed preferences for the domestic market in international portfolio allocation in the context of consumption risk, raises an interesting question: Does increasing financial integration lead to improved international risk sharing, i.e. do better diversified investment portfolios smooth countries’ consumption flows? My thesis tries to shed light on this question, composed by both finance and macroeconomic aspects. Thus, the study is able to discuss the importance of diversification of portfolio investments from another perspective. My work is composed of three main parts. Part one gives an overview of selected studies and presents the stand of literature. In a first step, I introduce analyses which discuss benefits from international portfolio diversification and give insight to research that document the home bias puzzle as an anomaly that is neither limited to special countries or country groups nor to a special type of investor (individual and institutional). In order to calculate the index of the home bias for a country, optimal portfolio weights are needed. Therefore, I introduce, evaluate, and discuss alternative approaches to calculate benchmark weights. Searching for an explanation for the home bias puzzle in international portfolio allocation economic and finance literature offers very different approaches. They reach from the institutional to the behavioral perspective. The institutional stream of explanations includes transaction costs, hedging of domestic risks, information asymmetries between national and foreign investors, as well as corporate governance aspects. Behavioral approaches discuss relative 1

As my thesis only focuses on equity holdings and does not consider bond or other asset classes, the results presented and discussed in part two and three are always referred to equity holdings.

INTRODUCTION

19

optimism with regard to the development of national positions, overconfidence, and familiarity as source of the overweighting of domestic positions in international portfolio allocation. The empirical analyses are conducted in part two and three of my thesis. Part two employs a geographical breakdown of cross-border portfolio investments and examines institutional and behavioral approaches to explain home bias. Therefore, I analyze the portfolio holdings of 26 holder countries in the time period between 2001 and 2011. This macroeconomic view allows me to examine the impact of the variables on the home bias by abstracting from special types of investors. As previous literature is predominantly focused on institutional or individual investors, this work enlarges research in this branch. Furthermore, this data set enables me to investigate on the structure of the home bias by distinguishing between a domestic and a foreign component. This idea goes back to Chan et al. (2005). The domestic component of the home bias (domestic bias) captures the extent to which a country overweights the domestic market. The foreign component of the anomaly (foreign bias) provides an answer to the question, how a foreign market is over- or underweighted in a national investment portfolio compared to the benchmark weight. The results of my sample countries indicate in a significant domestic overweighting and substantial underweighting of the respective target countries. To explain these phenomena I add cultural dimensions to the traditional explanatory variables. To my knowledge, there is no other study that combines those two – traditional and cultural – aspects by abstracting from special types of investors. In addition, this study applies these approaches to the domestic and foreign component of the home bias puzzle. I offer a large body of tests, by building numerous sub-samples to give evidence to the robustness of the results and add weight on the conclusions. In summary, this part of my thesis documents a substantial amount of both domestic and foreign bias in international portfolio allocation. I examine cross-country differences in investment behavior by analyzing economic and stock market development factors, capital control, investor protection as well as information availability and cultural aspects. As barriers towards international investment increase deadweight costs for foreign investors, I expect that if these costs are higher for foreign investors than for domestic investors, then foreign investors will hold less of that country’s equity than theoretically predicted. Furthermore, including cultural aspects in the empirical analysis may enable me to detect their impact on the international portfolio allocations around the world. Thus, this indicates that the cultural difference between home and target country seem to have impact on the investment decision. Hence, if a target country has a very different culture form that of the home country and the home country does not feel familiar with this target country, then investors will invest less,

20

INTRODUCTION

ceteris paribus, foreign bias is more negative. The empirical analysis confirms these expected relationships. Part three puts the home bias puzzle in a macroeconomic context. Although the concept of full risk hedging is different in macroeconomics and finance, it is one of the central topics in both streams of literature. While financial literature typically builds on the international Capital Asset Pricing Model (CAPM) as reference model, macroeconomists depart from the benchmark model of perfect markets. The international CAPM from finance literature predicts under standard assumptions that countries hold identical international portfolios of risky assets. On the other hand, in a setting of endowment economies under standard assumptions, perfect markets imply that consumption growth rates are equalized, i.e. risk is perfectly shared across countries. The empirical analysis includes 21 countries from the Organization for Economic Co-operation and Development (OECD) in the time period between 1988 and 2010. The concept of sharing consumption risk across economies (risk sharing) goes mainly back to Obstfeld and Rogoff (1996). Asdrubali et al. (1996) build on their work and identify two ways, the so called risk sharing channels, through which countries can share their consumption risk. I employ this approach and examine the impact of these channels in my country sample. Overall, the results show that the improved risk sharing is mainly due to income smoothing. This brings me to the question on the driving force behind this growing importance of the income smoothing. The analysis presents evidence that the growing fraction of foreign positions in international investment portfolios leads to income smoothing and thus improves consumption risk sharing. The conducted robustness tests add weight on these findings. My thesis concludes with a summary of my results and an outlook for future research.

21

A. Home Bias in International Investment Portfolios – A Literature Review 1.

Introduction

In the first part of this study I present recent contributions to finance literature on portfolio allocation with a special focus on home bias. Chapter 2 focuses on gains of international diversification and on the empirical evidence of the home bias puzzle. In detail, chapter 2.1 summarizes gains from international diversification considering two approaches, the mean-variance and the consumption-based approach. Section 2.2 documents empirical evidence of the home bias by grouping studies with different focuses such as different markets or different types of investors. The last category of empirical studies presented in this section analyzes the diversification opportunities within a country, the so called local bias. Subsequently, section 2.3 shows how the home bias is calculated. Any meaningful way to explain the home bias puzzle requires in a first step a correct description and characterization of the benchmark weights to which actual portfolio allocation is compared. Hence, I discuss different methodologies in chapter 3. Starting with the classical mean-variance approach in section 3.1, I briefly introduce the well-known international version of the Capital Asset Pricing Model (CAPM). While the latter approach is called the model-based approach, section 3.3 introduces two data-based approaches. The Bayesian mean-variance approach introduced by Pástor (2000) allows investors to have some degree of mistrust in the asset pricing model. Building on his model Garlappi et al. (2007) take a step further and introduce estimation risk in the standard mean-variance model. In section 3.4 the gravity model mainly building on Martin and Rey (2004) and the extension of Faruqee et al. (2007) are summarized. The chapter concludes with a discussion of the advantages and short-comings of the different methodologies. Chapter 4introduces and structures the different explanation attempts for the existence of the home bias puzzle in international portfolio allocation. I broadly group the explanations into two categories, the institutional and the behavioral perspective. Section 4.1 introduces traditional explanation approaches for the puzzle, such as transaction costs, hedging motives corporate governance, and information asymmetries between domestic and foreign investors. The behavioral perspective is discussed in section 4.2. These explanation attempts reach from relative optimism for the domestic market, to overconfidence aspects, herding and familiarity approaches. Section 4.3

22

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

compares the institutional and the behavioral perspectives and comes to the result that by far, none of the sources discussed in literature taken alone is able to capture the substantial amount of home bias still documented in international portfolio allocation. Section 5 summarizes the insights and concludes.

2.

International Diversification and Home Bias in Portfolio Allocation

2.1

Gains from International Diversification

Gains under a mean-variance approach During the last decades, capital markets in developed countries exhibited an increasing integration. Indeed, growing dependences of national economic cycles, the harmonization of national accounting standards, and the increasing co-operations of international stock exchanges are results of this development. In the same way it leads to an augmented correlation of international capital markets. However, there is still the possibility to realize a sustained rise of a portfolio’s risk adjusted return by international diversification. In a mean-variance approach, home bias can result from differences in expected returns between domestic and foreign investors and/or from the risks of the assets in the investment set. A large body of literature in finance studies the effect of international diversification in asset portfolios and considers models where investors adjust their expected returns of foreign assets downwards to eliminate market frictions such as transaction costs, information asymmetries and controls on international capital flows. The difference between the costs of investing in foreign assets and the expected returns of these assets may, ceteris paribus, result in the home bias in asset allocation.2 Markowitz’s (1952, 1959) pioneering work on portfolio optimization was subsequently extended to the international context by Grubel (1968), Levy and Sarnat (1970), Lessard (1973), and Grauer and Hakansson (1987). For instance, Grubel (1968) proves gains of international diversification both empirically, for the eleven major stock markets in the world, and theoretically. With his model he shows international capital flow to be a function of interest rate differentials and growth rates in total asset holdings in two countries. Hence, when interest rate differentials are ze2

Sercu and Vanpée (2007, p. 9-13) discuss in detail various mean-variance based models and evaluate each of them in turn.

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

23

ro or negative, capital should flow across countries and should not, in case they are positive (Grubel, 1968, p. 1299). Under the assumption that international capital markets are efficient like national ones, risk can be reduced by investing in markets with low correlation. Furthermore, expected returns can be increased simply by adding investments of emerging markets into the portfolio, which induces an increase of the risk-adjusted return and moves up the efficient frontier (Errunza 1983, p. 53-55). When the market portfolio is used as benchmark for evaluating the gains of portfolio diversification, then the benefits are still substantial. Tesar and Werner (1995, p. 475479) analyze the excess return on a portfolio of foreign securities compared to a portfolio of primarily domestic securities. Their results are consistent with other academic findings as they document substantial benefits in international diversification in the analyzed countries (Canada, Japan, U.K., and the U.S.). Errunza et al. (1999) show empirically that the costs of incomplete international diversification calculated according to the mean-variance method are overestimated. They analyze “home-made” diversification for the U.S. market and show that U.S. investors can mimic the foreign market index returns and allocate their portfolios in a way that they become international mean-variance efficient with portfolios of domestically traded multinationals, closed-end country funds and American Depository Receipts (Errunza et al., 1999, p. 2086-2088). However, they document that although the average gains from foreign-made diversification are insignificant, it should be considered that in economically instable market periods, international diversification provides a meaningful diversification opportunity that cannot be replicated in the home market (Sercu and Vanpée, 2007, p. 13). Furthermore, it should also be considered that Errunza et al. (1999) draw this conclusion from the viewpoint of an U.S. investor, so that it might be considered that investors from other countries may probably have less diversification opportunities within their home country to compose their international mean-variance efficient portfolio. Gains from local geographic diversification, that means a diversification within a country, have only been studied incipiently in academic literature (García-Herrero and Vazquez, 2007, p. 3.). Acharya et al. (2006) use data of Italian banks from 1993 to 1999 and explore how the returns of banks vary with the banks’ diversification level at different levels of bank risks. Furthermore, they analyze how increasing diversification in banks portfolios due to lending into new sectors impacts their risk. For instance, including new industries or countries in the portfolio may lead to decreasing effectiveness of that bank’s monitoring activities and may increase bank risk. They conclude that diversification across sectors and industries does not necessarily improve bank returns and could even be costly for high risk banks (Acharya et

24

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

al., 2006, p. 1382).3 Furthermore broader geographical presence of banks within the U.S. does not lead to higher returns or are characterized by lower risk. In detail, they show that geographic diversification leads to higher lending capacity of banks and consequently the whole banking system, but they fail to prove profits and loan performance of individual banks (Morgan and Samolyk, 2003, p. 15). As these findings suggest a strong co-movement of economic determinants within a country, it is feasible to expect that the benefits from geographical diversification within a country may be limited. Gains under a consumption-based approach In order to capture foregone gains of international diversification, the consumptionbased approach builds on the given production process, takes it as exogenously given, and determines the impact of optimal risk-sharing in investor’s consumption (Cole and Obstfeld, 1991, Tesar, 1993, and van Wincoop, 1994). The permanent increase in consumption typically indicates a welfare increase, while the implicit risk-free rate, the risk-adjusted growth rate, the rate of relative risk aversion, and the endowment uncertainty determine differences in the results documented in literature. Lewis (1996, p. 15) concludes: “Thus, the welfare gains depend both upon the utility under autarchy relative to the optimal portfolio, as well as the ratio of the value of domestic equity […] to world equity […].” Lewis (2000) employs data from 1969 to 1993 and calculates gains from international diversification from the U.S. perspective and compares the results from both the mean-variance approach and the consumption-based approach. She calculates the efficient frontier and then the gains from moving from the utility of a portfolio only composed by U.S. assets to the utility of the portfolio at the efficient frontier (Lewis, 2000, p. 8-12). Employing the mean-variance approach and choosing the parameter of relative risk aversion at a level of 2, she shows that the gains from international diversification are in between 10% (for an elasticity of substitution D = 5) and 28% (for an elasticity of substitution D = 2) of current wealth (Lewis, 2000, p. 11). Furthermore, the findings indicate that increasing levels of relative risk aversion lead to higher welfare gains. In a next step she compares these results with the gains from diversification resulting from the consumption-based approach. Again assuming a relative risk aversion of 2 the gains of diversification for U.S. investors resulting from moving from autarky to an open economy range from 0.04% to 0.25% of total current consumption. Lewis (2000) concludes that the difference in the estimated costs of under-diversification resulting from the mean-variance approach and the 3

Nonetheless, one could argue that banks may have a particular balance and portfolio structure and thus their findings may not be representative for other industries.

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

25

consumption-based approach come from the high variability of stock returns compared to low variability in consumption data. Thus, gains from international diversification may result from the benefits of reducing the variability of marginal utility over time. In the mean-variance methodology, marginal utility depends on stock returns, while it is estimated from consumption in the consumption-based approach. However, Sercu and Vanpée (2007, p. 16) state that the mean-variance approach of Lewis (2000) does not consider that stock returns are implicitly assumed to derive consumption. Overall, both the mean-variance and the consumption based approach document significant gains from international diversification. Thus, allocating international portfolio ignoring these benefits would reduce welfare gains. 2.2

Empirical Evidence of the Home Bias

An overview of selected studies examining home bias is given in Table 1. They are grouped according to their respective focus, as they largely differ from one study to another. The last group contains a special version of the home bias, called the local bias. Evidence in different markets Despite the well-documented benefits of international portfolio diversification presented in the previous section, there is still a strong preference for domestic assets in international markets. For eight countries (France, Italy, Germany, Japan, Spain, Sweden, U.K., and U.S.) Cooper and Kaplanis (1994, p. 46) compare the respective fraction on the world market portfolio to the fraction of domestic investments and show that the percentage of domestic investments is 43% to 99% higher than it should be. As a consequence of this capital market anomaly, securities and thus economies become more sensitive on local determinants. Although the U.S. equity market in 1989 comprises less than 48% of the global equity market, French and Poterba (1991) allocate for American equity traders nearly 94% of their funds to domestic securities. In Japan 98% and in the U.K. 82% of the stock holding is domestic (French and Poterba 1991, p. 222). They document that there is too little cross-border diversification in the international portfolios of the respective domestic investors. Hiraki et al. (2003) provide additional evidence for the home bias on the Japanese market. Evidence for different types of investors Goetzmann and Kumar (2008) examine the portfolio choices of U.S. retail investors for a sample period between 1991 and 1996. They show that investors hold underdiversified portfolios, whereat the amount of under-diversification is more pronounced

26

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

in retirement accounts. Their study investigates on determinants of retail investors on portfolio allocation and comes to the result that younger investors with a low education level, who are less sophisticated and have low income, exhibit higher levels of underdiversification. It is comprehensible to argue that individual investors may typically be less sophisticated than professional portfolio managers. This disadvantage would make portfolios of individual investors exhibit a stronger preference for domestic positions (higher extent of home bias in portfolio allocation). Lütje and Menkhoff (2007) follow this stream of argumentation and analyze the portfolio allocation of highly sophisticated German fund managers. However, in their examined sample there persists a substantial overweighting of domestic positions. Thus, these results show inefficient portfolio allocations. Furthermore, 62% of the interviewed German fund managers state that they prefer investments in geographically near markets. Country

Type of Investor

S tudy

S ample Period

Evidence in different markets France, Italy, Germany, Japan, Spain, Sweden, U.K., U.S.

institutional investors

Cooper and Kaplanis (1994)

1978 – 1987

Japan,U.K, U.S.

institutional investors

French and Poterba (1991)

1989

Japan

institutional investors

Hiraki et al. (2003)

1985 – 1998

Evidence for different types of investors U.S.

individual investors

Goetzmann and Kumar (2008)

1991 – 1996

Canada, France, Germany, Italy, Japan, Spain, U.K., U.S.

institutional investors (banks)

Lütje and M enkhoff (2007)

2003

Germany

institutional investors (fund manager)

García-Herrero and Vazequez (2007)

1995-2004

Finland

individual and institutional investors

Grinblatt and Keloharju (2001)

1995

Diversification within a country (local bias) U.S.

institutional investors

Coval and M oskowitz (1999)

1995

U.S.

individual investors

Ivković and Weisbenner (2005)

1991 – 1996

U.S.

individual investors

Seasholes et al. (2010)

1991-1996

Germany

institutional investors

Buch et al. (2005)

1995-1999

Table 1: Overview of selected studies on the empirical evidence of the home bias

27

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

García-Herrero and Vazequez (2007) use bank-level data for the 38 largest international banks incorporated in the G-7 (Canada, France, Germany, Italy, Japan, U.K., and U.S.) and Spain. Using the mean-variance portfolio model as normative benchmark they find that recent allocation of international bank portfolios show a substantial home bias. They suggest that gains from international diversification could also be considered in the second pillar of Basel II as in the first pillar only idiosyncratic risk of recipient countries are considered (Garcia-Herrero and Vazequez, 2007, p. 18). Their findings are qualitatively consistent with Buch et al. (2005), who also use the mean-variance portfolio model as benchmark. They use aggregated data in cross-border claims of banks between 1995 and 1999. Diversification within one country It has also been shown as a robust phenomenon across different types of investors and markets that in addition to the discussed home bias, geographically close positions within one country are overweighted. This phenomenon is called local bias and was first discussed in Coval and Moskowitz (1999, p. 2048). They hypothesize that the major fraction of the deviation from optimal portfolio weights in international portfolio allocation can be attributed to the local over-investments. Grinblatt and Keloharju (2001) employ a Finnish data set including both institutional and individual investors and document a substantial preference for stocks of companies with nearby headquarters. Analogous to their results, Ivković and Weisbenner (2005) investigate on the portfolio allocation of individual investors between 1991 and 1996 and document that one out of six U.S. individual investors owns only stocks of firms which are headquartered within 250 miles from their households. In a vein with their analysis, Seasholes et al. (2010) provide evidence for the local bias in international portfolio allocation on the U.S. equity market for the same sample period. In summary, this overview gives insights to the empirical evidence of the home bias puzzle. The selected studies document that the preference for domestic assets is limited neither to a special market nor to a type of investors and is thus an overall phenomena. Furthermore, it is also shown that there even exists a preference for assets from companies who are geographically close to the investor. 2.3

Measuring Home Bias in international portfolio allocation

Following large part of the literature the extent of home bias in country is measured as the relative difference between actual (EFGA ) and the optimal (HIGA ) weights of foreign assets in portfolio allocation: JKA = 1 − O

PQRS

TURS

V.

(1)

28

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

The home bias (JKA ) of country or later called the equity home bias WJKA is one minus the ratio of the actual and the optimal share of foreign equity holdings in the total equity holdings. It becomes zero if the portion of domestic investment equals the portion predicted by theory – in other words, according to the international CAPM the weight of domestic investments should be equal to the share of that country’s equity market in the total world equity market. In this case, where investors act according to theory, the diversification measure becomes zero. The other way around home bias will get to unity if all investments are allocated in the domestic market. While I discuss alternative methodologies to calculate optimal portfolio weights, actual portfolio holdings are determined by the share of foreign equity in the total equity portfolio of country . More specifically, EFGA is the ratio of its foreign equity holdings XWA and the total equity holdings (i.e. foreign and domestic). Domestic equity holdings are results of country ’s market capitalization (YFZ[A ) and the total domestic equity stocks held by foreign investors (X\A ): EFGA =

XWA . XWA + YFZ[A −X\A

(2)

International diversification does not need to be limited to corporate finance, as investments can be diversified through different asset classes, e.g. foreign direct investment, bank deposits, real estate, etc. As shown by Sørensen et al. (2007), home bias in bond markets is highly correlated with equity markets. They conclude that a separate analysis in their study with regard to this asset class does not shed more light in the respective puzzle. Following their findings this study only focuses on equity holdings and analyses the evolution of their allocation over time and leave other asset classes for future research. That is why in the following always the equity home bias (WJK) is intended. As any meaningful explanation of the home investments bias requires a correct description of the benchmark weights, I present different approaches for calculating optimal portfolio weights and discuss potential determinants for the home bias in a second step.

29

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

3.

Calculation optimal portfolio weights

3.1

Classical Mean-Variance Portfolio Model

The classical mean-variance framework is the common starting point for calculating the theoretically optimal portfolio weights. The model dating back to Markowitz (1952) and Sharpe (1954) builds on the assumption that investors are risk-averse and want to maximize the expected portfolio return while minimizing its variance. This leads to: c max `a b − `a e ` , _ 2

(3)

where ` is the g -vector of portfolio weights allocated to g assets. As there are domestic and foreign equity holdings g= 2 and b is the g-vector of expected returns, the scalar c is the relative risk aversion parameter and ∑ is the g × g variancecovariance matrix. Given the budget constraint it is assumed that `a i = 1, so that the solution of the portfolio problem becomes: 1 m' `∗ = e jb − kil c

(4)

where k is the expected return on the zero-beta portfolio, which corresponds to the optimal portfolio and i is the vector of ones.

Even though the mean-variance model has an eminent position in academic literature and the basic statement gained unlimited acceptance, practical application is still low. This is mainly due to the sensitivity to input assumptions and to the fact that expected returns are unobserved and thus actual returns may be used to calculate optimal weights in the portfolios. However, Merton (1980) states that realized returns do a bad job proxying expected returns. Since optimal portfolio weights are highly sensitive to changes in expected returns, imprecisely estimated expected returns have significant impact on the estimated portfolio allocation. Jeske (2001, p. 34) shows that this high sensitivity arises from the fact that returns in industrialized countries are highly positively correlated so that the covariance matrix is close to singular. Thus, this data-based approach often brings out extreme and volatile equity positions, which often cannot be realized due to legal or institutional reasons. Already marginal variations of the input assumptions can cause substantial fluctuations in portfolio

30

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

weights, which are not economically intuitive and can therefore hardly be understood by investors. 3.2

International Capital Asset Pricing Model

An alternative to the discussed data-based approach is an asset pricing model, such as the international CAPM of Solnik (1974b), which holds in a perfectly integrated world. The model assumes that all investors are of mean-variance type, hold the market portfolio and have homogeneous beliefs about the distributions of real asset returns. Furthermore, all investors face the same investment opportunities and there are neither transaction costs nor taxes. In this model-based approach optimal investment weights of a country in an international setting are given by the relative shares of domestic and foreign equities in the world market capitalization. The international CAPM leads to the well-known linear beta relationship between risk premium on the domestic portfolio and the expected excess return on the world market portfolio as benchmark. Wjno l − np = qo_ rWjn_ l − np ]

(5)

where no is the real return of the domestic market portfolio, np is the risk free rate,

qo_ ≡

uvwjxy ,xzl w{xjx| l

is the world beta of the domestic market and n_ is the return of the

world market portfolio. Equation (6) is the empirical counterpart of equation (5): no − np = } + qo_ ~n_ − np • + €

(6)

where } is the intercept and € the disturbance term. Predictions of the international CAPM are considered valid if estimates of the intercept }• are zero.

However, total market capitalization of stocks also contains not freely tradable assets, so that regular investors cannot hold the world market portfolio. La Porta et al. (2000), therefore, suggest employing the world float portfolio, which is the market capitalization-weighted portfolio of freely floated shares. Dahlquist et al. (2003) uses this float-adjusted world market portfolio and conclude that home bias can be reduced by this modification, but does not disappear. The practicability of the international CAPM is restricted by factors like e.g. transaction costs, legal regulation, taxes, currency risks or psychological effects, which hinder the development of an integrated capital market (Fischer and Keber 1997, p. 357, Solnik 2000, p. 138). Whereas all these determinants restrict the efficiency of capital

31

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

markets, they do not seem to have a significant role, as gains from international diversification should be much higher than the additional costs (Solnik 2000, p. 141). 3.3

Bayesian Portfolio Weights

Bayesian Mean-Variance Portfolio Weights To deal with estimation errors, literature adopted the Bayesian approach, where the unknown parameters are treated as random variables. A Bayesian decision-maker starts from a pre-specified prior over the parameters and combines it with observations from the data in order to construct a predictive distribution of returns. He can use an intercept different than zero – even if insignificant – to put into question the optimality of the portfolio prediction made by the international CAPM. Pástor (2000) states that the assumptions of the international CAPM are too strict and that investors would not accept the entire model. He develops the framework and constructs his model in a way that in case of mistrust in the international CAPM, data becomes informative and is involved in the decision of portfolio allocation. Values of standard errors of the intercept ‚ƒ stand for the degree of trust (i.e. the belief that the intercept }• is zero). If the values are low it indicates that there is a strong belief in the validity of the theoretical model and that the results in optimal portfolio weights will be close to that “model-based” approach. On the other side, high values indicate that the results will be closer to the “data-based” approach, so that in case of ‚ƒ → ∞, that means full mistrust in the international CAPM, it coincides with the optimal weights of the mean-variance framework. A prior (non-data) belief in a zero intercept and no mispricing is the starting point of the Bayesian approach. Depending on the chosen degree of mistrust in the model, this belief will then be updated by using returns data. The sample mispricing will in consequence shift accordingly towards the prior mean of } to obtain the posterior mean of } (Baele et al. 2007, p. 610). As now moments of predictive distribution are used to compute portfolio weights, the combination of data with model prediction leads to different estimates for the mean and variance covariance matrix of returns. The optimal weights are calculated as follows: ∑∗m' b†∗ ` = i′ ∑∗m' b†∗ ∗

(7)

32

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

b†∗ is the predictive mean and ∑* the predictive variance that replace the sample moments of the distribution of returns.4 Depending on the empirical performance of the model based on specific country data, a certain degree of mistrust in the theoretical international CAPM may result in optimal weights that are closer to the observed allocation in the country and thus imply a lower home bias than the deviation from the market capitalization share. Bayesian Multi-Prior Portfolio Weights However, the Bayesian decision-maker is assumed to have only a single prior or to be neutral to uncertainty (in the sense of Knight, 1921). It departs from the prediction of the international CAPM in proportion with the investors’ degree of mistrust in the model. Hence, larger mistrust in the international CAPM augments the relevance of historical return data in estimating optimal allocations, which become more volatile on the other hand. Garlappi et al. (2007, p. 42) argue that this approach brings out some short-coming such as difficulties in estimating moments of asset returns, the sensitivities of the portfolio weights to the choice of a particular prior and the experimental evidence that decision-maker are not ambiguity-neutral. Given those effects, it is important to characterize investors as decision-makers with multi priors who are averse to ambiguity and hence prefer robust portfolio rules.5 By extending the discussed meanvariance framework in this way Garlappi et al. (2007, p. 47-49) accomplish the problem of volatile data. Their multi-prior framework is defined as:

subject to

c max min `a b − `a e ` Š _ 2 ‹ Ob, b̂ , e

4

V≤Ž

(8)

(9)

Entering the utility function of the next period maximizing investor, the predictive density of returns is defined as follows:

[jn••' |Φl = “• [jn••' |D, Φl[jD|Φl”D , where [jn••' |Φl stands for the probability density function of excess returns conditional on Φ (the sample data). D is the set of parameters of the statistical model that describes the stochastic behavior of asset returns. For further explanation see Beale et al. (2007), p. 610-611. 5 Heath and Tversky (1991) show that ambiguity aversion is particularly strong when investors perceive their competence to be poor in assessing the relevant probabilities. See among others Chen and Epstein (2002) and Uppal and Wang (2003) for decision-making models that allow for multiple priors, where the decision maker is not neutral to ambiguity.

33

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

`a i = 1

(10)

where b̂ is the sample mean of asset returns. The solution of the max-min problem in the discussed Multi-Prior approach is a set of optimal weights with considerably smoother behavior compared to the pure data-based approach. 3.4

Gravity Model Approach

One of the short-comings of the basic international CAPM presented in section 3.2 is that it fails to capture the effect of high transaction costs. For the sake of completeness, I summarize the main aspects of the gravity model. This approach has long been one of the most successful empirical models in economics. The traditional framework goes back to Newton’s Law of Gravitation. More recently, Martin and Rey (2004) were the first to offer a gravity type model to examine financial markets in an international context. Their two-country model was extended to an g-country model by Faruqee et al. (2004).

In each country there are –A risk-averse immobile agents, which are identical in every country. In the first period, each agent ℎA ∈ ˜1, … , –A š in country is endowed with a risky project x›S and œ units of a freely traded good. They are free to choose whether to consume the good or to invest in the risky project by buying the shares. In the second period the agents are faced to • exogenously determined and equally likely states of nature. The risky project is an Arrow-Debreu security which pays dividends of žAŸ ”›S in state , ∈ ˜1, … , •š, where žAŸ = 1 if = and žAŸ = 0 if ≠ . This assumption captures the situation where the risky projects and assets are imperfectly correlated so that assets would become imperfect substitutes and thus diversification effects would improve safety. These dividends ”›S are the only source of consumption in the second period. In the first period, agents raise capital by selling shares of their projects and buying shares of other projects on the stock market. This implies that investing in one of these projects is equal to buying an Arrow-Debreu asset that yields a dividend payoff in only one state of nature. However, they assume that the total number of projects in the world G, is always smaller than the number of the total states of the world •. Thus, markets are incomplete and agents cannot eliminate all the risk by holding a portfolio of all traded assets (the market portfolio). 3.5

Discussion of the optimization frameworks

Baele et al. (2007, p. 617) calculate optimal portfolio weights and the amount of equity home bias in international portfolio allocation employing the discussed optimiza-

34

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

tion approaches. Their results build on models without short sales constraints but show that results are qualitatively similar when short sales are allowed. With regard to the international CAPM they confirm that investors overinvest in domestic assets, so that the equity home bias index ranges between 0.55 and 0.58 in Belgium and Austria to 0.98 in Greece, Hungary and 0.99 in Poland, and Turkey. Across their entire set of 25 countries they obtain an average amount of home bias of 0.80 in international equity portfolios. Employing the pure data-based approach reduces domestic overweighting substantially. Supposing that investors fully mistrust the international CAPM, the average home bias over time is 0.42, which corresponds to a drop of nearly 50%. As discussed by Merton (1980) this approach brings out extreme and volatile investment positions and deducing from the standard errors on the home bias measure, their results confirm these expectations. They obtain negative home bias measures for Belgium (-0.03), the Netherlands (-0.09), and the U.K. (-0.06), while the highest values are documented for the Czech Republic (0.88) and Japan (0.94). Across the country sample they document a standard error of 0.49. Employing the approach proposed by Pástor (2000), which allows a reasonable degree (‚ƒ( =0.05) of mistrust in the model-based approach, reduces home bias measures compared to the international CAPM to 0.51 with a standard error of 0.48. The partial reliance of this method on sample data leads to occasionally unstable optimal weights and, consequently, to varying home bias measures. A large part of this variance disappears with the Multi-Prior approach of Garlappi et al. (2007). They calculate an index value of 0.84, as the expected returns are restricted to a certain interval and investors minimize over the choice of expected returns. Using this approach they obtain higher values than with the pure data-based and the Pástor approach, but still lower than with the international CAPM. Overall, Baele et al. (2007, p. 619) document a decrease in the international CAPMhome bias for many countries and show that in some countries this decrease is more pronounced for a reasonable degree of mistrust in the international CAPM. Nonetheless, deviations from the model-based approach also worsen the extend of the puzzle when the domestic index underperforms the world benchmark (which in their sample is the case for Japan). However, in some emerging countries the amount of domestic overweighting in portfolio allocation is extreme and largely unaffected by the chosen measure. They conclude that the phenomenon of the home bias in international portfolio allocation remains substantial for the large part of the countries, even for a reasonable degree of mistrust in the model-based approach.

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

35

Faruqee et al. (2004) is one of the most recent empirical studies that examines the differences in optimal portfolio weights by comparing the international CAPM and the gravity model approach, which incorporates transaction costs and information asymmetries. They state that optimal country weights predicted by the gravity model are higher than with the international CAPM. This implicates higher values for the home bias in international portfolio holdings. None of the introduced approaches had convincing and sustainable success for large country samples. While the Bayesian approach in both versions lead to lower values of the home bias the gravity model brings out higher values than predicted by the international CAPM. However, the aim of my thesis is not to discuss different optimization frameworks or even extend them. Thus, I choose the traditional approach and employ the international CAPM as benchmark model for my empirical analyses in part two and three. Hence, if the theory holds, then investors invest in the international market portfolio where the weights of each country are the same as the countries’ share of total market capitalization.

4.

Potential Explanations for the Home Bias

4.1

Institutional Perspective

4.1.1 Overview In this chapter some of the explanations why the actual portion of domestic assets in investors’ portfolios is smaller than predicted by standard portfolio theory discussed in literature are presented. The explanation attempts for the home bias puzzle can roughly be divided into institutional and behavioral factors. Institutional factors have their source in the violation of the main assumptions of perfect markets and have a direct effect on portfolio allocation, while behavioral factors stem from the violation of the second key assumption, according to which investors are rational, e.g. they make portfolio choices by maximizing their expected utility. These factors influence asset allocation indirectly.6 This categorization is not free of vulnerability, as there are explanation attempts that may be categorized in both groups. One of these cases could be information asymmetry. This point is discussed in more detail later.

6

A very detailed overview of the explanations for home investment bias discussed in literature is made among others by Sercu and Vanpée (2007).

36

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

4.1.2 Transaction Costs Transaction costs are often mentioned as a reason for the preference of investors towards domestic positions. They include all costs associated with the acquisition or the holding of stocks such as fees, commissions or withholding taxes. Black (1974) and Stulz (1981) both study the effect of these direct barriers on portfolio choice, which make it costly to hold non domestic securities and thus lowers the profitability of diversification and induce an overweighting of domestic equity. Black (1974) constructs a model of capital market equilibrium with explicit barriers to international investment. To model them he uses proportional taxes that investors have to pay on their net holdings of risky foreign assets. These barriers to international investment hinder net investment in risky foreign assets. As investors in Stulz (1981) pay a tax proportional to the absolute value of their holdings of risky foreign assets, they get very different results. This model makes it difficult for investors to hold risky foreign assets or to hold them short (Stulz 1981, p. 923). Martin and Rey (2004) employ a two-country model with incomplete asset markets where the demand for non-domestic assets has a non-linear relationship with transaction costs such as banking commissions, transaction costs on the exchange rate, and other variable fees. They conclude that theoretically a significant amount of home bias in equity markets can be deduced from only tiny transaction costs. Although the conclusions deduced from the theoretical model of Martin and Rey (2004) are very appealing, transaction costs empirically are no reasonable determinants for the preference of domestic positions. If it would be the case that trading foreign assets is more expensive than domestic assets, then it would be expected that the amount of transactions in foreign assets to be lower than in domestic assets. However, Tesar and Werner (1995) follow this approach and show that the turnover rate on foreign equity is much higher than on domestic equity. The more recent work of Warnock (2002) suggests that their estimates of turnover rates on foreign equity were inaccurate and shows that they are comparable to those of domestic equities. He confirms the main finding of Tesar and Werner (1995) that transaction costs fail to explain the extent of the observed equity home bias. 4.1.3 Hedging of domestic risk A potential explanation for the overweighting of domestic assets in international portfolio allocation is that domestic positions hedge for home-country specific risks such as inflation risk, real exchange rate risk, and the risk from non-tradable goods, e.g. human capital and non-financial income.

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

37

Inflation Hedging Deviations from the purchasing-power parity (PPP) can cause an increased demand for assets that hedge inflation in the home country. With regard to the equity home bias puzzle, domestic equity can provide a hedge opportunity that does relatively a better job than foreign equities. Thus, it would be reasonable to assume that the possibility of inflation hedge can encourage investors to hold a bigger share of domestic positions. Although hedging inflation seems to be a plausible source of overweighting domestic equities, little evidence has been found in investment portfolios that confirms this theory. If this hedging motive was a rational reason for the investment bias towards domestic positions, then inflation has to be correlated significantly with equity returns. Cooper and Kaplanis (1994) test whether equity home bias is caused by investors trying to hedge inflation risk. They develop a model that combines deviations from the PPP with dead-weight costs for holding non-domestic positions. With their model the authors estimate deadweight costs, which are necessary for explaining the empirically proven bias towards domestic equities for different levels of investor risk aversion. Comparing their results with observable deadweight costs, they demonstrate that only by assuming very low levels of risk aversion their model is able to capture the observed home bias in equity.7 However, they do not find evidence that this factor alone is able to justify the missed benefits from internationally diversified investments, unless investors are significantly less risk averse than empirically measured. Their empirical analysis confirms this coherence only if investors are characterized with a very high level of risk tolerance and if equity returns are negatively correlated with domestic inflation. Hedging of consumption risk Foreign securities may also serve as better hedge against domestic consumption risk than domestic assets. Chue (2007) employs the Euler equation to capture the extent to which foreign securities hedge against domestic consumption. He finds that even though domestic stock market risks can be reduced by foreign equities, their impact on hedging domestic consumption risk is much poorer and concludes that he is not able to provide evidence that hedging domestic consumption risk can explain the documented amount of equity home bias in international portfolio allocation. Hedging of exchange rate risk If the PPP does not hold, it does not only create inflation risk, but also real exchange rate risk. Fidora et al. (2007) examine the impact of this factor on the equity home 7

Uppal (1993) develops a model where investor’s consumption is limited to their domestic capital stock and where transferring capital across countries is linked with proportional costs. He finds that in this case, a conventional level of risk aversion implies a preference for non-domestic assets.

38

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

bias. Their study not only focuses on stock holdings, but provides an interesting insight to the home bias in both equities and bonds. Their model-based approach implies that real exchange rate volatility increases the amount of home bias for assets with a relative low local currency return volatility. Their main finding is that currency risk is of secondary importance, as they get significant effects only for bonds. Hence, currency risk seems to have impact on the equity home bias, but is clearly not able to explain the pronounced dimension of this puzzle. Hedging of non-tradable wealth Another hedging motive may also be that domestic assets better protect from the risk of non-tradable wealth such as human capital or non-financial income. If investors put their investments on domestic assets in order to hedge this type of risk, nonfinancial and financial income is expected to be positive and associated with a decrease of the investment in risk assets. Baxter and Jermann (1997) detect a positive correlation between the retuns on human capital and domestic equities, which implies that reducing domestic holdings would hedge investors against human capital risks. In contrast, Bottazzi et al. (1996) document a negative correlation between those two factors, suggesting a possible explanatory power of hedging risks of human capital for the home bias puzzle. Massa and Simoniv (2006) find no support for this relationship and conclude that hedging the risk of non-financial income is not able to explain the home bias in equity markets. Overall, none of the presented studies can provide convincing results considering hedging motives as determinant for the home bias in international equity markets. In addition to explicit costs, investors face implicit costs, which arise from informational disadvantages to local investors. Both costs lower expected net returns on foreign investment. 4.1.4 Information asymmetry One type of indirect barriers is information asymmetry between domestic and foreign assets, which is one of the few explanations of the anomaly with empirical success. Many empirical studies employ regression analysis to examine the linkage between information asymmetries and international portfolio holdings by either regress actual portfolio holdings or differences between actual and optimal allocation according to theoretical predictions directly against proxies for asymmetric information such as regional and cultural factors. As there is a variety of literature, I will only discuss the main findings of some selected studies, while this overview is non-exhaustive. There is an information asymmetry between two countries when domestic investors have superior information about the domestic market compared to foreign investors.

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

39

Hence, domestic investors’ information costs on the domestic market will be lower than for the foreign markets. If there really exists an asymmetry in information, then obtaining information about companies or markets abroad will be more costly than information production of domestic companies or markets. Consequently, profitability of international diversification will be reduced. Indeed, if investors have different information whether they are domestic or foreign, risk-averse investors will put their wealth in stocks on which they have easily available information like domestic stocks, as they evaluate them as less risky. Gehrig (1993) develops a model, where even in equilibrium, investors have incomplete information. He concludes that a bias toward domestic assets arises automatically when investors have on average better information about domestic stocks. Hence, informational segregation seems to be an important contributing factor for explaining the equity home bias. Zhou (1998) employs a multi-asset inter-temporal general equilibrium model with differential information to explain home bias in equity markets. Ahearne et al. (2000) develop a model of international portfolio allocation based on Cooper and Kaplanis (1986) and show that country-specific proportional investment costs, which represent both explicit and implicit costs, go along with foreign investments.8 Brennan and Cao (1997) take the assumption that domestic investors are better informed as a given fact, and examine empirically the implications of this assumption for equity flows. Brennan et al. (2005) take a model-based approach where investors receive both public and private signals, while private signals are less precise for foreign investors and confirm the results of the earlier empirical work in Brennan and Cao (1997). Furthermore, they investigate on the linkage between informational disadvantages and the expectations of investors about a market (degree of bullishness). Van Niewerburgh and Veldkamp (2009) try to give theoretical insight to the question why increasing global information access does not reduce the extend of the home bias in international portfolio holdings. They develop a model where tiny information advantage is sufficient to induce significant home bias in case the amount of information investors can learn about, is strictly limited. Familiarity Sharing no common language or having different religions are determinants that could impact international portfolio holdings, as in this case investments would always be linked to translation and adaption to different languages or cultural habits and therefore lead to information costs (Grinblatt and Keloharju 2001). 8

A very interesting analysis is made by Suh (2005), who analyses the equity home bias by examining international portfolio recommendations by institutional investors. He shows that the advised portfolios exhibit home bias with regard to the level of portfolio holdings and to the frequency of portfolio adjustments. Specifically relative to benchmark portfolios, recommendations are heavily biased towards domestic markets.

40

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

Economic proximity Economic proximity captured by the Gross Domestic Product (GDP) or the market capitalization of listed companies could also have impact on investment decisions, as information on large economies and stock markets is more easily available and more widespread. Faruquee et al. (2004) investigate on this size bias and find confirming evidence on a large country sample. Sarkissian and Schill (2004) come to the same finding by analyzing the impact of overseas listing decisions. Geographic proximity A dominant factor to capture the impact of information asymmetries in portfolio allocation is the geographic proximity of the foreign market. Coval and Moskowitz (1999) examine the impact of distance on portfolio choice within the countries. They test local information advantage on mutual funds in a domestic setting and find out that funds investing in local businesses earn an abnormal return on investments located close to them. Especially for small funds with a concentrated portfolio allocation these abnormal returns are especially high. Furthermore according to their results the average fund shows a modest home bias, while some funds heavily overweighting local positions have higher excess/abnormal returns than the average fund. These results are in line with Coval and Moskowitz (2001). Grinblatt and Keloharju (2001) analyze portfolio allocation in the Finnish market and confirm distance and common language as important determinants for investment decisions. Also mergers and acquisitions (M&A) decisions show a strong local bias (Grote and Umber, 2006), while the fact that those investors seem to realize higher returns than geographically distant investors, confirms the intuition of information asymmetries between investors. Portes and Rey (2005) follow this intuition and conclude that the geography of information is a key determinant of the pattern of cross-border transactions, while they document only weak support for the diversification motive. However, being geographically close to a company apparently leads to better access to investment relevant information about that company. Performance Based-Approach If domestic investors have advantageous information about investments compared to foreign investors, they should realize higher returns on their investments than foreigners. Hau (2001) follows this performance-based explanation and analyzes German trading data and concludes that local investors outperform foreign investors as they base their investment decisions on better information than foreign investors. This finding is confirmed by empirical data from Korea and Indonesia showing that foreign investors trade at worse prices than domestic investors (Choe et al. 2005 and Dvořàk, 2005). On the other side, Grinblatt and Keloharju (2000) and Huang and

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

41

Shiu (2009) argue that due to better resources and a superior access to expertise foreign institutions (often professionally managed funds or investment banking houses) investors would be able to perform better than domestic investors. Analyzing the 16 largest Finnish stocks, Grinblatt and Keloharju (2000) show that over two years foreign investors outperform investments of domestic investors. Huang and Shiu (2009) confirm this finding for the Taiwanese market. Furthermore, Seasholes (2000) shows for the same market that while foreign investors buy ahead of good earnings announcements domestic investors do the opposite.9 Malloy (2005) documents for the U.S. that analysts located close to a company give more accurate forecasts. Bae et al. (2008) analyzes the relation between the precision of earnings forecasts made by foreign and local analysts in 32 countries. They call an analyst local when he is located in the same country as the respective company. As main measure of accuracy for an analyst they use his price-scaled absolute forecast error and subtract the average price-scaled absolute forecast error of all analysts who made earnings forecast. If a local analyst has better information then he should be able to predict earnings with more precision. They confirm the intuition in the case of univariate comparisons and when they control for different determinants of forecast accuracy it is explained with the fact that local investors have superior access to information as they are in direct contact to the companies’ representatives and are so able to better observe the company. They compare the performances of analysts belonging to local companies to those of foreign companies and conclude that they fail to show any significant difference in earnings forecasts. Hence, they find evidence for lower forecast errors for local stocks. However, employing information asymmetries as explanation for the home bias in international asset allocation brings some difficulties. If domestic investors have better or more investment relevant information or observe a signal about a specific asset, one would not only expect them to be faced to a lower return variance, but also their expected returns should be higher than those of foreign investors. Hence, if the information the domestic investor gets, indicates low expected returns, then informed investors should hold fewer domestic stocks than foreigners do. As this investment behavior differs from empirical results on portfolio allocation indicating continued home bias, Jeske (2001) argues information asymmetries fail to explain home bias in international asset allocation.

9

Froot et al. (2001) are in line with these studies, which see foreign investors as better informed and greater sophisticated. However, Shukla and van Inwegen (1995) confirm U.S. money managers’ better performance than U.K. money managers when picking U.S. stocks. Though, an outperformance of foreign investors cannot be documented for the Japanese market (see Kang and Stulz, 1997).

42

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

Another problem with information asymmetries as determinant for the overweighting of domestic assets in international portfolio allocation is that international diversification benefits are obtained through other asset classes so that the potential disadvantage of foreign investors could largely be avoided. Errunza et al. (1999) and Brealey et al. (1999) empirically test combinations of domestic traded multinational companies (MNC), American Depository Receipts (ADR) and closed-end country funds. Investing in a domestic company or stock will be accounted as an investment in favor of the equity home bias. In the case of MNC’s this is not necessarily true, as for many firms the business achieved on the domestic market represents just a fraction of the total business. Hence, it could be argued that international diversification and investments in domestic MNC are not necessarily conflicting actions. Jacquillat and Solnik (1978) show that investing in MNCs does not provide much portfolio diversification. Their empirical data shows that only 2% of the variance in U.S. multinational equity returns is due to the foreign markets they operate in. Rowland and Tesar (1998) and Cai and Warnock (2006) give this aspect more recent evidence, as these studies reflect the expansion of multinationals activity during the past 20 years. Rowland and Tesar (1998) find little evidence that investments in MNC’s rise international portfolio diversification. 4.1.5 Corporate Governance Corporate governance as well as political risk on the country level and transparency on the company-level are also discussed as potential drivers for the substantial home bias in international portfolio holdings. As there is an evident link between transparency and information asymmetries Pagano et al. (2001) and Ahearne et al. (2000, 2004) focus on the role of diverging standards in accounting, disclosure requirements, and regulatory frameworks across countries. When investors purchase equities from another country, they base their investment decision on information from published accounts. The problem arises when accounting principles and disclosure requirements differ greatly from the investor’s home country. Furthermore the credibility of this information is heavily determined by the regulatory framework, which can also vary across countries. In a nutshell, Ahearne et al. (2004) argue that companies who do not reduce information costs, e.g. information asymmetry to U.S. investors by getting listed on an U.S. exchange, will be underweighted in U.S. investment portfolios. Consequently, countries whose companies tend not to list on U.S. exchanges will heavily be underweighted in U.S. portfolios. This leads to an equity home bias at an aggregated level. Foreign investors are faced to much lower information costs if they are intended to invest in companies that follow adequate accounting and governance standards and

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

43

do protect minority shareholders. La Porta et al. (2000) document in their empirical analysis on the U.S. market that international dispersion of company ownership is significantly higher in countries where minority shareholders are well protected. Dahlquist et al. (2003, p. 3) show that differences in corporate governance across countries can give a contribution for solving to the home bias puzzle through the impact of the share ownership. In the same line, Gelos and Wei (2005) document a strong relationship between both government and corporate transparency and international investments in the respective country. Stulz (2005) introduces a twin-agency problem in order to cause overweighting of domestic stock in international portfolio allocation: The first agency problem stems from the agency problem of corporate-insider discretion (inside investors extract private benefits and expropriate in that way foreign investors). As state rulers have the authority to expropriate investors by introducing or increasing taxes or modify regulation framework to increase the state rulers’ own benefit, leads to the second agency problem. Stulz (2005) concludes from his empirical and theoretical analysis that the concentration of share ownership is higher in countries that protect investors poorly and the risk of state expropriation is high. As foreign investors typically do not own private investment relevant information on the domestic market or a domestic company, they therefore avoid companies with poor corporate governance in order to minimize their expropriation risk. Thus, the quality of a company’s governance should have impact on both the fraction of stocks held by foreign investors and the probability of new investors investing in that company (Giannetti and Simonov, 2006). 4.2

Behavioral Perspective

4.2.1 Overview As home bias still persists in portfolio allocation, institutional explanations do not seem to be sufficient to explain the capital market anomaly. All potential sources for home bias in international portfolio allocation described so far, build on traditional assumptions that model the behavior of individuals as perfectly rational. Psychologists and experimental economists have shown that people in experimental settings, for example have self-control problems and suffer from a wishful-thinking bias, so that they often base their decisions on the degree to which data resembles the model, rather than making appropriate calculations about the respective probability. However, since the seminal work of Kahneman and Tversky (1979) who developed the prospect theory, the theory of behavioral finance has become an established re-

44

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

search area. To explain international under-diversification in portfolio allocations, researchers have developed different behavioral approaches. The problem with behavioral-based approaches in order to explain home bias in international portfolio allocation is that the psychological effects are hard to measure and mainly difficult to distinguish. For instance, some of the explanations discussed in the following chapters overlap so that some lines of argumentation could also be transferred to other categories. Hence, the selected structure just tries to give a rough categorization of the existing literature, but is not concluding. 4.2.2 Relative Optimism Relative Optimism towards the domestic markets as reason for the home bias puzzle was first mentioned by French and Poterba (1991). They define relative optimism as a situation, where investors have more optimistic expectations about their home market than investors from abroad. In contrast, absolute optimism emerges when investors are more optimistic about their domestic market than about foreign stock markets. Mathematically, denoting by j”, ‹l, where ” stands for the investor sentiment in domicile for equities in market ‹ Relative Optimism

Absolute Optimism:

j”, ” l > j‹, ” l for all ‹ ≠ ”, j”, ” l > j”, ‹l for all ‹ ≠ ”.

(11)

(12)

French and Poterba (1991) made first empirical evidence and suggest that investors may be relatively more optimistic about their domestic markets and analyze the behavior of U.S., Japanese, and British investors. In their empirical analysis they calculate the set of investors’ expected returns which are necessary to explain existing patterns of their portfolio allocation.10 They show that for U.S. equities, U.S. investors’ expected annual return must be 110 and 240 basis points higher than those expected by British and Japanese investors. Shiller et al. (1996) analyze quarterly forecasts for one-year returns between 1989 and 1992 on the Nikkei 225 and the Dow Jones Industrial Average of Japanese and U.S. investors. In line with the results of French and Poterba (1991) they show that in every country, investors are more optimistic about the development of their domestic market. Similar results occur in the study of Strong and Xu (2003) by analyzing fund managers’ behavior in the same country sample as French and Poterba (1991). Despite that

10

Given the estimates of the return covariance matrix.

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

45

according to them the relative optimism is not able to explain the whole extent of equity home country bias, it is important to examine its role as a factor of the puzzle. 4.2.3 Overconfidence and investor’s competence Not just actual information advantage of domestic investors relative to foreign investors is discussed as potential explanation, but even just the perception of having more investment relevant information relative to investors from abroad can induce an overinvestment in domestic stocks. Hence, if all investors have the same information, so that objectively none of the groups has an actual information advantage, an overconfident investor perceives an information advantage for the investments he is familiar with. Barber and Odean (2001, 2002) investigate this issue and characterize overconfident investors as individuals who misjudge their competence in forecasting assets performance they are familiar with and overinvest in these assets. Kilka and Weber (2000) conduct a cross-country experimental study in the U.S. and Germany in order to test the hypothesis that overinvestments in the domestic market can be explained by the expectations of stock returns. They argue that domestic investors feel on average more competent in forecasting stock prices of their home markets, than from abroad (Kilka and Weber, 2000, p. 178). This perception can arise from true information asymmetries or from psychological biases. Generally, they expect this feeling to induce tighter, less dispersed probability distributions and see this presumption to correspond to the intuition that a rich set of detailed information enables the investor to make judgments more precisely.11 Their results confirm their expectations as the means of judged competence of U.S. participants is significantly higher for U.S. stocks than for German stocks, and vice versa. However, Dorn and Huberman (2005) conduct a study based on a questionnaire and do not find any convincing support for the notion of overconfidence to explain underdiversification of international portfolios. They conclude that it is mainly due to investor’s risk attitude that portfolios are underdiversified. Morse and Shive (2006) proxy overconfidence with measures of patriotism and document in their sample of 53 countries a high correlation between these measures and investors’ investment decisions. Other studies state that specific investor characteristics determine the composition of investment portfolios. For instance, Goetzmann and Kumar (2008) show in their analyses that more sophisticated (more experienced) investors tend to hold better diversified portfolios than less sophisticated investors. They find that the level of diversification increases with age and income as well as 11

Among others Yates et al. (1991) analyze the accuracy of probability judgments of experts and show that the probability distribution of experts, which feel more competent in a particular field is less dispersed than the probability judgments of investors who feel unfamiliar with this issue.

46

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

with wealth and education (Goetzmann and Kumar, 2008, p. 435). Graham et al. (2009) see the investor’s competence as reason for the under-diversification in the investment portfolio. They argue that an investor who classifies himself as competent will show less uncertainty about his subjective distribution of expected stock returns (Graham et al., 2009, p. 1096). 4.2.4 Herding Reisen and Williamson (1994, p. 17) document the tendency of portfolio managers to behave according to industry norms, which is due to a principal agency problem. Fund managers mimic the behavior of others to limit their personal risk. As a consequence in the case of a failure he will not be accounted personally. Keynes (1936, p. 157) defines “herding in financial markets is a situation where individuals act the same way without private information, without a planned direction just because the others do so.” In the context of the home bias herding behavior occurs when fund managers are remunerated with respect to other managers or if they incorporate possible reputation losses by separating from the benchmark into their decisions. Thus, herding can provoke the equity home bias only when it is the norm to overweight domestic positions in investments. If instead the norm is the opposite then herding would be an argument against the home bias. Hong et al. (2004) show that mutual fund managers tend to purchase a stock if other managers in the same city hold the same and find a strong link between participation in the stock market and social interaction. A market is considered more attractive by a social investor when more of his peers are also in the market. 4.2.5 Familiarity Huberman (2001) and Grinblatt and Keloharju (2001) investigate why investors often ignore the principles of portfolio theory and purchase positions they are familiar with. For this purpose, Huberman (2001, p. 667) analyzes the structure of the shareholders of the Regional Bell Operating Companies and shows that investors tend to invest in the company they are served from rather than other Regional Bell Operating Company’s stocks. Grinblatt and Keloharju (2001) suggest that the preference of investors for the domestic market in their Finnish sample could arise from information advantage due to language and cultural closeness or alternatively simply by patriotism towards domestic companies. They show that investors whose native language is Finnish, prefer holding and trading stocks of companies that publish their annual reports in Finnish rather than companies that have Swedish as company language and publish their reports in Swedish, and vice versa (Grinblatt and Keloharju, 2001,

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

47

p. 1069). In a line with their results Benartzi (2001) analyzes the familiarity on company stock and documents a positive correlation with the perceived safety of it. In his analysis he finds that employees of Coca-Cola allocate nearly three-quarter of their portfolio to their employer’s shares. The result underlines the additional risk of this investment with regard to the company’s possible bankruptcy: in this case the investor is not only exposed to the risk of losing retirement savings but also the loose his or her job (Benartzi, 2001, p. 1756). However, Massa and Simonov (2006) see familiarity-driven investment decisions as rational response to limited informational availability rather than to psychological effects. Furthermore, in their Swedish dataset they show that familiarity-based investment decisions allow investors to realize higher returns than if they hedged (Massa and Simonov, 2006, p. 667). 4.3

Discussion of the explanation attempts

A variety of explanation attempts war presented in the previous sections. Since nearly all direct limitations to foreign investments such as taxes or transaction costs decreased substantially or were even eliminated, they may fail to explain the anomaly or have only marginal impact explaining actual portfolio choices of investors. Thus, the most convincing explanation attempts for the home bias puzzle are information asymmetries and the behavioral approaches. Nowadays, researchers still discuss whether rational decision making (information asymmetry) or bounded rationality (behavioral finance) are the right approach to follow. For instance, Ke et al. (2010) investigate whether it is asymmetric information distribution or the fact of investing in the familiar that impact investment decisions of institutional investors. Their findings are not able to support the notion of the information-based explanation and therefore tend to conclude that the preference for domestic assets in portfolio allocation is driven by familiarity issues. In contrast, Massa and Simonov (2006) see familiarity-driven investment decisions simply as a rational answer to information constraints and not as a behavioral heuristic. In a line with their results, DeMarzo et al. (2004) show that the impact of familiarity on investment decisions depends on the degree the investor is informed, so that more informed investors are less affected by familiarity. They state that the availability of information to the investor drives investment decisions and that familiarity is a substitute for better information. These studies show how hard it is to distinguish clearly between the perspectives. It can easily be seen, that neither the intuitional nor the behavioral approaches have so far succeeded in solving the home bias puzzle. Overall, one single explanation for the still persisting home bias puzzle in international portfolio allocation may not exist, but all approaches discussed in the previous sections seem to have in some way

48

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

impact on the phenomenon. It can reasonably be assumed that not one approach taken alone but a mixture of institutional and behavioral aspects as well as cross-country differences have impact on international investment decisions and lead to the documented overweighting of domestic positions.

5.

Conclusion

This first chapter of my thesis gives insight to selected theoretical frameworks and empirical findings for the home bias in international portfolio allocation. Since this domestic overweighting leads to foregone welfare gains, I introduce in a first step two approaches in order to capture the extent of these gains from international diversification. Therefore, I summarize the findings of Lewis (2000) who compares the results of the mean-variance based and the consumption based approach to calculate gains from international diversification. Subsequently, selected studies are presented which focus on providing empirical evidence for the home bias in different markets or in the portfolios of different types of investors. Even though, intuitively one may expect home bias to be more pronounced in individual portfolio allocation, the presented empirical give proof to the contrary. The studies document that also institutional investors largely overweight domestic positions in their portfolio allocation. Overall, the selected studies furnish evidence that the home bias puzzle is neither a phenomena that is limited to some markets or to a special type of investor, but is an anomaly that is documented portfolio allocation independent of country or degree of sophistication. As it could be argued that home bias documented with substantial amount in international portfolio allocation, I summarize the main approaches for measuring optimal portfolio weights in order to calculate the anomaly. The comparison of the predicted weights from the different approaches indicates that the amount of home bias shifts significantly. Given a reasonable degree of mistrust in the international CAPM as stated by Baele et al. (2007, p. 619), the home bias in international portfolio allocation still remains substantial. While the Bayesian approaches measure less amounts of home bias, the gravity model brings out higher values than the international CAPM. As the focus of my thesis is not to examine optimal portfolio allocation or to identify shortcoming of the respective models, I follow the large part of literature and employ the model-based approach. Hence, in part two and three of this study the optimal portfolio weight of a country will always be determined by the share of that country on the world market capitalization.

HOME BIAS IN INTERNATIONAL INVESTMENT PORTFOLIOS – A LITERATURE REVIEW

49

Several explanations in academic literature are discussed in order to explain the home bias puzzle in international portfolio allocation. They reach from institutional approaches such as transaction costs, hedging motives, and corporate governance aspects to the most popular perspective, which capture different aspects of information asymmetries between domestic and foreign investors to behavioral aspects. The latter stream of literature discusses relative optimism, overconfidence, and familiarity as explanation attempts. Neither studies supporting the institutional nor the behavioral perspective have jet furnished convincing evidence for capturing the extent of the domestic overweighting with regard to investors’ investment decisions. Thus, the home bias puzzle in international portfolio allocation remains one of the major puzzles in finance literature. Hence, further research is essential to capture the driving force behind this anomaly. If researchers succeed and identify factors that determine home bias, implications for regulatory standards might be deducted. Furthermore, better insights to the anomaly may help lower forgone gains from international diversification and give answers to questions from branches such as corporate listing decisions.

51

B. Cultural Influences on Domestic and Foreign Bias in International Asset Allocation 1.

Introduction

The second part of my thesis analyses the cross-border investment behavior from 26 holder countries to 38 target countries, including both developed and developing countries. Therefore, the Coordinated Portfolio Investment Survey (CPIS) dataset from the International Monetary Fund (IMF) is used, which is available on yearly frequency from 2001 on. This rich panel dataset over-comes some shortcomings of previous studies in this field of research. Early papers such as Tesar and Werner (1995) use accumulated capital flows as proxy for countries portfolio holdings. Portes and Rey (2005) employ flow data to estimate this variable. However, Warnock and Cleaver (2003) show that both datasets may often fail to calculate accurate portfolio holdings on country level. This study extends the existing literature focused on investments on fund-level to a country-level analysis. This macroeconomic view allows to better explore determinants for the phenomenon and abstract from special types of investors. Second, I investigate the structure of the equity home bias by distinguishing domestic and foreign components (Chan et al., 2005). The domestic bias reflects the amount to which a country over-weights its home market while the foreign bias captures the over- or underinvestment in the respective foreign market relative to its market capitalization. Let me follow Chan et al. (2005) and illustrate this concept with the help of a simple example: Imagine there are only three countries and the world’s market capitalization is composed equally by these economies (each country holds one third). In aggregate, country 1’s local investors put 50% of their net wealth in their home market, 25% in country 2 and 25% in country 3. Hence, country 1’s investors now are characterized as domestic or home biased, as they overweight their home country, but not as foreign biased towards country 2 or 3 as they weight them according to their fraction on the world’s market capitalization. Country 2’s investors allocate one third of their net wealth in the domestic market, 23% in country 1 and 43.7% in country 3. These investors exhibit no domestic bias but only a foreign bias, underweighting country 1 and overweighting country 3. Finally, country 3 invests half of its net wealth domestically, 38% in country 1, and 12% in country 2 so that they exhibit a domestic bias. At the same time they overweight country 1 (less foreign bias) and underweight country 2 (more foreign bias).

52

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

International equilibrium settings imply that domestic and foreign biases arise from barriers to cross-border investments. These barriers exhibit deadweight costs for investors who want to in-vest in domestic equities (and the same the other way around for local investors). If deadweight costs for a foreign investor 1 for entering the local market 2 are higher than for the local investor 2 investing in his home market, then investor 1 will consequently hold less equities of country 2. Consequently, a higher amount of foreign bias (foreign investors put less on the domestic market) comes along with higher domestic bias (domestic investors put more on the domestic market). Distinguishing between those two types of biases allows me to analyze how they differ across countries and to detect how cross-border investment barriers impact both domestic and foreign bias. Third, the main contribution of this analysis is that it combines traditional aspects on portfolio holdings with cultural dimensions on a country-level analysis. As far as I know there is no study with this objective. The wide set of variables as potential sources of domestic and foreign bias is classified into six categories: economic development, stock market development, capital control, investor protection, familiarity and information availability, and culture. As the impact of these variables is expected to be in some cases symmetric and in others asymmetric, I will investigate on the impact of each category on the weights of a market’s equities in the portfolios of both domestic and foreign investors. For all analyzed 26 home countries between 2001 and 2011 the empirical findings document a strong bias towards the respective domestic market, while the extent differs across countries. The highest percentages of domestic investment are measured in India (99.75%) and Russia (99.61%). The regression analysis shows that the explanatory variables have in some cases asymmetric effects on domestic and foreign bias. With regard to the domestic bias it is shown that familiarity and cultural dimensions as well as economic and stock market development play a statistically significant role in order to explain the home bias puzzle. Turning to the foreign bias analysis I am able to document significant impact of these variables, but with asymmetric effects. The findings document a significant impact of familiarity on the domestic bias, so that domestic investors put more of their investments in the home country than foreign investors when it is remote from the rest of the world and does not share the same language with many other countries. This effect is also significant but asymmetric on the foreign bias. Also for the cultural dimension highly significant asymmetric results are documented. The findings indicate that foreign investors prefer in-vesting in culturally close economies, while domestic investors invest less. When a home country is economically developed, i.e. it has a high per capita GDP, a high real GDP growth rate or a high credit rating score foreign investors will put

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

53

more of their investments on that market while domestic investors invest less. In an analogous manner the results for the stock market development are interpreted, where foreign investors are more prone to invest in large markets, while domestic investors are not. Furthermore, the results indicate that factors such as capital control or investor protection have only slight impact on both domestic and foreign bias. Overall, the main results are robust against the conducted test.

2.

Data and Placement in Literature

2.1

Data

My key data set for this analysis is the cross-country equity holding data from the CPIS of the IMF. Either on security-by-security or on an aggregated basis data is collected under the guidance of the IMF. Based on national survey, the IMF produces a geographic breakdown of its residents’ aggregated holdings of securities for each country. For every economy, the collection system is structured in a way that under or double counting is minimized while all holdings are expressed in U.S. dollars. As portfolio holdings are considered and not direct investments it is important to recall how the IMF defines these investments. The definition is further elaborated in paragraph 362 in chapter XVIII of BPM5: “…[A] direct investment enterprise is defined in the Manual as an incorporated or unincorporated enterprise in which a direct investor, who is resident in another economy, owns 10 percent or more of the ordinary shares or voting power.” Nevertheless CPIS database is of high quality, there are some shortcomings: First, the country coverage is still incomplete, as some large countries did not participate from the beginning. Second, as any survey data on cross-border holdings, the data base does still suffer from the third-party holdings problem. Imagine the case, where a parent company in country 1 owns a foreign subsidiary in country 2. If this subsidiary invests in country 3 on behalf of the parent company, the CPIS will account this investment as country 2 holdings of country 3 but not as country 1’s holdings. Nevertheless these shortcomings, this data set is the best available database and is used in most of the studies in this branch of literature. I use this yearly data from 2001 to 2011 and investigate the geographically breakdown of 26 holder countries into 38 target countries. For analyzing determinants of the domestic and foreign bias, I employ numerous variables, which mainly are provided by the World Bank. The data sources from the explanatory variables will be described in detail in chapter 4 of this part. I do not employ the whole CPIS coverage as I excluded countries were data is not available over the whole time period. However, the selection of the 26 holder

54

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

countries captures more than 95% of the world market capitalization. With regard to the geographical breakdown of these countries investments, the selected 38 countries are the target market for 99% of holder markets’ investments, so that panel is expected to be representative. 2.2

Placement in Literature

In part one of my thesis I presented explanation attempts for the home bias puzzle international portfolio holdings. Recall, the institutional explanations employ transaction costs (Glassman and Riddick, 2001), currency risk hedging motives (Adler and Dumas, 1983) and information costs (Ahearne et al., 2004) as determinants for the home bias in equity portfolios. Behavioral approaches characterize investors as bounded rational and build on herding effects (Lauterbach and Reisman, 2004), relative optimism with regard to the home market (Strong and Xu, 2003, Solnik and Zuo, 2013), familiarity (Portes and Rey, 2005), overconfidence and other personal characteristics leading to under-diversification. Following these behavioral approaches investors value foreign stocks differently (Kang et al., 2010), use their information in different ways (Chen et al., 2009) or see foreign markets as more risky than they really are just because they are foreign (Hubermann, 2001). Behavioral explanations for the home bias employ proxies like language or geographic distance. Ke et al. (2010) examine the investment behavior of mutual funds with a local presence in the U.S. and come to the result that they allocate a higher fraction of their portfolio in U.S. firms than those that have no presence.. In addition, language and the country’s level of patriotism are considered as proxies for familiarity (Grinblatt and Keloharju, 2001). Familiarity has an important role for the investment decision in a foreign country. Analyzing foreign stock ownership of Japanese firms Kang and Stulz (1997) document in their survey a preference for larger firms, for the manufacturing industry and for companies with good accounting performances, low leverage, high market-to-book ratio, and low unsystematic risk. Recent contributions to which this empirical study is directly related are Chan et al. (2005), Beugelsdijk and Frijns (2010), Anderson et al. (2011) and Bekaert and Wang (2009). Chan et al. (2005) analyze on country-level mutual fund holdings in 26 countries for the years 1999 and 2000. They were the first to distinguish between domestic and foreign bias and to document a significant impact of stock market development and familiarity variables on both biases. In a similar vein, Anderson et al. (2011) also show on fund level that cultural aspects impact investment behavior directly. While they focus on home bias, Beugelsdijk and Frijns (2010) extend analyses on mutual fund-level to the foreign bias and document similar impacts.

Target Countries

BEL

BRA

CAN

CHL

DEN

FIN

FRA

GER HGK

IDA

JAP

KOR M AL

Home Countries ITA

NET

NOR

RUS

SIN

SOU

SPA

SWE

SWI

TUR

U.K.

U.S.

Table 2: International portfolio investment allocations

AUS 0.808 0.002 0.000 0.000 0.000 0.005 0.002 0.000 0.000 0.002 0.000 0.002 0.004 0.001 0.001 0.020 0.009 0.001 0.017 0.000 0.000 0.005 0.003 0.000 0.011 0.004 AUT 0.000 0.003 0.000 0.000 0.000 0.002 0.001 0.000 0.000 0.000 0.000 0.002 0.000 0.000 0.000 0.002 0.002 0.000 0.000 0.000 0.000 0.001 0.002 0.000 0.000 0.000 BEL 0.000 0.489 0.001 0.000 0.000 0.003 0.002 0.010 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.007 0.005 0.000 0.001 0.000 0.001 0.001 0.002 0.000 0.000 0.000 BRA 0.001 0.000 0.975 0.000 0.030 0.003 0.001 0.000 0.000 0.000 0.000 0.001 0.001 0.003 0.000 0.007 0.008 0.000 0.000 0.000 0.001 0.003 0.001 0.000 0.000 0.006 CAN 0.004 0.005 0.000 0.730 0.001 0.003 0.001 0.000 0.000 0.002 0.000 0.001 0.004 0.001 0.000 0.009 0.013 0.000 0.008 0.000 0.000 0.006 0.007 0.000 0.000 0.012 CHL 0.000 0.000 0.000 0.000 0.792 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 CHN 0.001 0.001 0.000 0.000 0.000 0.002 0.001 0.000 0.001 0.094 0.000 0.002 0.002 0.011 0.001 0.004 0.005 0.000 0.090 0.000 0.000 0.002 0.001 0.000 0.006 0.002 DEN 0.000 0.002 0.000 0.000 0.000 0.452 0.020 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.010 0.000 0.001 0.000 0.000 0.006 0.001 0.000 0.000 0.001 FIN 0.001 0.016 0.000 0.000 0.000 0.005 0.464 0.000 0.020 0.000 0.000 0.002 0.001 0.000 0.000 0.005 0.010 0.000 0.001 0.000 0.001 0.024 0.002 0.000 0.000 0.002 FRA 0.006 0.074 0.001 0.009 0.002 0.030 0.053 0.784 0.050 0.001 0.000 0.048 0.005 0.001 0.000 0.050 0.038 0.000 0.012 0.001 0.040 0.017 0.022 0.000 0.015 0.030 GER 0.005 0.072 0.001 0.009 0.001 0.098 0.045 0.042 0.541 0.001 0.000 0.039 0.004 0.001 0.000 0.040 0.034 0.000 0.008 0.000 0.010 0.020 0.033 0.000 0.015 0.040 GRE 0.000 0.004 0.000 0.000 0.000 0.001 0.001 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.002 0.002 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 HGK 0.007 0.003 0.000 0.000 0.001 0.004 0.003 0.000 0.000 0.813 0.000 0.001 0.003 0.007 0.005 0.010 0.005 0.000 0.025 0.000 0.000 0.004 0.002 0.000 0.010 0.004 HUN 0.001 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 IDA 0.002 0.001 0.000 0.000 0.000 0.003 0.003 0.000 0.000 0.001 0.997 0.000 0.001 0.003 0.000 0.003 0.003 0.000 0.022 0.000 0.003 0.001 0.000 0.000 0.000 0.002 IDO 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.001 0.000 0.006 0.000 0.000 0.000 0.000 0.000 0.000 0.000 IRE 0.001 0.004 0.000 0.000 0.040 0.004 0.041 0.016 0.020 0.007 0.000 0.173 0.002 0.001 0.000 0.030 0.009 0.000 0.002 0.009 0.006 0.012 0.013 0.000 0.019 0.002 ISR 0.000 0.040 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.001 ITA 0.002 0.008 0.000 0.000 0.000 0.030 0.005 0.010 0.030 0.000 0.000 0.488 0.002 0.000 0.000 0.012 0.012 0.000 0.003 0.000 0.005 0.005 0.005 0.000 0.009 0.003 JAP 0.014 0.019 0.000 0.010 0.000 0.027 0.020 0.010 0.030 0.007 0.000 0.010 0.863 0.004 0.001 0.050 0.036 0.000 0.065 0.000 0.002 0.024 0.014 0.000 0.040 0.030 KOR 0.003 0.002 0.000 0.000 0.001 0.004 0.002 0.000 0.000 0.002 0.000 0.001 0.001 0.932 0.001 0.008 0.009 0.000 0.022 0.000 0.000 0.003 0.002 0.000 0.009 0.001 M AL 0.001 0.030 0.000 0.000 0.000 0.001 0.001 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.959 0.001 0.001 0.000 0.018 0.000 0.000 0.000 0.000 0.000 0.000 0.000 M EX 0.000 0.000 0.000 0.000 0.002 0.002 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.002 0.000 0.001 0.000 0.001 0.002 0.001 0.000 0.000 0.002 NET 0.007 0.070 0.001 0.000 0.001 0.009 0.013 0.010 0.040 0.000 0.000 0.013 0.002 0.000 0.000 0.261 0.013 0.010 0.003 0.000 0.005 0.007 0.012 0.000 0.000 0.003 NOR 0.001 0.001 0.000 0.000 0.000 0.008 0.007 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.428 0.000 0.001 0.000 0.000 0.014 0.001 0.000 0.000 0.001 PHI 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.000 0.000 0.000 0.000 0.000 0.000 0.000 POL 0.000 0.000 0.000 0.000 0.000 0.002 0.001 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 POR 0.001 0.002 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.001 0.000 0.000 0.000 0.002 0.002 0.000 0.000 0.000 0.003 0.000 0.000 0.000 0.000 0.000 RUS 0.001 0.001 0.000 0.000 0.000 0.003 0.010 0.000 0.002 0.000 0.000 0.001 0.000 0.001 0.002 0.003 0.005 0.996 0.002 0.000 0.000 0.010 0.002 0.000 0.000 0.001 SIN 0.002 0.000 0.000 0.000 0.001 0.001 0.002 0.001 0.001 0.004 0.000 0.000 0.001 0.001 0.011 0.005 0.002 0.000 0.460 0.000 0.000 0.002 0.001 0.000 0.000 0.001 SOU 0.001 0.001 0.000 0.000 0.000 0.001 0.001 0.001 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.003 0.000 0.001 0.875 0.000 0.001 0.001 0.000 0.000 0.001 SPA 0.001 0.000 0.004 0.000 0.000 0.004 0.005 0.010 0.030 0.000 0.000 0.005 0.002 0.000 0.000 0.014 0.014 0.000 0.003 0.000 0.871 0.005 0.004 0.000 0.010 0.002 SWE 0.001 0.002 0.000 0.000 0.000 0.004 0.075 0.000 0.003 0.000 0.000 0.001 0.000 0.000 0.000 0.020 0.024 0.000 0.003 0.000 0.000 0.553 0.003 0.000 0.010 0.001 SWI 0.005 0.012 0.000 0.000 0.000 0.017 0.012 0.010 0.019 0.001 0.000 0.010 0.004 0.001 0.000 0.024 0.029 0.000 0.007 0.000 0.003 0.024 0.539 0.000 0.010 0.012 THA 0.000 0.000 0.000 0.000 0.000 0.002 0.001 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.001 0.002 0.001 0.000 0.008 0.000 0.000 0.000 0.000 0.000 0.000 0.000 TUR 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.000 0.000 0.000 0.000 0.004 0.000 0.986 0.000 0.001 U.K. 0.019 0.040 0.003 0.020 0.040 0.083 0.147 0.037 0.080 0.052 0.000 0.061 0.013 0.005 0.003 0.120 0.083 0.000 0.090 0.067 0.010 0.069 0.035 0.000 0.624 0.040 U.S. 0.105 0.100 0.009 0.240 0.090 0.170 0.065 0.072 0.140 0.018 0.010 0.139 0.100 0.026 0.011 0.281 0.180 0.000 0.123 0.050 0.040 0.174 0.300 0.010 0.160 0.809

AUS

This table contains the distribution of the 26 home countries average portfolio allocation across 38 target countries in the time period between 2001 and 2011: Australia (AUS), Belgium (BEL), Brazil (BRA), Canada (CAN), Chile (CHL), Denmark (DNK), Finland (FIN), Germany (GER), Hong-Kong (HGK), India (IDA), Italy (ITA), Japan (JAP), Korea (KOR), Malaysia (MAL), the Netherlands (NET), Norway (NOR), Russia (RUS), Singapore (SIN), South Africa (SOU), Spain (SPA), Sweden (SWE), Switzerland (SWI), Turkey (TUR), the United Kingdom (U.K.), the United States (U.S.).

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

55

56

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

Bekaert and Wang (2009) discuss different measures for domestic and foreign bias and get similar findings for their measures. I follow Chan et al. (2005) and analyze determinants for both home and foreign bias. Importantly, this analysis differs from theirs as they exclusively employ traditional determinants in their study, while I combine traditional aspects with cross-cultural variables. Given that Sørensen et al. (2007) compare home bias in equities and bonds and conclude that including bonds in the analysis does not shed more light on the home bias puzzle, I restrict this analysis on countries’ equity holdings. As the focus is not on detecting how home bias differs in different asset classes, only equity holdings are employed, while other asset classes are for future research.

3.

Calculation of the dependent variables

For calculating my dependent variables the country’s international portfolio investment allocations are employed. The domestic bias of home country is defined as the extent to which the actual domestic allocation deviates from the benchmark weight and is calculated as follows: © (13) ”¤ = log ¨ ∗ ª ©

while ”¤ is the domestic bias in equity markets in country , © is the actual allocation in the domestic market and ©∗ is the optimal value. As I use the international CAPM as benchmark model, the latter one is simply the fraction of country ′s stock market capitalization on the world market capitalization. For calculating the foreign bias score the log ratio of the actual weight of country in countries ’s portfolio to the optimal allocation is used. Hence, the foreign bias score is defined as follows: © (14) ‹¤ = log ¨ ∗ ª © ‹«n ≠

I expect most countries to have a large and positive domestic bias, so that in most cases © < ©∗ , i.e. investments in target countries are lower than according to the optimal allocation. Consequently, ‹¤ is expected to have negative values. Lower values for this variable indicate less foreign investments and so a higher foreign bias. According to Dahlquist et al. (2003) and Chan et al. (2005) also the float adjusted

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

57

world portfolio is constructed, where each country is weighted according to the amount of free-floating shares available for investors. The average foreign bias of foreign investors in in a home country is calculated by averaging the measure across all countries. The results are summarized in Table 3. For the domestic bias (column 1) the sample average is 3.805, while the lowest overweighting is documented in the U.S. with a domestic bias of 0.748. The highest values are documented in Turkey (5.667) and Chile (5.389). In the second column the results for the foreign bias are shown. They range between -1.712 in Russia to -3.262 in South Africa. For the whole sample an average amount of -2.332 for the foreign investment bias is documented. Overall, I investigate on cross-country differences in investment behavior characterized by economic and stock market development determinants, capital control, investor protection and also by familiarity and cultural aspects. I hypothesize that if a country is faced to barriers towards international investments, which as consequence lead to higher deadweight costs for foreign investors, then foreign investors will hold less of that country’s equity than theoretically predicted. More formally: Hypothesis 1: Foreign investors put less of their investments in countries with higher barriers to international investments, while domestic investors invest more. Introducing behavioral and cultural aspects into the analysis allows me to detect their impact on the international portfolio allocations around the world. I hypothesize that if a target country has a very different culture form that of the home country and the home country does not feel familiar with the target country, then investors will invest less, ceteris paribus, foreign bias is more negative. More formally: Hypothesis 2: Foreign investors invest less in culturally distant target countries they do not feel familiar with, while domestic investors invest more.

58

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

The table contains the values of the dom estic bias of dom estic investors and the average foreign bias of foreign investors in a target country, across our sam ple of 26 hom e countries. The dom estic bias measures the deviation from a county's actual dom estic investm ent weight to its optim al weight. The optimal weight is calculated according to that country's world m arket capitalization weight. The average foreign bias of foreign investors in a host country is m easured by averaging the foreign bias across all remaining countries. The table shows the average values for the sam ple period of 2001 to 2011.

Domestic Bias

Average Foreign Bias

AUS

3.674

-2.289

BEL

4.429

-2.389

BRA

4.147

-2.776

CAN

3.124

-2.322

CHL

5.389

-3.184

DEN

4.756

-2.079

FIN

4.627

-2.013

FRA

2.956

-2.226

GER

2.877

-2.265

HGK

3.708

-2.674

IDA

4.113

-2.151

ITA

3.45

-2.287

JAP

2.308

-2.382

KOR

4.143

-2.37

M AL

5.167

-2.591

NET

2.928

-2.18

NOR

4.653

-2.209

RUS

4.339

-1.712

SIN

4.339

-2.116

SOU

4.243

-3.262

SPA

3.619

-2.245

SWE

4.092

-2.269

SWI

3.19

-2.255

TUR

5.677

-2.452

U.K.

2.227

-2.002

U.S.

0.748

-1.942

M ean

3.805

-2.332

Table 3: Domestic Bias of Domestic Investors and Average Foreign Bias of Foreign Investors

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

59

4.

Explanatory Variables and Regression Framework

4.1

Explanatory Variables

Economic Development I suppose that investments in a particular country are related to the economic development of that country. Thus, several proxies to capture a country’s economic development are constructed. First, the per capita gross domestic product (GDP) in U.S. dollar (¬-IFEI) is employed. Table 4 shows the cross-sectional variation of the averaged measures from 2001 to 2011. With large distance the lowest ¬-IFEI is observed in India (U.S.$ 628.391), while the highest values are in Norway (U.S.$ 40,011.214), Switzerland (U.S.$ 37,446.197), and the U.S. (U.S.$ 37,094.887). The second variable is the real growth rate (¬-I¬) indicated in the next column. The values range from 0.401% in Italy, 0.640% in Japan, and 0.682% in Denmark to 7.412% in India and 5.650% in Singapore. The last variable for economic development is the respective credit rating score (RATING) provided by Fitch Ratings. Their scale based on letters is transformed into a numerical scale from 1 to 23, while the maximum value of 23 represents the best rating (AAA). The last column in this category contains the descriptive statistics for this variable: Countries with the lowest levels in credit rating are Turkey (11 which represents BB-), India (13 which represents BB+), Russia (14 which represents BBB-), and South Africa (15 which represents BBB). The highest score with 23.00 is documented on in several sample countries: Finland, France, Germany, the Netherlands, Norway, Switzerland, the United Kingdom, and the United States. Countries with high levels of economic development should have a greater ability to attract foreign investments in the country. Whether this development affects domestic and foreign bias similarly or differently depends on how investors (domestic and foreign) include this factor into their investment decisions: If the economic development of a country could help lower the deadweight cost for a foreign investor more than for a domestic investor, then more foreign investors would be drawn to invest in the country. Consequently, the share of domestic investors in local equities would be lower. In case the advancement of a country’s economic development could lower symmetrically the deadweight cost for both foreign and domestic investors, then our proxies for economic development would not have any different impact neither on foreign nor on domestic biases. However, I expect that a highly developed economy lowers its domestic bias and becomes more attractive for foreign investments, which will increase those countries foreign biases.

60

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

Stock Market Development For measuring stock market development I use the market capitalization of listed companies to GDP. The value of stocks traded to GDP is included, as this measure also captures the market’s liquidity. Furthermore, the Standard & Poor’s (S&P) Global Equity Index and the domestic credit provided by the banking sector is employed for the analysis. Other things being equal, stock market development in the target country is generally expected to be positively associated with higher investments. In other words, I hypothesize that investors are drawn by developed stock markets and will consequently invest more in this market of higher liquidity and lower transaction costs. All data in order to capture the stock market development in different countries is on annual periodicity and provided by the Financial Sector Database from the World Bank. The most common variable to proxy stock market development in home bias literature (Baele et al. 2007, Chan et al. 2005) and literature on market integration (Baele, 2005) is the relative size of the market (•®¯W) measured by the ratio of the market capitalization of the listed companies in that country to that country’s GDP. Listed domestic companies capture domestically incorporated companies listed on the respective country’s stock exchange at the end of the year. This does not include investment companies, mutual funds, or other collective investment vehicles. The values range from 28.679% in Turkey and 35.911% in Italy to 221.254% in Switzerland and 415.414% in Hong Kong.12 Market liquidity, i.e. the ability to buy and sell assets easily, is measured by the ratio of the total value of stocks traded to GDP (°E\±W). Liquidity is an important attribute of a stock market, as – in theory – liquid markets should improve capital allocation and improve long-term economic growth prospects. In the country sample the values of this variable reach from 16.344% in Chile to 374.892% in Hong Kong.

Furthermore, the (S&P) Global Equity Index jW²®g-W³l comprising the three indices S&P Frontier Broad Market Index, the S&P Global Broad Market Index, and the S&P/Industrial Finance Corporation of India (IFCI) is included. The S&P Global Equity Index series is constructed to capture the most liquid and investable stocks in developed, frontier, and emerging markets. While the S&P/IFCI is an index purely focused on the emerging market, the S&P Frontier and Global Broad Market Indices capture all index-eligible countries. Taking these three indices together, it offers a great benchmark to compare stock performance between emerging and developed countries. I employ changes of the index in percent rather than the absolute values as a high increase in the country’s market index indicates that the included companies 12

Again, also for the explanatory variables the float adjusted portfolio is constructed.

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

61

have a growing market capitalization. This offers investors a good investment opportunity, so that they are more willing to invest in their domestic market. In the same way it will also attract more foreign investors. The only countries with negative growth are Finland with -1.805 and Italy with -1.446. The sample average growth is 11.079%, while the Russia (+31.711%), Brazil (+27.180%), and India (23.004%) have the highest increase. However, with the advancement of the stock market development there is a similar discussion as I already had with the measures for economic development and expect the same impacts as for the economic development variables: It depends on the behavior of foreign and domestic investors and whether stock market development influences symmetrically or asymmetrically their domestic and foreign biases. Put the case that developed stock markets are generally characterized by lower deadweight costs for foreign investors than for domestic investors, this will lead to a situation where foreign investors will allocate a higher fraction of their portfolio in the local equity markets. As a result domestic investors will hold proportionately less of the local equities. Otherwise, if a developed stock market leads to lower deadweight costs for both foreign and domestic investors symmetrically, then it is expected that the measures do not have any impact on the investment biases. Capital Control Even though growing financial integration moves hand in hand with decreasing capital controls and many countries have greatly reduced them, some other countries still restrict international inward and outward flows. A trend towards increasing financial integration and financial openness is ought to lead to decreasing foreign investment bias. The variable used to capture capital control is an index constructed by the Economic Freedom Network that measures the imposed restrictions in capital flows (WX´WW-HY). This index assigns a lower rating to countries which impose severer restrictions on cross-border investments. If all investors, independently whether they are domestic or foreign, can act without being exposed to capital controls, then those countries will be rated with the maximum score of 10. If only some industries like banking, defense, and telecommunications are restricted for investments from abroad, then this country will be rated with a score of 8. A rating score of 5 will be assigned to a country where these investments are not restricted, but regulatory controls significantly slow the mobility of capital. In a country where both domestic investments of foreign investors and foreign investments of domestic investors are exposed to high capital control and require approval from government authorities the agency will rate it with a score of 2. A rating of 0 stands for the case where either foreign investments

62

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

of locals or domestic investments of foreigners are only allowed with government approval. The data are provided from the Economic Freedom Dataset (Gwartney et al., 2012). In this sample the country with the lowest index score for economic freedom is Russia with a value of 5.977, while Hong Kong gets the highest score with 8.891. High capital controls will prohibit or discourage foreign investors from investing in that country’s companies. Consequently, I expect the value of this country’s foreign investment bias to become more negative as it is exposed to higher measures of capital control. Investor Protection La Porta et al. (2000) claim that more developed financial markets are typically in countries, which are generally more developed and consequently typically known to better protect investors’ rights. This implicates that investors are averse to invest in counties with a low level of investor protection. The index was developed by Djankov et al. (2008) and updated by the World Bank and is available in the doing business data base of the World Bank. The authors also call it the “the anti-selfdeadling index” (Djankov et al. , 2008, p. 432). The measure captures the strength of minority shareholder protections against the misuse of directors of corporates assets for personal gain. For data collection they employ a questionnaire, which captures securities regulations, company laws, civil procedure codes and court rules of evidence. The index value is the simple average of the percentile rankings of these dimensions. The measure in order to capture this characteristic is the index of the strength of the investor protection j®I´HGl scaled from 0 to 10. Higher values indicate greater powers of shareholders to challenge transactions. The highest scores are documented in Singapore with 9.3 and Hong Kong 9.0 while countries such as Switzerland with 3.0, Norway with 4.4, and Russia with 4.7 have the lowest. With regard to the low values for Switzerland Djankov et al. (2008, p. 444) state that it seems to be extremely friendly to insiders, while it is hostile towards outsider shareholders. They add that they might have missed some important dimension or mechanisms for protecting minority shareholders. An alternative explanation attempt might be that the enormous investment resources of the banks in Switzerland could have inflated the value of the Swiss stock market artificially.

I expect higher informational asymmetries between domestic and foreign investors in counties with low levels of investor protection. The deadweight costs for foreign investors to allocate their wealth in these countries increase compared to those for domestic investors. Hence, investors from abroad will hold less of local equities, while domestic investors invest more in this market with poor investor protection.

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

63

Familiarity and Information Availability Brennan and Cao (1997) claim that investors have less information about foreign securities and consequently invest less in foreign countries and investors who are less familiar to a country associate an investment in that country with higher information costs, which discourages them from investing abroad. The study of Kang and Stulz (1997) indicates a strong preference of U.S. investors to invest in larger and more known companies from the manufacturing branch in Japan. This result confirms the notion that investors are reluctant in investing in companies they are not familiar with. Even within the U.S., mutual fund managers are more likely to invest in companies that have their headquarters close to their home city (Coval and Moskowitz, 1999). Grinblat and Keloharju (2001) analyze the Finnish market and show that investors prefer stocks from firms that have the same language and the same cultural background. Sarkissian and Schill (2004) analyze how geographic proximity influence overseas listing decisions and come to the result that (together with other proxies) geographic proximity to the foreign market plays a dominant role in making overseas listing decisions. The first variable is about the major or official languages in the analyzed countries. In the regression analysis I construct a dummy variable (\Eg¬±E¬W), that takes a value of 1 if country j and i have a common major or official language and is equal to 0 if they do not share the same language. Japan, South Korea, Russia, and Turkey do not share common languages, while the countries with the highest scores are Canada (0.278) and Australia (0.243). I hypothesize that investors are drawn by countries that share the same language with their home country.

The second variable -®•GEgFW is the distance in kilometers (km) between the capitals of two countries. It is no surprise that the most remote country is Australia (13,006.170 km). The average distance to the capitals of our sample is 6,931.350 km. Also very distant is Chile (12,646.889 km) and Brazil (10,976.993 km). All countries with the lowest distances are in Europe: The closest is Germany with an average distance to the other capitals of 4,768.901 km, followed by Denmark (4,807.225 km) and Sweden (4,870.375 km). I expect that increasing geographic distance will reduce investments in the respective country and will so lead to a more negative foreign bias. It is also convincible that a lack of investment relevant information also proxies for lack of familiarity. To capture information availability the number of internet users (per 100 people) (®gGW´gWG) provided from the World Bank is employed. The have the lowest value in India (0.273) and the highest in South Korea (27.577). Hence, the mentioned variables geographic proximity, language, and internet users capture the information flow between the countries. By quantifying the barriers that

64

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

foreign investors are faced to when investing in a foreign market, these variables capture the amount of information asymmetries between domestic and foreign investors and indicate differences in the deadweight costs for investors depending on whether they are domestic or foreign. Thus, I hypothesize that if there is a lower foreign investment bias in country towards country , then both countries are more familiar or geographically closer to each other, share the same language, and have a high number of internet users. The other way around: Put that a country is isolated from all the other economies then it will be exposed to high information costs which will in a next step discourage investors from this country to invest a fraction of their wealth in a foreign market. Consequently, they will invest it in the local market and will so not be able to enjoy benefits from international diversification. Cultural Dimensions Hofstede (2001, p. 9) defines culture as “the collective programming of the mind distinguishing the members of one group or category of people from another”. Although cross-cultural studies have multiple approaches towards culture, Hofstede dominates research in economics and international business with his 2001 study. His empirical work builds on cross-cultural survey on IBM employees and identifies dimensions in order to capture cross-cultural differences. Anderson et al. (2011) employed some of his cultural dimensions to document their impact on the home bias of mutual fund holdings and Beugelsdij and Frijns (2010) conduct a similar analysis on the foreign bias also based on mutual fund holdings. Individualism (®-°) (opposed to collectivism) captures the amount individuals are integrated into groups. Countries with a high degree of individualism are more autocratic and take decisions more individually (Spector et al., 2001). Kerr and Tindale (2004) follow by a long-standing literature focused on how the effect on risk behavior varies whether there is group-based (collective) and individual-based decision making. They show that individual decisions exhibit a higher variance than group decisions and that in high risk situations groups make more risk-averse decisions than individuals (Shupp and Williams, 2008). Based on the definition of individualism of Hofstede (2001), Chui et al. (2010) claim that there is a linkage between individualism and overconfidence: They document that decisions in individualistic societies (where decisions are made individually), are driven by more overconfidence. Building on their results Van den Steen (2004) argues that overconfident investors, i.e. those who are more confident on their capabilities compared to those of the others, overestimate the precision or the quality of their information. I hypothesize that individualistic economies will exhibit more international diversification (less home bias)

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

65

and will so have higher (less negative) foreign investment bias than countries with low individualism scores. In this sample the countries with less individualism are South Korea (18), Singapore (20), and Chile (23), while the average values of the sample is about 57.692. The most individualistic countries are the United States (91), Australia (90), and the United Kingdom (89). Uncertainty avoidance (±E®) quantifies the extent to which a country’s culture influences its population to feed comfortable/uncomfortable in uncertain and novel situations where people do not know the outcome and indicates their willingness to handle these situations. Despite Hofstede (2001) proclaims to not equalize uncertainty avoidance and risk avoidance, countries with a low uncertainty score typically have a relatively risk-tolerant market-based financial system compared to countries with lower levels which typically have a risk adverse bank-based financial system. Traditional approaches which model investment behavior see risk aversion as a central topic, while the focus of behavioral approaches is on the heterogeneity of investor’s risk attitude. Mostly heterogeneity in investor’s risk attitude is examined along multiple dimensions like education, gender, or age. Dorn and Huberman (2005) show that a shifting in self-reported risk aversions helps explaining variations in the actual risk taking, which they proxy by portfolio concentration and volatility. Using a sample of German families in order to analyze their economic decision making process, Dohmen et al. (2012) confirm a correlation between risk and risk attitudes. They find strong evidence that the willingness to take risk and the willingness of parents to trust others is projected on their children. This finding implies that risk attitudes and ceteris paribus the decision on how to allocate investments, are persistent over time and across individuals. Furthermore on the country level, data show that if an economy has high scores for uncertainty avoidance, also the level of life insurance is high (Chui and Kwok, 2009). Overall, I hypothesize that countries with low scores of uncertainty avoidance are more willing to take risks and so exhibit only low preference for the domestic market (low home bias) and consequently invest more abroad. The other way around, countries with high levels of uncertainty avoidance tend to be more risk averse and classify investments in foreign countries as more risky than those in the domestic market. The country with the lowest value of uncertainty avoidance is by far Singapore (8), while Russia (95), Belgium (94), and Japan (92) have the highest scores for this proxy. The overall cultural distance (F-) was first introduced by Kogut and Singh (1988) and is based on the culture dimensions of Hofstede. According to Shenkar (2001) the shortcoming of this measure is that the difference in the score on each dimension is equal, so I follow Beugelsdijik and Frijns (2010) and employ the Euclidean distance versions of the index.

66

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

Formally:

4

F- = µe ¶ ·=1

~ F· − F· •

‚2·

2

¸

(15)

where F- is the cultural distance between home country and target country , F· is the index for the –th cultural dimension of country , while F· is the respective score for country . ‚2· is the variance of the –th index. In this calculation four dimensions are included to calculate the cultural distance between home and target country: the individualism j®-°l, the uncertainty avoidance j±E®l as well as two other Hofstede dimensions the power distance jI-®l and the masculinity jYE•l.13 This measure enables me to quantify how cultural proximity between home and target country impacts investment decisions. Geographical distance is negatively associated with foreign investments, because investors feel unfamiliar to this market or perceive informational disadvantage with increasing distance. This effect may be transferred to the cultural distance between home and target countries. Culturally distant countries are unfamiliar to investors as they may have different legal systems, incentive structures, or decision making processes (Hofstede, 1980 and Morosini et al., 1998). In this sample the home countries that invest in culturally distant economies are Denmark and Sweden with scores of 3.349 and 3.309 while I find the lowest scores Brazil (2.060) and South Africa (2.071). Thus, I expect portfolio allocations in culturally distant countries to be lower as they are considered as less attractive, ceteris paribus, and this leads to higher amounts of home bias in culturally distant countries from the rest of the world. Hence, foreign bias in culturally distant target countries is more negative as their weight in that country’s investment portfolio is lower. Other Variables This last category of explanatory variables includes two measures in addition to the variables introduced above. The first measure is the ratio of domestic credit provided by the banking sector to GDP (-F´W-®G). This measure is often used to capture the size and the development of a country’s banking market (Driessen and Laeven, 2007 and Beck et al., 2010). Countries with developed banking markets are ought to have a 13

The cultural dimensions from the analyzed countries are statistically correlated with other country data. For instance, individualism is associated with national wealth and the mobility between social classes. Uncertainty avoidance is correlated with Roman Catholicism and the legal obligation for citizens to carry identity cards. Power distance is linked to the inequality of income distribution in a country and to the use of domestic politics’ violence. There is a negative relationship between masculinity and the percentage of democratically elected women in governments.

67

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

more evolved deepness in their banking markets which typically offers better diversification opportunities. Hence, institutional and regulatory environment in a country impact the banking sector’s development. By far, the highest ratio is observed in Japan with 313.292%, while the lowest value is in Russia (28.246%). Including -F´W-®G allows quantifying frictions which should decrease the deepness of markets and consequently lead to less developed banking sectors and stock markets. Thus, it is expected that this variable might influence both depending variables positively. The second variable is the correlation between the returns of two countries (FH´´). For pair of holder and target country ( and ), where is the home country and is the target country, the correlation coefficient is computed by using markt index returns provided by Datastream. Hence, investing in countries whose equity returns are negatively correlated with the returns of the domestic market generates additional benefits of international diversification. Thus, I expect the correlations to have negative impact on foreign bias and a positive impact on domestic bias. 4.2

Regression Framework

The following econometric model is used to estimate the domestic bias in the country sample by employing an OLS regression: ”¤ = } + q1 WF- + q2 •Y- + q3 FFHgG´ + q4 ®I´HG + q5 F±\G + q6 XEY +q7 HGJW´ °E´®EK\W• + € ,

(16)

where dbi is the average degree of domestic bias in country and € is the um. WF-A is a matrix of variables that measures country ’s economic development. These variables include the gross domestic product (GDP) per capita j¬-IFEIl, the real GDP growth rate j¬-I¬l, and the country’s credit rating score j´EG®g¬l. •Y- controls for the stock market development and includes variables like the market capitalization of listed companies to GDP j•®¯Wl, the total value of stocks traded to GDP j°E\±Wl, and the growth rate of the S&P Equity Index jW²®g-W³l.

FFHgG´H\ captures the amount of economic freedom jWX´WW-HYl, while ®I´HG measures the strength of investor protection.

XEY controls for familiarity and information availability. It includes a dummy variable which indicates whether two countries use a common language j\Eg¬±E¬Wl, a variable indicating the geographic distance j-®•GEgFWl, and the number of internet users per 100 people j®gGW´gWGl.

68

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

CULT is a matrix of three cultural dimensions: the first indicates the amount of individualism j®-°l, the second captures the extent of uncertainty avoidance j±E®l, while the latter variable measures the cultural distance between two countries jF-l.

HGJW´ °E´®EK\W• controls for two other variables that have potential power to explain both domestic and foreign bias. This matrix includes the domestic credit provided by the banking sector j-F´W-®Gl and the correlation between the returns of two countries (FH´´). A very similar econometric model is employed in order to estimate the foreign bias in the portfolio allocation by employing an OLS regression: ‹¤ = } + j1 − ”¤ l + q1 WF- + q2•Y- + q3FFHgG´ + q4 ®I´HG + q5F±\G + q6XEY + q7HGJW´ °E´®EK\W• + € ,

(17)

while ‹¤ indicates the weight a home country over-/ underweights a target country, there is nearly the same structure as in equation (16), except for one modification: If investors are overinvested in their home market, then the remaining fraction of their portfolio that can be invested in foreign markets becomes proportionally smaller. So this additional explanatory variable j1 − ”¤ l is included in the regression, which controls for the automatic impact of the domestic bias on the foreign bias. However, one could argue that the explanatory variables are correlated. This problem of multicollinearity between two or more independent variables makes it impossible to determine which of the variables account for variance in the dependent variable. For detecting multicollinearity in the data a pairwise correlation analysis, the principal component or the auxiliary regression approach could be conducted. I employ the latter methodology which regresses each independent variable on the other remaining variables and calculate the respective values for the ´2 . In a next step multicollinearity is detected by employing the X-Test. If the result is statistically significant, then multicollinearity is excluded. Compared to the other approaches to detect multicollinearity the advantage of this method is that it allows to detect which variable provokes the effect. Controlling for multicollinearity in my data ensures that the observed effect can be ascribed to the explanatory variables.

69

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

For each of the home countries this set of explanatory variables from six categories is employed: (i) Economic development: gross domestic product (GDP) per capita, real GDP growth, the country's credit rating, (ii) Stock market development: ratio of stock market capitalization to GDP, stocks traded value as a percentage of GDP, growth rate of S&P Global Equity Index in percent, domestic credit provided by banking sector as percentage of GDP, (iii) Capital control: capital control restrictions, (iv) Investor protection: strength of investor protection, (v) Familiarity and Information Availiability: average common language dummy variable, average distance in kilometers, and internet users per 100 people, (vi) Cultural dimensions: indivualism, uncertainty avoidance, and the average cultural distance.

ECONOMIC DEVELOPMENT

GDP per capita GDP Growth

S TOCK MARKET DEVELOPMENT

Credit Rating

M arket Capitalization to GDP

Stocks Traded S&P Global Total Value to Equity Indices GDP Growth

CAPITAL CONTROL

INVES TOR PROTECTION

Index of capital control restrictions

Investor protection 5.7

AUS

24,021.53

2.954

22

116.226

93.986

14.028

8.149

BEL

24,072.34

1.451

21

63.915

28.247

5.768

7.517

7

BRA

4,168.60

3.559

12

54.476

26.815

27.18

6.185

5.3

CAN

25,237.07

1.94

23

114.047

83.215

11.808

8.257

8.7

CHL

6,114.36

4.088

18

107.245

16.344

18.759

7.855

6.1

DEN

31,084.19

0.682

23

62.448

49.73

12.103

7.876

6.3

FIN

26,481.19

1.885

23

94.586

119.395

-1.805

7.838

5.7

FRA

22,699.99

1.175

23

78.479

78.942

0.8

7.377

5.3

GER

24,356.39

1.192

23

44.546

63.89

7.187

7.631

5

HGK

31,804.43

4.201

21

415.414

374.892

9.776

8.891

9

IDA

628.391

7.412

13

66.692

60.386

23.004

6.5

5.4

ITA

19,591.36

0.401

20

35.911

49.553

-1.446

7.047

8.3

JAP

39,066.73

0.64

21

76.369

87.755

1.338

7.691

6

KOR

14,290.70

4.124

19

75.197

148.393

20.848

7.277

7

MAL

4,651.84

4.659

17

137.049

41

13.917

6.746

5.4

NET

25,698.63

1.336

23

89.496

132.498

2.86

7.814

5.7

NOR

40,011.21

1.511

23

54.712

60.784

16.45

7.492

4.4

RUS

2,551.19

4.861

14

59.892

35.902

31.711

5.967

4.7

SIN

28,505.91

5.65

23

181.082

118.335

10.158

8.638

9.3

SOU

3,481.22

3.494

16

197.873

97.022

17.702

6.933

8

SPA

15,550.40

1.919

22

85.167

132.974

6.2

7.481

5

SWE

31,231.97

2.354

23

102.534

121.695

10.298

7.54

5.8

SWI

37,446.20

1.759

23

221.254

217.565

4.997

8.238

3

TUR

4,863.33

4.414

11

28.679

41.398

20.39

6.108

5.5

UKD

27,957.86

1.697

23

127.586

173.569

2.495

8.245

8

USA

37,094.89

1.585

23

122.296

250.606

1.533

8.159

8.3

2.729

20.058

108.199

104.034

11.079

7.517

6.301

Mean

21,256.23

Table 4: Descriptive Statistics for the explanatory variables from 2001 to 2011

70

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

For each of the home countries this set of explanatory variables from six categories is employed: (i) Economic development: gross domestic product (GDP) per capita, real GDP growth, the country's credit rating, (ii) Stock market development: ratio of stock market capitalization to GDP, stocks traded value as a percentage of GDP, growth rate of S&P Global Equity Index in percent, domestic credit provided by banking sector as percentage of GDP, (iii) Capital control: capital control restrictions, (iv) Investor protection: strength of investor protection, (v) Familiarity and Information Availiability: average common language dummy variable, average distance in kilometers, and internet users per 100 people, (vi) Cultural dimensions: indivualism, uncertainty avoidance, and the average cultural distance. FAMILIARITY AND INFORMATION CULTURAL DIMENS IONS

AVAILABILITY

Dummy Language Average Distance

OTHER VARIABLES

Internet

IDV

UAI

Average Cultural Distance

Domestic credit to GDP

Average Return Correlation

AUS

0.243

13,006.17

12.065

90

51

2.503

123.839

BEL

0.194

4,906.24

19.557

75

94

2.395

111.802

0.457 0.336

BRA

0.028

10,976.99

2.981

38

76

2.06

84.834

0.458

CAN

0.278

8,139.30

21.506

80

48

2.268

192.837

0.589

CHL

0.056

12,646.89

5.378

23

86

2.566

79.023

0.407

DEN

0.139

4,807.23

24.732

74

23

3.349

187.216

0.398

FIN

0.028

4,915.94

20.151

63

59

2.407

81.362

0.375

FRA

0.111

5,000.36

17.099

71

86

2.282

116.917

0.5

GER

0.139

4,768.90

16.546

67

65

2.169

134.306

0.524

HGK

0.222

7,987.60

23.273

25

29

2.57

149.727

0.499

IDA

0.222

6,562.86

0.273

48

40

2.284

62.695

0.298

ITA

0.167

5,032.62

11.875

76

75

2.265

121.016

0.416

JAP

0

8,357.89

17.358

46

92

2.968

313.292

0.436

KOR

0

7,794.42

27.577

18

85

2.481

95.094

0.357

M AL

0.056

8,511.90

2.808

26

36

3.111

127.017

0.235

NET

0.028

4,895.75

24.159

80

53

2.843

184.07

0.536

NOR

0.028

4,939.27

21.024

69

50

2.98

80.893

0.517

RUS

0.000

4990.799

3.401

39.000

95.000

2.834

28.246

0.235

SIN

0.278

8694.842

15.555

20.000

8.000

3.170

79.127

0.485

SOU

0.194

9223.548

0.594

65.000

49.000

2.071

177.369

0.546

SPA

0.056

5513.736

12.585

51.000

86.000

2.078

177.959

0.436

SWE

0.056

4870.375

22.437

71.000

29.000

3.309

123.653

0.495

SWI

0.194

4926.362

22.302

68.000

58.000

2.255

174.127

0.446

TUR

0.000

5143.760

17.155

37.000

85.000

2.139

52.694

0.416

UKD

0.222

5030.807

17.206

89.000

35.000

2.723

179.108

0.585

USA

0.222

8570.555

1.283

91.000

46.000

2.536

224.836

0.552

M ean

0.122

6931.350

14.649

57.692

59.192

2.562

133.195

0.443

Table 4: continued

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

5.

Empirical Results

5.1

Results for the Domestic Bias

71

In this section I empirically test the implication of the predetermined factors in domestic and foreign bias. In all the conducted tests, the observations on the measure of the domestic bias are regressed against each set of the previously described explanatory variables. The same approach is employed on the foreign investment bias analysis. Beside the whole time period I always run separate time sub-samples. These results do not differ substantially. The reason for this effect is twofold: On the one side the respective weights of the countries do not change over time. On the other side the large part of the explanatory variables are fairly stable during the sample period. Thus, the reported results in all subsequent sections are based on averaged values of the dependent variables (domestic bias and foreign bias) against the averaged predetermined variables for the sample period between 2001 and 2011. The results for the domestic bias are reported in Table 5. The variables are grouped in six categories. The first column indicates the predicted signs of the coefficients and from column two on the estimates and the t-ratios of the explanatory variables categories are shown separately. The last column contains the estimates taking all variables into the regression. It is evident that some groups of predetermined variables have more influence on the cross-sectional variation of the domestic bias than others. Judging from the ´2 (Z” . ´ ( l the results show that the variables capturing stock market development have the highest explanatory power reaching an ´2 of 17,56% (Z” . ´ ( = 17.54%). The lowest explanatory power is reported for the capital control variable. It has an ´2 of 3.74% and an Z” . ´ ( 3.70%. The variable •®¯W has a negative coefficient of 0.0010 and has a ¾-stat of -1.9016), which I interpret in the following way: countries with small stock markets have smaller deadweight costs investing in the domestic market than foreign investors have and therefore invest more in the domestic market. As predicted, the W²®g-W³ has a positive coefficient of 0.017 with a ¾-ratio of 2.4537, as high index growth rates increase the willingness of investors to put a large fraction of their wealth in the local market and so leads to an increasing domestic bias. All variables bear the predicted negative signs. These findings do not contradict the hypothesis that deadweight costs for countries with a developed stock market for investing in foreign markets are lower. Hence, they have smaller domestic exposures and with that lower measures for domestic bias. The variables in order to capture cultural aspects have the predicted signs and help explain 12.26% of the cross-sectional variation of ”¤ . The estimates for uncertainty

72

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

avoidance (±E®) and cultural distance (F-) are statistically significant at the 10% level, while no statistical significance could be documented for ®-°. The results support the intuition as investors from countries with a high ±E® tend to exhibit higher levels of domestic bias. Also, investors who are culturally distant from other countries put more of their investments on their local market.

The strength of investor protection j®I´HGl has an ´2 of 12.12% (Z” . ´ ( of 11.98%). This variable has a positive coefficient of 0.0199 with a ¾-ratio of 1.5691. When a country strongly enforces its laws, domestic investors’ confidence in that country’s regulatory system increases. Hence, their willingness to put more investment in the local market augments and thus supports the hypothesized relationship.

With regard to information availability the results have the predicted negative sign for the number of internet users per 100 people j®gGW´gWGl: the coefficient is 0.0143 with a ¾-ratio of -0.2043. As the internet makes information from all over the world more easily available, a higher number of internet users increases the quantity and quality of investment relevant information from other countries. Unfortunately, no statistical significance could be reported for this variable. Both common language j\Eg¬±E¬Wl and (log) geographic distance j-®•GEgFWl are statistically significant variables at the 5% level. They have the predicted impact on the preference of domestic investors on local equities. The findings indicate that investors from countries that share the same language with many other countries or are geographically close to other economies exhibit lower deadweight costs investing aboard. Thus, they are more willing to invest in foreign markets and so lower their domestic bias. As there is a negative coefficient for common language, countries that share on average the same language with many other countries, have a lower exposure on the domestic market. Furthermore, it is documented that geographic distance has a positive impact on domestic bias, which indicates that more remote countries are subject to invest more domestically. Hence, these results support the notion that countries prefer domestic investments when they feel less familiar with the other economy. This does not contradict the hypothesis that measures for domestic bias increase when domestic investors are faced to increasing deadweight costs when investing abroad. For the whole category an ´2 of 8.31% (Z” . ´ ( of 8.31%) is reported.

The category of the other variables captures 8.12% of the cross-sectional variation of the domestic bias in the sample countries (Z” . ´ ( of 8.10%). Only FH´´ has slight explanatory power with an estimate of the coefficient of -0.0324 and a ¾-ratio of 1.202. No statistical significance could be documented for the domestic credit provided by the banking sector.

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

73

For the data set the economic development of an economy is only able to explain 5.79% of the documented domestic bias in that country’s equity portfolio. In this set of explanatory variables only ´EG®g¬ is statically significant at a 5% level. This finding suggests that the degree of development of an economy only plays a little role in domestic portfolio allocation.

With regard to the influence of capital control on domestic portfolio allocation WX´WW-HY has a negative coefficient of -0.1586, but no statistical significance could be documented. Thus, it could not be confirmed that a higher level of economic freedom, for instance due to less capital control reduces the country’s domestic exposure and, ceteris paribus, leads to a decreasing amount of home bias in portfolio allocation. This evidence is not consistent with the hypothesis that lower deadweight costs for domestic investors investing abroad affects the domestic bias measure positively. Taken all variables together it is interesting to evaluate changes in the explanatory power of each independent variable indicated in the last column. The variable -F´W-®G is now significant at the 5% level, while °E\±W loses significance. A similar change can be observed for WX´WW-HY. ®I´HG is now even significant at the 5% level. In this joint regression also the significance of the cultural dimensions increases, so that F- now has a ¾-stat of 2.6084. Given that the sample of analyzed countries includes 26 home and 38 target countries one could argue that it is not feasible to include all explanatory variables in one single regression. In order to prevent this objection, I follow Chan et al. (2005) and also employ an alternative approach: Therefore in a first step the fitted values of the six categories are computed. The index of the domestic bias is subsequently regressed against these fitted values jointly. These results show that the coefficients from the stock market development and cultural categories are statistically significant at the 1% level, while investor protection and familiarity have a statistical significance at the 5% level. The impact of economic development is only significant at the 10%, while no significance could be documented for capital control.

74

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

In this OLS regression results domestic bias db i , the ratio of the actual share of foreign equity positions in that country’s international investment portfolio to the optimal value according to the world market capitalization share is employed. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING), the foreign direct investments (FDI), and the trade of that (TRADE); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level. S et of predetermined variables coeff

t-stat

Joint Regression coeff

t-stat

ECONOMIC DEVELOPMENT c

0.8608

1.2048

1.0195

1.351 *

GDPCAP

-

-0.0022

-1.8011 *

0.0000

-1.6561 *

GDPG

-

-0.0725

-0.4762

-0.0539

-0.7254

RATING

-

-0.0336

-2.0872 **

-0.0042

-1.9103 **

R²

0.0579

Adj R²

0.0577

Fitted Values

0.665

1.9738 *

S TOCK MARKET DEVELOPMENT C

0.5337

1.3199 *

-

-0.001

-1.9016 *

-0.0043

-1.5489 *

VALUE

-

-0.0009

-1.9002 *

-0.0065

-0.1733

EQINDEX

+

0.0167

SIZE

2.4537 **

R²

0.1756

Adj R²

0.1754

Fitted Values

0.3026

C

1.8914

0.3968

-0.1586

-0.1174

0.0061

2.1205 **

3.6931 ***

CAPITAL CONTROL EFREEDOM

-

R²

-0.0625

-1.6782 *

-0.0038

-2.2988 **

0.0374

Adj R²

0.037

Fitted Values

-0.2965

-0.3435

INVES TOR PROTECTION C IPROT

+

0.5712

1.6234 *

0.0199

1.5691 **

R²

0.1212

Adj R²

0.1198

Fitted Values

0.5694

2.666 **

Table 5: Explaining the Domestic Bias in countries' equity portfolios

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

In this OLS regression results domestic bias db i , the ratio of the actual share of foreign equity positions in that country’s international investment portfolio to the optimal value according to the world market capitalization share is employed. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING), the foreign direct investments (FDI), and the trade of that (TRADE); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level. Set of predetermined variables coeff

t-stat

Joint Regression coeff

t-stat

FAMILIARITY C

0.9714

1.1819 *

LANGUAGE

-

-0.0376

-2.5915 **

-0.0342

-2.353 **

DISTANCE

+

0.0152

2.282 **

0.0081

2.6974 **

INTERNET

-

-0.0143

R²

-0.2043

-0.0023

-1.6253 *

-0.0002

-1.0268 *

0.0831

Adj R²

0.0831

Fitted Values

0.1989

2.5228 ** CULTURAL DIMENS IONS

C IDV

-

0.7718

0.9694

-0.0013

-0.1346

UAI

+

0.0015

1.5144 *

0.0012

2.1919 *

CD

+

0.0234

1.5631 *

0.0617

2.6084 **

R²

0.1229

Adj R²

0.1226

Fitted Values

0.5065

3.6563 *** OTHER VARIABLES

0.6512

0.7854

DCREDIT

+

0.0014

1.0668

0.0016

2.3809 **

CORR

+

0.0324

1.0221 *

0.0201

2.0015 **

C

R² Adj R² Fitted Values

0.0812

0.5862

0.081 0.2012

0.5755 2.1689 **

Table 5: continued

75

76

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

In this section the regression results for the foreign bias ‹¤AŸ are presented in Table 6. In the second row the coefficients of domestic bias are reported and document a strong negative relationship. An increase of the domestic bias measure induces a significant decrease in the foreign bias in all specification models.

5.2

Results for the Foreign Bias

The results for the foreign bias are not that different from those of the domestic bias regression. Hence, it can be seen that compared to the former regression the explanatory power of the variables is higher. Also judging from the ´2 higher explanatory power for both the single groups as well as for the joint regression and the fitted values is observed. Despite a part of the explanatory power of our regression will be due to the domestic bias measure, I attribute large predictive power to the other explanatory variables. Thus, a detailed look in order of its magnitude across the six regression categories is given as an indication of the relative powers of the independent variables to explain foreign bias. The largest impact on the foreign bias departs from the category of the stock market development. For this category highest explanatory power compared to the other cat2 egories is documented with an ´2 of 39.65% (Z” . ´ of 39.07). All variables bear the predicted sign and are statistically significant at the 1% level. The results are consistent with the earlier finding on the domestic bias where a larger market capitalization of listed firms is associated with lower levels of this measure. This indicates that countries put more investments (exhibit less negative ‹¤ ) in large, highly visible markets j•®¯Wl, which trade a high value of stocks to GDP j°E\±Wl, and exhibit high growth rates of equity indices jW²®g-W³l, Judging from the ´( (28.91%), the group with the second largest predictive power is the economic development (Z” . ´ ( of 28.34%). All explanatory variables are significant at least at the 5% level. Compared to the results in Table 5, it can be deducted that domestic investors are concerned about the economic development in the target country, as in this regression the economic development can explain more of the cross-sectional variation of the foreign bias than of the domestic bias. This finding is intuitive as investors are more prone to invest in highly developed economies which realize higher GDP per capita j¬-IFEIl, have higher real GDP growth j¬-I¬l and are characterized with a high credit-worthiness indicated by high rating scores j´EG®g¬l.

The proxies for the cultural dimensions are clearly more important in the foreign bias analysis compared to the previous regression. An Z” . ´ ( of 27.20% is reported for this category. Individualism j®-°l has a coefficient of 0.0138 with a ¾-statistics of

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

77

9.8308, while the estimates for uncertainty avoidance j±E®l -0.0035 and cultural distance jF-l -0.2192 are both significant at the 1% level. It is very interesting to see that foreign investors are more concerned about cultural aspects than domestic investors, whereas the cultural distance, i.e. whether a domestic country is on average distant to other countries (in case of the domestic bias) or whether investors prefer investing in foreign markets they feel culturally close (in case of the foreign bias), is of high importance in both cases.

The domestic credit to GPD j-F´W-®Gl indicating a developed banking market is highly significant as the second variable in this category FH´´. They are able to explain 14.4% of the cross-sectional variation of the foreign bias documented in international portfolio allocation. Together with stock market and economic development as well as cultural dimensions, familiarity plays an important role for international portfolio allocation. For all explanatory variables a statistical significance at the 1% or 5% level is reported. While the dummy for common official language j\Eg¬±E¬Wl has a positive sign, there is a negative sign for the logarithm of geographic distance between home and target country j-®•GEgFWl. The proxy for information availability j®gGW´gWGl has a coefficient of 0.0262 with a ¾-ratio of 6.6194. Consistent with the earlier findings these results give evidence that countries tend to invest more in countries they share a common language with and that are geographically closer to their own country. Moreover, they are more prone to put more of their investments in countries where they have easily available investment relevant information. The whole category is able to explain 18.78% of the documented foreign bias (Z” . ´ ( of 18.43%). It is no surprise that there is also a positive and highly significant coefficient for the variable capturing investor protection j®I´HGl. The strength of investor protection is able to explain 11.56% of the cross-sectional variation of the foreign bias. A higher score in the index of the strength for investor protection thus leads to an increasing (less negative) foreign bias, as investors put more of their investments into this target country.

The results of the foreign bias regression confirm the predicted impact of capital control jWX´WW-HYl on investment behavior. The estimate of the coefficient is 0.4883 with a ¾-ratio of 2.2496. This finding suggests that countries with fewer restrictions imposed on capital flows are more attractive for foreign investors and so experience a less negative foreign bias (´2 of 9.64% and Z” . ´ ( of 9.46%).

However, when all independent variables are estimated jointly (last column), previous findings are confirmed. The explanatory variables for economic development, stock market development, and investor protection variables remain statistically sig-

78

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

nificant at the 1% level or at least at the 5% level. However, WX´WW-HY loses significance and is now only statistically significant at the 10%. With regard to the cross-cultural dimensions it is very interesting to see that ®-° and ±E® lose some explanatory power, while the measure for the cultural distance (F-) between home and target country remains significant at the 1% level. This implies that it is rather the cultural distance between home and target country that determines how much weight an investor puts in a foreign market. The characterization of the home or target country (i.e. whether they make decisions more individually or not or whether they do not feel comfortable in uncertain situations), however, does not have substantial explanatory power of investors’ portfolio allocation. As in the previous analysis, the fitted explanatory variables from the first stage regression are also employed in this case. The results indicate statistically significant regression estimates at a 1% level for the categories economic and stock market development and culture. Also strong and statistically significant impact is reported for familiarity and investor protection, while no statistical significance could be found for capital control. The results discussed in the previous two paragraphs, i.e. for the domestic and foreign bias analysis, are directly related to the findings of Portes and Rey (2000), Chan et al. (2005), and Anderson et al. (2011). Portes and Rey (2000) document for a sample of 14 developed countries between 1989 to 1996 that stock market capitalization (stock market size), a market’s openness (ratio of total trade to GDP), distance, and the transaction efficiency mainly impact equity transaction flows. Chan et al. (2005) analyze mutual fund portfolio allocations and document the

79

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

In this OLS regression results the foreign bias FB ij, the log ratio of the share of country j in the investment portfolio of the home country i (i≠j) to the world market capitalization weight of country j, as dependent variable is employed. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING), the foreign direct investments (FDI), and the trade of that (TRADE); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of domestic bias exhibited in a country on the foreign bias, the factor (1-db i ), which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level.

c (1-dbi) GDPCAP GDPG RATING R² Adj R² Fitted Values C (1-dbi) SIZE VALUE EQINDEX R² Adj R² Fitted Values C (1-dbi) EFREEDOM R² Adj R² Fitted Values C (1-dbi) IPROT R² Adj R² Fitted Values C (1-dbi) LANGUAGE DISTANCE INTERNET R² Adj R² Fitted Values

+ + +

+ + +

+

+

+ +

Regression coeff t-stat ECONOMIC DEVELOPMENT -3.8883 -10.7461 *** -1.2222 -8.7644 *** 0.0023 4.6521 *** 0.0362 2.1117 ** 0.0642 4.4422 *** 0.2891 0.2834 0.1635 3.6609 *** S TOCK MARKET DEVELOPMENT -2.1597 -17.3262 *** -1.2268 -9.1598 *** 0.0061 8.9351 *** 0.0088 4.1653 *** 0.0069 2.8297 ** 0.3965 0.3907 0.095 3.7196 *** CAPITAL CONTROL -5.1984 -15.9991 *** -1.1806 -7.8968 *** 0.4883 2.2496 ** 0.0964 0.0946 0.0667 0.6575 INVES TOR PROTECTION -2.6801 -16.2037 *** -1.172 -7.6696 *** 0.1726 8.4118 *** 0.1156 0.1137 0.2032 2.4439 ** INFORMATION AVAILABILITY -1.7932 -9.2208 *** -1.4567 -9.8279 *** 0.9174 9.4938 *** -0.0199 -2.5658 ** 0.0262 6.6194 *** 0.2758 0.272 0.2548 3.6631 ***

Joint Regression coeff t-stat -2.3905 -1.3578 0.0018 0.0265 0.0818

-4.469 -11.049 3.0928 2.6723 4.9971

*** *** *** ** ***

0.0043 0.0065 0.0008

5.5489 *** 8.8675 *** 2.9185 **

0.3654

1.3137 *

0.1137

5.8094 ***

0.5722 -0.0278 0.0279

7.1675 *** -3.8008 *** 1.7418 *

Table 6: Explaining the Foreign Bias in countries' equity portfolios

80

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

In this OLS regression results the foreign bias FB ij, the log ratio of the share of country j in the investment portfolio of the home country i (i≠j) to the world market capitalization weight of country j, as dependent variable is employed. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING), the foreign direct investments (FDI), and the trade of that (TRADE); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of domestic bias exhibited in a country on the foreign bias, the factor (1-db i ), which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level.

c (1-dbi) IDV UAI CD R² Adj R² Fitted Values C (1-dbi) DCREDIT CORR R² Adj R² Fitted Values

+ -

+ -

Regression coeff t-stat CULTURAL DIMENS IONS -2.0415 -11.9648 *** -1.1043 -7.4079 *** 0.0138 2.8308 ** -0.0035 -3.859 *** -0.2192 -5.9296 *** 0.1878 0.1843 0.1674 2.1591 ** OTHER VARIABLES -2.6542 13.549 *** -1.4124 -8.1565 *** 0.0046 8.4012 *** -1.4562 -9.568 *** 0.1436 0.1432 0.2658 2.5015 **

Joint Regression coeff t-stat

0.0009 -0.0012 -0.1551

2.5732 ** -2.3025 ** -5.6892 ***

0.0029 -1.2555

2.0587 ** -2.4124 **

0.6125 0.6110

Table 6: continued

impact of stock market size, market’s openness, and distance between home and target country on international equity flows. Anderson et al. (2011) use a similar data set of mutual funds and show that cross-cultural variables provide additional perspective on foreign asset allocation, which they put on a par with geographic distance in terms of economic importance. My study addresses a similar issue, but employs macroeconomic data to identify impacts on international portfolio holdings on a more generalized view and thus the findings are not restricted to a special type of investor. Furthermore, a longer time period is used, which allows to abstract form cyclical short-time fluctuations. However, more importantly the findings of Portes and Rey (2000), Chan et al. (2005), and Anderson et al. (2011) support and add weight on my results.

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

5.3

81

Robustness Test

5.3.1 Additional Country Controls Broadly, the results presented above are in a line with the expectations. However, I perform several tests to check for their robustness. The results are reported in Table 7. I include some additional variables in order to capture economic and stock market development, and the cultural dimension. In the category of the economic development two variables are included. As one would expect economically developed countries to be characterized by high levels of variables that capture the economic interaction with other economies, I introduce the variable G´E-W into the analysis. This measure scales the average imports and exports by GDP. Furthermore, the ratio of foreign direct investments in stocks to GDP jX-®l is included. The data is provided by the World Bank. It is feasible to expect that investors are more prone to put their wealth in economies that are characterized by a large turnover and thus may seem to be more attractive for investments. Hence, I add the turnover ratio jG±´gl to the group of stock market development. This variable is expected to have negative impact on domestic bias and a positive on the foreign bias and is defined as the ratio of the total value of shares traded divided by the average market capitalization. Also the default probability of a country’s banking system may impact investment decisions. If investors prefer investing in target markets with stable banking systems, I would expect them to invest more in countries with a high level of banking stability. Thus, I introduce the Bankscope variable ¯ − •FH´W calculated as a weighted average of the z-scores of the individual banks in a country. A banks z-scores compares its reserves (returns and capitalization) to the volatility of those returns. Hence, higher z-scores indicate high stability and consequently high safety of investments in that country (Boyd and Nicoló, 2005, Laeven and Levine 2009). With regard to the cultural aspects the distribution of emotional roles between genders could bias portfolio allocation. The survey study in Hofstede (2001) shows that the values of women are more similar across societies than men’s values and that “men’s values from one country to another contain a dimension from very assertive and competitive and maximally different from women’s values on the one side, to modest and caring and similar to women’s values on the other” (Hofstede, 2011, p. 12). Hofstede names the assertive pole ‘masculine’ and the modest or caring pole ‘feminine’. In feminine countries women and men are modest and caring in the same way, while in masculine countries women are more assertive and competitive, but not as much as the men. Hence, in masculine countries a genders’ value gap is observed.

82

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

This table shows the results of the OLS regression by employing additional variables. In the first group domestic bias dbi isused, log ratio of the actual share of foreign equity positions in international investment portfolio to the optimal value according to the world market capitalization share. The second group employs the foreign bias fb ij , the , the log ratio of the share of country j in the investment portfolio of the home country i (i≠j) to the world market capitalization weight of country j, as dependent variable. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING), the foreign direct investments (FDI), and the trade of that (TRADE); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), the growth rate of S&P Global Equity Index (%) (EQINDEX), the turnover ratio to GDP (TURNOVER), and the default probability of a countrie’s banking sector (Z-SCORE); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor Protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), the average cultural distance (CD), and the masculinity (MAS); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of domestic bias exhibited in a country on the foreign bias, the factor (1-db i ),which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level. Domestic Bias

Foreign Bias

coeff

t-stat

coeff

0.8608

0.252

-3.406 -1.2208

ECONOMIC DEVELOPMENT GDPCAP GDPG RATING FDI TRADE

0.6845 -0.0018 -0.0725 -0.0586 -0.068 -0.0272

1.7475 1.1901 -0.1592 2.1571 -2.4874 -1.2851

* *

S TOCK MARKET DEVELOPMENT SIZE VALUE EQINDEX TURNOVER Z-SCORE

0.2485 -0.007 -0.0021 0.0701 -0.001 0.0553

3.363 -2.2702 -1.9002 2.4327 -1.1995 3.4887

CAPITAL CONTROL EFREEDOM

-0.2965 -0.1586

-0.3435 -0.1174

c (1-dbi)

INVES TOR PROTECTION IPROT

t-stat -13.2909 *** -8.7237 ***

** ** *

0.102 0.0023 0.0362 0.0642 0.013 0.003

3.856 4.6521 2.1117 4.4422 1.2062 2.9224

*** *** ** *** * **

*** ** * ** * ***

0.035 0.0052 0.0081 0.0021 0.0007 -0.0724

3.072 4.1857 5.7782 5.2587 0.5737 -0.1319

*** *** *** ***

0.185 0.4883

1.754 * 2.2505 *** 1.002 8.4118 ***

0.5694 0.0199

1.666 * 1.5691 **

0.024 0.1726

INFORMATION AVAILABILITY LANGUAGE DISTANCE INTERNET

0.5647 -0.0376 0.0152 -0.0143

1.0154 * -2.5915 ** 2.282 ** -0.2043

0.367 0.9174 -0.0199 0.0262

2.984 9.4938 -2.5658 6.6194

*** *** ** ***

CULTURAL DIMENS IONS IDV UAI CD M AS

0.2404 -0.0033 0.001 0.0035 -0.0234

5.0454 *** -1.8135 * 1.7143 * 0.7916 -0.5631

0.164 0.0138 -0.0034 -0.0035 0.0218

1.978 2.7265 -2.9336 -3.859 2.9373

** ** *** *** **

OTHER VARIABLES DCREDIT CORR R² Adj R²

-0.2321 0.0016 0.0201 0.5827 0.5786

-1.9854 ** 2.3809 ** 2.0015 **

0.3517 0.0046 -1.4562 0.6413 0.6401

Table 7: Regression Results – Additional Variables

2.1279 ** 8.4012 *** -9.568 ***

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

83

A high score of this index indicates that the country is driven by competition, achievement, and success, while lower scores stand for a dominant value of life quality or caring for others. The fundamental issue here is the motivation of the people living in these countries, whether they want to be the best (masculine) or like what they do (feminine). For capturing this aspect I introduce the variable masculinity jYE•l (opposed to feminity) into the regression.

The findings of this additional analysis are almost consistent with the results. Overall, including these additional variables brings out results, which are almost in a line with the previous findings. However, judging from the ´2 , their including into the regression analysis does not add particular value to the specification. 5.3.2

Income Analysis

High and middle income countries In this section the possibility is considered that a country’s income level could affect their willingness to invest abroad and so their portfolio allocation. Therefore I employ the definition of the World Bank. Their main criterion for classification is the per capita gross national income (GNI). Based on this figure they classify economies as low income, middle income, and high income, which is set each year on July 1st and remains fixed during the World Bank’s fiscal year. The GNI per capita is displayed in U.S. dollars and is calculated using the World Bank Atlas method. Therefore, they employ the Atlas conversion factor, which is the average of a country’s exchange rate for that year and the exchange rates for the two previous years. These exchange rates are adjusted for the difference between the respective country’s inflation, and that in the G-5 countries (France, Germany, Japan, U.K., and U.S.). Since 2001 this group includes the Euro Zone, Japan, U.K. and U.S. A country’s rate of inflation is calculated by the change on its GDP deflator. These countries’ inflation rates, representing international inflation, are measured by the change in the special drawing rights (SDR) deflator “calculated as a weighted average to the countries’ […] GDP deflators in SDR terms, the weights being the amount of each country’s currency in one SDR unit” (World Bank, Website on World Bank Atlas Method).14 As both the composition of the SDR and the relative exchange rates for each currency change, weights vary over time. The SDR deflator is calculated in a first step in SDR terms and then converted to U.S. dollars employing the SDR to dollar Atlas conversion factor, which is then applied to a country’s GNI. In order to get the GNI per capita, the resulting GNI in U.S. dollars is divided by the midyear population. In case the 14

World Bank Group, Washington, DC: http://data.worldbank.org/about/country-classifications/world-bankatlas-method.

84

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

official exchange rates cannot be used, then the World Bank uses an alternative estimate of the exchange rate.15 Countries which are defined as high income countries are Australia (AUS), Belgium (BEL), Canada (CAN), Chile (CHL), Denmark (DEN), Finland (FIN), France (FRA), Germany (GER), Hong Kong (HKG), Italy (ITA), Japan (JAP), Korea (KOR), Netherlands (NLD), Norway (NOR), Russia (RUS) , Singapore (SIN), Spain (SPA), Sweden (SWE), Switzerland (SWI), United Kingdom (U.K.), and United States (U.S.), while countries with (upper and lower) middle income are Brazil (BRA), Malaysia (MYS), India (IND), South Africa (SOU), and Turkey (TUR). All values are expressed in U.S. dollars. Data is provided by the World Bank national accounts data, and OECD national Accounts data files. 0.00

10,000.00

20,000.00

30,000.00

AUS

40,000.00

50,000.00

BEL

36,735.45 5,754.55

CAN

34,919.09

CHL

7,653.64

DNK

48,078.18

FIN

38,560.91

FRA

34,615.45

GER

35,198.18

HKG

30,270.00 870.00

ITA

29,927.27

JPN

38,064.55

KOR MYS

70,000.00

33,648.18

BRA

IDA

60,000.00

17,169.09 5,983.64

NLD

39,984.55

NOR RUS

66,321.82 6,141.82

SIN SOU

30,986.36 4,786.36

SPA

25,247.27

SWE

42,315.45

SWI TUR U.K. U.S.

56,714.55 7,030.00 37,122.73 45,273.64

Figure 1: Average GNI per capita (Atlas Method) between 2001 and 2011

15

World Bank Group, Washington, DC: http://data.worldbank.org/about/country-classifications/world-bankatlas-method.

85

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

This table shows the OLS regression results by focusing on the impact of income level. In the first group domestic bias db i is used, log ratio of the actual share of foreign equity positions in international investment portfolio to the optimal value according to the world market capitalization share. The second group employs the foreign bias fbij, the log ratio of the share of country j in the investment portfolio of the home country i, (i≠j) to the world market capitalization weight of country j, as dependent variable. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING), the foreign direct investments (FDI), and the trade of that (TRADE); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor Protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of domestic bias exhibited in a country on the foreign bias, the factor (1-dbi), which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level. Domestic Bias High Income Middle Income Countries Countries coeff t-stat coeff t-stat 1.4079 1.4558 * 0.8865 0.5432

c (1-dbi)

Foreign Bias High Income Middle Income Countries Countries coeff t-stat coeff t-stat -1.7464 -3.5806 *** -1.5456 -4.5685 *** 0.8248 -6.5062 *** 0.7458 -3.5018 ***

Economic Development GDPCAP GDPG RATING

0.7565 1.9058 * -0.0058 -1.6001 * -0.0169 -0.3424 -0.0509 -2.2981 **

0.6125 1.0475 * -0.0128 -1.1999 * -0.043 -1.2368 * -0.0119 -2.1017 **

0.1925 0.002 0.0381 0.1019

2.1243 4.0503 2.4591 6.4968

** *** ** ***

0.1233 0.0012 0.0244 0.0192

S tock Market Development SIZE VALUE EQINDEX

0.4806 6.6929 *** -0.0042 -1.4369 * -0.0092 1.9428 * 0.0133 2.1704 **

0.3272 3.4814 *** -0.002 -1.6691 * -0.0015 -1.7713 * 0.0011 2.6312 **

0.2316 0.0044 0.0073 0.0052

2.5698 5.7648 9.1344 3.7802

** *** *** ***

0.2316 3.2146 *** 0.0078 4.0236 *** 0.0025 10.1344 *** 0.0481 2.8525 **

Capital Control EFREEDOM

-0.4416 -0.3602 -0.0828 -0.6954

-0.2458 -0.5578 -0.1049 -0.1564

0.1324 0.4464

0.725 2.456 **

0.1712 0.4154

0.3893 0.1155

2.8819 ** 6.0238 ***

0.2144 0.1587

0.17 3.3276 *** 0.0045 1.0302 * -0.0015 -1.3498 * -0.1433 -5.4071 ***

0.1847 0.0019 -0.0025 -0.1365

4.0054 1.012 -2.5683 -6.4071

*** * ** ***

0.2877 0.5488 -0.0197 0.0312

2.4875 7.1928 -7.5281 6.8253

** *** *** ***

Investor Protection IPROT

0.4184 0.037

Culture IDV UAI CD

0.2114 1.4075 * -0.002 -0.6241 0.0015 1.3781 * 0.0186 1.6035 *

Familiarity LANGUAGE DISTANCE INTERNET Other Variables DCREDIT CORR R² Adj R²

2.1417 ** 1.1435 **

0.2488 4.2814 *** -0.0238 -2.4905 ** 0.001 2.1468 ** -0.0106 -0.3624 *** 0.1985 0.0044 0.0257 0.4008 0.3985

2.0654 ** 0.6179 1.6513 *

0.4731 0.0218

1.2456 * 1.4921 **

0.3651 2.9876 *** -0.0067 -0.4679 0.0062 1.3476 * 0.0054 1.484 * 0.1846 -0.0007 0.0026 0.0022

3 *** -1.055 * 2.5729 ** 0.6817

0.2344 2.1041 ** 0.5488 7.1928 *** -0.009 -2.5281 ** 0.0312 2.1873 **

0.2136 0.0065 0.0133 0.4796 0.4792

2.653 ** 0.5745 1.1246 *

0.2412 0.0067 0.0054 0.5347 0.5256

2.5601 ** 3.1955 *** 3.8451 ***

0.2514 0.0052 0.0065 0.5413 0.5412

2.9874 3.2505 2.9066 2.4968

*** *** *** **

0.375 -2.0005 ** 2.0735 ** 7.1212 ***

2.7316 ** 7.5946 *** 2.8499 **

Table 8: Explaining the Domestic and Foreign Bias – Income Analysis

In order to classify income groups I follow the categorization of the World Bank: economies with an income of 1,035 U.S. dollars or less are categorized as low in-

86

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

come, economies with an income between 1,036 U.S. dollars and 4,085 U.S. dollars are in the category lower middle income, and income between 4,086 U.S. dollars and 12,615 U.S. dollars characterizes countries in the upper middle income group, while high income economies have an income of 12,616 U.S. dollars or more. In this country sample high income countries are Australia, Belgium, Canada, Chile, Denmark, Finland, France, Germany, Hong Kong, Italy, Japan, Korea, the Netherlands, Norway, Russia, Singapore, Spain, Sweden, Switzerland, U.K., and U.S. The upper middle income countries Brazil, Malaysia, South Africa, and Turkey are grouped with the lower middle income economy India in a group called middle income group. Figure 1 illustrates the average GNI of the analyzed countries between 2001 and 2011.Table 8 shows the results for analysis on foreign bias in both high income and (upper and low) middle income countries. The findings confirm our previous results. Financial Centers In the previous analyses income level and geographic sub-samples in order to control for special effects in these countries were analyzed. It was shown that the main findings of the analysis are robust. But what would happen if only countries were examined which are known as financial centers who offer investors attractive investment possibilities? Recently, the understanding of the activities of offshore financial centers (OFCs) became of increasing importance as they largely capture global financial flows. The Executive Board of the Financial Stability Forum (FSF) mandated in the year 2000 the IMF with the compliance of internationally supervisory standards. Zoromé (2007, p. 7) provides in his IMF working paper a clear definition of what constitutes an OFC and takes financial services to nonresidents into account, which is on a scale far exceeding the size of the respective economy: „An OFC is a country or jurisdiction that provides financial services16 to nonresidents on a scale that is incommensurate with the size and the financing of its domestic economy.”17 In order to control whether this fact impacts the results I follow Zoromé (2007) and employ his IMF database (Zoromé, 2007, p. 21-22). The findings confirm the previous results. Figure 2 illustrates the ratios of net exports of financial services to GDP in % for the whole sample. In the analyzed country sample Hong Kong, Singapore, Switzerland, and the U.K. are considered as financial centers and are thus examined in a separate regression analysis. Table 9 reports the results for the analysis of the fi-

16 17

Referring to the definition of financial services of the United Nations et al. (2002, paras. 3.108, 3.99). The definition and the data of Zoromé (2007) explicitly exclude insurance services as most balance of payment submissions were not able to provide information to exclude freight insurance.

87

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

nancial centers for both domestic and foreign bias. The findings show no significance difference between financial and non-financial centers. This figure illustrates the ratios of net exports of financial services to GDP in % for the whole country sample: Australia (AUS), Belgium (BEL), Brazil (BRA), Canada (CAN), Chile (CHL), Denmark (DEN), Finland (FIN), France (FRA), Germany (GER), Hong Kong (HGK), India (IDA), Italy (ITA), Japan (JAP), South Korea (KOR), Malaysia (MAL), Netherlands (NET), Norway (NOR), Russia (RUS), Singapore (SIN), South Africa (SOU), Spain (SPA), Sweden (SWE), Switzerland (SWI), Turkey (TUR), United Kindgom (U.K.), United States (U.S.). Data are provided from the IMF. 2.393

2.10 1.860

1.60

1.481

1.10

0.900

0.60 0.231 -0.076 -0.312 0.003 0.075 -0.048 -0.053 -0.092

0.052

0.10

0.141 0.098 0.006 -0.034 0.030 -0.032 -0.016 -0.008

0.033 0.087 0.000

0.135 -0.035

-0.40

Figure 2: Ratio of net exports of financial services to GDP in %

The European Monetary Union In the introduction of my thesis I outlined the recently discussed banking union in the European Monetary Union (EMU). Hence, I run a separate analysis on the EMU countries in order to detect whether the cohesions documented in the previous chapters are also valid in this sample. Countries within the EMU in the employed country sample are: Belgium, Finland, France, Germany, Italy, the Netherlands, and Spain. Table 10 reports the findings for this analysis. A statistically significant result at the 5% level could be documented for ´EG®g¬ in order to explain the domestic bias, while all variables from the matrix of the stock market development are significant at least at the 10% level. Analogue to the previous findings no statistically significance could be found for the capital control.

88

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

This table shows the OLS regression results by focusing on the impact of financial centers. In the first group domestic bias dbi is used, log ratio of the actual share of foreign equity positions in international investment portfolio to the optimal value according to the world market capitalization share. The second group employs the foreign bias fbij, the log ratio of the share of country j in the investment portfolio of the home country i, (i≠j) to the world market capitalization weight of country j, as dependent variable. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING), the foreign direct investments (FDI), and the trade of that (TRADE); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor Protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of domestic bias exhibited in a country on the foreign bias, the factor (1-dbi), which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level. Domestic Bias Foreign Bias Financial Centers Non-Financial Financial Centers Non-Financial Centers Centers coeff t-stat coeff t-stat coeff t-stat coeff t-stat c 0.431 1.659 * 0.238 1.302 * -3.865 -8.173 *** -2.821 -6.658 *** -1.514 -4.124 *** -1.013 -5.849 *** (1-dbi) Economic Development GDPCAP GDPG RATING

0.602 1.541 * -0.002 -1.041 * -0.044 -0.242 -0.004 -2.257 **

0.557 1.981 * -0.001 -1.514 * -0.091 -0.571 0 -2.658 **

0.094 0.004 0.013 0.056

3.214 3.753 2.725 4.002

*** *** ** ***

0.187 0.003 0.042 0.075

3.614 2.344 2.005 3.375

*** ** ** ***

Stock Market Development SIZE VALUE EQINDEX

0.655 3.012 *** -0.003 -2.127 ** -0.002 -2.445 ** 0.042 2.714 **

0.527 2.577 ** -0.002 -2.742 ** -0.004 -2.599 ** 0.067 2.52 **

0.033 0.005 0.01 0.006

6.101 3.579 3.754 3.124

*** *** *** ***

0.046 0.008 0.052 0.006

3.45 3.413 3.901 3.123

*** *** *** ***

Capital Control EFREEDOM

-0.318 -0.11 -0.334 -1.125 *

-0.232 -0.984 -0.435 -1.139 *

0.023 0.571

0.518 2.715 **

0.031 0.461

0.216 2.015 **

Investor Protection IPROT

-0.145 -2.482 ** 0.052 1.065 *

-0.375 -1.443 * 0.052 1.065 *

0.312 0.141

3.458 *** 2.455 **

0.411 0.155

3.659 *** 2.865 **

Culture IDV UAI CD

0.357 1.563 * -0.001 -0.455 0.007 1.847 * 0.051 2.469 **

0.543 -0.007 0.008 0.045

Familiarity LANGUAGE DISTANCE INTERNET

0.386 4.545 *** -0.036 -2.548 ** 0.005 2.655 ** -0.057 -0.686

0.413 3.578 *** -0.016 -2.745 ** 0.006 2.456 ** -0.027 -0.456

Other Variables DCREDIT CORR R² Adj R²

0.2 0.041 0.002 0.4 0.398

2.614 ** 0.013 1.1

0.21 0.501 0.003 0.456 0.451

1.986 * -0.12 1.751 * 2.005 **

2.419 ** 0.842 1.232 *

0.325 4.451 *** 0.01 2.451 ** -0.008 -3.858 *** -0.157 -3.681 ***

0.357 3.421 *** 0.239 2.046 ** -0.007 -3.005 *** -0.149 -3.985 ***

0.124 2.659 ** 0.845 4.684 *** -0.002 -2.456 ** 0.054 5.542 ***

0.155 2.442 ** 0.546 4.569 *** -0.003 -2.895 ** 0.046 4.857 ***

0.241 2.412 ** 0.018 2.518 ** -0.871 -4.515 *** 0.502 0.5

0.223 2.888 ** 0.042 3.982 *** -0.49 -3.875 *** 0.495 0.491

Table 9: Explaining Domestic and Foreign Bias – Financial Centers

89

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

This table shows the OLS regression results by focusing on the impact of the EMU countries. In the first group domestic bias dbi is used, log ratio of the actual share of foreign equity positions in international investment portfolio to the optimal value according to the world market capitalization share. The second group employs the foreign bias fbij, the log ratio of the share of country j in the investment portfolio of the home country i, (i≠j) to the world market capitalization weight of country j, as dependent variable. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING), the foreign direct investments (FDI), and the trade of that (TRADE); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor Protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of domestic bias exhibited in a country on the foreign bias, the factor (1-dbi), which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level. Domestic Bias c (1-dbi) Economic Development GDPCAP GDPG RATING S tock Market Development SIZE VALUE EQINDEX Capital Control EFREEDOM Investor Protection IPROT Familiarity LANGUAGE DISTANCE INTERNET Culture IDV UAI CD Other Variables DCREDIT CORR R² Adj R²

coeff 0.1884 0.5422 -0.0054 -0.021 -0.0867 0.2898 -0.0006 -0.0027 0.012 -0.242 -0.2925 0.6298 0.1285 0.2033 -0.0283 0.0021 -0.0319 0.6048 -0.0008 0.001 0.0492 0.2557 0.0657 0.0512 0.5948 0.5913

Foreign Bias

t-stat -3.1792 *** 1.5756 -1.0306 -0.4833 2.3377 3.4513 -1.1507 -1.8363 2.5875 -0.7523 -0.5449 2.8136 2.0302 2.6856 -2.0957 2.5512 -0.6962 3.9857 -0.3183 1.6842 2.251 2.6545 1.0748 1.5644

* * ** *** * * **

** ** ** ** ** *** * ** ** * *

coeff -1.916 -0.5874 0.2655 0.0071 0.0742 0.1092 0.4516 0.0082 0.0097 0.0421 0.1658 0.5278 0.2895 0.1017 0.1457 0.8423 -0.0001 0.0313 0.2012 0.0188 -0.0009 -0.2142 0.1985 0.0868 0.0081 0.415 0.406

t-stat -2.2542 -4.1918 2.6546 3.9753 2.8909 5.2368 3.8457 3.3067 9.633 3.9636 0.1546 1.6874 2.469 3.1651 2.5457 5.0098 -2.1022 4.8826 3.568 1.3406 -1.5485 -3.556 2.5469 3.1076 3.4155

Table 10: Explaining Domestic and Foreign Bias – the European Monetary Union

** *** ** *** ** *** *** *** *** *** * ** *** ** *** ** *** *** * * *** ** *** ***

90

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

This table shows the OLS regression results by excluding the U.K. and the U.S. In the first group domestic bias dbi is used, log ratio of the actual share of foreign equity positions in international investment portfolio to the optimal value according to the world market capitalization share. The second group employs the foreign bias fbij, the log ratio of the share of country j in the investment portfolio of the home country i, (i≠j) to the world market capitalization weight of country j, as dependent variable. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING), the foreign direct investments (FDI), and the trade of that (TRADE); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor Protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of domestic bias exhibited in a country on the foreign bias, the factor (1-dbi), which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level.

c (1-dbi)

Domestic Bias coeff t-stat 0.976 1.059 *

Foreign Bias coeff t-stat -1.242 -3.594 *** -1.259 -8.346 ***

Economic Development GDPCAP GDPG RATING

0.705 -0.003 -0.071 -0.019

1.196 -1.786 -1.467 -2.747

* * * **

0.146 0.002 0.038 0.06

2.471 4.809 2.106 4.17

** *** ** ***

S tock Market Development SIZE VALUE EQINDEX

0.313 -0.002 -0.002 0.018

3.12 -1.144 -1.082 2.489

*** * * **

0.085 0.005 0.01 0.005

3.103 3.619 8.914 2.951

*** *** *** **

Capital Control EFREEDOM

-0.109 -0.164

-0.286 -0.535

0.314 0.479

1.175 * 2.291 **

-0.36 0.026

-2.099 ** 2.373 **

0.102 0.173

2.038 ** 8.388 ***

Culture IDV UAI CD

0.349 -0.004 0.002 0.035

1.561 * -0.408 1.664 * 1.489 *

0.169 0.004 -0.001 -0.269

4.517 1.636 -1.73 -7.353

*** * * ***

Familiarity LANGUAGE DISTANCE INTERNET

0.22 -0.038 0.013 -0.016

4.597 *** -2.397 ** 2.549 ** -0.766

0.155 0.911 -0.001 0.03

2.499 4.399 -3.099 2.543

** *** *** **

Other Variable DCREDIT CORR

0.246 0.001 0.022

2.505 ** 0.734 1.09 *

0.265 0.005 -0.879

2.71 ** 3.427 *** -4.327 ***

R² Adj R²

0.416 0.413

Investor Protection IPROT

0.55 0.549

Table 11: Explaining Domestic and Foreign Bias – Excluding the U.K. and the U.S.

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

91

Judging from the ´² for the domestic bias analysis the explanatory power of the regression is higher and captures 59.48% of the cross-sectional variation of the domestic bias. On the other side the respective values for the foreign bias analysis only reach a level of 41.50% (Z” . ´ ( of 40.60%).

Excluding UK and US Since the U.K. and the U.S. dominate international equity portfolio investments, I am particularly concerned that the results could be driven by the inclusion of these large economies. In order to check the robustness of the country sample, I conduct another test excluding both countries and re-estimate the model.

In the domestic bias regression I am now able to document statistical significance for all the three variables at the 1% level for the group of economic development variables, even realizing a higher Z” . ´ ( compared to the main regression result. With regard to the other categories or the joint regression nearly unchanged results except for the cultural dimensions in the joint regression are obtained. With regard to the latter category the only variable with statistical significance is F-. The findings for the foreign bias regression are also robust excluding those two large economies. Like in the domestic bias analysis I have the strongest changes within the cultural dimensions, where only F- remains statistically significant at the 1% level in the joint regression. Table 11 indicates the results. 5.3.3

Geographic Analysis

In this subsection I run an analysis to test whether the results are robust against geographic regions. Therefore again the World Bank Database is employed. Table 12 indicates their country classification. In this analysis data from South Africa and India are not considered, as they could not be categorized into one of the three groups. Table 13 shows the results. Starting from the first group Europe & Central Asia highly significant results for all the variables are obtained and document a ´2 of 55.3% (Z” . ´ ( of 54.8%). For the sample of the Americas countries this values are slightly lower (´2 of 53.8% and Z” . ´ ( of 53.4%). In the group of the six countries in the East Asia & Pacific sample the highest value with a ´ ( of 58.5% is indicated (Z” . ´ ( of 55.6%). Overall, the findings in the geographical analysis are consistent with the previous results on the foreign bias in international portfolio allocation and so underline their robustness.

92

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

This table shows the geographical classification of the country sample by employing the OECD classification. Countries from North and Latin America are grouped into the category Americas. As South Africa and India cannot be grouped to the other countries geographically they are excluded in this analysis. Country

Geographical Classification

Australia Belgium Brazil Canada Chile Denmark Finland France Germany Hong Kong Italy Japan Korea, Rep. M alaysia Netherlands Norway Russia Singapore Spain Sweden Switzerland Turkey United Kingdom United States

East Asia & Pacific Europe & Central Asia Americas Americas Americas Europe & Central Asia Europe & Central Asia Europe & Central Asia Europe & Central Asia East Asia & Pacific Europe & Central Asia East Asia & Pacific East Asia & Pacific East Asia & Pacific Europe & Central Asia Europe & Central Asia Europe & Central Asia East Asia & Pacific Europe & Central Asia Europe & Central Asia Europe & Central Asia Europe & Central Asia Europe & Central Asia Americas

Table 12: Geographical Classification of country sample

5.3.4

Financial Crisis

In addition, I test in a next step whether our results are robust with respect to the sample period which includes the financial crisis. Even though the most severe period of the financial crisis began with the collapse of Lehman Brothers in September 2008, I follow the definition of the National Bureau of Economic Research and see the start of the first stage of the financial crisis already in the fourth quarter of 2007. According to their opinion this recession lasted 18 month and reaches its trough in June 2009. However, it is not expected that the period of the financial crisis affects cross-country portfolio allocation. Thus, the years 2007 to 2009 are excluded from the analysis and the model is re-estimated. Table 14 summarizes the results and confirms this intuition. With regard to both domestic and foreign bias, results do not change significantly. There is still statistical significance at the 5% level for ´EG®g¬, W²®g-W³, ®I´HG, \Eg¬±E¬W, and -®•GEgFW.

93

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

This table shows the OLS regression results by building geographic sub-samples. In the first group domestic bias db i is used, log ratio of the actual share of foreign equity positions in international investment portfolio to the optimal value according to the world market capitalization share. The second group employs the foreign bias fb ij , the log ratio of the share of country j in the investment portfolio of the home country i ,(i≠j) to the world market capitalization weight of country i, as dependent variable. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of domestic bias exhibited in a country on the foreign bias, the factor (1-db i ), which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level. Domestic Bias Europe & Central Asia

Americas

East Asia & Pacific

coeff

t-stat

coeff

t-stat

coeff

0.544

1.355 *

1.535

0.816

2.19

Economic Development GDPCAP GDPG RATING

0.692 -0.001 -0.084 -0.008

1.825 * -1.357 * -0.718 -2.198 **

0.847 -0.003 -0.059 -0.027

1.874 * -2.391 ** -0.531 -2.269 **

0.864 -0.003 -0.059 -0.054

0.862 -1.382 * -1.063 * -1.968 **

S tock Market Development SIZE VALUE EQINDEX

0.616 -0.002 -0.002 0.015

3.996 -2.315 -2.141 2.35

0.834 -0.004 -0.001 0.011

3.833 1.822 -2.478 2.075

0.802 -0.002 -0.003 0.01

3.798 -2.062 -1.086 2.907

Capital Control EFREEDOM

-0.226 -0.236

-0.366 -1.007 *

0.364 -0.128

0.364 -0.637

0.49 -0.153

0.488 -0.133

Investor Protection IPROT

-0.377 0.052

-1.43 * 1.065 *

0.393 0.068

1.391 * 2.396 **

0.268 0.028

1.064 * 1.199 *

Culture IDV UAI CD

0.424 -0.004 0.002 0.03

1.5 * -0.834 1.016 * 2.846 ***

0.227 -0.001 0.001 0.017

1.22 * -0.27 1.798 * 2.862 **

0.192 -0.002 0.006 0.022

1.191 * 0.24 1.785 * 2.834 **

Familiarity LANGUAGE DISTANCE INTERNET

0.442 -0.041 0.003 -0.027

3.28 *** -2.338 ** 2.193 ** -0.969

0.886 -0.052 0.007 -0.003

2.031 ** -2.811 ** 4.142 *** -0.487

0.64 -0.108 0.003 -0.005

3.638 *** -2.349 ** 2.595 ** -0.945

Other variables DCREDIT CORR

0.176 0.053 0.033

2.315 ** 0.523 1.394 *

0.212 0.08 0.011

2.459 ** 0.615 1.234 *

0.188 0.24 0.032

2.012 ** 0.218 1.275 *

R² Adj R²

0.451 0.45

c

*** ** ** **

0.466 0.46

*** * ** **

0.387 0.376

Table 13: Explaining Domestic and Foreign Bias– Geographic Analysis

t-stat 1.4 *

*** ** * ***

94

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

This table shows the OLS regression results by building geographic sub-samples. In the first group domestic bias db i is used, log ratio of the actual share of foreign equity positions in international investment portfolio to the optimal value according to the world market capitalization share. The second group employs the foreign bias fb ij , the log ratio of the share of country j in the investment portfolio of the home country i ,(i≠j) to the world market capitalization weight of country i, as dependent variable. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor protection: strength of investor protection index (IPROT), (v) Information Availability: average common language dummy (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dimensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of domestic bias exhibited in a country on the foreign bias, the factor (1-db i ), which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level. Foreign Bias Europe & Central Asia coeff c (1-dbi)

-3.45 -1.875

t-stat

Americas coeff

-14.256 *** -4.185 ***

-2.472 -1.309

East Asia & Pacific

t-stat

coeff

-3.282 *** -3.532 ***

-2.36 -1.792

t-stat -3.31 *** -3.433 ***

Economic Development GDPCAP GDPG RATING

0.146 0.001 0.045 0.084

3.745 2.344 2.507 3.539

*** ** ** ***

0.141 0.002 0.104 0.095

3.127 2.179 2.611 2.912

*** ** ** ***

0.133 0.003 0.06 0.129

3.851 3.162 2.439 4.041

*** *** ** ***

S tock Market Development SIZE VALUE EQINDEX

0.02 0.009 0.012 0.005

3.611 5.051 3.939 3.306

*** *** *** ***

0.055 0.004 0.007 0.009

3.017 4.791 5.973 3.678

*** *** *** ***

0.058 0.004 0.008 0.008

3.46 7.871 5.689 2.493

*** *** *** **

Capital Control EFREEDOM

0.039 0.477

0.379 *** 2.319 **

0.045 0.341

0.542 2.267 **

0.031 0.352

0.69 2.047 **

Investor Protection IPROT

0.39 0.147

3.773 *** 2.771 **

0.254 0.124

2.124 ** 2.358 **

0.329 0.12

2.462 ** 2.798 **

Culture IDV UAI CD

0.291 0.018 -0.002 -0.131

3.869 2.719 -2.875 -4.83

*** ** *** ***

0.155 0.002 -0.001 -0.085

3.022 2.953 2.789 -3.891

*** ** *** ***

0.302 0.004 -0.001 -0.279

3.546 2.362 2.068 -4.477

*** ** ** ***

Familiarity LANGUAGE DISTANCE INTERNET

0.162 0.754 -0.004 0.029

2.177 6.351 -2.496 6.042

** *** ** ***

0.154 0.794 -0.002 0.041

2.579 2.917 -2.232 2.759

** *** ** ***

0.177 0.385 -0.001 0.04

2.458 3.159 -2.35 4.269

** *** ** ***

2.789 ** 3.072 *** 3.715 ***

0.295 0.01 0.007

2.416 ** 3.942 *** 3.012 ***

0.287 0.007 0.004

Other variables DCREDIT CORR

0.301 0.013 0.004

R² Adj R²

0.553 0.548

0.538 0.534

Table 13: continued

0.585 0.556

2.549 ** 2.796 ** 3.746 ***

95

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

This table shows the OLS regression results excluding the period of the financial crisis between 2007 and 2009. In the first group domestic bias dbi is used, log ratio of the actual share of foreign equity positions in international investm ent portfolio to the optimal value according to the world market capitalization share. The second group em ploys the foreign bias fbij , the log ratio of the share of country j in the investm ent portfolio of the hom e country i ,(i≠j) to the world market capitalization weight of country i, as dependent variable. There are seven categories of explanatory variables: (i) Economic development: log GDP per capita (GDPCAP), real GDP growth (GDPG), and country credit rating (RATING); (ii) Stock market development: stock market capitalization scaled by GDP (SIZE), stocks traded values to GDP (VALUE), and the growth rate of S&P Global Equity Index (%) (EQINDEX); (iii) Capital Control: capital control restrictions (EFREEDOM), (iv) Investor protection: strength of investor protection index (IPROT), (v) Information Availability: average com mon language dumm y (DUMLANGUAGE), average log geographical distance (DIST), and internet users per 100 people (INTERNET), (vi) Cultural dimensions: Hofstede’s dim ensions like individualism (IDV), uncertainty avoidance (UAI), and the average cultural distance (CD); (vii) Other Variables: domestic credit provided by banking sector to GDP (%) (DCREDIT), and the correlation between the markets indices of the home and the target country (CORR). Controlling for automatic impact of dom estic bias exhibited in a country on the foreign bias, the factor (1-dbi), which is the actual share of foreign equity holdings to the optimal value is included. The t-ratios are based in standard errors adjusted for heteroskedasticity using the method introduced by White (1980); *, **, *** are the significance levels on the 10%, 5%, and 1% level.

Domestic Bias

Foreign Bias

coeff

t-stat

0.259

1.494 *

-3.4 -1.239

-4.078 *** -3.686 ***

Economic Development GDPCAP

0.542 -0.003

1.281 * -1.575 *

0.231 0.002

3.087 *** 2.264 **

GDPG RATING

-0.018 -0.028

-0.41 -2.028 **

0.032 0.072

2.185 ** 4.677 ***

S tock Market Development SIZE

0.214 -0.002

3.899 *** -2.681 **

0.13 0.005

2.872 ** 3.668 ***

VALUE EQINDEX

-0.001 0.015

-2.261 ** 2.144 **

0.007 0.007

3.797 *** 2.17 **

Capital Control EFREEDOM

-0.751 -0.154

-0.589 -2.356 **

0.329 0.372

0.32 2.17 **

Investor Protection IPROT

-0.315 0.018

-2.593 ** 1.233 *

0.339 0.101

3.878 *** 2.833 **

Culture IDV

0.397 -0.004

1.131 * -0.172

0.14 0.01

3.822 *** 2.492 **

c (1-dbi)

UAI CD

coeff

t-stat

0.002 0.017

1.955 * 2.229 **

-0.006 -0.286

-4.407 *** -3.331 ***

Familiarity LANGUAGE

0.172 -0.034

3.487 *** -2.449 **

0.189 0.921

2.645 ** 3.427 ***

DISTANCE INTERNET

0 -0.014

2.644 ** -0.787

-0.047 0.034

-2.305 ** 4.038 ***

0.202

2.651 **

0.288

2.81 **

0 0.01

0.441 1.541 *

0.004 -0.457

3.752 *** -4.856 ***

Other Variables DCREDIT CORR R² Adj R²

0.482

0.506

0.48

0.504

Table 14: Explaining Domestic and Foreign Bias – Excluding the Financial Crisis

96

6.

CULTURAL INFLUENCES ON DOMESTIC AND FOREIGN BIAS IN INTERNATIONAL ASSET ALLOCATION

Conclusion

This part of my thesis investigates on the equity home bias puzzle and therefore analyzes its components, the domestic and the foreign bias. The empirical analysis is based on the portfolio holdings of 26 home countries distributed across 38 target countries in the time period between 2001 and 2011 and mainly builds on the Coordinated Portfolio Investment Survey of the International Monetary Fund. More specifically, I ask which frictions cause domestic (overweighting of the domestic market) and foreign biases (over- or underweighting of the foreign country) in countries’ equity portfolios. In contrast to earlier studies this analysis is not limited to a special type of investor, but detects possible determinants for the cross-sectional variation of the biases across countries on a macroeconomic view. I provide robust evidence for home bias in all of our sample countries, and more interestingly, it is documented that there is a substantial variation between the extent of the domestic and the foreign bias across countries. This evidence encourages me to detect whether this finding can be explained by common factors. For this purpose I employ familiarity and cross-cultural variables which I expect to provide an additional perspective on both domestic and foreign bias. Hence, it is shown that dimensions of cultural differences among countries help explain cross-sectional variations in international portfolio allocation. My results indicate that familiarity and information availability exhibit significant, but asymmetric, impact on domestic and foreign bias, so that if a home country is more remote from the rest of the world and does not share the language with the other country, then foreign investors put less of their investments in this market, while domestic investors invest more. Having a look into the category of cultural dimensions shows that it is not necessarily the culture of the target country that impacts investors’ over- or underweighting, but it is mainly due to cultural distance between the home and target countries that determines the portfolio weight in the investing country. Furthermore I integrate categories of predetermined variables, which reach from economic and stock market variables to capital control and investor protection. The finding shows that economic and stock market development has significant, but asymmetric, impact on the two analyzed biases in international investment portfolio allocation. The findings suggest that country-specific, survey-based measures of cultural aspects in economies can help explain deviations in portfolio allocations and add significant value to traditional approaches in finance literature.

97

C. Is increasing Financial Integration related to improved International Risk Sharing? 1.

Introduction

This third part of my thesis puts the home bias puzzle in a macroeconomic context. The growing financial integration often discussed in academic literature is expected to lead to better international risk sharing. Thus, I try to shed light on the long-term co-movement of these two phenomena. Therefore, I first measure both the deviations from the international CAPM portfolio allocation, and the deviations from perfect risk sharing allocation, called unshared consumption risk. The data sample covers the time period between 1988 and 2010. Analyzing such a relatively long sample period allows me to better compare the evolution of both phenomena including the globalization period typically allocated in the 1990s. Furthermore, I examine the driving force behind the evolution of the international risk sharing and empirically link the risk sharing discussion to the cross-holdings of international equity holdings. It is important to examine the link between these two phenomena, i.e. international risk sharing and home bias, and financial integration on long-term data as a long sample period may be able to better capture the amount of potential welfare benefits from globalization on financial markets (see Lucas, 1987). In case of permanent idiosyncratic macroeconomic shocks, benefits from consumption smoothing are expected to be huge. I will analyze empirically to what extent international risk sharing has actually increased during the considered time period. Furthermore it is important to study the linkage of those phenomena, as earlier studies found it hard to detect significant improvements in consumption risk sharing associated to the growth in international asset positions. For this purpose, I explicitly condition on international asset holdings including data. In line with Artis and Hofmann (2012) I argue that focusing on the long-term allows me to detect the increase in international risk sharing better. Again, I expect that focusing on a longer term allows individuating the impact of the risk sharing channels more sharply. As long as macroeconomic shocks are not persistent, countries are indifferent whether they insure their consumption through international capital income flows or through consumption by savings and dis-savings.

98

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

For the theoretical background please consider these identities: ¬g® = ¬-I + nÁ •Á + n •Â , F = ¬g® − ¬g•

(18)

(19)

where ¬g® stands for Gross National Income (later denoted as income, or ®gF), ¬-I is the Gross National Product (later denoted as output, or Ã), •J is the stock of foreign assets owned by domestic investors, nJ is the rate of return of these assets, and •X and nX is the stock of national assets owned by foreign investors and their return, respectively. F stands for the total final consumption, including both private and government consumption and is defined as the difference between the ¬g® and the Gross National Saving (¬g•). Both phenomena, the international risk sharing and the home bias, have been discussed quite separately in literature (Lewis, 1999). Home bias and international risk sharing do not need to be closed phenomena: if agents do not smooth their income streams through the cross-ownership of assets they can smooth consumption through inter-temporal asset trading, i.e. through savings and dis-savings. By the logic of Permanent Income Theory, this behavior may be optimal if macroeconomic shocks are not permanent. Another case where full international equity portfolio diversification will not lead to smooth income flows is when the total equity portfolio is small relative to GDP. This study is directly related to recent studies by Lane and Milesi-Ferretti (2007), Sørensen et al. (2007), and Artis and Hoffmann (2012). Lane and MilesiFerretti (2007) show a very strong increase in international asset holdings during the 1990s. Sørensen et al. (2007) investigate whether economies with comparatively high shares of foreign assets tend to realize better international factor income flows. Artis and Hoffmann (2012) follow Sørensen et al. (2007) and employ long-run level data as they expect them to give better insights on the linkage of increasing risk sharing and higher international asset positions. My study goes in a line with both approaches and use long-run growth rates to shed light on the effect of financial globalization. Matsumoto et al. (2009) analyze long-term volatility on consumption shares for measuring risk sharing. Both studies document an increase in international consumption risk sharing. But building on consumption volatility confronts them with limitations: consumption volatility does not permit them to study the roles of the risk sharing channels and consequently they cannot detect how increasing international asset holdings affect long-term risk sharing. Studying level data as Artis and Hoff-

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

99

mann (2012) poses the problem of non-stationarity, including the possibility that the data is co-integrated. From theoretical point of view it should not be possible to get consumption risk insurance against macroeconomic permanent idiosyncratic shocks through intertemporal asset trade (borrowing and lending), but only by de-coupling consumption and income (international diversification). This linkage is notably growing since financial globalization typically allocated in the 1990s.18 The main contribution of this analysis is that, employing the risk sharing channel framework, the main driver for the increase in international risk sharing is the rising importance of international capital income flows across all analyzed OECD countries. As far as I know there is no other study that examines the effect of financial globalization on international risk sharing by explicitly condition on the equity home bias proxying financial integration in the income channel regression of risk sharing. This work empirically provides evidence on the increasing importance of international income flows for international consumption risk sharing and adduces the missing link between this increasing role of international capital flows and the decreasing equity home bias by explicitly condition on this diversification measure.

2.

Data

I examine the patterns of international capital income flows and international risk sharing on the long-run. Therefore I analyze the development of the fraction of foreign portfolio investments, GNI, final consumption and GDP for a group of 21 OECD countries on annual frequency during the sample period between 1988 and 2010. I focus on these countries due to data availability and do not consider nonindustrialized countries. Several empirical papers document that the extent of idiosyncratic risk shared is higher for industrialized countries than for emerging markets (among others see Kose et al., 2007). As my focus is not on the differences in the amount of risk sharing among different groups of countries but on the documentation of the evolution of risk sharing and equity home bias over time, I only consider a group of industrialized countries. The selected countries are: Austria, Belgium, Canada, Chile, Denmark, Finland, France, Germany, Greece, Israel, Italy, Japan, Korea, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, U.K., and the U.S. 18

For theoretical models see Crucini (1999), Heathcote and Violante (2007), Artis and Hoffmann (2006), and Artis and Hoffmann (2008).

100

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

To measure domestic overweighting in international equity portfolio data from the International Monetary Fund (IMF), precisely from the Coordinated Portfolio Investment Surveys (CPIS), is employed. These surveys were conducted for investor countries using consistent guidelines for quantifying equity and bond holdings across countries. High quality data on international asset positions are hard to come by. As IMF surveys seem to contain the best available, I use this high quality data source which is also standard in this branch of research. As CPIS data are only available in 1997 and from 2001 on, I further use the Lane and Milesi-Ferretti (2007) database to obtain a complete time series, which allows capturing three economic crises and the development of the equity home bias on a low-frequency. The sum of stock market capitalization and foreign equity held by a country minus the amount of this country’s equity held by foreigners is the size of the total equity portfolio of that country. Foreign owned equity is taken by CPIS, whose surveys focus on assets and not on liabilities. Hence, they are likely to capture some amount of noise.19 Beside the shortcoming described in part two of my thesis, using IMF data leads to the problem of incomplete reporting and the difficulty of tracking the increasing complexity of international financial transactions. This point is particularly acute for economies in the Middle East, sub-Saharan Africa, and small financial centers. However, my focus is on industrialized countries, so that this work is not exposed to this shortfall. The second problem arises from growing complexity of financial instruments and the financial structure of companies, which make the measurement of external positions more and more complicated. The growth in cross-border transactions may overstate the amount of financial interdependence in some cases. For instance, a U.S. financial institution sets up a mutual fund in an offshore center. Say, this center in turn buys U.S. assets, whose shares are purchased by euro-area residents. Rather than simply measuring equity inflows in the U.S. from Europe, the data set will register an equity inflow in the offshore center from the euro-area. Respectively, the data will record a corresponding outflow from the offshore center to the U.S. These problems have also to be taken into account for the interpretation of the data, a global perspective is crucial for understanding recent trends of the integration in financial markets and international borrowing and lending. For estimating international risk sharing I need the macroeconomic variables: GDP, GNI, final consumption, consumer prices (CPI), and population which are all provided from the World Bank, precisely from the World Development Indicators. All variables are in per capita terms, deflated by the price index for personal consumption expenditure in constant (2005) international $. Per capita values of real GDP, GNI, 19

As data from CPIS do not allow identifying exchange listed equity, they are not quite consistent with data from stock market capitalizations.

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

101

and final consumption are deflated by the implicit private consumption deflator of the respective country. Building on that, I calculate per capita growth rates. As my focus is to know how much the purchasing value of GDP gets insured internationally, I do not use quantity indices. The Purchasing Power Parity (PPP) adjusted aggregate GDP growth rates are calculated as follows: For each country GDP is deflated with the respective price index normalized in 2005. In a next step using U.S. $ exchange rates these results are translated in PPP-adjusted values, which are then aggregated to OECD-wide real GDP. The stock market capitalization of a country is defined as the value of publicly traded equity listed on the stock market.

3.

Theoretical Background and Placement and Literature

3.1

International Portfolio Holdings and the Equity Home Bias

I introduce the method for calculating the equity home bias in international portfolio holdings already in section 2.3 of part one. Recall, employing the international CAPM as benchmark model every country should be weighted according to its market capitalization in the optimum. The equity home bias, thus is defined as one minus the actual weight of foreign equity to the theoretically optimal value. To ensure that the widely documented home bias is also present in my sample, I calculate the home bias for the 21 countries for the years 1990, 2000 and 2010. Table 15 illustrates the results. Except for Canada, Israel, and Norway, in each country exhibits a clear decrease of the equity home bias. For all the analysed countries an average decrease from 1990 to 2010 of 37.01% is documented.20 The value of Austria and Finland even gets reduced to 0.537 and 0.0864 in 2010. On average, the values decreased from 0.8408 in 1990 to 0.7388 in 2000 to 0.5916 in 2010. The last two rows show the descriptive statistics for the robustness tests conducted in every analysis in the subsequently sections. The impact of the home bias in equity portfolio is examined separately for small but financially open economies (SOE) such as Austria, Belgium, Denmark, the Netherlands, Sweden, and Switzerland in this sample. The last row shows the domestic overweighting in equity portfolios for the OECD sample excluding these SOE. The decreasing amount of home bias in the analyzed portfolio is consistent across all country sub-samples. 20

This value results by excluding Canada, Israel and Norway. When they are included the Equity Home Bias compared to 1990 decreases by 28%.

102

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

Figure 3 illustrates the evolution of the home bias for the whole sample period between 1988 and 2010.It shows that the domestic overweighting is more pronounced when the SOE countries are excluded. This finding is intuitive as the SOE are characterized as financially open economies and, thus exhibit a less than average amount of home bias in their portfolio holdings. This table shows the index of the equity hom e bias for the sample countries for the years 1990, 2000, and 2010. The home bias in country’s equity portfolio is calculated as follows: 1- ACT/OPT. ACT is equal to the ratio of total foreign equity holdings of country i to the total equity portfolio, where the total equity portfolio is equal to the stock m arket capitalization plus the foreign equity held m inus the am ount of country i’s equity held by foreigners. OPT is the ratio of the stock m arket capitalization of country i to the stock m arket capitalization of the world (employing the CAPM as benchm ark m odel). Data Source: Foreign Equity Holdings. Domestic Equity held by Foreigners are from the IMF and from the database of Lane and MilesiFerretti (2006), while the stock m arket capitalization is from the World Bank.

Austria Belgium Canada Chile Denmark Finland France Germany Greece Israel Italy Japan Korea Netherlands Norway Portugal Spain Sweden Switzerland United Kingdom United States Mean all Mean S OE Mean all w/o S OE

AUT BEL CAN CHL DEN FIN FRA GER GRE ISR ITA JAP KOR NDL NOR POR SPA SWE SWI U.K. U.S.

1990 0.7388 0.6359 0.8712 0.973 0.8863 0.9902 0.8611 0.8536 0.9766 0.5862 0.9766 0.9684 0.9963 0.6537 0.5862 0.9152 0.9626 0.897 0.655 0.7692 0.9028

2000 0.1946 0.5719 0.7376 0.8593 0.5897 0.8133 0.8095 0.6132 0.976 0.8088 0.976 0.8987 0.9664 0.5239 0.8088 0.8455 0.7998 0.6692 0.5956 0.6852 0.772

2010 0.0537 0.4635 0.883 0.7704 0.5111 0.0864 0.6116 0.5194 0.6971 0.8077 0.6971 0.8069 0.8981 0.2475 0.8077 0.3504 0.8765 0.5602 0.5483 0.5588 0.6675

0.8408 0.7218 0.9002

0.7388 0.5648 0.8258

0.5916 0.456 0.6594

Table 15: Home Bias in international equity portfolios in 1990, 2000, and 2010

Furthermore, data from the SOE fluctuate much more than from the other sample countries. An interesting pattern of this figure is that the evolution of the home bias across these country sub-samples is quite similar. It can be seen that in 1991 the first local minimum is reached, after that the amount of home bias increases slightly. For the SOE the decrease from 1995 to 2004 is more pronounced but similar to the other samples. In 2005 all three groups touch a local minimum and increase or remain stable in the following two years. From 2007 on a severe drop in the domestic over-

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

103

weighting in portfolio allocation is observed until the minimum jet documented is reached in 2009. In the latter year the OECD countries exhibit a home bias of 0.5407, the SOE of 0.3376 and the OECD without the SOE of 0.6220. A very thrilling issue is that after this year the value across both sample and sub-samples increases with the highest annual growth during the sample period. I detect a growth rate of 9.401% for the whole OCED sample, while it is as one would expect more pronounced for the SOE (17.714%) and less distinct for the OECD excluding the SOE (7.597%). This figure shows the evolution of the home bias in the analyzed equity portfolios during the sample period between 1988 and 2010. The index of the equity home bias is measured as one minus the ratio of actual fraction of foreign equity holdings in the equity portfolio to the optimal value. As benchmark model the international CAPM is employed, stating that every country should be weighted according to the fraction of its market capitalization on the world market capitalization. The solid line indicates the equity home bias for the whole sample of 21 OECD countries; the dotted line shows the values for the small but financially open economies (Austria, Belgium, Denmark, the Netherlands, Sweden, and Switzerland); the line with the crosses captures the value for the OECD sample excluding the SOE. Data source: IMF and Lane and Milesi-Ferretti (2006), and World Bank.

1.00 0.90 0.80 0.70 0.60 0.50 0.40 0.30 1990

1995 MEAN

2000 Mean SOE

2005

2010

Mean w/o SOE

Figure 3: Evolution of the Home Bias in international equity portfolios 1988 – 2010

104 3.2

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

International Risk Sharing

3.2.1 Theory The basic theory of international risk sharing goes back to Obstfeld and Rogoff (1996). Here I just outline the basic ideas for endowment economies with one homogeneous tradable good.

In period ¾ per capita output of country is an exogenous random variable with a commonly known probability distribution.21 In each country there is a representative consumer who is risk averse and maximizes his expected utility. Consumers within each country have identical Constant Relative Risk Aversion (CRRA) utility functions and access to a complete set of Arrow-Debreu markets for contingent claims. Optimal consumption then satisfies the full risk sharing relation F¾ = · FÄ ¾ , where · Ä is a country specific constant, F¾ is world per capita consumption, and F¾ is country per capita consumption in period ¾. A testable implication is that consumption growth rates are identical for all countries, i.e. ∆ log F•A = } + q∆Æ«ÇF•È + Ž•A ,

(20)

where } is a constant and Ž¾ is an error term due to taste shocks or noise. After controlling for world consumption growth j∆Æ«ÇF•È l, the consumption growth rate of a country j∆Æ«ÇF•A l should not be a function of countries output growth. Obstfeld (1994) and Canova and Ravn (1996) conducted regression based tests for full risk sharing at country level.22 If consumption growth rates are identical in all countries, there would be perfect consumption risk sharing (or full consumption smoothing). Market equilibrium is then defined as a state where each country consumes a constant country-specific fraction of the world output.23 This is a market equilibrium, in which countries with a more stable country-specific output than world output will get compensated for accepting higher variance in consumption and therefore take on more risk by being allocated a 21

22

Balli et al. (2011, p.4) show that non-separabilities in the utility function between consumption and leisure or non-tradable output are not considered. For these issues see Canova and Ravn (1996).

First tests of full risk sharing were run by Cochrance (1991), Mace (1991), and Townsend (1994) using individual-level data. Backus et al. (1992), Baxter and Crucini (1995), and Stockman and Tesar (1995) in International Real Business Cycle literature examined notably the prediction that the correlation of consumption across countries should be equal to unity. 23 In this setting world output is also world consumption. For further details see Obstfeld and Rogoff (1996) who show that under the described assumptions (complete Arrow-Debreau markets and identical Constant Relative Risk Aversion utility functions), across all consumers the rate of consumption growth and therefore also across countries should be identical.

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

105

larger average share of world output. Vice versa, countries with a less stable output than world output take on less risk and as consequence have a minor average share of world output (Sørensen et al., 2007, p. 594). Asdrubali et al. (1996) identified channels through which countries can share their risk. This approach is discussed and examined in detail in section 4. Artis and Hoffmann (2006) developed a general framework that allows them to study the link between consumption risk sharing, home bias in international portfolio allocation, and net capital income flows. They use this framework to motivate their alternative way of measuring consumption risk sharing based on relative log-levels of consumption and output. They assume that in each period the representative consumer in country consumes a fraction 0 < b•A < 1 of his income: F•A = Ê•A ®gF•A

(21)

where F and ®gF denote per capita values of consumption and income respectively.24 The parameter ʾ captures an array of country-specific effects, e.g. the rates of return on country ’s wealth. According to the definition of income in national accounts, the income of country is equal to the sum of its output and its net claims to output produced in the rest of the world. ®gF¾ = þ + gX®¾

(22)

®gF¾ = ËÃ∗¾ + j1 − Ëlþ

(23)

where gX®¾ is net factor income from abroad, i.e. the net claims on flows of foreign output of country . In order to link international income flows to the structure of the countries’ asset portfolio in a tractable manner they build on Crucini (1999) and assume that countries trade perpetual claims to their respective output streams. In this model, every country allocates its wealth between a claim to domestic assets and a world mutual fund of foreign securities.25 As income constitutes the dividend from wealth, per capita income must be the weighted average of dividends paid on domestic and foreign assets. The securities dividends are just foreign and domestic output, so that per capita income in country is

This formulation is consistent with permanent income models, where the effect of discounting and uncertainty on consumption is captured by b•A , given today’s income. 25 Shiller (1993) first suggested such assets. 24

106

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

where Ë is the share of foreign assets in the wealth portfolio of country and þ denotes the per capita output of country and the asterisk indicates the average of per capita outputs across all countries. Under these assumptions, net factor income flows are given by gX®¾ = þ − ®gF¾ = Ë OÃ∗¾ − þ V

(24)

This formulation captures the idea that countries with higher portfolio shares of foreign assets will achieve more risk sharing through income smoothing (Sørensen et al., 2007). In frictionless markets, countries would want to diversify completely, which amounts to selling their national output to the world mutual fund. Hence, under complete diversification it is expected to be Ë = 1, which would imply claims to a country’s entire output and so claims to labour income and other non-tradeable output components. Furthermore, frictions in financial and goods markets are expected to drive Ë away from unity. The parameter Ë indicates how close observed income flows are to the income flows one would observe if countries could completely diversify their idiosyncratic risk by investing all their wealth in a world. The parameter Ë indicates which fraction of the income is effectively derived from domestic and foreign sources is estimated from the data. Plugging (24) into (21) leads to,

F¾ = ʾ ®gF¾ = ʾ rËÃ∗¾ + j1 − Ëlþ ].

(25)

This equation implies that country will be able to decouple income and consumption through intertemporal savings and dis-savings. In the context of the Permanent Income Hypothesis (PIH), consumption should always be equal to the permanent component of income defined as ®gF¾I

n 1 Æ = W Ìe Í Î Ï®gF¾+Æ ÐÑ 1+n ¾ 1+n ∞

Æ=0

(26)

where n is the countries real interest rate and W¾ is the expectations operator. Hence, according to PIH, F¾ = ®gF¾I. Assume for expository purposes that income follows a stationary E´j1l with autoregressive coefficient Ò, 0 < Ò < 1. Then

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

F¾ =

n ®gF¾ 1+n −Ò

107

(27)

so that in this simple case, ʾ = n j1 + n − Òl is a country-specific constant reflecting country ’s return on wealth and the persistence of its income process. −1

3.2.2 Measuring International Risk Sharing and empirical evidence In this analysis international risk sharing is measured based on growth rates. Using growth rates rather than absolute values for calculating the aggregated values is due to the fact that aggregated fluctuations cannot be eliminated by risk sharing. The amount of deviation from perfect consumption smoothing is measured through this regression: ∆ log

F•A ÕA A = } + q log ¨∆ Ó Ô ∗ ∗ ª + Ž• F• Õ

(28)

where F¾ is the final consumption per capita growth rate of country in year ¾, F∗¾ is the respective (log) value for the group, }Ó is the constant and Ž¾ stands for the residual. Equation (28) relates relative consumption to relative output and implies that in a model with complete international financial markets and under perfect risk sharing the left-hand side will be zero. Hence, the co-movement coefficient q± which can be interpreted as the portion of non-diversified idiosyncratic risk faced will also become zero – the lower the coefficient the higher the international consumption risk sharing, as more consumption is buffered against output fluctuations. The smaller q± becomes, the lower is the average co-movement of the idiosyncratic consumption growth with the idiosyncratic output growth in this country. This equation is similar to Mace (1991), Asdrubali et al. (1996), Cochrane (1991) and Crucini (1999) on differenced data and reminiscent to Artis and Hoffmann (2012) on level data. The cross-sectional analysis underlines the evolution of the increasing risk sharing. It can easily be seen that fluctuations from small but financially very open economies (SOE) are much higher than for the other analyzed countries. Hence, the average values for these economies are slightly higher. On average 0.2204 of the consumption risk is shared during the sample period, while the minimum is reached in 2000 with even a negative value for the unshared risk of β = -0.085. The highest value is measured at the end of the sample period in 2010 with a β = 0.5979.

Figure 4 illustrates the evolution of the risk sharing in consumption j1 − β l from 1988 to 2010 for all three geographic sub-samples. To smooth the highly fluctuating

108

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

year-by-year estimates for the unshared risk β the coefficients are filtered. Therefore, the Hodrick-Prescott filter with a smoothness parameter λ = 6 is used. For the OECD countries the β = 0.0216 is documented for the beginning of the sample period, while the maximum is reached 2009 with 0.490. On average over the whole period I calculate for OECD countries a risk sharing of 0.2044. It can easily be seen that This figure shows the cross-sectional estimates for (1-ß U,t ). The black solid line indicates the crosssectional estimates for the risk sharing regression based on growth rates for the whole analyzed sample of 21 OECD countries. The dotted line displays estimation results for small but financially very open economies (SOE). The SOE group includes Austria, Belgium, Denmark, the Netherlands, Sweden, and Switzerland. The dashed line is the sequence of cross-sectional estimates for the OECD sample excluding SOE. All sequences are filtered using a Hodrick-Prescott filter with a smoothness parameter of λ =6 .

0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 -0.10 1990

1995 OECD

2000 SOE

2005

2010

OECD w/o SOE

Figure 4: International Consumption Risk Sharing from 1988 to 2010

fluctuations for small but financially very open economies (SOE) are much higher than for the other analyzed countries. Hence, the average values for these economies are slightly higher. On average 0.2204 of the consumption risk is shared during the sample period, while the minimum is reached in 2000 with even a negative value for the unshared risk of β = -0.085. The highest value is measured at the end of the sample period in 2010 with a β = 0.5979.

Excluding the highly fluctuating values for the SOE countries smoothes the unshared risk. By analyzing OECD countries without the small economies (SOE) the analysis

109

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

brings out very similar results. The evolution of this sample mimics that of the whole OECD country sample across the analyzed time period between 1988 and 2010. I now estimate pooled risk sharing regressions on differenced data. Table 16 presents the first regression results and suggests that there is a lack of long-run risk sharing in international data. As pooled estimate is potentially more susceptible to unobserved heterogeneity across countries, I always run fixed-effect panel regressions. Panel I shows the results from pooled regression for our whole OECD sample. First, I run the regression for the whole period considered (1988-2010) and then construct sub-samples including both fifteen and ten year periods. Furthermore, geographic sub-samples are built to emphasize the cross-sectional dimension of the data. In Panel II indicates the results for small but financially very open economies (SOE) namely Austria, Belgium, Denmark, the Netherlands, Sweden, and Switzerland in our case. The parameters are estimated from pooled OLS regression of equation (28): The first column captures the results for the whole sample period from 1988 to 2010. The following columns report the results for the various time subsamples. Panel I includes all analyzed 21 OECD countries and Panel II comprises small but financially open economies (SOE), which in this country sample are Austria, Belgium, Denmark, the Netherlands, Sweden, and Switzerland. Panel III indicates all analyzed OECD excluding the small but financially open countries (SOE). Fixedeffects are included. *, **, and *** stand for 10%, 5%, and 1% significance. Panel I: OECD All t-Stat

0.974 12.391 ***

1988-1990

1990-2000

2000-2010

0.976 0.824 0.635 8.252 *** 22.956 *** 13.528 ***

1988-1995 0.976 22.961 ***

1995-2002 0.733 13.53 ***

2002-2010 0.654 9.876 ***

Panel II: S OE All ßU

t-Stat

0.83 20.408 ***

1988-1990

1990-2000

2000-2010

0.861 0.809 0.777 9.794 *** 11.784 *** 11.107 ***

1988-1995

1995-2002

0.81 0.771 12.689 *** 11.671 ***

2002-2010 0.685 8.549 ***

Panel III: OECD w/o S OE All ßU

t-Stat

0.973 18.571 ***

1988-1990 0.975 17.977 ***

1990-2000 0.823 21.583 ***

2000-2010 0.623 12.486 ***

1988-1995 0.975 9.915 ***

1995-2002 0.728 19.125 ***

Table 16: Consumption Risk Sharing Pooled Regression

2002-2010 0.608 8.548 ***

For the analyzed 21 OECD countries (Panel I) unshared risk β exceeds 0.976 with a t-Stat of 22.961 in the 1988 to 1995 sample. Between 1993 and 1998 the estimates of the coefficient decrease to 0.733 with a t-Stat of 13.530. The respective value in the last sub-period from 2002 to 2010 is 0.654 with a t-Stat of 9.876. A similar increase in consumption risk sharing is observed with estimates decreasing from 0.824 with a

110

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

t-Stat of 22.96 in the 1990-2000 period compared to 0.635 with a t-Stat of 13.528 in the 2000-2010 period. This decrease of the unshared risk β is robust across the geographic sub-samples. Panel II shows a similar evolution in the SOE sample. The estimates of the coefficients decline from 0.810 with a t-Stat of 12.689 in the period between 1988 and 1995 to 0.771 with a t-Stat 11.671 (1995-2002) and 0.685 (tStat=8.549) in the sample of from 2002-2010. The results for the whole country sample excluding these SOEs changes the results indicated in panel III only marginally.

4.

Patterns of Risk Sharing and International Asset Positions

4.1

Channels of Risk Sharing

In this section the channels through which the increase in international risk sharing has occurred are examined. Building on the risk sharing channel approach of Asdrubali et al. (1996), I show that international capital income flows are the main drivers of improved risk sharing and give evidence that increasing risk sharing can mainly be traced back to the growth in international investment positions. They state that consumption risk sharing can be conducted in two ways: The first channel shares consumption risk through income smoothing by de-coupling income (GNI) from output (GDP). This channel is called the ex-ante channel. The second channel leads to risk sharing as it de-couples consumption from income, through saving and dis-saving. I refer to this channel as the ex-post channel. Following theory it should not be possible to insure against permanent idiosyncratic shocks through the consumption channel, but only through the ex-ante income channel. Changes in the long-run risk sharing parameter β are expected to be associated mainly to the ex-ante channel. To test this prediction the estimate of the risk sharing parameter introduced above is decomposed and the importance of the income channel (ex-ante channel) is measured by running the following regression: ¨∆ log

YÕ INCÕ YÕ − log = α + β log ª ¨∆ ª ¨∆ ª + ϵÕ , Õ YÕ∗ INCÕ∗ YÕ∗

(29)

where INC denotes the logarithm of income growth, Y the respective value for the output. αÕ is a country or state-level fixed effect, β captures the amount of risk shared through this channel, while ϵ stands for the residuals. Recall, outputs minus

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

111

income are the net foreign investments.26 Hence, the more than average increase of the net foreign investments is best explained by the more than average output growth. The risk shared through the consumption channel (ex-post channel) is measured by the regression: ¨∆ log

INCÕ CÕ YÕ − log = α + β log ª ¨∆ ª ¨∆ ª + ϵÕ Õ INCÕ∗ CÕ∗ YÕ∗

(30)

where C denotes the logarithm of consumption growth, Y the respective value for the output. β stands for the amount of risk shared through the ex-post channel, αÕ is a country or state-level fixed effect, and ϵ stands for the residuals. Keep in mind, that the left-hand side is the value for gross national savings. It can be seen, that the idiosyncratic growth of these savings is best explained by systematic output growth. By construction, the sum of the two channels must always be equal to one minus the coefficient from our risk sharing regression calculated above (unshared risk): β + β = j1 − β l.

(31)

However, if taste shocks are sufficiently persistent or even permanent, ex-post consumption smoothing though borrowing and lending will not be appropriate for sharing consumption risk of an economy. So it is to expect that the increasing importance of the ex-ante income channel for risk sharing over time is confirmed by the data. I estimate the regressions by pooled OLS, controlling for both country and time fixed effects. Table 17 reports the results. The regression outcome lines up with theoretical predictions stating that risk sharing in the long-run should be associated with the exante channel and not with consumption smoothing. The results for the OECD countries as well as for the SOE clearly evidence this impact: in the 1980-1995 period risk sharing is quite low, as both ex-ante and ex-post channels are very small. This effect changes turning to the post-1995 period. This relatively high amount of risk sharing in international data is mainly driven by increasing capital income flows. The ex-post consumption smoothing channel also increases moderately, but reaches small values compared to the ex-ante channel. Panel I shows that over the whole period, β reaches a value of 0.201 with a t-Stat of 3.480 compared to the estimate of β with a value of 0.026 and a t-Stat with 3.327. 26

Leaving small positions of the national accounts out of consideration.

112

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

As shown above, also a number of tests are provided to illustrate that the increase in the ex-ante channel is indeed a robust feature of the data. Examining the sub-samples more closely shows the increasing importance of the income smoothing channel, as the estimates rise to 0.368 with a t-Stat of 3.69 in the 10-year period in panel I. Fixed effects are included. The parameters are estimated from pooled OLS regression of equation (29) and (30) described above: equation (29) (ex-ante) and equation (30) (ex post). Panel I reports results for analyzed OECD countries, Panel II includes small open economies (SOE), whereas Panel III documents channels of risk sharing for analyzed OECD countries without SOE. Numbers in parentheses are t-values, where *, **, and *** stand for 10%, 5%, and 1% significance. Panel I: OECD All 1990-2000 2000-2010 1988-1995 1995-2002 2002-2010 0.201 0.042 0.368 0.005 0.1291 0.291 ßI t-stat 3.48 *** 2.831 ** 3.689 *** 2.193 ** 4.591 *** 3.191 *** 0.483 0.774 0.475 0.423 0.421 0.588 R2 ßC t-stat R2

ßI t-stat R2 ßC t-stat R2

ßI t-stat R2 ßC t-stat R2

0.026 3.327 *** 0.46 All 0.104 3.867 *** 0.518 0.17 4.177 *** 0.745 All 0.101 4.015 *** 0.581 0.027 3.269 *** 0.443

0.018 4.889 *** 0.597 1990-2000 0.089 2.771 ** 0.434 0.109 2.784 *** 0.405

0.031 7.765 *** 0.381

0.024 2.541 *** 0.599

Panel II: S OE 2000-2010 1988-1995 0.294 0.067 3.499 *** 2.194 ** 0.514 0.493 0.113 3.181 *** 0.32

0.109 2.974 *** 0.795

Panel III: OECD w/o S OE 1990-2000 2000-2010 1988-1995 0.045 0.171 0.005 4.147 *** 3.657 *** 1.898 ** 0.744 0.486 0.32 0.107 4.638 *** 0.5

0.138 7.561 *** 0.329

0.025 2.528 *** 0.587

0.027 4.015 *** 0.389

0.027 7.444 *** 0.498

1995-2002 2002-2010 0.127 0.246 5.403 *** 3.703 *** 0.455 0.515 0.111 3.828 *** 0.474

0.131 3.18 *** 0.578

1995-2002 2002-2010 0.104 0.154 4.115 *** 3.141 *** 0.3 0.397 0.102 3.547 *** 0.39

0.127 7.157 ***

Table 17: Channels of Risk Sharing

Also analyzing geographic sub-samples offers the possibility to investigate whether the impact is more pronounced for different country groups. In the OECD countries the ex-post channel only grows marginally, whereas the income smoothing ex-ante channel increases from 0.005 (t-Stat 2.193) in the sample 1988-1995 to 0.291 (t-Stat 3.191) in group 2002-2010. Again, panel II displays the results for the SOE and also provides highly significant results. Furthermore, the same evolutions can be measured for this sample as the estimates for the risk shares through the ex-ante income

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

113

channel in both our time sub-sample increase significantly from 0.089 (t-Stat 2.771) to 0.294 (t-Stat 3.499) and from 0.067 with a t-Stat of 2.19 to 0.246 with a t-Stat of 3.70. Panel III shows the respective results for the whole country sample excluding the SOE. The previous findings are confirmed. For the whole sample period as well as for the sub-sample the results are highly significant at a 1% level or at least at the 5% level. 4.2

The Increase in International Risk Sharing and the growth in Gross International Asset Positions

4.2.1 International financial integration proxied by the Equity Home Bias This section of my thesis focuses on the documented increasing importance of the exante channel. Therefore, it is examined whether the driving force behind this effect is the increasing internationalization of the asset ownership. International portfolio diversification is captured by the previously introduced index of the equity home bias, which is included in the measure of the risk sharing channels. Thus, both ex-ante and ex-post risk sharing become a function of the diversification measure. I run a pooled OLS regression including both time and country fixed effects: ¨∆ log

YÕ INCÕ YÕ ∗ ª − ¨∆ log ∗ ª = αÕ + κ ¨∆ log ∗ ª + ϵÕ YÕ INCÕ YÕ

(32)

while coefficient κ stands for the average co-movement of the countries’ idiosyncratic growth of net foreign investments with the idiosyncratic GDP growth during the time-period considered and is defined as follows κ = κ& + κ' jt − tl̅ + κ( jEHB Õ − ÙÙÙÙÙÙÙ EHBÕ l (33)

where t̅ indicates the middle year of the sample period, EHB Õ the amount of the equiÙÙÙÙÙÙÙÕ is the (un-weighted) average across ty home bias in country i at time t, and EHB countries at time t. Plugging (33) in (32) leads to:

114

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

YÕ INCÕ ª ¨∆ ª − log YÕ∗ INCÕ∗ ÙÙÙÙÙÙÙÕ l ∗ χÕ + ϵÕ = αÕ + κ& ∗ χÕ + κ' ∗ jt − tl̅ ∗ χÕ + κ( ∗ jEHB Õ − EHB ¨∆ log

(34)

where κ& captures the idiosyncratic output growth and κ' measures the average yearby-year amount of increasing income risk sharing. The parameter κ( quantifies how much higher than average the equity home bias lowers the amount of income Fixed effects are included. The parameters are estimated from pooled OLS regression of equation (34) described above: κ0 captures the importance of idiosyncratic output growth, κ 1 stands for the time-trend of the risk sharing, κ 2 stands for the impact of the more than average amount of equity home bias. Numbers in parenthesesare t-values, where *, **, and *** stand for 10%, 5%, and 1%. OECD SOE OLS w/o SOE coeff t-stat coeff t-stat coeff t-stat R2 R2 R2 all 0.0564 2.8047 *** 0.5287 0.0073 2.0652 ** 0.4289 0.0645 2.9063 *** 0.3158 κ0 κ1 0.0025 1.0103 * 0.0073 0.0652 0.0026 0.9756 κ2 1990-2000 κ0 κ1 κ2 2000-2010 κ0 κ1

-0.1845 -3.4775 ***

-0.6574 -3.1108 ***

-0.2401 -3.6925 ***

0.0234 3.728 *** 0.5082 -0.0074 -1.5922 * -0.2823 -2.7473 ***

0.2071 2.6504 ** 0.3263 -0.0121 -0.6759 -0.5751 -2.8667 ***

0.0386 2.1059 ** 0.5203 -0.0075 -1.529 * -0.2074 -2.9291 ***

0.0454 0.0071

5.3213 *** 0.5876 0.6291

0.0155 0.0375

2.4724 ** 0.5366

κ2 1988-1995 κ0 κ1

-0.3422 -4.6287 ***

-0.9866

0.0221 4.8512 *** 0.5124 -0.0073 -1.3418 *

0.2122 2.4851 ** -0.0105 -0.6172

κ2 1995-2002 κ0 κ1 κ2

-0.2781 -3.5848 ***

2002-2010 κ0 κ1 κ2

0.0387 3.1255 *** 0.5366 0.0054 1.0984 * -0.2945 -4.1158 *** 0.0459 0.007 -0.3469

3.7895 *** 0.5004 1.0488 * 4.8988 ***

0.4127

-7.621 ***

0.8995 0.3414

0.3628

-0.339 -2.9633 *** 0.3874

-0.535 -3.0189 *** 0.0345 0.0001 -0.6124

0.082 0.004

0.0346 2.5742 ** -0.007 -0.9085

0.3615

-0.4022 -3.1565 ***

2.7156 ** 0.3984 0.2789 4.5682 ***

0.0716 1.8452 * 0.3848 0.0061 0.8475 -0.4091 -4.6182 ***

0.0155456 2.4518 ** 0.4088 0.0401 0.1322 -0.9874 -3.9515 ***

0.0864 1.9751 * 0.4557 0.0054 0.3623 -0.4981 -4.9312 ***

Table 18: Income Channel as a function of the Equity Home Bias

smoothing and consequently the obtained international risk sharing. Including the time trend allows me to guard against the decreasing trend of the equity home bias measure spuriously capturing trend changes in risk sharing maybe caused by other

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

115

developments in international markets. Subtracting the respective aggregate value is crucial as the aggregate growth cannot be insured. As the equity home bias is expected to have a negative effect on the international risk sharing and the ex-ante smoothing, κ( should have negative loadings.

Table 18 displays results from income smoothing (ex-ante channel) as a function of equity home bias for the OECD countries, as well as for the respective country subsamples. For the time trend I find a slightly significant coefficient of 0.0025 (t-Stat 1.0103). For the equity home bias coefficients κ( in the time period from 1988-2010 the results indicate high significance robust across countries and time. For the whole sample and period I estimate a coefficient for κ( of -0.1845 (t-Stat -3.4775), which implies that in the analyzed countries lowering EHB by 0.1 will increase risk sharing through the income channel by about 0.1845. Panel II documents the same significant impact for SOE with even more negative coefficients. In this sample the estimate for the coefficient κ( is -0.6574 with a t-Stat of -3.1108. The picture that can be conducted from Table 18 clearly indicates that the growth in international cross-holding in equity leads to an increasing importance of the income channel and thus to better risk sharing. Hence, it can be seen that these findings cannot reject the hypothesis that the main driver behind the income smoothing is the growing amount of foreign equity in countries portfolio allocation. Across the time sub-samples the relationships are similar to the documented impact for the whole analyzed time period. 4.2.2 International Financial Integration proxied by Foreign Equity Holdings to GDP

As the theoretical benchmark weights for portfolio allocation may be non-optimal, I perform a similar regression employing the ratio of foreign equity holdings to GDP as an alternative measure of financial integration. This ratio does not include any theoretical benchmarks and is included in this analysis in order to test for robustness of the previous findings. I use this ratio rather than total equity holdings to GDP as if total equity holding may be small relative to GDP the ratio of foreign holdings to GDP may be more relevant for macroeconomics consumption risk sharing. Figure 5 illustrates the average values for the sample during the 1988 and 2010. Thus, the regression is modified as follows: ¨∆ log

YÕ INCÕ YÕ ∗ ª − ¨∆ log ∗ ª = αÕ + η ¨∆ log ∗ ª + ϵÕ , YÕ INCÕ YÕ

(35)

116

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

ÙÙÙÙÙÕ l, where FE Õ stands for the (log) ratio of with η = η& + η' jt − tl̅ + η( jFE Õ − FE ÙÙÙÙÙÕ is the respective averforeign equity holdings to GDP of country i in year t and FE age value for the group. The interaction terms based on the ratio of foreign equity holdings to GDP are performed in the same way as in the analysis for income risk sharing. Recall, η& captures the idiosyncratic output growth and η' stands for the average year-by-year amount of increasing income risk sharing, while the parameter η( quantifies how much higher than average foreign equity holdings to GDP increases the amount of income smoothing. I expected the latter explanatory variable to have positive impact on international risk sharing and, thus, the signs for the estimates of η( are predicted to be positive. This figure shows the evolution of the foreign equity holdings to GDP in the analyzed equity portfolios during the sample period between 1988 and 2010. The solid line indicates the equity home bias for the whole sample of 21 OECD countries; the dotted line shows the values for the small but financially open economies (Austria, Belgium, Denmark, the Netherlands, Sweden, and Switzerland); the line with the crosses captures the value for the OECD sample excluding the SOE. Data source: IMF/Lane and Milesi-Ferretti (2006), and World Bank.

0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 1990

1995 OECD

2000 SOE

2005

2010

OECD w/o SOE

Figure 5: Evolution of the Foreign Equity Holdings to GDP

Table 19 shows results running regression (35) with income smoothing as a function of the foreign equity holdings to GDP, while I do not disclose further time subsamples as in the analyses before. They lead to the same conclusions. With regard to the estimates of η( , the impact of this measure for financial integration on interna-

117

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

tional consumption risk sharing is as expected positive. This relationship is intuitively, as an increasing ratio of foreign equity to GDP in a portfolio may lead to income smoothing and then in a next step to an improved sharing of consumption risk. For the whole sample the results bring out a coefficient of 0.0174 with a t-Stat of 6.8015. However, no impact of time could be documented, while the coefficient of η& is 0.0491 is highly significant at a 1% level. These findings are robust across both geographical and time sub-samples. Again it can be seen that the impact of the foreign equity holdings to GDP is much higher for SOE than for the whole sample. The coefficient reaches a value of 0.1226 and is highly significant at the 1% level. Fixed effects are included. The parameters are estimated from pooled OLS regression of equation (35) described above: η 0 captures the importance of idiosyncratic output growth, η 1 stands for the time-trend of the risk sharing, and η 2 stands for the impact of the more than average amount of foreign equity holdings to GDP. Numbers in parentheses are t-values, where *, **, and *** stand for 10%, 5%, and 1% significance.

all η0 η1 η2 1990-2000 η0 η1 η2 2000-2010 η0 η1 η2 1988-1995 η0 η1 η2 1995-2002 η0 η1 η2 2002-2010 η0 η1 η2

OECD t-stat

2

coeff

3.525 *** 0.1689 6.8015 ***

0.349

0.0535 0.0162 0.1226

0.0024 2.8199 *** -0.0094 -1.1953 * 0.0027 5.0097 ***

0.3943

coeff 0.0491 0.0049 0.0174

0.0408 0.0009 0.0845

R

4.7891 *** 0.0788 4.3888 ***

0.5843

0.0018 3.1232 *** -0.0083 -1.2416 * 0.0016 3.5615 ***

0.41

SOE t-stat

2

coeff

5.499 *** 0.3123 1.1362 * 3.7293 ***

0.0513 0.0002 0.0183

0.1876 1.6062 ** -0.0182 -0.9792 0.1752 2.572 ** 0.0237 0.0041 0.3954

R

OLS w/o SOE t-stat

R2

2.9022 *** 0.5257 0.062 0.6785

0.2754

0.0012 3.0388 *** 0.4969 -0.0101 -1.0956 * 0.0017 4.0232 ***

3.0424 *** 0.4236 0.055 4.6254 ***

0.064 3.7108 *** 0.5971 -0.0023 -0.1828 0.088 2.4369 ***

0.2846 2.3465 ** 0.2488 -0.0158 -0.0547 0.1643 2.8465 ***

0.0028 -0.0075 0.0024

0.0085 2.9842 *** 0.5294 -0.0015 -1.0004 0.0182 3.5142 ***

0.0105 0.0002 0.0095

3.5413 *** 0.1546 4.5146 ***

0.3457

0.1875 -0.0201 0.2713

2.5741 ** 0.2561 0.0843 3.9521 ***

0.0445 0.0012 0.0867

3.9945 *** 1.0232 * 4.8456 ***

0.5838

0.2744 3.1057 *** 0.3645 -0.0136 -0.8153 0.4015 3.8412 ***

3.2154 *** 0.5181 -0.812 3.9152 ***

0.0246 3.0156 *** -0.0154 -0.7212 0.0924 4.5165 ***

Table 19: Income Channel as a function of Foreign Equity to GDP

0.53

118

5.

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

Conclusion

Financial integration should make it easier for countries to trade international assets. But this access to international markets may also lead to shifts in the patterns of the underlying risks. For example, changing structures in industries and altering patterns of specialization, may lead to less symmetric business cycles. As a consequence, financial integration is expected to induce a requirement for countries to share more risk (Kalemi-Ozcan et al., 2003). One way to shed light on this issue is to focus on the long-run evolution of the data. Lucas (1987) discussed in his seminal work that welfare effects provided by the elimination of consumption risk may be small if idiosyncratic shocks are permanent. At the same time, long-run shifts or trends in output or consumption are less likely to be affected by changes in the correlation of countries’ business cycles. In a next step this implies that improvements in international consumption risk sharing may show up first and empirically most robustly in a long-run perspective. This approach allows documenting an increase in international risk sharing among industrialized countries. More than 37% of the consumption risks sharing among the OECD country sample gets shared across countries by the end of our sample period. In order to examine how countries share their consumption risk, I employ the approach of Asdrubali et al. (1996). They identified two channels through which economies smooth their consumption flows. Risk can either be shared ex-ante through the income channel by decoupling income from output or ex-post through the consumption channel by decoupling consumption from income. After showing increasing international consumption risk sharing, these channels are empirically examined to shed light on this improvement. The results document that the main channel through which this improvement seems to be realized is the ex-ante income smoothing channel, i.e. the growing importance of international capital income flows, rather than ex-post consumption channel (through borrowing and lending). But what is the driving force behind the growing impact of the ex-ante channel for consumption risk sharing? For answering this question, I explicitly condition on the index of the equity home bias as measure for financial integration in the regression. Employing the often used measure of the equity home bias to capture financial integration brings me to the problem of the optimal benchmark weights discussed in section 3 of part one of my thesis. As I suspect that the results may be driven by the employed benchmark model, I subsequently rerun the model using the foreign equity holding to GDP of a country as

IS INCREASING FINANCIAL INTEGRATION RELATED TO IMPROVED INTERNATIONAL RISK SHARING?

119

measure for financial integration. As this dependent variable is simply calculated a ratio and not a deviation from an optimal value it is free of these problems. The regression results indicate that there is significant impact of the home bias in international equity portfolios and the ex-ante channel for international consumption risk sharing. This finding is robust for all our sub-samples and sub-periods. Given these results, I cannot reject the hypothesis that higher cross-holdings of assets (decreasing equity home bias) leads to increasing international consumption risk sharing. It can be argued that both phenomena are endogenous reactions to common driving forces such as the liberalization of financial markets or technological innovations. As I cannot solve this endogenous problem my findings are not able to reject the hypothesis that the growth in international portfolio equity holdings is the driving force behind the documented improvements in long-run consumption risk insurance. Employing the foreign equity holdings to GDP as dependent variables leads to the similar results. Thus, this robustness test adds weight on our previous findings and documents that the results are not driven by the international CAPM as a benchmark model used to calculate the home bias in international equity holdings.

121

Summary My thesis examines the home bias puzzle in international equity portfolios, its determinants, and its impact on international consumption risk sharing. In detail, this study gives detailed insights on the structure of the equity home bias on a country-level as well as on the linkage of this phenomenon on the macroeconomic perspective of international risk sharing in consumption. In the first part of my thesis I present an overview of selected literature on the home bias in international portfolio holdings. At first, attempts to capture gains from international diversification and empirical evidence for the puzzle are shown. As the home bias is measured as a deviation of the actual weight of domestic position in the portfolio holdings from the optimal weight, I subsequently introduce and discuss theoretical frameworks in order to measure the optimal benchmark weights. Specifically, I introduce the traditional International CAPM, the Bayesian approach also with a modification, and the gravity model approach. These approaches bring out very different optimal portfolio weights and thus, very different amounts of the home bias in international portfolio holdings are measured. Baele et al. (2007) focus on the identification of the model that brings out the optimal theoretical weight. As until now, none of these attempts had convincing and empirically consistent results, I decided to choose the traditional model-based approach, i.e. the international CAPM as benchmark model. Even though my aim is not to examine optimal portfolio weights, it should be considered that the findings of my empirical analysis could differ when alternative benchmark models are employed. Subsequently, different explanation attempts are presented. These approaches can be divided into two groups. On the one hand, the institutional perspective discusses transaction costs, hedging motives, and information asymmetries between domestic and foreign investors as well as investor protection aspects. On the other hand, the behavioral explanation attempts employ the relative optimism with regard to the domestic market, overconfidence and investor’s competence, herding behavior, as well as familiarity to capture the extent of home bias in international portfolio holdings. The first part of my thesis concludes with a discussion of the explanation attempts. The following parts two and three contain my empirical analyses. Part two analyzes the structure of the home bias, by examining the domestic and the foreign component separately. This approach builds on Chan et al. (2005). While the domestic bias captures the amount of domestic positions are overweighted, the foreign bias gives in-

122

SUMMARY

sight how much a foreign markets is over- or underweighted in a domestic portfolio. One innovation of this analysis is, that studies yet restricted to fund-level analysis, now are examined on a macroeconomic country-level analysis. This complementary approach allows me to shed light on the structure of the home bias in equity portfolios from another perspective. However, the major contribution of this analysis is with regard to the explanation attempts for the home bias puzzle. I combine traditional approaches with cultural aspects which were mainly discussed separately in literature. The results indicate that foreign investors put less of their wealth in countries where they are faced to high barriers to international capital flows than investors from the domestic market. In addition, the findings indicate that foreign investors invest a smaller fraction of their portfolio in countries which are culturally distant to their home countries. Overall, these results make evidence that adding cultural dimensions such as the cultural distance between home and target country to traditional variables improves the analysis significantly and helps capturing the substantial amounts of home bias documented in international equity portfolios. The third part of my thesis goes a step further and examines the equity home bias puzzle in the context of long-term international consumption risk sharing. For this purpose, the study tries to answer the question whether the increasing financial integration and thus, less home bias in international portfolio allocation is related to the improved sharing of consumption risk. Therefore, I employ the risk sharing-channel approach mainly building on Asdrubali et al. (1996) and so have the possibility not only to document increasing risk sharing between economies, but also to analyze the driving force behind this evolution. The findings show that the documented improved international consumption risk sharing is in large part due to the growing amount of risk shared through the ex-ante income channel rather than through the ex-post channel. Examining this way for sharing consumption risk in detail, brings to light that the latter effect could be tracked back to a better diversification of portfolio holdings, i.e. decreasing home bias, and thus growing financial integration. I find evidence that both anomalies, i.e. the equity home bias puzzle and the international consumption risk sharing are linked. Furthermore, I cannot exclude that this strong relationship is due to the liberalization of international financial markets. My thesis contributes to the finance literature on portfolio diversification and the home bias puzzle on the country-level and its behavioral explanation attempts. Large part of research in this field analyzes portfolio holdings whether of individual or institutional investors. Hence, they run the risk that their results could be related the characteristics of a special type of investor. Again, my approach abstracts from these

SUMMARY

123

potential characteristics with regard to the investment behavior. In this third part of my thesis a very long time period was employed, which also allows abstracting from short-time fluctuations. Overall, my thesis documents that the overweighting of domestic positions in international portfolio allocation is still substantial, even though information availability has improved and financial markets are liberalized. This persisting anomaly may increase economies deadweight costs. Hence, better insights on the determinants and sources of this puzzle are incessant to better capture negative impacts on economies and helping prevent them. For future research it might be interesting to examine the costs of underdiversification and forgone returns of international investments as well as introducing a return based approach. The letter perspective may shed more light on the return chasing effect on investment decision. Being able to detect a linkage between the home bias puzzle in international portfolio allocation and investment strategies, such as positive-feedback trading, would add significant value to both streams of literature.

124

References Acharya, V., I. Hasan, A., Saunders. 2006. Should Banks be Diversified? Evidence from Individual Bank Loan Portfolios. Journal of Business 79, 1355-1412. Adler, M., B. Dumas. 1983. International portfolio choice and corporation finance: A synthesis. Journal of Finance 38, 925-984. Ahearne, A.G., W.L. Griever, F.E. Warnock. 2000. Information costs and home bias: an analysis of U.S. holdings of foreign equities. Working Paper. Ahearne, A.G., W.L. Griever, F.E. Warnock. 2004. Information costs and home bias: an analysis of U.S. holdings of foreign equities. Journal of International Econom-ics 62, 313-336. Anderson, C.W., M. Fedenia, M. Hirschey, H. Skiba. 2011. Cultural influences on home bias and international diversification by institutional investors. Journal of Banking and Finance 35, 916-934. Artis, M.J., M. Hoffmann. 2006. The Home Bias and Capital Income Flows between Countries and Regions. CEPR Discussion Papers 5691. Artis, M.J., M. Hoffmann. 2008. Financial Globalization, International Business Cycles and Consumption Risk Sharing. Scandinavian Journal of Economics 110, 447-471. Artis, M.J., M. Hoffmann. 2012. The Home Bias, Capital Income Flows and Improved Long-Term Consumption Risk Sharing between Industrialized Countries. Journal of International Finance 14, 481-505. Asdrubali, P., B.E. Sørensen, O. Yosha, O. 1996. Channel of interstate risk sharing: United States 1963-1990. Quarterly Journal of Economics 111, 1081-1110. Backus D.K., P.J. Kehoe, F. E. Kydland. 1992. International Real Business Cycles. Journal of Political Economy 45, 477-505. Bae, K.-H., R.M. Stulz, H. Tan. 2008. Do local analysts know more? A cross-country study of the performance of local analysts and foreign analysts. Journal of Financial Economics 88, 581-606. Baele, L, C. Pungulescu, J.T. Horst. 2007. Model Uncertainty, Financial Market Integration and the Home Bias Puzzle. Journal of International Money and Finance 26, 606-630.

REFERENCES

125

Baele, L. 2005. Volatility spillover effects in European equity markets. Journal of Financial and Quantitative Analysis 40, 373-401. Balli, F., S. Kalemli-Ozcan, B.E. Sørensen. 2011. Risk Sharing through Capital Gains. Working Paper. Barber B.M., T. Odean. 2001. Boys will be boys: Gender, Overconfidence, and common stock investment. Quarterly Journal of Economics 116, 261-292. Barber B.M., T. Odean. 2002. Online Investors: Do the slow die first? Review of Financial Studies 15, 455-487. Baxter, M., M.J. Crucini. 1995. Business Cycles and the Asset Structure of Foreign Trade. International Economic Review 36, 821-54. Baxter, M., U.J. Jermann. 1997. The International Diversification Puzzle is worse than you think. American Economic Review 87, 170-180. Bayoumi, T. 1990. Saving-investment Correlations: Immobile Capital, Government Policy and Endogenous Behavior? International Monetary Fund 37, 360-387. Beck. T., L.F. Klapper, J.C. Mendoza. 2010. The typology of partial credit guarantee funds around the world. Journal of Financial Stability 6, 10-25. Bekaert, G., X. Wang. 2009. Home Bias revisited. Working Paper. Benartzi, S. 2001. Excessive Extrapolation and the allocation of 401(k) accounts to company stock. Journal of Finance 56, 1747-1764. Beugelsdijk, S., B. Frijns. 2010. A cultural explanation of the foreign bias in international asset allocation. Journal of Banking and Finance 34, 2121-2131. Black, F. 1974. International Capital Market Equilibrium with Investment Barriers. Journal of Financial Economics 1, 337-352. Bottazzi, L., P. Pesenti, E. van Wincoop. 1996. Wages, profits and the international portfolio puzzle. European Economic Review 40, 219-254. Boyd, J.H., G. De Nicoló. 2005. The Theory of Bank Risk Taking and Competition Revisited. Journal of Finance 60, 1329-1343. Brealey, R.A., I.A. Cooper, E. Kaplanis. 1999. What is the International Dimension of International Finance? European Finance Review 3, 103-119. Brennan, M.J., H.H. Cao, N. Strong, X. Xu. 2005. The Dynamics of International Equity market Expectations. Journal of Financial Economics 77, 257-288. Brennan, M.J., H.H. Cao. 1997. International portfolio investment flows. Journal of Finance 52, 1851-1880.

126

REFERENCES

Buch, C.M. J. Kleinert, A. Lipponer, F. Toubal. 2005. Determinants and effects of foreign direct investment: Evidence from German Firm-Level Data. Economic Policy 20, 52-110. Cai, F., F.E. Warnock. 2006. International Diversification at Home and Abroad. Working Paper. Canova, F., M. Ravn. 1996. International Consumption Risk Sharing. International Economic Review 37, 573-601. Chan, K., V. Covrig, L. Ng. 2005. What determines the domestic bias and foreign bias? Evidence from mutual fund equity allocations worldwide. Journal of Finance 60, 1495-1534. Chen, L. S.A. Johnson, J. Lin, Y. Liu. 2009. Information, sophistication, and foreign versus domestic investors’ performance. Journal of Banking and Finance 33, 1636-1651. Chen, Z., L. Epstein. 2002. Ambiguity, Risk, and Asset Returns in continuous time. Econometrica 70, 1403-1443. Choe, H. B.-C. Kho, R.M. Stulz. 2005. Do Domestic Investors have an Edge? The Trading Experience of Foreign Investors in Korea. Review of Financial Studies 18, 795-829. Chue, T.K. 2007. Is Equity Home Bias Really Suboptimal? AFA 2007 Chicago Meeting Paper, Working Paper. Chui, A.C., C.C. Kwok. 2009. Cultural practices and life insurance consumption: An international analysis using GLOBE scores. Journal of Multinational Financial Management 19, 273-290. Chui, A.C., S. Titman, K.C. Wei. 2010. Individualism and Momentum around the World. Journal of Finance 65, 361-392. Cochrane, J.H. 1991. A Simple Test of Consumption Insurance. Journal of Political Economy 99, 957-76. Cole, H., M. Obstfeld. 1991. Commodity trade and international Risk Sharing: How much do Financial Markets matter? Journal of Monetary Economics 28, 3-24. Cooper, I., E. Kaplanis. 1986. Costs to crossborder investment and international equity market equilibrium. Recent Developments in Corporate Finance, J. Edwards, J. Franks, C. Mayer, and S. Schaefer (eds.), Cambridge University Press, Cambridge.

REFERENCES

127

Cooper, I., E. Kaplanis. 1994. Home Bias in Equity Portfolios, Inflation Hedging and International Capital Market Equilibrium. The Review of Financial Studies 7, 45-60. Coval, J. D., T.J. Moskowitz. 2001. The geography of investment: Informed trading and asset prices. Journal of Political Economy 109, 811-841. Coval, J.D., T.J. Moskowitz. 1999. Home bias at home: Local equity preference in domestic portfolios. Journal of Finance 54, 1-39. Crucini, M.C. 1999. On International and National Dimensions of Risk Sharing. Review of Economics and Statistics 81, 73-84. Dahlquist, M., L. Pinkowitz, R.M. Stulz, R. Williamson. 2003. Corporate Governance and the Home Bias. Journal of Financial and Quantitative Analysis 38, 87110. Deutsche Bundesbank. 2013. Monthly Report March 2013. Monthly Report 65. Djankov, S. R. La Porta, F. Lopez-de-Silanes, A. Shleifer. 2008. The law and economics of self-dealing. Journal of Financial Economics 88, 430-465. Dohmen, T. A. Falk, D. Huffmann, U. Sunde. 2012. The Intergenerational Transmission of Risk and Trust Attitudes. Review of Economic Studies 79, 645-677. Dorn, D., G. Huberman. 2005. Talk and Action: What Individual Investors Say and What They Do. Review of Finance 9, 437-481. Driessen, J., L. Laeven. 2007. International portfolio diversification benefits: Crosscountry evidence from a local perspective. Journal of Banking and Finance 31, 1693-1712. Dvořàk, T. 2005. Do Domestic Investors have an Information Advantage? Advantage from Indonesia. Journal of Finance 60, 817-839. Errunza, V. 1983. Emerging Markets: A New Opportunity for Improving Global Portfolio Performance. Financial Analysts Journal 39, 51-58. Errunza, V. K. Hogan, M.-W. Hung. 1999. Can the Gains from International Diversification be achieved without trading abroad? Journal of Finance 54, 2075-2107. Faruqée, H. S. Li., I.K. Yan. 2007. The Determinants of International Portfolio Holdings and Home bias. International Monetary Fund, Working Paper. Fidora, M., M. Fratzscher, C. Thimann. 2007. Home Bias in Global Bond and Equity Markets: The Role of Real Exchange Rate Volatility. Journal of International Money and Finance 26, 631-655.

128

REFERENCES

Fischer, E.O., C. Keber. 1997. Risikoanalyse international Aktienportfeuilles: Eine empirische Untersuchung. Zeitschrift für Betriebswirtschaft 67, 333-360. Flandreau, M. 2006. Home Bias, Nineteenth Century Style. Journal of European Economic Association 4, 634-643. French, K., J.M. Poterba. 1991. Investor Diversification and International Equity Markets. American Economic Review 81, 222-226. Froot, K.A., P.G. O’Connell, M.S. Seasholes. 2001. The Portfolio Flows of International Investors. Journal of Financial Economics 59, 151-193. García-Herrero, A., F. Vazquez. 2007. International Diversification Gains and Home Bias in Banking. International Monetary Fund 07/281, Working Paper. Garlappi, L., R. Uppal, T. Wang. 2007. Portfolio Selection with Parameter and Model Uncertainty: A Multi-Prior Approach. Reviews of Financial Studies 20, 40-81. Gehrig, T. 1993. An Information Based Explanation of the Domestic Bias in International Equity Investment. Scandinavian Journal of Economics 95, 97-109. Gelos, R.G., S.-J. Wei. 2005. Transparency and International Portfolio Holdings. Journal of Finance 60, 2987-3020. Giannetti, M., A. Simonov. 2006. Which Investors Fear Expropriation? Evidence from Investors’ Portfolio Choices. Journal of Finance 61, 1507-1547. Glassman, D.A., L.A. Riddick. 2001. What causes home asset bias and how should it be measured? Journal of Empirical Finance 8, 35-54. Goetzmann, W.N., A. Kumar. 2008. Equity Portfolio Diversification. Review of Finance 12, 433-463. Graham, J.R., C.R. Harvey, H. Huang. 2009. Investor Competence, Trading Frequency, and Home Bias. Management Science 55, 1094-1106. Grauer, R.R., N.H. Hakansson. 1987. Gains from International Diversification: 196885 Returns on Portfolios of Stocks and Bonds. Journal of Finance 42, 721-739. Grinblatt, M., M. Keloharju. 2000. The Investment Behavior and Performance of various Investor Types: a study of Finland’s’ unique data set. Journal of Financial Economics 55, 43-67. Grinblatt, M., M. Keloharju. 2001. How Distance, Language, and Culture Influence Stockholdings and Trade. Journal of Finance 56, 1053-1073.

REFERENCES

129

Grote, M.H., M.P. Umber. 2006. Home Biased? A spatial analysis of the domestic merging behavior of US firms. Johann-Wolfgang-Goethe-Universität Frankfurt am Main, Working Paper. Grubel, H.G. 1968. Internationally Diversified Portfolios: Welfare Gains and Capital Flows. American Economic Review 58, 1299-1314. Gwartney, J., R. Lawson, J. Hall. 2012. Economic Freedom Dataset. Economic Freedom of the World: 2012 Annual Report of the Fraser Institute. Availible at: http://www.freetheworld.com/datasets_efw.html. 06.04.2014. Hau, H. 2001. Location Matters: An Examination of Trading Profits. Journal of Finance 56, 1959-1983. Heath, C., A. Tversky. 1991. Preference and Belief: Ambiguity and Competence in Choice under Uncertainty. Journal of Risk and Uncertainty 4, 5-28. Heathcote, J.S., G.L. Violante. 2007. Consumption and Labour Supply with partial Insurance: An Analytical Framework. Federal Reserve Bank. Hiraki, T., A. Ito, F. Kuroki. 2003. Investor Familiarity and Home Bias: Japanese Evidence. Asia-Pacific Financial Markets 10, 281-300. Hofstede, G. 1980.Culture’s Consequences: International Differences in WorkRelated Values. Beverly Hills CA, Sage Publications. Hofstede, G. 2001. Cultures consequences (2nd ed.). Thousand Oaks, CA, Sage Publiacation. Hofstede, G. 2011. Dimensionalizing Cultures: The Hofstede Model in Context. Online Readings in Psychology and Culture 2. Availible at: http://dx.doi.org/10.9707/2307-0919.1014. 09.03.2014 Hong, H., J.D. Kubik, J.C. Stein. 2004. Social Interaction and Stock-Market Participation. Journal of Finance 59, 137-163. Huang, R.D., C.-Y. Shiu. 2009. Local Effects of foreign ownership in an emerging financial market: Evidence from qualified foreign institutional investors in Taiwan. Journal of Financial Management 38, 567-602. Seasholes. M. 2000. Smart Foreign Traders in Emerging Markets. Working Paper, UC-Berkeley, Berkeley, CA: Huberman, G. 2001. Familiarity Breeds Investment. Review of Financial Studies 14, 659-680.

130

REFERENCES

Ivković, Z., S. Weisbenner. 2005. Local Does as Local Is: Information Content of the Geography of Individual Investors‘ Common Stock Investments. Journal of Finance 60, 267-306. Jacquillat, B., B. Solnik. 1978. Multinational are poor tools for Diversification. Journal of Portfolio Management 4, 8-12. Jeske, K. 2001. Equity Home Bias: Can Information Cost explain the Puzzle? Federal Reserve Bank Economic Review 86, 31-42 Kahneman, D., A. Tversky. 1979. Prospect Theory: An Analysis of Decision under Risk. Econometrica 47, 263-292. Kalemli-Ozcan, S., B. Sørensen, O. Yosha. 2003. Risk Sharing and Industrial Specialization: Regional and International Evidence. American Economic Review 93, 903-18. Kang, H.C., D.W. Lee, K.S. Park. 2010. Does the difference in valuation between domestic and foreign investors help explain their distinct holdings of domestic stocks? Journal of Banking and Finance 34, 2886-2896. Kang, J.-K., R.M. Stulz. 1997. Why is there a home bias? An analysis of foreign portfolio equity ownership in Japan. Journal of Financial Economics 46, 3-28. Ke, D., L. Ng, Q. Wang. 2010. Home Bias in foreign Investment Decisions. Journal of International Business Studies 41, 960-979. Kerr, N.L., R.S. Tindale. 2004. Group Performance and Decision Making. Annual Review of Psychology 55, 623-655. Keynes, J.M. 1936. The General Theory of Employment, Interest and Money. London, Macmillan. Kilka, M., M. Weber. 2000. Home Bias in International Stock Return Expectations’. Journal of Behavioral Finance 1, 176-192. Knight, F., 1921, Risk, Uncertainty and Profit, Houghton Mifflin, Boston. Kogut, B., H. Singh. 1988. The Effect of National Culture on the Choice of Entry Mode. Journal of International Business Studies 19, 411-432. Kose, M.A., E.S. Prasad, M.E. Terrones. 2006. How does Financial Globalization Affect Risk Sharing? Patterns and Channels. International Monetary Fund 07/238, Working Paper. La Porta, R., F. Lopez-de-Silanes, A. Shleifer, R. Vishny. 2000. Investor protection and corporate governance. Journal of Financial Economics 58, 3-27.

REFERENCES

131

Laeven, L., R. Levine. 2009. Bank governance, regulation and risk taking. Journal of Financial Economics 93, 259-275. Lane, P.R., G.M. Milesi-Ferretti. 2007. The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970-2004. Journal of International Economics 73, 223-250. Lauterbach, B., H. Reisman. 2004. Keeping Up with the Joneses and the Home Bias. European Financial Management 10, 225-234. Lessard, D.R. 1973. International Portfolio Diversification: A Multivariate Analysis for a Group of Latin American Countries. Journal of Finance 28, 619-633. Levy, H., M. Sarnat. 1970. International Diversification of Investment Portfolios. American Economic Review 60, 668-675. Lewis, K.K. 1996. Consumption, Stock Returns, and the gains from international Risk-Sharing. Working Paper. Lewis, K.K. 1999. Trying to explain Home Bias in Equities and Consumption. Journal of Economic Literature 37, 571-608. Lewis, K.K. 2000. Why do stocks and consumption imply such different gains from international risk sharing? Journal of International Economics 52, 1-35. Lucas, R.E. 1987. Models of Business Cycles. Oxford, UK. Blackwell. Lütje, T., L. Menkhoff. 2007. What Drives Home Bias? Evidence from Fund Managers’ Views. Discussion Paper 296, University of Hannover. Mace. B. 1991. Full Insurance in the Presence of Aggregate Uncertainty. Journal of Political Economy 99, 928-956. Malloy, C.J. 2005. The Geography of Equity Analysis. Journal of Finance 60, 719755. Markowitz, H. 1952. Portfolio Selection. Journal of Finance 7, 77-91. Markowitz, H. 1959. Portfolio Selection. Efficient Diversification of Investment. John Wiley & Sons. New York, Yale University Press. Martin, P., H. Rey. 2004. Financial super-markets: size matters for asset trade. Journal of International Economics 64, 335-361. Massa, M., A. Simonov. 2006. Hedging, Familiarity and Portfolio Choice. Review of Financial Studies 19, 633-685. Matsumoto, A., R.P. Flood, N.O. Marion. 2009. International Risk Sharing During the Globalization Era. International Monetary Fund 09/209, Working Paper.

132

REFERENCES

Merton, R.C. 1980, On estimating the Expected Return on the Market: An Explanatory Investigation. Journal of Financial Economics 8, 1-39. Morgan, D.P., K. Samolyk. 2003. Geographic diversification in Banking and its Implications for Bank Portfolio Choice and Performance, Federal Reserve Bank of New York, Working Paper Morosini, P., S. Shane, H. Singh. 1998. National Cultural Distance and Cross-Border Acquisition Performance. Journal of International Business Studies 29, 137158. Morse, A., S. Shive. 2006. Patriotism in your portfolio. Journal of Financial Markets 14, 411-440. Obstfeld, M. 1994. Risk-Taking, Global Diversification, and Growth. American Economic Review Association 84, 1310-1329. Obstfeld, M., K. Rogoff. 1996. Foundations of International Macroeconomic. MIT Press, Cambridge, MA. Pagano, M., O. Randl, A.A. Röell, J. Zechner. 2001. What makes stock exchanges succeed? Evidence from cross-listing decisions. European Economic Review 45, 770-782. Pástor, L. 2000. Portfolio Selection and Asset Pricing Models. Journal of Finance 55, 179-223. Portes, R., H. Rey. 2000. The Determinants of cross-border equity flows. CEP Discussion Papers, Center’s Globalization Programm, LSE. Working Paper. Portes, R., H. Rey. 2005. The Determinants of cross-border equity flows. Journal of International Economics 65, 269-296. Reisen, H., J. Williamson. 1994. Pension Funds, Capital Controls and Macroeconomic Stability. OECD Development Centre Working Paper 98, OECD Publishing. Rowland, P.F, L.L. Tesar. 1998. Multinationals and the gains from international diversification. Review of Economic Dynamics 7, 789-826. Sarkissian, S., M.J. Schill. 2004. The Overseas Listing Decision: New Evidence of Proximity Preference. Review of Financial Studies 17, 769-809. Seasholes, M.S., M. Tao, Z. Yang. 2010. Individual Investors and Portfolio Choice. Journal of Finance, 65, 1987-2010. Sercu, P., R. Vanpée. 2007. Home Bias in international Equity Portfolios: A Review. Working Paper.

REFERENCES

133

Sharpe, W.F. 1954. Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk. Journal of Finance 19, 425-442. Shenkar, O. 2001. Cultural Distance Revisited: Towards a More Rigorous Conceptualization and Measurement of Cultural Differences. Journal of International Business Studies 32, 519-535. Shiller, R.J. 1993. Macro Markets: Creating Institutions for Managing Society’s largest Economic Risk. Clarendon Lectures in Economics Series. Oxford University Press. Shiller, R.J., F. Kon-Ya, Y. Tsutsui. 1996. Why did the Nikkei Crash? Expanding the Scope of Expectations Data Collection. Review of Economics and Statistics 78, 156-164. Shukla, R.K., G.B. van Inwegen. 1995. Do Locals perform better than Foreigners?: An Analysis of UK and US Mutual Fund Managers. Journal of Economics and Business 47, 241-254. Shupp, R.S., A.W. Williams. 2008. Risk preference differentials of small groups and individuals. Economic Journal 118, 258-283. Solnik, B. 1974a. Why not diversify internationally rather than domestically. Financial Analysts Journal 51, 89-94. Solnik, B. 1974b. An Equilibrium Model of the International Capital Market. Journal of Economic Theory 8, 500-524. Solnik, B. 2000. International Investments. Addison-Wesley Ed. 4, Reading, Mass. Solnik, B., L. Zuo. 2013. Relative Optimism and Home Bias. Working Paper. Sørensen, B.E., Y. Wu, O. Yosha, Y. Zhu. 2007. Home bias and international risk sharing: Twin puzzles separated at birth. Journal of International Money and Finance 26, 587-605. Spector, P.E., C.L. Cooper, J.I. Sanchez, M. O’Driscoll, K. Sparks, P. Bernin, A. Büssing, P. Dewe, P. Hart, L. Lu, K. Miller, L.F.R de Moraes, G.M. Ostrognay, M. Pagon, H. Pitariu, S. Poelmans, P. Radhakrishnan, V. Russinova, V. Salamatov, J. Salgado, S. Shima, O.L. Siu, J.B. Stora, M. Teichmann, T. Theorell, P. Vlerick, M. Westman, M. Widerszal-Bazyl, P. Wong, S. Yu. 2001. Do national levels of individualism and internal lucus of control relate to well-being: an ecological level international study. Journal of Organizational Behavior 22, 815832.

134

REFERENCES

Stockman, A.C, L.L. Tesar. 1995. Tastes and Technology in a Two-Country Model of the Business-Cycle: Explaining International Co-movements. American Economic Review 85, 168-185. Strong, N., X. Xu. 2003. Understanding the Equity Home Bias: Evidence from Survey Data. Review of Economics and Statistics 85, 307-312. Stulz, R.M. 1981. On the Effects of Barriers to International Investment. Journal of Finance 36, 923-934. Stulz, R.M. 2005. The Limits of Financial Globalization. Journal of Finance 60, 1595-1638. Suh, J. 2005. Home Bias among Institutional Investors: A Study of the Economist quarterly portfolio poll. Journal of the Japanese and International Economies 19, 72-95. Tesar, L.L. 1993. International Risk-Sharing and Non-Traded Goods. Journal of International Economics 35. 69-89. Tesar, L.L., I.M. Werner. 1995. Home bias and high turnover. Journal of International Money and Finance 14, 467-492. Townsend. R.M. 1994. Risk and Insurance in Village India. Econometrica 62, 539591. United Nations. 2002. Manual on Statistics of International Trade in Services. United Nations Publications, New York. Available at: unstats.un.org/unsd/publication/Seriesm/Seriesm_86e.pdf. 09.03.2014. Uppal, R. 1993. A General Equilibrium Model of International Portfolio Choice. Journal of Finance 48, 529-553. Uppal, R., T. Wang. 2003. Model Misspecification and Underdiversification. Journal of Finance 58, 2465-2486. Van den Steen, E. 2004. Rational overoptimism (and other biases). American Economic Review 94, 1141-1151. Van Niewerburgh, S., L. Veldkamp. 2009. Information Immobility and the Home Bias Puzzle. Journal of Finance 64, 1187-1215. Van Wincoop, E. 1994. Welfare gains from international risk sharing. Journal of Monetary Economics 34, 175-200. Warnock, F.E. 2002. Home Bias and High Turnover reconsidered. Journal of International Money and Finance 21, 795-805.

REFERENCES

135

Warnock, F.E., C. Cleaver. 2003. Financial centers and the geography of capital flows. International Finance 14, 467-492. Yates, J.F., L.S. McDaniel, E.S. Brown. 1991. Probabilistic Forecasts of Stock Prices and Earnings: The Hazard of Nascent Expertise. Organizational Behavior and Human Decision Processes 49, 60-79. Zhou, C. 1998. Dynamic Portfolio Choice and Asset Pricing with differential Information. Journal of Economic Dynamics and Control 22, 1027-10151. Zoromé , A. 2007. Concept of Offshore Financial Centers: In Search of an Operational Definition. International Monetary Fund 07/87, Working Paper.

Studienreihe der Stiftung Kreditwirtschaft an der Universität Hohenheim Bände 1 - 11 sind nicht mehr lieferbar. Band 12: Axel Tibor Kümmel: Bewertung von Kreditinstituten nach dem Shareholder Value Ansatz, 1994; 2. Aufl.; 1995. Band 13: Petra Schmidt: Insider Trading. Maßnahmen zur Vermeidung bei US-Banken; 1995. Band 14: Alexander Grupp: Börseneintritt und Börsenaustritt. Individuelle und institutionelle Interessen; 1995. Band 15: Heinrich Kerstien: Budgetierung in Kreditinstituten. Operative Ergebnisplanung auf der Basis entscheidungsorientierter Kalkulationsverfahren; 1995. Band 16: Ulrich Gärtner: Die Kalkulation des Zinspositionserfolgs in Kreditinstituten; 1996. Band 17: Ute Münstermann: Märkte für Risikokapital im Spannungsfeld von Organisationsfreiheit und Staatsaufsicht; 1996. Band 18: Ulrike Müller: Going Public im Geschäftsfeld der Banken. Marktbetrachtungen, bankbezogene Anforderungen und Erfolgswirkungen; 1997. Band 19: Daniel Reith: Innergenossenschaftlicher Wettbewerb im Bankensektor; 1997. Band 20: Steffen Hörter: Shareholder Value-orientiertes Bank-Controlling; 1998. Band 21: Philip von Boehm-Bezing: Eigenkapital für nicht börsennotierte Unternehmen durch Finanzintermediäre. Wirtschaftliche Bedeutung und institutionelle Rahmenbedingungen; 1998. Band 22: Niko J. Kleinmann: Die Ausgestaltung der Ad-hoc-Publizität nach § 15 WpHG. Notwendigkeit einer segmentspezifischen Deregulierung; 1998. Band 23: Elke Ebert: Startfinanzierung durch Kreditinstitute. Situationsanalyse und Lösungsansätze; 1998. Band 24: Heinz O. Steinhübel: Die private Computerbörse für mittelständische Unternehmen. Ökonomische Notwendigkeit und rechtliche Zulässigkeit; 1998. Band 25: Reiner Dietrich: Integrierte Kreditprüfung. Die Integration der computergestützten Kreditprüfung in die Gesamtbanksteuerung; 1998. Band 26: Stefan Topp: Die Pre-Fusionsphase von Kreditinstituten. Eine Untersuchung der Entscheidungsprozesse und ihrer Strukturen; 1999. Band 27: Bettina Korn: Vorstandsvergütung mit Aktienoptionen. Sicherung der Anreizkompatibilität als gesellschaftsrechtliche Gestaltungsaufgabe; 2000. Band 28: Armin Lindtner: Asset Backed Securities – Ein Cash flow-Modell; 2001. Band 29: Carsten Lausberg: Das Immobilienmarktrisiko deutscher Banken; 2001. Band 30: Patrik Pohl: Risikobasierte Kapitalanforderungen als Instrument einer marktorientierten Bankenaufsicht – unter besonderer Berücksichtigung der bankaufsichtlichen Behandlung des Kreditrisikos; 2001. Band 31: Joh. Heinr. von Stein/Friedrich Trautwein: Ausbildungscontrolling an Universitäten. Grundlagen, Implementierung und Perspektiven; 2002. Band 32: Gaby Kienzler, Christiane Winz: Ausbildungsqualität bei Bankkaufleuten – aus der Sicht von Auszubildenden und Ausbildern, 2002.

Band 33: Joh. Heinr. von Stein, Holger G. Köckritz, Friedrich Trautwein (Hrsg.): E-Banking im Privatkundengeschäft. Eine Analyse strategischer Handlungsfelder, 2002. Band 34: Antje Erndt, Steffen Metzner: Moderne Instrumente des Immobiliencontrollings. DCFBewertung und Kennzahlensysteme im Immobiliencontrolling, 2002. Band 35: Sven A. Röckle: Schadensdatenbanken als Instrument zur Quantifizierung von Operational Risk in Kreditinstituten, 2002. Band 36: Frank Kutschera: Kommunales Debt Management als Bankdienstleistung, 2003. Band 37: Niklas Lach: Marktinformation durch Bankrechnungslegung im Dienste der Bankenaufsicht, 2003. Band 38: Wigbert Böhm: Investor Relations der Emittenten von Unternehmensanleihen: Notwendigkeit, Nutzen und Konzeption einer gläubigerorientierten Informationspolitik, 2004. Band 39: Andreas Russ: Kapitalmarktorientiertes Kreditrisikomanagement in der prozessbezogenen Kreditorganisation, 2004. Band 40: Tim Arndt: Manager of Managers – Verträge. Outsourcing im Rahmen individueller Finanzportfolioverwaltung von Kredit- und Finanzdienstleistungsinstituten, 2004 Band 41: Manuela A. E. Schäfer: Prozessgetriebene multiperspektivische Unternehmenssteuerung: Beispielhafte Betrachtung anhand der deutschen Bausparkassen, 2004. Band 42: Friedrich Trautwein: Berufliche Handlungskompetenz als Studienziel: Bedeutung, Einflussfaktoren und Förderungsmöglichkeiten beim betriebswirtschaftlichen Studium an Universitäten unter besonderer Berücksichtigung der Bankwirtschaft, 2004. Band 43: Ekkehardt Anton Bauer: Theorie der staatlichen Venture Capital-Politik. Begründungsansätze, Wirkungen und Effizienz der staatlichen Subventionierung von Venture Capital, 2006. Band 44: Ralf Kürten: Regionale Finanzplätze in Deutschland, 2006. Band 45: Tatiana Glaser: Privatimmobilienfinanzierung in Russland und Möglichkeiten der Übertragung des deutschen Bausparsystems auf die Russische Föderation anhand des Beispiels von Sankt Petersburg, 2006. Band 46: Elisabeth Doris Markel: Qualitative Bankenaufsicht. Auswirkungen auf die Bankunternehmungsführung, 2010. Band 47: Matthias Johannsen: Stock Price Reaction to Earnings Information, 2010. Band 48: Susanna Holzschneider: Valuation and Underpricing of Initial Public Offerings, 2011. Band 49: Arne Breuer: An Empirical Analysis of Order Dynamics in a High-Frequency Trading Environment, 2013. Band 50: Dirk Sturz: Stock Dividends in Germany. An Empirical Analysis, 2015. Band 51: Sebastian Schroff: Investor Behavior in the Market for Bank-issued Structured Products, 2015. Band 52: Helena Kleinert: The International Diversification Puzzle: Home Bias in Countries’ Investment Portfolios, 2016.