Educated Investors and the Equity
Home Bias
Ciaran Jaras
10824952
MSc Finance – Asset Management
Amsterdam Business School
University of Amsterdam
Master’s Thesis
Supervisor: Ms D. G
üler
Statement of Originality
This document is written by Ciaran Jaras who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Acknowledgements
I would like to use this opportunity to express my gratitude towards my supervisor, Derya Güler, who offered me invaluable support in writing this thesis. Her feedback was clear and constructive, and made it possible for me to undertake this research. I would also like to thank the market intelligence team at the CFA Institute, who made available all the data which I requested. Without such data, this research would not have been possible.
Abstract
This paper seeks to understand the role of education amongst investors on their investment decisions. More specifically, we investigate whether countries with a larger proportion of CFA trained portfolio managers are synonymous with a lower equity home bias. Indeed, we find that this is the case. Furthermore, we find that countries with a greater level of education amongst portfolio managers, have a smaller likelihood of having a home bias. Additionally, we conclude that countries with a relatively lower level of financial development experience a stronger relationship between education and the home bias. We did not find sufficient evidence to suggest that a significant increase in CFA trained portfolio managers leads to a reduction in the level of home bias within a given country. Our data consists of a panel of forty countries, spanning the time period 2007-2016.
Table&of&Contents& 1.#Introduction#...#1# 2.#Theoretical#Framework#...#3# 2.1#Equity#Home#Bias#...#4# 2.2#Institutional#Factors#...#5# 2.3#Behavioral#Factors#...#7# 2.4#Selected#Independent#Variables#...#8# 2.5#Literature#Review#of#Economic#Models#and#Results#...#13# 3.#Hypotheses#...#14# 3.1#OLS#...#14# """"""""""3.1.1"The"effect"of"education"on"equity"home"bias"during"the"financial"..."14" """"""""""3.1.2"The"effect"of"education"on"equity"home"bias"during"the"financial"crisis"..."15# ##########3.1.3"The"effect"of"education"on"equity"home"bias"in"developed"and"non>developed"countries"..."16# 3.2#Probit#...#17# 3.3#DifferenceNinNDifferences#...#18# 4.#Data#...#19# 4.1#Equity#Home#Bias#...#20# 4.2#Education#...#20# 4.3#Alternative#Independent#Variables#...#21# 4.4#Data#Limitations#...#21# 4.5#Construction#of#Variables#...#21# 4.6#Summary#Statistics#...#22# 4.7#Matrix#of#Correlations#...#25# 5.#Methodology#...#26# 5.1#Measure#of#Home#Bias#...#26# 5.1.1"CAPM"..."26" 5.1.2"ICAPM"..."27" 5.2#OLS#...#29# 5.2.1"Student"t>Test"..."30" 5.2.2"Hausman"Test"..."31" 5.2.3"Forecasted"Direction"of"Coefficients"in"OLS"Regression"..."33" 5.3#Probit#...#33# 5.4#DifferenceNinNDifferences#...#35# 5.5#Tobit#Regression#...#38# 6.#Results#...#39# 6.1#Hausman#Test#...#39# 6.2#Pooled#OLS#...#40# 6.3#Probit#...#42# 6.4#DifferencesNinNDifferences#...#43# 7.#Robustness#Checks#...#44# 7.1#Random#Effects#...#45# 7.2#Tobit#Model#...#45# 7.3#BreuschNPagan#test#...#46# 8.#Conclusion#...#46# References#...#50# Appendix#...#55#
1.! Introduction
A large body of research is dedicated to the concept of portfolio diversification. In particular, researchers have identified the potential financial gains of diversifying equity holdings internationally, and have tried to understand why most investors opt to forgo such benefits. An equity home bias arises when an investor over-weights domestic equity in their portfolio, relative to the optimal level. To this day, the rationale behind this investor behaviour remains partly unexplained. The concept of an equity home bias can be traced back to the work of Markowitz (1952) and French and Poterba (1991), the former developing the idea of optimal portfolio diversification and the latter applying those findings to an international context. Indeed, French and Poterba (1991) were followed by countless other researches, including Ahearne et al. (2004) and De Moor and Vanpee (2012), with each attempting to improve the understanding of the mechanisms which underpinned the home bias. With each additional piece of research, there comes a new factor which seeks to best explain the existence of the bias. Collectively, these factors can be categorised as “institutional “- those factors relating to any institutional barriers, i.e. transaction costs, which could impact home bias, and “behavioural”- any factor which relates to the behaviour of specific investors which could impact their investment decision.
Recent years have seen a shift of focus in home bias research, with contemporary researches proposing investor specific, or behavioural causal factors. One such factor that has been shown to influence investment decisions is the level of education in a country. Bose et al. (2015) found that higher levels of university enrolment, amongst other measures of nationwide education, contributed to a lower home bias. Furthermore, Kimball and Shumway (2010) found that education was a contributing factor to investment sophistication, which itself had a negative impact on home bias. Previous research has also highlighted some channels by which education could indirectly impact the home bias. Christelis at al. (2010) considered education under the wider spectrum of “financial literacy” and concluded that a higher literacy was consistent with better quality financial decision-making. One such financial decision is the decision to invest with a globally diversified portfolio, or in other words, to avoid a large home bias. A similar conclusion was reached by Jappelli and Padula (2013) who found that a more financially literate population was more able to produce returns from their assets, for instance by employing an efficient investment strategy. Education can be categorized as a behavioural factor, as it directly impacts the ability of an investor to make an informed investment decision. However, education, as defined in this paper, has not yet been considered as a factor which could impact the home bias. Therefore, this paper will build on the findings of existing papers, and will seek to explain cross-country differences in the
equity home bias by varying levels of education amongst portfolio managers. By considering the findings of Christelis at al. (2010), we propose that better educated portfolio managers should be better placed to make efficient investment decisions, due to a greater knowledge, cognitive ability and experience. More specifically, a portfolio manager that is a CFA charter holder is less likely to invest with a home bias, and those that do will have a less severe home bias compared to a portfolio manager that is not CFA trained. Data on the geographical breakdown of CFA charter holders and their occupations was supplied directly by the CFA Institute.
The contribution of this paper to the existing field of academic literature can be divided into three primary elements. Firstly, this paper provides a new measurement of education, which stems from an alternative outlook on the relationship between education and the home bias. Previous research has considered education as the general level of education in a country’s population. This would imply that the general public are equally and collectively responsible for the investment decisions within a country, for instance by influencing portfolio managers to invest domestically or otherwise. Alternatively, this paper proposes that the level of education of the portfolio manager themselves is a more relevant measurement. This stems from the fact that in the case of large institutional investors, for instance hedge funds, the portfolio manager has a great level of independence in their investment decisions, and are therefore not influenced by the general public. With this in mind, this paper proposes a proxy for education based on the number of portfolio managers in each country which are CFA charter holders, scaled by a number of country-specific variables. The CFA accreditation is regarded as the gold standard for investment professionals. We therefore suggest that a country with a greater proportion of CFA trained portfolio managers should have a lower equity home bias, after controlling for other factors. The second contribution of this paper is the inclusion of a Difference-in-Differences model, which seeks to explain the impact of education on home bias across time. Whereas previous literature has only considered the cross-country impact of education, this paper will consider changes in the level of home bias within countries over time. More specifically, it will consider whether an event –defined as a large increase in CFA charter holders- can explain a reduction in the home bias in a given country. The added benefit of this approach is particularly relevant to policy makers, as it quantifies the added benefit to each country of an increase in CFA trained portfolio managers. For instance, as found in Bose et al. (2015) the effects of education in reducing the home bias varies greatly across countries (categorised by extent of financial development), so the inclusion of a Difference-in-Differences model to investigate potential heterogeneity in the causal effect of education has precedence.
The third and final major contribution of this paper is the inclusion of a Probit model. This econometric model enables us to develop a greater understanding of the relationship between education and equity home bias, by testing of the causal effect of education on the probability of a country having a bias. The traditional approach entails testing for the causal effect of the level of education on the level of home bias. In other words, the extent to which a change in education causes a change in equity home bias. Following Karlsson and Norden (2007) this paper proposes that differing levels of a variable can impact the probability of having a home bias. More specifically, we investigate the extent to which an increased level of education can decrease the probability of a country having a high equity home bias.
In addition to these primary contributions, we apply the methods of Bose et al. (2015) in testing for heterogeneity of the causal effect of education across time and across country-specific factors. More country-specifically, we investigate whether the relationship between education and the home bias remains constant during crisis and non-crisis periods. Furthermore, we control for the differing levels of financial development of the countries in our sample, and investigate whether the effect of education in reducing the level of home bias is greater in countries with a higher level of financial development.
This paper will continue with the following structure. Section 2 will give an outline of the existing literature on the topic of home bias. Section 3 will introduce the hypotheses that will be tested in this paper. Section 4 gives a brief description of the data collection process and some of the issues encountered. Section 5 presents the methodology that is central to this research. In section 6 the results of the econometric tests will be reported and section 7 will include robustness tests of our models. Finally, section 8 will report a conclusion to the research undertaken.
2.! Theoretical Framework
The following section will introduce the most relative existing literature on the field of equity home bias, as well as a description of where this paper fits into it. This paper proposes that the quality of education amongst a country’s portfolio managers is a relevant to the level of equity home bias of that country. Crucially, this paper differentiates from previous studies by considering the education of specific individuals, rather than an entire population, as having a significant causal effect on home bias. This section will continue with brief outline of the literature on equity home bias, before giving a detailed breakdown of the potential causal factors.
This section will continue, firstly by describing the extant theoretical literature relating to equity home bias and the various causal factors. The description will focus on those variables,
both behavioural factors and institutional factors, which are widely accepted to significantly influence equity home bias. From the discussed factors, the ones most relevant to our research, due to their nature and contemporariness, are selected as additional independent variables for use in our econometric testing. Following these selections, a further brief overview of the literature relevant to each variable will be given.
The second part of this section will emphasise the technical aspects of the existing literature. That is, we will offer a review of the econometric methodology used in the relevant literature, as well as discussing their key findings.
2.1 Equity Home Bias
The phenomenon of the equity home bias has been of great interest to academics over the last sixty years. The principle of home bias can be traced to the pioneering work in the field of modern portfolio theory, by Markowitz (1952). This work developed the idea of optimal diversification within a portfolio, which would later be applied to an international context. Indeed, Levy and Sarnat (1970) build on this work by considering international diversification. Working from the viewpoint of an American investor, they find that investors would benefit from investing in countries, often developing countries, who’s stock markets have a low correlation with the US. This gain comes from an improved risk-return combination, due to a diversification-driven reduction in risk. Arguably, it was French and Poterba (1991) who first addressed the issue of home bias in great detail, and tried to identify the factors which were responsible. The authors take the benefits of international diversification as granted, and they seek to explain the existence of a strong preference from domestic stocks. They conclude that this is due to investors’ choices, as opposed to any institutional barriers. Daly and Vo (2013) consider equity home bias in Australia, and attempt to pinpoint significant causal factors. They conclude that factors such as market size, capital controls and trade all contribute to the existence of equity home bias (Daly and Vo, 2013). Finally, researchers have tried to quantify the potential benefits, which are otherwise forgone, of international diversification. Schröder (2003) considered the optimal portfolio for four potential investors; one British, one French, one German and one American, and compared it to the actual portfolios. They found that the optimal portfolio for a German investor is a 100% investment in the world portfolio, which would yield an additional 3% annual return. Similar positive gains were found for a British investor, and to a lesser extent, a French investor. Looking from the perspective of a wider range of investors, Driessen and Laeven (2007) conclude that investors from less developed, higher risk countries have more to gain from international diversification, especially if this diversification takes place outside of that
country’s region. Furthermore, they find that this diversification benefit varies over time, and reduces in relation to a decrease in country-specific risk.
Among the wealth of existing literature, several alternative causal factors have been mooted. These factors can be broadly divided amongst two classes, institutional and behavioral. Institutional factors include transaction costs, capital controls, hedging strategies, information asymmetry, financial development and alternative diversification opportunities. On the contrary, behavioral explanations include familiarity, patriotism and inattention. 2.2 Institutional factors
2.2.1 Financial liberalization
French and Poterba (1991) highlight a number of constraints imposed on investors which limit the size of their foreign stock holdings. For instance, they mention that Japanese investors cannot hold more than 30% of their portfolio in foreign stocks, whilst France face similar regulation. This clearly poses issues for investors who would otherwise opt for a greater level of international diversification. Mondria and Wu (2010) consider country-level financial liberalization over a period of sixteen years, and conclude that an increase in liberalization coincides with a gradual decrease in home bias. However, they point out that for many countries, despite a reduction due to financial liberalization, home bias remains relatively high. This is due to the existence of numerous alternative causal factors. Furthermore, Park and Mercado (2014) find evidence from the markets of emerging Asia. They conclude that an increasing financial integration, both globally and regionally, has a significantly negative impact on the level of home bias.
2.2.2 Transaction costs
One aspect of transaction costs relates to the cost of trading in a foreign currency. Schoenmaker and Bosch (2007) and Coeurdacier and Martin (2009) consider the effect of the introduction of the EMU on transaction costs. They found that costs fell significantly after the introduction and, as a result, equity home bias within the EMU fell. As a result of the lower transaction fees, investors within the EMU also invested less outside of the EMU, leading to an effective “regional bias”. Another argument relating to transaction costs concerns liquidity costs. The logic being that markets with lower liquidity entail higher illiquidity costs. French and Poterba (1991) suggest that because of the disparity in illiquidity costs, investors should favor investing in liquid markets. This in itself goes some way to explain the home bias, as countries with large investment assets tend to be more liquid, hence investing within themselves.
2.2.3 Hedging strategies
Literature on the effect of hedging has identified two potential channels of causality of an equity home bias. Cooper and Kaplanis (1994) suggest that investors feel inclined to hedge against domestic inflation. Indeed, they find that the chosen method of doing so to construct a large portion of their equity portfolio using domestic stocks. However, they concede that in many cases the role of inflation hedging is superseded by other market imperfections, such as transaction costs. Fidora et al. (2007) investigate the relationship between real exchange rate volatility and home bias across asset classes. They find that a high exchange rate volatility leads to a larger home bias for financial assets, as investors hedge against the risk.
2.2.4 Information asymmetry
Ahearne et al. (2004) argue that while many of the earlier tested variables concerning direct barriers to investment proved to be statistically significant, they were not economically motivated. Rather, the authors suggest that a lack of availability to and a poor quality of financial information is more inductive of a home bias. That is to say that US investors favor US stocks due to the relative lack of information surrounding foreign stocks. Barron and Ni (2008) consider a heterogeneous effect of asymmetric information on the home bias. They conclude that managers of larger funds are more able to collect information relating to foreign assets relative to a small fund, and this is reflected in a greater tendency of larger funds to invest internationally. Dziuda and Mondria (2012) argue that investors are generally better informed of domestic markets, therefore making them relatively less risky and leading to a home bias.
2.2.5 Financial development
De Moor and Vanpee (2012) find that a greater level of financial development within a country is conducive with greater inflow of investment into the country, which as a result reduces the domestic investors’ home bias. Furthermore, Driessen and Laeven (2007) argue that the benefits of international diversification are reducing with the financial development of a country. That is, the purpose of international diversification is to mitigate exposure to country-specific risk. However, financially developed countries tend to imply a lower risk, therefore leading to domestic investors to hold a significant portion of their portfolio in domestic stocks. Bose et al. (2015) control for this effect by including a variable relating to market turnover. The argument is, that a larger market turnover implies a greater liquidity, hence a greater financial development. As previously discussed, a greater market liquidity should lead to a larger home bias.
2.2.6 Alternative diversification opportunities (Size)
Schoenmaker and Bosch (2008) suggest that a larger relative market capitalisation of a country signifies greater investment opportunities available to domestic investors. They conclude that this availability allows investors to achieve similar diversification benefits without having to invest in foreign equity. Furthermore, Errunza et al. (1999) investigate whether the gains of international portfolio diversification can be replicated by following a solely domestic investment strategy. They concluded that in the case of the US, benefits attained from international diversification, in excess of domestic diversification, were “economically and statistically insignificant”. This implies that US investors do not need to look beyond the confines of the US stock market in order to achieve an optimally diversified portfolio. Therefore, we might expect market size to be positively related to the home bias. However, as highlighted by Bose et al. (2015), larger stock markets are more appealing to foreign investors. With a greater inflow of foreign investment, domestic home bias would decrease. Therefore, the net effect of market size on equity home bias is unclear.
2.3 Behavioral factors 2.3.1 Familiarity
Heath and Tversky (1991) introduce the concept of familiarity bias as part of a wider facet of human behavior. The authors suggest that over time, humans acknowledge that they are more successful in situations in which they are familiar with the issues, as opposed to being unfamiliar. The authors tailor this to investor behavior by concluding that investors will prefer to invest in a small number of familiar stocks at the expense of an optimally diversified portfolio. Hiraki et al. (2003) draw the same conclusion with respect to money managers in Japan. Huberman (2001) considers the case of Regional Bell Operating Companies (RBOCs) in the US. The author finds that the shareholders of each company tend to be residents of the area that the company operates, suggesting that the familiarity of the company amongst residents has a positive impact on stock ownership.
2.3.2 Patriotism
Patriotism, and the wider facet of culture, have long been mooted as a possible factor in explaining the home bias. Benos and Jochec (2013) found that patriotism was a determining factor in the construction of investment portfolios. Specifically, they find that the returns of “patriotic” stocks (stocks which include America(n) or USA in their name) earn positive abnormal returns during conflict periods, as the demand of investors increases for such stocks.
Furthermore, Morse and Shive (2011) use country-level data related to levels of patriotism, and find that such levels are strongly positively related with country-level home bias. Vosilov (2015) found evidence of a joint effect of patriotism and familiarity in explaining the home bias.
2.3.3 Inattention
Less sophisticated investors are subject to greater attention constraints (Barber and Odean, 2008). The authors find that inattentive investors are more likely to hold salient stocks in their portfolio due to their inability to monitor all available stocks in the market. As domestic stocks tend to be more salient, there is a tendency for these investors to have a home bias. Karlsson and Norden (2007) also draw the conclusion that there is a negative relationship between investor sophistication and the home bias. Large institutional investors have greater resources, and therefore lesser attention constraints, than “unsophisticated” investors (Barber and Odean, 2008). Indeed, Darvas and Schoenmaker (2017) find that investment funds characterized by larger assets tend to have a lower home bias. Having taken into account of the findings of the existing literature, the following variable shave been constructed to control for effects on the home bias.
2.4 Selected independent variables 2.4.1 Education
The direct relationship between education and equity home bias is one that hasn’t yet been fully investigated. Bose et al. (2015) went some way to address this literary void by considering the extent to which a higher quality of education within a country’s population could reduce the home bias. Specifically, they considered three measures of education, relating to school enrolment rates and attained aptitude tests scores. They conclude that, after controlling for other variables, education did indeed explain a relatively lower equity home bias. Karlsson and Norden (2007) considered education as one of many investor specific factors, which could impact the home bias within Swedish mutual funds. In similarity to Bose et al. (2015), this paper measured education with respect to school enrolment rates. The authors concluded that these measures of education could not significantly explain differences in the home bias. Kimball and Shumway (2010) consider whether the general sophistication of an investor is related to the home bias. Firstly, they proxy a measurement of financial education using dummy variables. They conclude that a financial education is positively related to investor sophistication, and further conclude that greater investor sophistication can explain a lower equity home bias.
All of these existing papers contribute significantly to the understanding of the relationship between education and investment decisions. However, the existing literature falls short in fully capturing the effects of all relevant forms of education. The aforementioned papers all consider education to be a measurement of the general level of education within a country. This implies that all portfolio allocations will be made by, or under the instruction of, an average individual in a given country. In other words, portfolio managers solely operate under the instructions of their clients, and have no free rein to make their own investment decisions. In reality, there is a second mechanism by which education can affect investment decisions. Whist the former mechanism may relevant to cases in which shareholders have influence over the investment strategies, for instance the 401K pension plan, it is not relevant to those cases, such as hedge funds, in which portfolio allocation independence is given to the portfolio managers. In that case, the importance of the level of education of the portfolio manager takes precedence. In order to proxy for the education level of portfolio managers, a variable which scales the number of CFA charter holders per country can be created. Based on data provided by the CFA institute, it is apparent that the position of Portfolio Manager is the most popular occupation for charter holders, and is therefore a suitable measure for education. The CFA is arguably the gold standard of professional qualifications within the finance industry and is, therefore, a requirement for the majority of portfolio management positions. However, with CFA charter holders being so sought after, it could be the case that smaller firms, or firms in less developed economies are less able to attract charter holders. As a result, such firms would be forced to employ less qualified portfolio managers. If this is the case, we could expect to see less sophisticated investment strategies, hence a larger equity home bias, among countries with a lower relative number of CFA charter holders.
2.4.2 Financial Literacy
It can be argued that, in comparison to education, financial literacy is a broader topic for which education is a proxy. Christelis at al. (2010) found that cognitive ability, which included proxies for financial literacy, was positively related to both stock market participation and high-quality investment decisions. Cole et al. (2011) consider the role of financial literacy in the context of demand for broader financial services. They conclude that greater financial literacy leads to a greater demand for financial services. This finding is echoed by van Rooij et al. (2011) who found that financial literacy negatively impacted stock market participation. They also concluded that the level of financial literacy impacted the quality of financial decision making. Furthermore, Lusardi (2012) argues that both numeracy levels and financial literacy impact financial decision making, with respect to asset choices and debt management.
Jappelli and Padula (2013) argue that financial literacy enables individuals to increase the returns on their wealth, for instance by making better informed investment decisions.
2.4.3 Trade
Following Bose at al. (2015), the ratio of the sum of exports and imports to GDP is taken as a proxy for trade. The authors argue that a larger value of trade would imply a greater exposure to cross-border shocks. Hence, a country with a high trading volume should hedge risk by holding international stocks, therefore improving portfolio diversification. Therefore, we would expect to see a negative coefficient of trade on equity home bias. On the contrary, Schoenmaker and Bosch (2008) pose an alternative hypothesis. They argue that there are two potential economic mechanisms which could dominate the effect of trade on the home bias. Firstly, if a country has a greater trading volume, its investors have a lesser need to diversify their investments, as firms in that country are already diversified through cross-border business operations. Thus, we could see a positive relationship between trade and equity home bias. Alternatively, they propose that investors in countries which have greater relative international trade are themselves, intrinsically, more internationally minded, and therefore we should see a negative coefficient of Trade on equity home bias (Schoenmaker and Bosch, 2008).
2.4.4 Financial Openness
Following Bose et al. (2015), this variable is included to control for varying levels of financial liberalisation both between and within countries. Similar in some respects to Trade, Financial Openness proxies the degree to which a country is open to international capital movements and holdings. Chinn and Ito (2006) propose an index, the Chinn-Ito index, which quantifies the degree of a countries openness using several dummy variables which relate to the stringency of regulations on international capital flow, as reported by the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER). A greater level of financial openness implies that investors have greater access to and can reap greater benefits from international diversification (in the absence of trade restrictions). As a result, we would expect to see a negative coefficient of the Financial Openness variable on the home bias. 2.4.5 Availability
Following on from Schoenmaker and Bosch (2008), we include the variable Availability to control for cross-country differences in the ratio of domestic market capitalisation to GDP. Schoenmaker and Bosch (2008) argue that a relatively larger stock market provides more
investment opportunities for domestic investors, with a greater number of domestic stocks being available. Therefore, these investors have a lesser need to look internationally for investment opportunities, as they are able to diversify their portfolio sufficiently using varied domestic stocks. The findings of Cai and Warnock (2006) support this line of thinking. Their evidence from the US equity market suggests that even after controlling for other variables, there is a strong institutional investor preference for the stocks of domestic multinationals. They conclude that this is due to the fact that investors are able to create a “foreign exposure” without ever having to hold foreign securities, due to the international operations of the domestic firms into which they do invest. Therefore, we should expect to see a positive correlation of Availability and equity home bias.
2.4.6 Euro
The majority of existing literature in the field of equity home bias has identified foreign currency risk as a contributing factor. The logic being that foreign exchange fluctuations provide an extra layer of risk to investors, which moderate any potential gains from internationally diversified holdings. Following on from Bose et al (2015) and De Moor and Vanpee (2012), foreign currency risk is controlled for using the dummy variable Euro, which signifies whether that country has adopted the Euro as its currency. The choice of proxy is particularly useful for our dataset, as a significant share (22.5%) of those countries have adopted the Euro. Taking into account the elimination of currency exchange risk (through a common currency), we should expect Eurozone countries to exhibit a relatively lower home bias, and therefore to see a negative correlation of Euro and the level of equity home bias. 2.4.7 Emerging Markets
The combined percentage share of emerging markets’ market capitalisation, relative to world market capitalisation, was suggested by Amadi (2004) as a measure for new investment opportunities to investors. The logic being that a relatively larger emerging market will provide greater investment alternatives to investors, that would otherwise only invest domestically or within other developed economies. This proves to be particularly relevant when considering the recent global financial crisis. Didier et al. (2011) show that the root cause of the crisis lay amongst the developed economies, and that emerging markets, which had a lower relative exposure, performed better than developed economies during this period. Furthermore, emerging economies such as Brazil, Russia, India and China (BRICs) were able to recover relatively quickly during the post-crisis period, being attributed by Didier at al. (2011) to high quality financial policy. With this in mind, we might expect emerging
economies to be of increasing interest to investors, hence we would see a negative impact of emerging market share on the equity home bias.
2.4.8 Crisis
Following from Bose et al. (2015) include the variable Crisis to control for the recent global financial crisis of 2008-2009. There are two mechanisms by which the financial crisis could impact the equity home bias. Firstly, a publication by the ECB showed that equity home bias increased during the crisis in the US and Japan, due the phenomenon of the flight to quality, and the greater perceived risk of holding foreign securities. On the contrary, the equity home bias in Eurozone, on the whole, fell. This could be attributed to the aforementioned currency exchange risk, which is eliminated in the case of the shared currency. The second mechanism by which the crisis could affect equity home bias is through its interaction with education. Bose et al. (2015) argue that in times of financial crisis, well informed financial decision making becomes more important. That is, a well-informed, highly-skilled investor will be more likely to identify and execute optimal trading strategies when working under strict pressure. Therefore, we should expect to see the effect of education become stronger during the crisis period and therefore a negative relationship between the interaction of crisis and education with equity home bias.
2.4.9 Professionalism
The measure of professionalism is proxied by the ratio of the assets of institutional investors to GDP, at the country level. This follows from Schoenmaker and Bosch (2008) who argued that the average institutional investor will invest with greater sophistication that the average non-institutional investor. Furthermore, Barber and Odean (2008) propose that short sale constraints and limited attention capability inhibit the ability of non-institutional investors to invest to the same degree of quality as institutional investors. In addition, Gigerenzer and Gaissmaier (2011) highlight the phenomenon of heuristics in decision making; the process whereby people simplify complex problems, consciously or unconsciously, due to their own cognitive or time constraints. With regards to investors, the resources and expertise of institutional investors outweighs that on non-institutional investors, such that they are more likely to make better informed investment decisions. From this, it could be argued that a relatively larger proportion of assets under the control of (sophisticated) institutional investors, should relate to a more optimally diversified portfolio, and therefore lower equity home bias.
2.5 Literature review of econometric models and results
This section will provide a brief overview of the existing literature that is most relevant to this study. More specifically, it will emphasize the econometric methodology that was utilised in the research, and how the methodology in our research builds on this. Furthermore, we will provide a more in-depth review of the results obtained in the existing literature, with a focus on our primary independent variable, education.
Bose et al. (2015), the most relevant paper to this research in terms of the central research focus, employs an OLS linear regression as its baseline model. In addition to a pooled OLS model, it also considers random effects and fixed effects variations. In addition, they make use of Instrumental Variables, or IV, regressions, which reduce the issue of omitted variable bias within a model. The primary focus of Bose et al. (2015) is the effect of education, in the sense of the general level across a country’s population, on a country’s home bias. The paper concludes that three measurements of education, tertiary education, mathematics test scores and financial skills all significantly reduce the level of home bias across countries. They also conclude that the equity home bias levels of less developed countries are more sensitive to changes in education levels.
Schoenmaker and Bosch (2008) focus on the changes in the levels of home bias in European countries following the adoption of the Euro. More specifically, they investigate which factors had a significant impact on the reduction in home bias. In order to address the question, they employ an OLS linear regression model, with equity home bias as the dependent variable. From their investigation, they conclude that the professionalism of institutional investors, the level of trade and the market capitalisation of a country, all have a significantly negative impact on the level of home bias.
Karlsson and Norden (2007) sought to explain the existence of equity home biases in Swedish mutual funds by a number of investor specific characteristics. They applied a multinomial logit model in order to determine the effect of these various characteristics on the probability of a given mutual fund having a home bias in their portfolios. They found that investor specific factors, such as age and gender, all impacted the likelihood of an equity home bias. In contrast to Bose et al. (2015), they found that education levels did not significantly affect the existence of a home bias. This finding may be, in part, due to their choice of measurement of education, which referred only to primary school, high school and university. Whereas our measurement of education, CFA charter holder, has strong relevance to the quality of an investor’s decision making, the distinction between high school and university education (particularly if unrelated to finance) is likely to yield small differences.
Daly and Vo (2013) investigated the causal effect of several variables on the level of equity home bias in Australia. The conducted an OLS linear regression of home bias on numerous independent variables. They concluded, in line with several preceding research papers, that both capital controls and transaction costs had a significantly positive impact on the level of home bias. Furthermore, they also find that trade and market liquidity significantly reduce the level of home bias.
3.! Hypotheses
This section outlines the hypotheses that will be tested in this paper. Initially, the hypotheses of the primary OLS model will be discussed. Following this, we will outline the hypotheses tested in the alternative models.
3.1 OLS
3.1.1 The effect of education on equity home bias
Our first three hypotheses are tested using an OLS linear regression model. The full OLS model (discussed further in section 5.2) can be formulated as follows:
!"#$,& = ()+ (+!,-./0123$,&4+ 4 (567/,8$,&+ (9:132;83$,&+ (<=>/1?/@1?10A$,&
+ (BC72D8EE123/?1EF$,&+ (G!-72$+ 4 (H!F87I13I&+ (JK1L8$,&+ (M:13N10$
+ (+)O71E1E$,&+ (++:13,8>$,& + (+5O71E1E$,&∗ !,-./0123$,&+ (+9:13,8>$,&
∗ !,-./0123$,&4 + 4 -$&
For a further explanation of the variables, see section 5.2. The following three hypotheses refer directly to equation (3.1).
The purpose of this research is to ascertain whether or not financial education affects the aggregate investment decision making within a country. In particular, we seek to understand whether better trained portfolio managers can explain a lower equity home bias. Bose et al. (2015) concluded that countries with a more highly educated population tend to have a lower equity home bias. Logic would suggest that a CFA charter holder would have a thorough understanding of the benefits of international diversification, and an accurate perception of the risks involved. Firstly, with respect to the understanding of the benefits of diversification, Ahearne et al. (2004) concluded that investors who did not have access to quality information which would enlighten them to the benefits, were more likely to invest with a home bias. On the other hand, Barron and Ni (2008) address the issue of risk perceptions, arguing that large institutional investors have a greater knowledge of international investments, for instance, through past experience. As such, where more naïve investors may deem all international
investment to be overly risky, better informed investors will have a more detailed understanding, and will be able to select the investments which they deem to be less risky.
Building on this existing literature, this research proposes that countries with a greater number of CFA trained portfolio managers, a proxy for the education of investors, should have a lower equity home bias. With that in mind, our primary hypothesis can be formulated by the following:
Hypothesis I: Equity home bias is lower in countries with a greater level of education amongst investors.
This will be tested against the null hypothesis, which suggests that education does not have an effect on equity home bias. This can be formulated numerically as follows:
"):4(+= 0
"+:4(+< 0
Where (+ refers to the coefficient in equation (3.1). The alternative hypothesis suggests that
the effect of education on the level of equity home bias will be negative, hence (+ < 0. This
follows from the logic that a better educated investor will make more optimal investment decisions.
3.1.2 The effect of education on equity home bias during the financial crisis Figure 3.1.2: Timeline of data sub-samples
Following on from Bose et al. (2015), two additional hypotheses will be tested relating to the effect of education on equity home bias. These hypotheses will compare the effect over different time periods and over two sub-samples of countries. Specifically, one time period relating to the financial crisis (2007-2010) and one non-crisis period, (2001-2016) Furthermore, we compare the effect of education on the home bias across two sets of countries, separated
(3.1.1)
2007 2010 2016
based on their financial development. Firstly, we propose the effect of education on lowering the equity home bias is more pronounced during the financial crisis. Evidence from the ECB concludes that home bias, on the whole, rose during the crisis. This was attributed to the flight to quality, as investors attached a greater risk perception to foreign equity. However, as mentioned in the previous section, perceptions of risk are not always accurate. In fact, in an economic downturn a naïve investor may allow general economic sentiment to influence their view on risk, meaning that they view foreign equity as being riskier that they are. On the contrary, a well–educated investor should be able to make a more informed judgement with regards to the risk of holding foreign equity. In other words, in times of economic crisis, the importance of being a well-educated investor is intensified, as investors are under greater pressure to identify and follow the most beneficial strategies (Bose et al., 2015).
An interaction variable of education and crisis is added to the regression model. If the coefficient on that term is significantly negative it would suggest that education becomes more important during a period of crisis, and that there is an additional causal effect for during crisis years compared to the effect of the whole sample, (+. Hence, we formulate our second
hypothesis.
Hypothesis II: The effect of education in lowering equity home bias is stronger during crisis years.
This hypothesis will be tested against the null hypothesis that the effect of education on reducing the home bias is homogenous across crisis and non-crisis periods. This can be expressed numerically as:
"):4(+5= 0
"+:4(+5< 0
Where (+5 refers to the coefficient in equation (3.1)
3.1.3 The effect of education on equity home bias in developed and non-developed countries.
In addition to testing for heterogeneity of the effect across different periods, we also follow Bose et al. (2015) in testing for changes in the effect across two sub-sets of countries. Namely, these sub-sets are based on the financial development of a country. A country is denoted as being financially developed if its GDP per capita ratio is equal to at least the average of the data set. Bose et al. (2015) use the logic that non-financially developed countries possess a
greater initial home bias, so are likely to see a stronger effect of education in reducing the bias, as there is greater capacity for further international investment.
On the contrary, we suggest that less developed countries tend to face larger barriers to international investment. These can include stricter regulations on foreign equity holding and other administrative costs, which tend to be more liberal in developed countries Bose et al., 2015). As a result, it is reasonable to expect that the effect of education on the home bias will be dampened in a non-financially developed country. This would arise from the fact that regardless of the ability of the investor to identify diversification benefits, if the costs of investing internationally outweighed the benefits, an investor would be forced to invest domestically. In other words, the investor is not necessarily constrained by their own ability, but by the institutional barriers to investing internationally from their country. In similarity to Hypothesis II we include an interaction variable of education and a dummy relating to the status of financial development. Based on the preceding argument, we propose the following hypothesis:
Hypothesis III: The effect of education in lowering equity home bias is relatively stronger in financially developed countries.
In keeping with Hypothesis II, this is tested against the null hypothesis that the effect of education is homogenous across both sub-sets of countries. This can be formulated as:
"):4(+9 = 0
"+:4(+9< 0
Where (+5 refers to the coefficient in equation (3.1). 3.2 Probit
This field of research is complicated by the fact that there is no consensus on how to define the equity home bias. For instance, whilst the levels of home bias varied greatly across our sample, they were all positive. That would suggest that all countries have a home bias. In fact, Norway had an average home bias of approximately 20% across the ten-year period, while the Philippines had close to 100%.
This paper seeks to understand how education can impact the home bias. In this section, we investigate whether education can explain the existence of a home bias. As all countries have some level of bias, we categorise countries using the same method used to distinguish
between levels of financial development. For this purpose, we denote a country as being home biased if its average home bias over the period is at least as great as the sample average of 72.6%. In order to answer this question, we use the following probit model (Stock and Watson, 2014):
Pr !"# = 1 O:=, W5, … , WY = Φ(()+ (+O:=+, (5W5, … , (YWY)
Where W5, … , WY refer to the additional regressors. A more detailed explanation is
offered in section 5.3. In order to understand the effect of education on the probability of having a large home bias, we construct the following hypothesis:
Hypothesis IV: The probability of having a home bias is lowered by a greater level of education amongst investors.
This is tested against the null hypothesis that education does not impact the probability of having a home bias. Alternatively, this can be expressed as:
"):4(+ = 0
"+:4(+< 0 Where (+ refers to the coefficient in equation (3.2) 3.3 Difference-in-differences
This paper builds on the existing literature by investigating whether an event, in this case a significant increase in CFA trained portfolio managers, can explain a reduction in the home bias on a country level. In particular, we again separate the countries into two sub-sets, this time based on the increase in CFA charter holders. Countries which experienced an increase in the scaled CFA portfolio manager variable of at least 100% between the years 2010-2013 were earmarked as the treatment group. Those countries which experienced no change were allocated to the control group.
Addressing the relationship between education and home bias from this standpoint provides an interesting proposition. Whereas past research has only sought to explain the relationship across countries, this approach allows us to test the relationship within countries. Bose et al. (2015) were able to partially explain a negative impact of education on the home bias across countries, however their paper, in addition to entirety of research on this topic, did not
(3.2)
investigate whether an increase in the level of education in a given country leads to a reduction in home bias. Given the theory discussed both before and within this section, we propose that after controlling for other variables as well as the given time period, a significant increase in education should lead to a reduction of the home bias within that country. The model used to test this proposition is:
]$& = 4 ()4+ 4 (+4W$&+ 4 (5^$+ (9_&+ (<`+$&+ ⋯ + (9bc`c$&+ 4 -$&
Where W$&4refers to the “treatment effect”. See section 5.4 for a more detailed explanation of
the model.
In order to test for the effect of an increase in education on the level of home bias within a country, we formulate our final hypothesis is as follows:
Hypothesis V: Countries that experience on increase in the education levels of investors also experience an abnormal reduction in equity home bias.
This is tested against the null hypothesis that an increase in education does not have an abnormal impact on home bias. Alternatively, this can be expressed numerically:
"):4(+= 0
"+:4(+< 0
Where (+ refers to the coefficient of the treatment variable in equation (3.3) 4.! Data
This section will give an overview of the data collection process, including a full listing of data sources for each of the variables and the measurements used to construct certain variables. This section will conclude with a brief discussion on some of the issues faced due to limitations in the available data.
The data used in this paper has been collated from a number of sources. These include the databases of the IMF, the World Bank, the OECD, Chinn and Ito as well as the archives of the CFA Institute. The data spans the period of 2007-2016, with all variables consisting of annual observations. Using the available data, a sample of 40 countries was constructed for the purpose of testing. The following sections will give a more in-depth description of the data
(3.3)
collection process, as well as providing some remarks with regards to the issues faced through data limitations.
4.1 Equity home bias
The data used to construct the dependent variable, equity home bias, is taken from the Coordinated Portfolio Investment Survey (CPIS) belonging to the IMF. This survey provided a breakdown of the foreign equity holdings of each country, as well as the amount of equity of each country that was held by foreign investors. Data on the annual market capitalisation of each country was taken from the World Bank.
The variable was constructed by applying the methodology of the International Capital Asset Pricing Model (ICAPM), which implies that the optimal proportion of domestic equity in an investor’s portfolio, is equal to the proportion of domestic market capitalisation to world market capitalisation. Therefore, the equity home bias can be defined as any surplus in domestic equity holdings over the optimal level. This can be formulated as:
!"#$ =_!6!$
$ −
eO$
`eO$
Where the first term relates to the proportion of domestic equity in an investor’s portfolio (domestic equity to total equity), and the second term represents the domestic country’s market capitalisation, relative to the world market capitalisation. The derivation of this measurement is discussed in greater detail in section 5.1.2.
4.2 Education
A proxy for education was created using data on the number of CFA trained portfolio managers per county, scaled by a number of variables. The data relating to CFA charter holders is not publically available, but was provided in full by the market research division of the CFA Institute. A breakdown was available on the number of CFA charter holders in each country, as well as their job positions.
The country-level CFA data was scaled by the following variables: GDP, population and Assets under management of institutional investors. Data on GDP was retrieved from the World Bank, whilst data on the assets of institutional investors was taken from the OECD.
4.3 Alternative independent variables
Data on exports and imports is used to form a proxy for trade. This data is taken from the World Bank. Also taken from the world bank is the market capitalisations of all emerging markets. The website of Chinn and Ito provides the full data set for the Chinn-Ito index, a measurement for financial openness. Data on financial literacy is taken from the 2016 survey of Standard & Poor.
4.4 Data limitations
The issues faced in the data collection process can be categorized by two factors; data availability and data quality. Data availability, or lack thereof, is most prominent in the data used to construct the dependent variable, equity home bias. The Coordinated Portfolio Investment Survey (CPIS) is a voluntary survey, meaning that some parties choose not to participate, or choose to only provide limited data. For instance, data was widely available on the equity holdings in each country that were owned by foreign investors. However, a large portion of countries chose not to disclose the scale of their foreign equity holdings in other countries. This made the measurement of home bias impossible for these countries, meaning that they were excluded from the sample.
Data quality posed an altogether different issue. For instance, there are cases in which full data was available for a number of countries, but they were still dropped from the sample. In line with previous literature, countries which are synonymous with lenient tax practices, such as Ireland, Luxembourg and Panama, are all excluded from the sample. This is due to the fact that a significant portion of the investments into these countries are redirected from another country, meaning that they are not representative of general investment practice, and therefore unsuitable for this sample.
4.5 Construction of variables
The dependent variable, Home Bias represents the equity home bias for each country at each point in time over the ten-year period. CFApopulation and CFAGDP represents the education of investors, which is proxied by the number of CFA charter holders employed as portfolio managers, scaled by population and GDP respectively. The additional independent variables include the following: Trade, which is the sum of exports and imports scaled by GDP, Financial Openness, which is the degree of financial openness of a country derived from the Chinn-Ito index, Availability is the domestic equity market capitalization scaled by GDP, Professionalism is the assets of institutional investors scaled by GDP, Euro indicates whether a country adopts the Euro as its currency, Emerging Share is the relative aggregate equity
market capitalization of emerging markets to the world market, Size is defined as the relative size of a country’s equity market capitalization relative to the world market, Financial Literacy gives an indication of the percentage of the population of a country that is financially literate, Crisis Period is a dummy variable which indicates whether the observation is from the period 2007-2010 and Financial Development is a dummy variable which signifies whether a country is financially developed.
4.6 Summary statistics
Table 1 reports the summary statistics of the entire sample. It can be seen that the average equity home bias is the sample is 73%, however this varies greatly across time and countries. The lowest reported home bias in the period was 16.2%, which belonged to Norway in 2015. On the contrary, home bias was as high as 99.8% in the Philippines in 2009.
Table 1: Descriptive Statistics
The CFAGDP variable had a mean of .000453. Again, this varied significantly across the sample. It was as low as .0000213 in Russia in 2013, whilst it reached .0046 in Hong Kong in 2009.
A similar variation exists for investor professionalism. It averages .278 across the sample, yet falls as low as .00011 in Greece in 2007 and reaches 1.72 in The Netherlands in 2016.
The euro variable shows that close to one quarter of the countries in our sample are adopters of the Euro, whilst the crisis variables can be interpreted as showing that 40% of our sample period is in the period of the financial crisis of 2007-2010.
1
CFApopulation and CFAGDP have a mean close to zero due to the nature of the scale. However, they are not exactly equal to zero, as suggested by the table.
Variable Obs Mean Std.Dev. Min Max
Home Bias 400 .73 .21 .27 1.00 CFApopulation1 400 0 0 0 0 CFAGDP 400 0 0 0 .01 Trade 359 .76 .68 .17 4.20 Financial Openness 360 .71 .33 0 1 Availability 394 .97 1.65 .06 12.54 Professionalism 323 .28 .36 0 1.72 Euro 400 .23 .42 0 1 Emerging Share 400 .15 .02 .11 .18 Size 394 .02 .06 0 .42 Financial Literacy 400 44.08 14.14 24 71 Crisis Period 400 .40 .49 0 1 Financial Development 400 .42 .49 0 1
In line with Bose et al. (2015), we provide summary statistics of the full sample, as well as of two sub-samples based on the financial development of the country, with a t-test for the difference in means of the sub-samples. Following King and Levine (1993) financial development is proxied by GDP per capita. A country will be denoted as being financial developed if it has an average of above average GDP per capita, and non-financially developed otherwise.
Table 2 displays the descriptive statistics for the two sub-samples. Columns 1 and 2 provide data for financially developed firms, while columns 3 and 4 relate to non-financially developed firms.
The first row of Table 2 states that the mean level of equity home bias for financially developed countries is 57%, which is significantly different (at 1% significance) that the 85% average of non-financially developed. This is in line with the expectations drawn from the literature review. De Moor and Vanpee (2012) argued that a greater financial development would stimulate investment into a country, therefore leading to a lower home bias of domestic investors. With this in mind, the summary statistics support our expectations.
Table 2: Descriptive Statistics of Sub-samples
Financially developed Non-Financially developed
Variable Mean Std.Dev. Mean Std.Dev. Difference
Home Bias .57 .17 .85 .14 .28*** CFApopulation 0 0 0 0 -.00*** CFAGDP 0 0 0 0 -.00*** Trade .94 .94 .63 .34 -.31*** Financial Openness .98 .07 .52 .30 -.46*** Availability 1.49 2.37 .60 .53 -.89*** Professionalism .43 .44 .14 .17 -.29*** Euro .38 .49 .11 .32 -.27*** Emerging Share .15 .02 .15 .02 0 Size .04 .09 .01 .02 -.03*** Financial Literacy 57.23 9.40 34.65 8.17 -22.58*** Crisis Period .41 .49 .39 .49 -.01 Financial Development 1 0 0 0 -1
In addition to the equity home bias variable, the majority of the variables within the two sub-groups are significantly different at the 1% level. For instance, our two measures of education, CFApopulation and CFAGDP are both significantly greater in financially developed countries than non-financially developed, suggesting a greater quality of
education of investors in financially developed countries. In addition, size is significantly larger in the financially developed group, suggesting a strong positive correlation between market capitalisation and GDP per capita. Furthermore, financial developed countries appear to be significantly more financially open, based on the Chin-Ito index. In addition, availability, professionalism, financial literacy and openness to trade all appear to be significantly greater in financially developed countries. Lastly, Euro is significantly greater in the financially developed group. This is intuitive as Euro a dummy variable, which is equal to one if a country uses the Euro as its main currency. Given that Eurozone countries tend to be financially developed, we would expect to see a larger portion of the “Euro” countries in the financially developed sub-sample.
On the other hand, there are two variables which are not significantly different across the two sub-samples. Again, this is intuitive as in both cases these variables refer to a macroeconomic tendency, rather than country-specific. More specifically, Emerging Share refers to the relative share of market capitalisation of emerging markets. We would expect this to be the same for both sub-samples. In addition, Crisis is a dummy variable, equal to one if the observation is in the period 2007-10. Given that crisis relates to a four-year period within a ten-year sample, we would expect to see a mean of 0.40 for both sub-samples. The small discrepancy between the means is due to a small number of missing observations.
In summary, it is apparent that differing levels of financial development constitute a significant difference in the means of the majority of our variables. In essence, financially developed countries are more open, with a greater level of education and financial literacy. Furthermore, whilst an equity home bias is present in both sub-groups, it is significantly larger amongst non-financially developed countries.
4.7 Matrix of correlations
Table 3 provides a table of correlations for our proxies for education and our other independent variables. In keeping with the “rule of thumb” from existing literature, we pay particular attention to correlations of a magnitude of 0.5 or greater (bold in the table). Large correlations
between independent variables can be an indication of multicollinearity, which lead to a reduction in significance of variables, or in some cases, changes in signs of the variables. In order to control for any possible multicollinearity, we will systematically remove independent variables from the model which are highly correlated and compare the results of the numerous separate regressions.
Table 3: Matrix of correlations
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (1) CFApop 1.000 (2) CFAGDP 0.920 1.000 (3) Trade 0.621 0.706 1.000 (4) FinOpen 0.330 0.184 0.222 1.000 (5) Professionalism 0.707 0.844 0.749 0.108 1.000 (6) Availability 0.482 0.353 0.079 0.249 0.212 1.000 (7) Euro -0.183 -0.203 0.049 0.422 -0.184 -0.129 1.000 (8) Emergingshare 0.003 -0.002 -0.023 -0.036 0.042 0.033 -0.009 1.000 (9) Size 0.221 0.183 -0.098 0.096 0.128 0.249 -0.136 0.007 1.000 (10) FinLit 0.339 0.209 0.098 0.645 0.035 0.492 0.207 -0.014 0.133 1.000 (11) Crisis -0.033 0.030 -0.058 0.031 0.011 -0.029 0.033 -0.173 -0.011 0.009 1.000 (12) FinDev 0.470 0.334 0.211 0.665 0.245 0.400 0.298 -0.026 0.247 0.794 0.017 1.000
5.! Methodology
This section will outline the methodology used in this paper. Firstly. The choice of the measure of equity home bias will be discussed and justified. Next, our motivation for the choice of econometric models used to test our five hypotheses will be reviewed. In addition to an explanation of the motivation for their usage, an in-depth explanation of the technical foundations of each model will be given.
5.1 Measure of home bias
In keeping with the majority of existing literature on this topic, this paper measures equity home bias by applying the International Capital Asset Pricing Model (ICAPM). The ICAPM is an extension of the traditional CAPM developed by Sharpe (1964) and Lintner (1965). This extension applied the existing CAPM theory to an international context, by taking into account foreign exchange risk and the broader benefits of international diversification. 5.1.1 CAPM
The introduction of the CAPM signified the birth of modern asset pricing (Rossi, 2016). Its role is to analyse the trade-off between (systematic) risk and return for tradable assets. In essence, the model derives the fair price for an individual asset based on two factors; the time value of money (given by the risk-free rate) and the systematic risk of the asset (given by the beta). The theory follows that individual assets are combined into a portfolio to make a more efficient investment. Following from the work of Markowitz (1952) which introduced the concept of an efficient frontier, Sharpe (1964) concluded that the most efficient portfolio, based on the risk-return trade-off, was the market portfolio.
The CAPM can be displayed in equation form as follows: !" = $ !%& + $ ("(!*− $ !%&)
Where !" is the expected returns of an asset, !%& is the market risk-free rate, !* is the
expected return of the market and (" represents the risk of a given asset, based on the
relationship between its past returns and the past returns of the market. More specifically, an asset’s beta is defined as follows:
(" = $
-./$(!", !*)$ 123$(!*)
