Financial literacy as a determinant of equity home and foreign bias

Transcription

1 Financial literacy as a determinant of equity home and foreign bias Matthias Feldhues This version: February 2017 Abstract Contributing to the solution of the home bias puzzle, this research establishes financial literacy as a determinant of holdings of foreign equities. Relying on country-level data for 58 domestic and 100 destination countries, I analyze the relationship between domestic investors financial literacy and international diversification in their holdings of foreign equities. Moreover, I determine whether financial literacy moderates the established effect of informational frictions on holdings of foreign equities in financial markets. Results confirm both hypothesized relationships. Home bias is lower in countries whose residents possess relatively higher financial literacy. Furthermore, the negative impact of information costs as expressed by geographical distance on cross-border equity investment is lower when the domestic population has comparatively high financial literacy. Correspondingly, informational linkages between countries shown to lower information asymmetry accounted for by higher values of bilateral imports are associated with markedly higher bilateral holdings of foreign equities if domestic investors exhibit comparatively high financial literacy. JEL classification: F30, G11, G15, I20 Keywords: Financial Literacy, International Finance, Portfolio Choice Preliminary draft: Please do not cite or distribute without permission. WHU Otto Beisheim School of Management, Allianz Endowed Chair of Finance, Burgplatz 2, Vallendar, Germany, Tel.: , matthias.feldhues@whu.edu

2 1 Introduction Equity portfolios of investors across countries suffer from severe underdiversification because they are tilted towards domestic securities. This bias towards holding domestic securities despite the well documented benefits of international portfolio diversification (Levy and Sarnat, 1970) is known as the home bias puzzle in the literature on international and macro-finance and the integration of financial markets. It is the subject of studies focusing on equilibrium theory in international financial markets and behavioral determinants of portfolio choice. Research on explanations for geographical patterns in international bilateral equity investment is based on the work of Solnik (1974) and Adler and Dumas (1983) and assumes that in equilibrium, homogeneous investors hold equities in proportion to their respective market capitalization, enabling them to maximize risk-adjusted returns. Deviations from these optimal market portfolio weights result in underdiversification. Urged by evidence that households neither save enough (Bernheim et al., 2001; Campbell, 2006) nor hold well diversified portfolios of risky assets (Calvet, Campbell, and Sodini, 2007), investor-level research analyzing the relationship between investor characteristics and measures of portfolio diversification shows that while financial illiteracy is widespread around the world (van Rooij, Lusardi, and Alessie, 2011a; Lusardi and Mitchell, 2011a), higher financial literacy is positively associated with diversification of the equity portfolio by ownership of foreign equities. This is attributed to the ability of financially literate individuals to gather and process information at a lower cost, allowing them to make better informed decisions (Lusardi and Mitchell, 2014; van Rooij, Lusardi, and Alessie, 2012; Christelis, Jappelli, and Padula, 2010). This research expands the understanding of equity home bias and the deviations from optimal portfolio weights observed in international financial markets by linking the established argument on informational frictions as a barrier to foreign equity investment (Coeurdacier and Rey, 2013) with evidence that populations across countries vary in their ability to gather and process financial information due to heterogeneity with regard to financial literacy. This evidence is provided by data from the Standard & Poor s Ratings Services Global Financial Literacy 1

3 Survey which aims to assess individuals aptitude to make financial decisions based on the four dimensions risk diversification, inflation, numeracy, and interest compounding. Carried out in 148 countries throughout 2014 and published in 2015, the study offers the most extensive and consistent coverage of differences in financial literacy across countries up to today. To investigate the implications of variation in financial literacy for international equity portfolio diversification across countries, I furthermore rely on country-level data on holdings of foreign equities. Taken together, these datasets enable the first extensive analysis of the relationship between financial literacy and international portfolio diversification across countries. The empirical setting comprises annual data on domestic investors level of financial literacy as well as foreign portfolio investment for a panel of 58 domestic and 100 destination countries from 2001 to This paper extends the current scope of the literature on the covariates of cross-border equity investment in two ways. First, it proposes the level of financial literacy of a country s residents as a determining factor for the magnitude of holdings of foreign equities. By positing that financial literacy enables investors to perceive the variability of return of foreign equities more precisely, an effect shown by Burke and Manz (2014) for inflation expectations, I modify the theoretical argument that investors perceive foreign assets as riskier than domestic assets, leading to home bias (Gehrig, 1993). I hypothesize that residents of countries with an on average higher level of financial literacy hold a higher proportion of foreign equities in their equity portfolios. Second, this research argues that the relationships between bilateral holdings of foreign equities and information costs (Portes and Rey, 2005) as well as information asymmetry (Lane and Milesi-Ferretti, 2008; Van Nieuwerburgh and Veldkamp, 2009) are moderated by financial literacy because costs for information gathering and processing decline with increasing financial knowledge (van Rooij, Lusardi, and Alessie, 2012). I juxtapose the cases when domestic investors exhibit high and low financial literacy and hypothesize that countries whose residents are more financially literate are less affected by information costs and information asymmetry. Both dimensions of the hypothesized relationship between financial literacy and international 2

4 portfolio diversification are confirmed empirically in this analysis. First, higher levels of financial literacy are correlated with lower overall investment in domestic equities as evidenced by a lower home bias ratio. The results suggest that the gains from financial literacy can be economically meaningful. If, for example, the Belgian and Spanish populations were as financially literate as the Danish, their aggregate equity portfolio holdings would exhibit approximately 9.9 and 13.7 percent points less home bias. Moreover, this finding is robust when adding proxies for differences in economic development, educational attainment and culture across countries to the econometric model. Second, financial literacy influences the extent to which informational frictions in international financial markets affect bilateral holdings of foreign equities across pairs of domestic and destination countries. The negative relationship between information costs measured by bilateral distance and the foreign bias ratio capturing the magnitude of deviations from optimal portfolio weights is less pronounced if domestic investors have a high degree of financial literacy. Correspondingly, informational linkages between countries shown to lower information asymmetry accounted for by higher values of bilateral imports are associated with markedly higher bilateral holdings of foreign equities if domestic investors exhibit comparatively high financial literacy. By analyzing the impact of financial literacy on investors holdings of domestic and foreign equities, this paper answers a call for research by Calvet, Campbell, and Sodini (2009) suggesting to correlate investors behavior with their financial literacy. Due to its broad country sample, it additionally offers information about the convergence of developed and emerging markets and helps to assess the costs caused by inequality as proposed by Bekaert and Harvey (2002). Moreover, assuming market portfolio weights as a yardstick, the selection of the dependent variables allows unambiguously answering the question if financial literacy comes along with superior economic decision-making as called for by Guiso and Viviano (2015). I contribute to literature and practice in the following four ways. First, I advance research on frictions in financial markets identifying determinants of geographical patterns in international portfolio holdings of foreign equities and relate financial literacy to investors preference for domestic equities and lack of international portfolio diversification. This is of interest for 3

5 researchers, practitioners and policy-makers since findings with regard to the magnitude of home bias and foreign bias provide information on the size of risk stemming from the transmission of shocks in the financial system (Karolyi and Stulz, 2003) and pertain to variation in the cost of equity capital across countries (Lau, Ng, and Zhang, 2010). Furthermore, equity home bias is of relevance for the question whether stocks are priced globally or locally (Karolyi and Stulz, 2003). When the population of a country exhibits a high preference for domestic stocks, the importance of global determinants of the demand for domestic financial assets and their prices is reduced. Second, this paper provides a comparative international perspective on financial literacy. It uses consistent data on financial literacy for a set of domestic countries which is considerably larger compared to prior studies, most of which are singe country studies. Additionally, it accounts for a large number of emerging markets as source countries of bilateral equity investment in detail. Both will help to understand where educational programs or policy measures to mitigate low financial literacy can be most beneficial with regard to countering deadweight losses from underdiversification as well as enhancing the functioning of financial markets, e.g. by making investors less susceptible to fraud (Jappelli, 2010). Third, the research design enables the precise measurement of both, financial literacy and portfolio diversification. With the exception of Gaudecker (2015), earlier research either proxies for financial literacy using socio-demographic data such as wealth and general education or lacks asset-level holding data enabling a precise measurement of portfolio diversification. Fourth, the findings of this study may contribute to political debate regarding the design of pension schemes. A shift in policy-making around the world replaces defined benefit pension schemes with defined contribution schemes. In its course, financial decisions have to be made by individuals rather than by companies and government. As a consequence, individuals assume responsibility for saving and investing for retirement, exposing them to additional risk because their future financial well-being depends on investment decisions made years or decades in advance (Lusardi and Mitchell, 2014; Christelis, Jappelli, and Padula, 2010). The remainder of this paper is organized as follows. Section 2 reviews the literature on 4

6 equity home bias and the relationship between financial literacy and portfolio choice. Section 3 delineates the theoretical motivation. Section 4 presents the empirical strategy and data. Section 5 is devoted to the results of the analysis, additional robustness checks and followed by a discussion. Section 6 concludes. 2 Literature review The information-based explanation for equity home bias is rooted in informational frictions in financial markets causing heterogeneity across investors. These frictions put foreign investors at a disadvantage since they are not as well-informed regarding the payoffs of domestic equities as domestic investors. Examining bilateral equity holdings of Japanese, U.S. and UK investors to a set of six developed equity markets for as early as 1989, French and Poterba (1991) suggest that the overweighting of domestic stocks they discern may, amongst others, stem from the fact investors penalize foreign stocks by attributing additional risk to them rather than assessing their risk purely based on the standard deviation of their returns. Gehrig (1993) and Brennan and Cao (1997), for example, use noisy rational expectations models with two countries and one domestic stock for each to show how informational frictions can lead to home bias. In comparison to models with homogeneous information, frictions arise because domestic investors observe signals on the future performance of the domestic stock with higher precision and thus perceive the foreign stock as riskier, inducing them to tilt their portfolio towards domestic equities. While information provision is exogenous in these models, more recent work focuses on the endogenous acquisition of information (Van Nieuwerburgh and Veldkamp, 2009). In this context, investors deliberately obtain information on a specific stock to lower the variance with which its payoffs are perceived. As domestic investors are endowed with a small information advantage regarding the domestic asset, focusing their learning on this becomes rational, increasing the informational advantage and resulting in home bias. Subsequent to French and Poterba (1991), numerous studies have investigated the link between holdings of foreign equities and information asymmetries empirically. Kang and Stulz (1997) 5

7 rely on the argument of Merton (1987) that investors hold equities they know about and for Japanese stocks find support for their hypothesis that companies which export more should be better known abroad and hence have a higher proportion of foreign ownership. This is echoed by Chan, Covrig, and Ng (2005), who suggest that information costs decrease with familiarity which they proxy for using bilateral trade, geographical distance and common languages spoken. They confirm the negative association between information costs and holdings of foreign equities for a sample of 26 domestic countries. However, the literature on the role of bilateral trade in modeling equity home bias is not unequivocal. Obstfeld and Rogoff (2001) illustrate how equity home bias can emerge due to frictions in product markets. According to their reasoning, foreign equities are held as a hedge against output shocks in foreign countries the domestic country imports from. In the absence of imports from a foreign country, there is no need for domestic residents to hold its stocks. Lane and Milesi-Ferretti (2008) discuss the robustness of this finding and underscore the relevance of bilateral trade as a covariate of holdings of foreign equities in their empirical analysis, pointing out that bilateral trade can be regarded as a more general proxy for informational frictions. A related line of research emphasizes the relevance of geographical distance to capture information asymmetry stemming from remoteness, inhibited interaction between economic agents, ensuing information costs, and cultural barriers, amongst others. In this literature, distance is negatively associated with stock ownership, trades and excess mutual fund returns. The latter is shown for U.S. funds which tilt their holdings towards companies headquartered close-by (Coval and Moskowitz, 2001), while the former two are documented for Finnish investors (Grinblatt and Keloharju, 2001). Several later studies offer corroborating evidence that this finding extends to the global capital market for both, mutual fund equity allocations (Chan, Covrig, and Ng, 2005) and country-level bilateral equity holdings (Portes and Rey, 2005; Coeurdacier and Guibaud, 2011). Since distance can also be understood as a proxy for transportation costs in international trade, it is possible that the effect of distance on the holding of foreign financial assets goes through its influence on trade in goods. Aviat and Coeurdacier (2007) disentangle this potential overlap and find that the magnitude by which distance impacts bilateral foreign financial claims 6

8 of banks reporting to the Bank of International Settlements is cut in half once they control for the influence of trade. Moreover, a reduction of informational frictions may result from the harmonization of accounting standards. Seeking to explain variation in the bilateral U.S. holdings of foreign equities across destination countries, Ahearne, Griever, and Warnock (2004) uncover that destination country portfolio weights are positively associated with the proportion of the destination country s market capitalization accessible to U.S. investors domestically due to cross-listings via American Depositary Receipts (ADR) enforcing adherence to U.S. Generally Accepted Accounting Principles (GAAP). One may argue that an ADR simply makes a foreign company better known among U.S. investors (Karolyi and Stulz, 2003). However, immediate evidence on the preference of U.S. institutional investors for companies whose accounting standards conform with U.S. GAAP (Bradshaw, Bushee, and Miller, 2004) and a positive relationship between the adoption of the International Financial Reporting Standards (IFRS) in Europe during the last decade and ownership of foreign stocks by mutual funds (DeFond, Hu, Hung, and Li, 2011) supports the notion that comparability in accounting standards lowers information costs and facilitates cross-border equity investment. Besides informational frictions, the literature offers further explanations contributing to the overweighting of domestic securities and lack of international portfolio diversification observed in international financial markets. 1 Early research focuses on the hedging of non-tradable income and real exchange rate risk which leads to variation in optimal portfolios across investors in frictionless markets. However, with the exception of real exchange rate risk (Fidora, Fratzscher, and Thimann, 2007), there is little empirical support for these models (Sercu and Vanpe, 2007). A sizable part of the literature models heterogeneity across domestic investors resulting in deviations from optimal portfolio weights implied by an International CAPM based on equilibrium capital market models which introduce explicit and implicit barriers as a tax on cross-border equity investment (Cooper and Kaplanis, 1986). While explicit barriers such 1 Lewis (1999) and Karolyi and Stulz (2003) provide an excellent overview of the foundational literature on equity home bias. Sercu and Vanpe (2007) and Coeurdacier and Rey (2013) add to this by reviewing more recent advancements in this strand of research. 7

9 as capital controls, restrictions to foreign ownership and dividend withholding taxes appear too small to explain the observed levels of home bias subsequent to the liberalization of financial markets towards the end of the 20th century (Cooper and Kaplanis, 1986; French and Poterba, 1991), there is empirical support for a multitude of implicit barriers. In addition to informational frictions, these contain especially differences in legal and regulatory frameworks across countries, such as corporate governance and transparency (Dahlquist, Pinkowitz, Stulz, and Williamson, 2003; Gelos and Wei, 2005; Kho, Stulz, and Warnock, 2009; Giofré, 2014) and quality of institutions (Daude and Fratzscher, 2008). To investigate variation in the magnitude of bilateral holdings of foreign equities and transactions therein, a part of this research relies on gravity models borrowed from the literature on international trade in goods (see for example Portes and Rey (2005), Aviat and Coeurdacier (2007), Coeurdacier and Guibaud (2011)). Nevertheless, they also successfully explain bilateral trade in financial assets relying on two masses, usually domestic and destination country equity market capitalization, and multilateral resistance terms such as distance to account for relative information and transaction costs (Anderson and van Wincoop, 2003; Feenstra, 2004). Finally, a number of behavioral biases are in line with investors preference for domestic securities. In this research, investors dispose of the same information but use it differently owing to heterogeneity in beliefs about the information content in public signals (Dumas, Lewis, and Osambela, Dumas et al.). In addition to patriotism (Morse and Shive, 2011) and uncertainty avoidance (Aggarwal, Kearney, and Lucey, 2012), the literature on behavioral biases established familiarity as a determinant of home bias, suggesting that investors hold stocks of companies they know. Hence, it is difficult to empirically distinguish familiarity from informational frictions (Coeurdacier and Rey, 2013). Huberman (2001) uncovers that customers of a U.S. Regional Bell Operating Companies (RBOC) usually hold shares of this company rather than of one of the other RBOCs and attributes this to investors considering unknown stocks as more risky. Keloharju, Knupfer, and Linnainmaa (2012) confirm this result for a dataset including all Finnish investors. Døskeland and Hvide (2011) discern that Norwegian investors allocate a substantial fraction of their equity portfolio to stocks within their sector 8

10 of employment and attributes this to overconfidence, since they do not earn abnormal returns. Pool, Stoffman, and Yonker (2012) document that mutual fund managers tilt their portfolios towards stocks from their home state and understand this as indicative of familiarity heuristics. Their results as furthermore suggest that fund managers do not benefit from an information advantage with regard home state stocks because the proportion of fund portfolios invested in stocks from the home state does not earn higher returns than the stocks headquartered in the state where the fund is located or other stocks. Solnik and Zuo (2012) model a home-biased representative investor who is foreign averse consistent with familiarity and exhibits an utility function borrowed from regret theory, due to which she overweights domestic stocks to limit the potential for regret. They find empirical support for their hypothesis that expected returns on domestic stocks are lower when home bias is high based on the notion that domestic investors valuation of domestic stocks will be high when they have a large preference for domestic shares. Literature establishing financial literacy as a determinant of portfolio choice links individuals abilities and education to economic behavior with regard to portfolio diversification. These microeconomic investor-level analyses confirm that low educational attainment, financial literacy and cognitive abilities are correlated with committing investment mistakes. Financially illiterate individuals and households hold poorly diversified equity portfolios, earn lower riskadjusted returns and are less successful at accumulating wealth. They are also less likely to hold stocks at all. The theoretical argument linking financial literacy and international portfolio diversification posits that individuals with higher financial literacy incur lower costs for gathering and processing information, enabling them to make better informed decisions with regard to investments and saving (Lusardi and Mitchell, 2014; van Rooij, Lusardi, and Alessie, 2012; Christelis, Jappelli, and Padula, 2010). However, isolating the effect of financial literacy on portfolio diversification empirically is not without complications. First, despite strong evidence that the statistical effect runs from financial literacy to economic behavior, concern that survey results measuring financial literacy are subject to reverse causality since ownership of foreign equities provides an incentive to 9

11 become more financially literate may remain (Lusardi and Mitchell, 2014). Moreover, low levels of financial literacy are associated with overconfidence, higher risk aversion and differences in time preferences (Christelis, Jappelli, and Padula, 2010). In their analysis, van Rooij, Lusardi, and Alessie (2012) account for these considerations and emphasize the robustness of the impact of financial literacy on wealth accumulation for Dutch investors. Finally, data comprising information on financial literacy of investors and their asset-level portfolio holdings at the same time are rare. In a recent study, Gaudecker (2015) is the first to combine both for Dutch households. He confirms that households whose financial literacy is below median achieve comparatively lower risk-adjusted returns and shows that this effect is even larger in case these households make financial decisions on their own instead of seeking advice. Further studies can be divided into two groups based on the data used to account for financial competence and portfolio composition. The first group a single country studies which precisely explore the diversification properties of individuals portfolios of risky assets based on asset-level holding data. These are obtained from Finnish or Swedish administrative databases (Grinblatt, Keloharju, and Linnainmaa, 2011; Calvet, Campbell, and Sodini, 2009, 2007; Karlsson and Nordén, 2007; Grinblatt and Keloharju, 2001) and U.S. brokerage firms (Christelis and Georgarakos, 2013; Korniotis and Kumar, 2011; Goetzmann and Kumar, 2008; Bailey, Kumar, and Ng, 2008). However, these studies do not contain data isolating the influence of financial literacy. Instead, they proxy for competence using data on schooling, income and wealth as well as experience due to ownership of risky securities and document a positive relationship with measures of portfolio diversification. Grinblatt, Keloharju, and Linnainmaa (2011) represent an exception. They account for heterogeneity in ability among Finnish individual investors using data on IQ and reveal a markedly positive correlation with portfolio Sharpe ratios. The second group of studies employs precisely identified data on financial literacy collected with survey questions specifically designed for this purpose. In their case, asset holding data do not allow for an accurate measurement of portfolio diversification or performance. Guiso and Jappelli (2008) construct a diversification index accounting for the portfolio shares invested 10

12 in mutual and exchange traded funds and individual stocks as well as the number of stocks in the portfolio. Kimball and Shumway (2007) use a dummy variable indicating the inclusion of an internationally-diversified mutual fund in the portfolio. The authors of both studies report a positive relationship between financial literacy and portfolio diversification. Graham, Harvey, and Huang (2009) pursue a different theoretical focus. They relate self-attributed competence to cross-border investment based on the hypothesis that investors who feel more knowledgeable about the benefits and risks of holding foreign securities have a higher willingness to act based on their judgment. Their results suggest that portfolios of investors who perceive themselves as more competent are less bias towards domestic equities and that these investors exhibit higher levels of education. Related research investigates the relation of individuals cognitive and educational traits to wealth accumulation and stock market participation. A limited number of these studies dispose of survey data on financial literacy. van Rooij, Lusardi, and Alessie (2011a) show that the likelihood of stock ownership among the Dutch population rises for higher levels of financial literacy. Lusardi and Mitchell (2011a) and van Rooij, Lusardi, and Alessie (2011b) focus on retirement planning in the Netherlands and the U.S., respectively. They document a significant positive relationship between financial literacy and the likelihood to accumulate wealth to provision for retirement. 3 Theoretical motivation The main idea of this paper is to substantiate whether domestic countries whose residents are on average more financially literate exhibit less bias towards domestic equities and a higher degree of international portfolio diversification. I argue that the ability to gather and process information in financial markets is heterogeneous across populations of different countries because they differ in terms of financial literacy. Hence, domestic countries will be asymmetrically affected by informational frictions. Facing frictions of a given size, residents of countries with on average higher financial literacy should be expected to make better 11

13 informed decisions resulting in lower home bias and a higher degree of international portfolio diversification. To develop my hypotheses, I use the international capital market equilibrium model with barriers to foreign investment established by Cooper and Kaplanis (1986). First, I briefly summarize the main characteristics of their theoretical framework. Subsequently, I motivate why financial illiteracy represents a barrier to cross-border equity investment. Finally, I hypothesize how domestic investors financial literacy can be expected to influence holdings of foreign equities by moderating the impact of information costs and information asymmetries. 3.1 Model summary The model of Cooper and Kaplanis (1986) enables studying why investors hold considerably more domestic equities than predicted by the world market portfolio. It has two main features. First, it enables distinguishing barriers to bilateral investment which are faced by the investor in her domestic country from barriers present in the destination country. Second, barriers impeding foreign investment can originate from both, measurable costs such as capital flow restrictions or withholding taxes and intangible costs such as poor investor protection and information costs for foreign investors. Chan, Covrig, and Ng (2005) adopt this model to analyze the relationship between foreign equity investment and variables capturing amongst others economic development, stock market development, familiarity, quality of governance and capital flow restrictions. Hence, only a brief overview of the theoretical framework is given here and the reader is referred to Cooper and Kaplanis (1986) for an in-depth understanding. In their work, the economy consists of a set of countries, each populated by one representative investor i who maximizes wealth for a given level of variance and who incurs deadweight costs domestically and in the destination country j when holding securities of country j. Deadweight costs are modeled as a tax proportional to the net investment in securities of foreign country j. Moreover, investors have homogeneous expectations about gross real returns on risky assets, risk aversion and the same covariance matrix of asset returns. They do not have access to a risk-free asset and consume an identical bundle of goods for which purchasing power parity is 12

14 assumed to hold. In this model, deadweight costs which do not equal zero lead to deviations of investors portfolio weights from those implied by an International CAPM. The deviation rises with increasing marginal deadweight costs, indicating higher levels of home bias. More specifically, home bias originates domestically if deadweight costs for investments in country i are smaller for the domestic investor i than the weighted average deadweight cost all investors face when investing into i. In this case, deadweight costs can for example be imagined as a withholding tax levied by country i which exclusively targets foreign investors or as a case of asymmetric information because information on companies from country i is difficult to obtain for foreigners. Home bias is induced abroad if the average weighted deadweight cost investor i faces across all destination countries is larger than the deadweight cost she faces when investing domestically. This translates to for example on average worse investor protection and higher risk of expropriation abroad than domestically. 3.2 Hypotheses My research postulates that financial illiteracy represents an intangible deadweight cost which impedes the holding of foreign equities. Intuitively, this is appealing since research measuring financial literacy confirms that across countries, between 50 and 70 percent of the population is ignorant about the benefits of international portfolio diversification (van Rooij, Lusardi, and Alessie, 2011a; Lusardi and Mitchell, 2011a). Moreover, Bailey, Kumar, and Ng (2008) empirically confirm that investment in foreign equities is positively associated with U.S. investors understanding of the benefits of portfolio diversification. This understanding rises for higher levels of financial literacy. Kimball and Shumway (2007) obtain the same result and underscore this reasoning. Theoretically, the main argument that financial illiteracy inhibits investment in foreign equities is supported by the model of portfolio choice of Gehrig (1993). It establishes that investors perceive foreign assets as riskier than domestic assets, leading to variation in the perceived variance-covariance matrices of investors across countries (also see Kang and Stulz (1997)). 13

15 As a result, investors in different countries hold dissimilar equity portfolios which deviate from the market portfolio and are biased towards domestic equities. The link between the model of Gehrig (1993) and variation in bilateral barriers to foreign equity investment is established in the literature (Giofré, 2014). I make the critical assumption that for higher levels of financial literacy, deadweight cost associated with holding foreign risky assets are expected to decline because financial literacy facilitates a more precise observation of the risk of foreign securities due to lower costs for information gathering and processing. This view is supported by Burke and Manz (2014) who show in an experiment that economic and financial literacy helps individuals to make more precise inflation forecasts because they select more relevant information and deploy the information they are provided with more effectively than individuals with comparatively low economic and financial literacy. In terms of the theoretical framework of Cooper and Kaplanis (1986), financial literacy is accordingly expected to lower home bias because it facilitates a more precise observation of the risk of foreign equities, leading to a decrease of the average weighted deadweight cost to investor i through a reduction of the deadweight cost on foreign equities. Hence, I formulate my first hypothesis: Hypothesis I: Equity home bias is lower in domestic countries whose population has a comparatively high level of financial literacy. Hypothesis I is tested against the null hypothesis that domestic investors level of financial literacy has no influence on the portfolio weights of domestic and foreign equities. Lower levels of home bias indicate that a higher proportion of wealth is held in foreign equities as a whole. However, the home bias ratio does not enable studying whether foreign equities of different countries are held in proportion to their respective weights in the optimal portfolio. Hence, even in cases when investors hold domestic equities in proportion to their weight in the world market portfolio, benefits of international portfolio diversification with regard to the reduction of portfolio risk may not be fully exploited because actual foreign country weights 14

16 deviate from optimal weights (Chan, Covrig, and Ng, 2005). To discern whether a comparatively higher level of education in finance is positively related to a lower deviation of observed portfolio weights from optimal portfolio weights of individual foreign countries, I study the foreign bias ratio. The foreign bias ratio towards foreign country j increases when the observed portfolio weight of country j equities in the portfolio of investor i approximates the weight of country j equities in the optimal portfolio. Furthermore, the foreign bias ratio captures bilateral investment between pairs of domestic and destination countries. This enables observing variation in the size of informational frictions between countries across country pairs and by moderating proxies for these frictions by financial literacy allows for a better understanding of the transmission channel through which financial literacy affects economic behavior in financial markets. As van Rooij, Lusardi, and Alessie (2012) emphasize, the costs investors incur for gathering and processing information decrease with increasing financial knowledge. This directly affects explanations of geographical patterns in bilateral foreign equity investments which relate differences in bilateral equity holdings to information costs and information asymmetry. Portes and Rey (2005) assume a positive correlation of information costs and distance, for instance due to weaker business relationships, cultural distance and longer travel distance, hampering information transmission. Thus, investors with high financial literacy can be expected to be less affected by information costs in comparison to less financially literate peers. For example, investors with high financial literacy may better compensate for a lack of knowledge about companies and the economy of a particular foreign country because they are able to build up this information more quickly and at a lower cost. Variation in the ability to gather and process information may have a corresponding influence on the information advantage hypothesis established by Van Nieuwerburgh and Veldkamp (2009). The authors develop a two country setting in which the domestic investor, endowed with a small information advantage with regard to the payoff of domestic asset, chooses to learn more about the domestic asset rather than about the foreign asset. Since the foreign investor behaves accordingly, both extend their comparative advantage with regard to the respective domestic 15

17 asset. The specialization leads to home bias and is associated with expected excess returns stemming from the difference between expected asset payoff and its price. Learning more about the domestic asset than the foreign investor facilitates observing information signaling a high payoff of the domestic asset while instead of reflecting this, its price only accounts for the amount of information the average investor observes. Aviat and Coeurdacier (2007) suggest the magnitude of bilateral trade as main proxy for information asymmetry in international financial markets. According to them, the effect of distance on bilateral holdings of foreign equity runs through its influence on trade in goods. In the context of the model of Van Nieuwerburgh and Veldkamp (2009), bilateral trade allows for an information advantage concerning the trade partner country. The larger the advantage, the higher the incentive for the domestic investor to specialize in a specific destination country and hold its stock. I argue that this relationship is moderated by financial literacy due to differences in the ability to gather and process information so that the effect of an information advantage on bilateral foreign equity holdings is larger for countries whose residents are highly financially literate in comparison to countries where financial literacy is on average low. Ensuing, I propose the following two hypotheses: Hypothesis II: The relationship between bilateral equity investment and information costs is moderated by domestic investors financial literacy. When the domestic population disposes of high financial literacy, the negative association between bilateral country distance and domestic investors foreign bias is less pronounced than in cases where the domestic population exhibits low financial literacy. To disentangle potential overlap between geographical distance and bilateral trade in goods as proxies for informational frictions, hypothesis three focuses on the moderating effect of financial literacy on the association of bilateral holdings of foreign equities and bilateral trade. Hypothesis III: The relationship between bilateral equity investment and information asymmetry 16

18 is moderated by domestic investors financial literacy. When the domestic population disposes of high financial literacy, the positive association between bilateral trade and domestic investors foreign bias is more pronounced than in cases where the domestic population exhibits low financial literacy. Hypotheses II and III are tested against the null hypothesis that the relationships between bilateral equity investment and proxies for informational frictions are the same irrespective of differences in domestic residents financial literacy across countries. 4 Empirical strategy and data 4.1 Cross-border equity investment To analyze cross-border portfolio holdings of foreign equities, this paper makes use of aggregate bilateral asset holding data gathered from the International Monetary Fund s Coordinated Portfolio Investment Survey (CPIS). For each participating domestic country, the CPIS provides a breakdown of holdings of foreign equities by destination country measured in U.S. dollars. Since 2001, end-of-year data are available on an annual basis for a broad set of up to 67 domestic countries. One initial survey conducted in 1997 included only 29 domestic countries because several major investing countries did not participate. The CPIS data are widely accepted in the literature studying the geography of international equity portfolio holdings (Solnik and Zuo, 2012). But although they are not subject to the same imprecisions confounding estimates of holdings of foreign equities elicited from capital flow data (Tesar and Werner, 1995), there are drawbacks requiring precaution (Lane and Milesi- Ferretti, 2008). First and foremost, the data are affected by the holding of equities through a third party. In case that an investor from country A holds foreign equities of country B through a custodian located in country C and data on holdings are not gathered using an end-investor survey but rather based on information collected from custodians in country A, 17

19 the assets held through C are not apportioned to the holdings of foreign equities of country B. This is of particular interest since offshore and onshore financial centers are home to financial intermediaries attracting significant foreign investment which is rather invested through than in these countries. This distorts holdings data because investment into these countries is overstated. Second, the underreporting of assets represents another caveat. For example, the Bahamas exclusively report equities held through banks but not those held through mutual funds. As noted by Lane and Milesi-Ferretti (2008), underreporting may also be due to capital flight in the past e.g. in Latin America and is probable in the case of assets held in offshore centers due to tax considerations. Hence, in concordance with the existing literature, financial centers, most notably Luxembourg and Ireland, are excluded from the sample of domestic and destination countries. 2 For the analysis conducted in this paper, 100 destination countries are considered, covering more than 90 percent of average world equity market capitalization over the period from 2001 to Table 1a and b provide an overview of the destination countries and their respective average equity market capitalization and weight in the value-weighted global equity market portfolio over the sample period. To quantify home bias, this paper relies on the home bias ratio as defined by Coeurdacier and Rey (2013) and Solnik and Zuo (2012). As the former point out, in a fully integrated world financial market with homogeneous investors, a simple International CAPM would predict that the representative investor from any given country holds equities from all foreign equity markets and the domestic equity market in proportion to their relative weight in the world market portfolio. Hence, the fraction of her portfolio allocated to domestic equities should mirror the share of domestic equities in the world market portfolio. The home bias ratio is used to express the magnitude of deviations of the actual share of domestic equities held in the portfolio from the share of domestic equities in the world market portfolio. The ratio is defined as an unilateral measure as follows: HB i,t = 1 Share of foreign equities in equity holdings of country i t Share of foreign equities in world market portfolio t (1) 2 For a detailed discussion see Fidora, Fratzscher, and Thimann (2007), Lane and Milesi-Ferretti (2008) as well as Solnik and Zuo (2012). 18

20 where indices i and t denote domestic country and year, respectively. If HB i,t equals 1, the residents of country i do not hold foreign equities at all. If foreign and domestic equities are held in proportion to their weights in the world market portfolio, HB i,t amounts to 0. Values between 0 and 1 indicate home bias. For example, a value of 0.90 for HB i,t suggests that the residents of country i only hold 10 percent of the foreign equities they are expected to hold based on an International CAPM. Finally, HB i,t is negative in case that the portfolio weight of foreign equities exceeds the weight predicted by an International CAPM. The foreign bias ratio captures if domestic investor i holds equities from destination country j in proportion to their weight in the optimal portfolio. It measures the size of deviations by relating the weight of equities from destination country j in the portfolio of the domestic investor i to the weight of equities from j in the world market portfolio. Hence, this variable is specified as a bilateral measure for pairs of domestic and destination countries. The foreign bias ratio is calculated as follows: ( ) Share of equities of country j in equity holdings of country it FB ij,t = log Share of equities of country j in world market portfolio t (2) where indices i, j and t denote domestic country, destination country and year, respectively. Due to the log transformation, the foreign bias ratio equals 0 if the weight of destination country j in the equity portfolio of country i matches the weight implied by the world market portfolio. In case the portfolio weight of equities from country j exceeds the optimal weight, the foreign bias ratio will assume values larger than 0. In contrast, if the optimal weight is undercut, resulting values will be negative. It may also be the case that domestic investors do not hold equities from country j. Since the foreign bias ratio is not defined in this case, it is assumed that bilateral equity investment amounts to one U.S. dollar. Hence the log transformation can be applied without loss of information. 19

21 4.2 Financial literacy To measure financial literacy, this paper uses data from the Standard & Poor s Ratings Services Global Financial Literacy Survey which aims to assess individuals aptitude to make financial decisions based on four dimensions. These are understanding of risk diversification, inflation, numeracy and interest compounding. Carried out in 148 countries throughout 2014 by Gallup as part of its Gallup World Poll and published in 2015, the study offers the most extensive coverage of the differences in financial literacy across countries up to today. Using interviews, it surveys in excess of 150,000 adults and data meeting the originator s quality standards is available for 143 countries. To gather the data, survey participants across countries are evaluated with regard to their answers to five questions. Their answers are then mapped to the four measurement dimensions. Per country, each dimension indicates the fraction of adults answering correctly. Ensuing, the aggregate financial literacy score is constructed as the fraction of adults per country providing correct answers for at least three of the four dimensions. The five questions are (mapping to dimensions is indicated in brackets): 1. Suppose you have some money. Is it safer to put your money into one business or investment, or to put your money into multiple businesses or investments? (risk diversification) 2. Suppose over the next 10 years the prices of the things you buy double. If your income ALSO doubles, will you be able to buy less than you can buy today, the same as you can buy today, OR more than you can buy today? (inflation) 3. Suppose you need to borrow $100. Which is the lower amount to pay back: $105 or $100 plus three percent? (numeracy) 4. Suppose you put money in the bank for two years and the bank agrees to add 15 percent per year to your account. Will the bank add MORE money to your account the second year than it did the first year, or will it add the same amount of money both years? (interest compounding) 5. Suppose you had $100 in a savings account and the bank adds 10 percent per year to the account. How much money would you have in the account after five years if you did not remove 20

22 any money from the account? (interest compounding) To ascertain a high quality of results, Gallup performed face-to-face interviews in countries where telephone coverage was lower than 80 percent of the population, taking into sampling units stratified by population size, geography or both. Furthermore, data are weighted based on base sampling and post stratification weights to obtain a representative sample for each country. The former evens out potential effects of household size on selection probability. The latter employs country-level data on gender, age, and, as far as reliable information is accessible, education or socioeconomic status to counterbalance sampling and nonresponse error. Methodically, the Global Financial Literacy Survey is a continuation of the work by Lusardi and Mitchell (2014) to measure financial literacy (Lusardi and Mitchell, 2008, 2011b; van Rooij, Lusardi, and Alessie, 2011a). An overview of related studies is given in Lusardi and Mitchell (2014). Hastings, Madrian, and Skimmyhorn (2012) describe the evolution of data collection with regard to financial literacy. 4.3 Stylized facts The sample of this analysis stems from combining the CPIS dataset with the data on financial literacy. This yields a sample of 58 domestic countries for which data on home bias, foreign bias and financial literacy are available. Descriptive statistics are displayed in Table 2. The analysis comprises the years from 2001 until Table 3a and 3b provides an overview of the domestic countries in the sample, averages for both, home and foreign bias ratio as well as financial literacy for each domestic country. It is immediately apparent that home bias is prevalent in all sample countries. It is lowest in the Netherlands with an average value of 0.33 and highest in India, Bolivia and Turkey with an average value of about These three countries are also among those with the lowest rating attained for financial literacy, together with Romania. As evidenced in Table 4a, home bias ratio and financial literacy exhibit a strong negative correlation of which is statistically 21