Culture's impact on institutional investors' trading frequency

International Review of Financial Analysis 31 (2014) 34–47

Contents lists available at ScienceDirect

International Review of Financial Analysis

Culture's impact on institutional investors' trading frequency Eli Beracha a, Mark Fedenia b, Hilla Skiba a,⁎ a b

Department of Economics and Finance, University of Wyoming, United States Department of Finance, Investment and Banking, Wisconsin School of Business, United States

a r t i c l e

i n f o

Article history: Received 23 January 2013 Received in revised form 12 August 2013 Accepted 4 October 2013 Available online 19 October 2013 JEL classification\: G11 G15 G23 Z10

a b s t r a c t This paper examines how cross-cultural differences influence institutional investors' trading frequency within their own portfolio. We find evidence that as cultural distance between the investors and their stock holdings increases, institutions trade with lower frequency. Findings are consistent with our hypothesis that trading frequency and cultural distance are negatively related due to increasing difficulty of interpreting investment environments in culturally distant foreign markets. We also show that traders from different cultural backgrounds behave differently when faced with information asymmetry that cultural differences generate. Specifically, we show that ambiguity aversion and lower trust relate to lower trading frequencies at home and abroad. © 2013 Elsevier Inc. All rights reserved.

Keywords: Trading frequency Institutional investor Culture Home bias Ambiguity aversion Trust

1. Introduction Culture's impact on economic exchange is a new and emerging field in international finance research. Many recent papers have documented that cross-cultural psychology is an important determinant and driver of many observed finance phenomena. As cultural differences between investors' or firms' home markets and foreign target markets increase, access to information, interpretation of information, and understanding of business and market environment becomes increasingly difficult. Cultural differences between countries have been shown to impact trade flows, foreign direct investment, portfolio flows (Aggarwal, Kearney, & Lucey, 2012; Felbermayr & Toubal, 2010; Guiso, Sapienza, & Zingales, 2009), success of cross-border mergers (Ahern, Daminelli, & Fracassi, in press; Chakrabarti, Gupta-Mukherjee, & Jayaraman, 2009), and portfolio performance (Choi, Fedenia, Skiba, & Sokolyk, 2013). In this paper, we contribute to the literature of culture and finance. Specifically, we examine culture's influence on trading frequency of institutional investors. The theory on how often rational investors

⁎ Corresponding author at: Department of Economics and Finance, University of Wyoming, 1000 E. University Ave., Laramie, WY 82072, USA. Tel.: +1 307 766 4199. E-mail address: [email protected] (H. Skiba). 1057-5219/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.irfa.2013.10.002

should trade stocks differs from much of the evidence documented in the literature. Investors trade more frequently than theory suggests, and of all factors, investor competence (Graham, Harvey, & Huang, 2009), overconfidence, self-attribution bias (Barber & Odean, 2000, 2001), and familiarity (Chan & Covrig, 2012; Coval & Moskowitz, 2001) appear to be the most consistent in explaining frequent and often excessive trading. The purpose of this paper is to study trading frequencies from a new angle. We study trading behavior of institutional investors in international setting. We test how cultural difference between an investor's home market and the markets of the investor's stock holdings (target markets) affect the trading frequencies inside each investor's own portfolio. We argue that a manager from a culturally distant market has a lower ability to access and process information compared to his or her home market or a target market that is culturally close to his or her home market. Therefore, we hypothesize that the institutional investor's trading frequency inside the manager's own portfolio will vary, so that the turnover will be the highest in the investor's home market and culturally close markets, where information asymmetry is the lowest. As cultural distance increases and the information asymmetry increases, we expect to observe declining trading volumes in the same portfolio. To further study the effect of cultural characteristics on trading frequency, we also test whether the level of ambiguity aversion, trust, and overconfidence of the investors' home country relate to trading frequency.

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

In addition to the cultural distance, we also examine the impact of geographical distance between the investor's home market and target markets on trading frequency. In a recent study by Chan and Covrig (2012), the authors document that an increase in geographical distance between an investor's home market and target markets increases trading volumes because investors are less familiar with distant target markets. Building on Chan and Covrig's (2012) findings, we argue that geographical distance is directionally dependent and that across longitudes, as fewer business hours overlap, information barriers between home markets and target markets increase. Therefore, we hypothesize that as longitudinal distance increases, the trading volumes will decrease, so that the effect of longitudinal distance is similar to the effect of cultural distance on trading volumes. Consistent with our expectations, the main finding of our analysis suggests that institutional investors trade with higher frequency in their home market and in markets that are culturally close to their home country. The results provide new evidence for observed excessive trading that takes place in financial markets, and that trading behavior changes, not only from individual to individual and from institution to institution, but also inside investors' own holdings. The decline in trading frequency is especially rapid as cultural distance increases between each investor's home market and the target markets. This suggests that, on average, investors believe that they are unable to benefit from more frequent trading in target markets where they may be informationally disadvantaged. This finding is consistent with some prior papers that provide evidence on negative relationship between the barriers to obtain information and trading and market participation activity (Coval & Moskowitz, 2001; Graham et al., 2009). The difference in trading frequencies is large in magnitude. On average, investors' home market turnover exceeds culturally close target markets' turnover by a factor of 7 and the culturally most distant target markets by as much as 21 times. In addition, we find evidence that cultural ambiguity aversion is related to lower trading frequency and that cultural trust is related to higher trading frequency. Finally, we document that geographical distance matters much more, when investor moves further away from the target market across time zones. Longitudinal distance is negatively related to trading frequencies, and it is statistically and economically significant, whereas latitudinal distance is not a significant determinant of trading frequency in most of our analyses. Our paper makes several contributions to the literature. First, our findings reveal that observed excessive trading in financial markets is also a market specific phenomenon, so that investors' trading frequency changes within investors' own portfolio. Second, we document that “home bias” exists not only with respect to asset allocation, as it has been long recognized, but also with respect to trading frequency. Investors' trading frequency appears to be the highest in markets that are culturally closest to their home market, where information asymmetry is the lowest. Third, our results add to the new and emerging stream of literature on cross-cultural psychology and investor behavior. Specifically, we focus on cultural distance between markets as a determinant of economic exchange, as opposed to many recent papers that have linked cultural characteristics of individual countries to explain firm and investor behavior (for example Beugelsdijk & Frijns, 2010; Chui, Titman, & Wei, 2010). Our fourth and final contribution is that we study culture and trading frequency in a large sample of institutional investors from 37 international markets. Our extensive holdings' dataset allows us to observe portfolio allocation of institutions at the security level across the global market place in 46 target markets. The holdings data used in the study are the most detailed international portfolio dataset of which we are aware. The remainder of the paper is organized as follows. Section 2 reviews the literature and develops our hypotheses, Section 3 details the data and methodology used in this study, Section 4 presents and discusses the results and Section 5 concludes.

35

2. Literature review and hypotheses development 2.1. Excessive trading Grossman and Stiglitz (1980) use a rational expectation framework and argue that investors only trade if the marginal benefit from trading exceeds its marginal cost. In practice, however, the finance literature provides evidence that overconfident investors trade more often than the rational expectation framework suggests. Benos (1998) shows that overconfident investors who compete with rational investors using limit orders may enjoy higher profits due to a “first mover advantage”. Contrary to Benos (1998), Caballé and Sákovics (2003) develop a theoretical model to illustrate that overconfident investors will suffer from trading too much. In their influential study, Barber and Odean (2000) show that individual investors who trade stocks most frequently earn returns significantly lower in statistical and economic terms than the market average. Moreover, the authors document that these lower returns are accompanied by investing in stocks that are riskier, on average. The hypothesis that overconfidence can explain, at least in part, excessive trading is supported in their study. Barber and Odean (2001) strengthen their overconfidence argument by showing that male traders trade more often and earn lower returns than female traders. The authors establish a link from psychology and gender overconfidence to trading volumes, so that females exhibit lower levels of confidence in areas like finance and therefore turn their portfolios over less frequently. Graham et al. (2009) study the relationship between investor competence and international asset allocation and trading frequency. Their research builds on the seminal work on the “competence effect” by Heath and Trevsky (1991). Heath and Trevsky's (1991) natural experiment suggests that when people feel more knowledgeable about a subject matter, they are more willing to bet on their own judgment instead of betting on a lottery that carries the same probability of winning. Graham et al. (2009) point out that in financial markets, investors are constantly making decisions based on subjective probabilities. The authors document that individual investors with specific characteristics feel more competent about their ability to understand financial information than others, which in turn, translates into higher trading volumes. Evidence on competency and trading frequency is also documented by Grinblatt, Keloharju, and Linnainmaa (2012) who show that individuals' level of competency is related to trading behavior and that more competent investors, proxied by their IQ, outperform less competent investors. The better than average effect with relation to trading is also examined by the literature. Dorn and Huberman (2005) observe clients of a German retail broker and show that investors who perceive themselves as more knowledgeable trade more. Glaser and Weber (2007) use survey data from an online broker's investors, and find that investors who think that they have better investment skills and believe to perform above average (but in actuality have average or below average performance) trade more than other investors. The literature on trading behavior in non-home markets is young and only few papers, complementary to ours, have explored the topic recently. Lai, Ng, Zhang, and Zhang (2013) study trading behavior by global mutual funds, but focus on the momentum patterns rather than trading frequency. The authors find that funds generally follow the same trading strategies at home and abroad, but their buy and sell intensities vary with the location of trades. That is, the momentum investing in local stocks is mostly affected by market momentum anomaly and window-dressing, and that in foreign stocks the information environment of the host country plays a leading role. Chan and Covrig (2012) find that global mutual fund investors rebalance their portfolios more often in less familiar environments, where familiarity is measured by geographical distance, trade flows, and commonality in language.

36

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

2.2. Culturally rooted behaviors and financial decision making Culture is defined as a system of shared values, beliefs, and attitudes that influence individual perceptions, preferences, and behaviors. National culture has been a topic in many recent studies in the field of financial economics, but frequently culture has been defined or measured in order to help explain variations in institutions or legal practices rather than individual investor behavior (e.g., Stulz & Williamson, 2003). In contrast, a recent series of papers by Guiso, Sapienza, and Zingales (2004, 2006, 2008) and Guiso et al. (2009) shows that cultural differences in how people trust others help explain stock market participation and other facets of portfolio investment. In this study, we investigate a link between cultural similarity and trading frequency. We employ cultural proxies from Hofstede (1980, 2001), which is one of the most influential works in cross-cultural psychology. The study identifies primary dimensions of culture and differences in thinking, values, and social behaviors among people from more than 50 nations. Hofstede's survey-based evidence shows that countries' cultural attributes can be measured in five primary dimensions: Uncertainty Avoidance, Individualism, Power Distance, Masculinity and Long-Term Orientation. Appendix A defines each of these primary dimensions in more detail. Some of Hofstede's cultural dimensions have been used recently in the international economics and finance literatures to explain economic behavior. Uncertainty avoidance has been linked to lower level of foreign equity investment (Beugelsdijk & Frijns, 2010), higher levels of under-diversification and home bias (Anderson, Fedenia, Hirschey, & Skiba, 2011), and to disproportionately slower growth in industries characterized by high information asymmetries (Huang, 2008). Individualism has been linked to self-attribution biases, trading activity, momentum patterns in stock returns (Chui et al., 2010), and higher levels of foreign equity investment (Beugelsdijk & Frijns, 2010). Hofstede's measure for masculinity has been linked to investor overconfidence and, as a result, to lower levels of under-diversification (Anderson et al., 2011). In this study we employ Hofstede's primary dimensions of culture in order to compute the cultural distance between institutional investors' home markets and target markets. We also examine the correlation between individual primary dimensions of culture and trading frequency. A few prior papers in finance literature have used cultural distance as an explanatory variable for investor and firm decision making. Aggarwal et al. (2012) and Anderson et al. (2011) show that cultural similarity relates to higher levels of foreign portfolio investment. Chakrabarti et al. (2009) and Ahern et al. (in press) study cultural distance and success of cross-border mergers. Choi et al. (2013) document better performance by institutional investors in culturally similar target markets. 2.3. Hypotheses development The main contribution of our paper is to explore whether culture influences institutional investors' trading behavior in a cross-country setting. We begin with a general hypothesis that bilateral cultural distance between the investors' home markets and target markets of the securities in the investors' portfolio reduces trading frequency. The hypothesis is motivated by literature on asymmetric information and investor competence. Previous literature shows that all financial transactions involve asymmetric information, and that the largest part of transaction costs stem from the asymmetric information, adverse selection, and agency costs (Hart, 1995). Especially foreign market financial transactions involve high levels of asymmetric information and because of it, transaction costs can be high enough for investors to shy away from foreign markets completely (Christelis & Georgarakos, 2013). Also, market-based transactions involve more asymmetric information in general as opposed to institution-based financial intermediation (Aggarwal & Goodell, 2009). In addition to the marketbased information asymmetry, differences in the market environment

across countries influence investors' ability to interpret information effectively. Findings from Arrow (1974) and Akerlof (1997) support the notion that the ability to communicate effectively declines as the cultural and societal distances between the investor and the target increase. Therefore, in cross-border transactions, there is more asymmetric information involved compared to transacting in the home markets, especially in non-home markets that are culturally distant. As a result, we expect investors to take on more passive stock positions when they operate in environments where they suffer from higher information asymmetry (Christelis & Georgarakos, 2013). This expectation is also consistent with Van Nieuwerburgh and Veldkamp's (2010) theory on optimal, but under-diversified, portfolios when information acquisition is costly. Consistent with the findings of Dorn and Huberman (2005) and Glaser and Weber (2007), we expect that investors will perceive themselves as more knowledgeable and competent in culturally close environments and in their home markets and to transact more often despite the information asymmetry. Formally, we test a hypothesis that relates bilateral cultural distance between the investors' home market and target markets to trading frequency: H1. Bilateral cultural distance is negatively related to institutional investors' trading frequency. We also test the effect of three cultural characteristics of the investors' home markets on trading behavior. First, we hypothesize that agents from ambiguity averse backgrounds will be more sensitive to information asymmetry that arises from the market-based trading and/or from bilateral cultural differences. We expect that not only information asymmetry, but also investors' attitude towards that information asymmetry relate to trading frequency. Aggarwal and Goodell (2009) assert that culture and social values of a country are an important determinant of financial intermediation, so that cultures with high levels of ambiguity aversion have a preference for financial intermediation that is institution- rather than market-based. Given that cultural ambiguity aversion relates to less market-based financial systems, cultural ambiguity aversion may also influence investors' willingness to transact in information asymmetric equity market and reduce trading frequency. More specifically, if cultural ambiguity aversion is high, investor may be less willing to act in the financial markets, which translates into lower trading frequency. The effect of cultural ambiguity aversion on trading frequency should be magnified, when the cultural distance between the investor and the target market is large. In several recent finance papers, scholars have proxied cultural ambiguity aversion with uncertainty avoidance from Hofstede (1980, 2001). For example, as mentioned in the previous section, uncertainty avoidance has been linked to lower levels of foreign equity investment (Beugelsdijk & Frijns, 2010), higher levels of under-diversification and home bias (Anderson et al., 2011), and to disproportionately slower growth in industries characterized by high information asymmetries (Huang, 2008). Formally, we test: H2a. Cultural ambiguity aversion reduces trading frequency. Second cultural dimension we relate to trading behavior is cultural trust. This hypothesis postulates that agents from less trusting backgrounds will be more sensitive to information asymmetry that may arise from the market-based trading environment and/or from bilateral cultural differences. Specifically, when cultural trust or trust in certain markets is comparatively low, we should observe lower trading frequency in the equity markets due to agents' reduced willingness to act on asymmetric information. According to Franks, Mayer, and Wagner (2006), countries with higher levels of trust have more market-based financial intermediation. Additionally, trust has been shown to relate to higher rates of market participation and savings

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

rates (Guiso et al., 2004). Trust in others has been linked to higher trade flows, foreign direct investment, and portfolio investment (Guiso et al., 2009). Formally, we test: H2b. Cultural trust increases trading frequency. The third cultural dimension we relate to trading behavior is cultural overconfidence. At individual level, overconfidence has been shown to have a positive effect on trading frequency because overconfident investors are more willing to act on subjective probabilities in the markets, as documented before by Barber and Odean (2001). In crosscountry setting, overconfidence has been proxied by cultural individualism and by cultural assertiveness (Anderson et al., 2011; Chui et al., 2010) of the country, so that more individualistic and more assertive countries' investors exhibit higher levels of overconfidence. More formally, we hypothesize: H2c. Cultural overconfidence increases trading frequency. In addition to culture and cultural distance, we study the impact of geographical distance on trading frequency. Our final hypothesis relates both longitudinal and latitudinal distances between investors' home market and target markets to trading frequency. Chan and Covrig (2012) and Coval and Moskowitz (2001) have documented a significant relationship between geographical distance and trading frequency and asset allocation. In this study we examine the relation between geographical distance and trading frequency in more detail. We test the impact of both latitudinal and longitudinal distances on trading frequency and portfolio allocation. We believe that both longitudinal and latitudinal distances will make information acquisition more difficult and result in lower trading frequency, but for different reasons. Longitudinal distance will make information acquisition more difficult because of non-overlapping trading and business hours, consistent with Portes and Rey (2005). Latitudinal distance will make information acquisition more difficult because of institutional under-development in the countries which are located near the equator (and latitudinally far from most of the investor countries in our sample). Beck, Demirgüç-Kunt, and Levine (2003) document that colonialization of the countries that are located near the equator was merely resource extraction-based and left these countries with weaker institutions, lower transparency, and weaker rule of law. We also presume that longitudinal and latitudinal distances matter with respect to portfolio allocation because of increased information asymmetry. More formally, we test our final hypothesis: H3. Longitudinal and latitudinal distances are negatively related to institutional investors' trading frequency and proportion of asset allocation. 3. Data and methodology 3.1. Data We obtain institutional holdings data from Factset Company. The holdings data comprise all of 13-F filings and similar filings from each institutional investor's home country where such information is reported. The raw sample of this dataset contains security level holdings information for over 10,000 institutions at the institutions' family level. At the portfolio level, the dataset contains more than 35,000 different portfolios, and a total of 36,070,466 institutionsecurity-time observations. This makes this dataset the most extensive institutional cross-border holdings dataset of which we are aware. The institution-security-time observations span the period between the fourth quarter of 1999 and the first quarter of 2010. For each holding, we observe how many shares of stock each institutional investor holds in its portfolio and the associated total market value (in USD) of the holding. The data on holdings at the

37

security level allow us to estimate the institutions' turnover based on the end-of quarter reported holdings in institutions' domestic markets, as well as turnovers for each individual international market in which the institutional investor has invested. Each security is matched with the company characteristics, which we also obtain from Factset. These characteristics include the name of the firm, industry, country domicile, stock exchange and the country of the stock exchange, and share type data. Our main variable of interest in this study is the cultural distance between different markets and primary dimensions of culture of the investors' home markets. To proxy for cultural distance between the investor country and the target market, we rely on Hofstede's and GLOBE's primary dimensions of culture, which are commonly used in the social sciences. We repeat all our analyses using the cultural distance measure from both, GLOBE's1 and Hofstede's cultural studies, but, for brevity, we only report the main results with Hofstede dimensions.2 The reason why we choose to report the results of Hofstede cultural dimension is that it is the most cited cross-cultural study in social psychology and widely used in international business literature. For selected robustness checks we report the results using GLOBE dimensions as well. In addition to these two datasets we also use bilateral trust in others in a subset of European economies as a proxy for cultural distance. Bilateral trust data are from Guiso et al. (2009) while country specific trust data are from World Heritage Foundation.3 Other data points for the analyses are collected from several different sources. Geographical distance and countries' coordinates, are from the Centre d'Etudes Prospectives et d'Informations Internationales (CEPII).4 We base our east–west and north–south distances, as well as the time zone differential analysis on the country's capital cities' coordinates and we compute the distances using the Latitude/Longitude Distance Calculator that is available at National Hurricane Center.5Using the CEPII dataset we are also able to determine whether the investor and target markets share a common language. The common legal standard indicator is from Djankov, La Porta, Lopez-deSilanes, and Shleifer (2008).6 We compute countries' market capitalization weights based on FTSE investable indexes. For the trading costs associated with each market we use a dataset from Elkins McSherry LLC. Elkins McSherry LLC dataset is based on more than 24 million institutional investors' transactions worldwide. The average trading cost for each market includes commissions, fees, market impact of trades, as well as the total cost of trading. From Compustat Global's Security Daily we collect the securities' closing prices, market values, and daily volumes. The volumes are used to compute a liquidity measure for each market since market liquidity may affect investors' turnover. Finally, our macroeconomic control variables, which include GDP and GDP per capita, are obtained from World Bank. The master dataset we use in the analyses is constructed in the following way. First, we merge the holdings data based on security level identifiers to the security data. Of the 36,070,466 holdings observations from 1999:4 to 2010:1, we are able to merge 35,243,928 or 97.7%to securities that report their country domicile and country exchange. Then, we restrict the sample to those securities that have price and market value information available. After we compute

1 The GLOBE cultural dimensions are from a study by House, Hanges, Javidan, Dorfman, and Gupta (2004). GLOBE incorporates data from 17,000 managers in 951 organizations in 62 countries. GLOBE replicates the Hofstede study in some ways, but expands the cultural dimensions to nine: power distance, uncertainty avoidance, institutional collectivism and in-group collectivism, assertiveness and gender egalitarianism, future orientation, humane orientation, and performance orientation. 2 The general results are similar under both measures. 3 World Heritage Foundation's trust data: http://www.worldvaluessurvey.org. 4 Centre d'Etudes Prospectives et d'Informations Internationales: www.cepii.fr. 5 Latitude/Longitude Distance Calculator: www.nhc.gov. 6 Legal Origin data is obtained from Andrei Shleifer's website: http://scholar.harvard. edu/shleifer/.

38

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

turnovers in each security (as shown in Eq. (1a) and (1b) in Section 3.2.1), we aggregate the turnovers based on the security's country of exchange, so that the resulting dataset has institutional investor-target country-time observations. Next, we restrict our analysis to those investors who are domiciled in countries for which our macroeconomic and bilateral variables are available. The final dataset results in 952,822 institutional investor-target country-time observations. The turnover data are computed for 10,635 investor families on quarterly basis. The investors included in the final sample are domiciled in 38 markets and hold securities in 46 target markets. 3.2. Methodology 3.2.1. Institutional investors' turnover in a target market The main dependent variable in the study is the portfolio turnover in market J for each investor i. Turnover is calculated relative to investor's i holdings in market J as well as relative to investor's i holdings in all nonhome markets. We measure turnover in these two distinctive ways to help clarify whether home bias in trading is compounded by home bias in holdings. We follow Barber and Odean's (2001) method of computing portfolio turnover based on the quarterly holdings' data. The quarterly portfolio turnover consists of one-half of the quarterly purchase turnover and one-half of the quarterly sales turnover, so that each target market J's turnover is computed separately. In each quarter, we identify the securities held by each investor in the beginning of quarter t. The sales, S, are computed as the number of shares sold in quarter t times the beginning of quarter price, p, of security j in USD divided by the total beginning of the quarter market value, MV, invested by investor i in market J, or in all non-home markets, J ≠ I, in USD. To calculate the quarterly purchase turnover, we take the shares purchased, B, during quarter t, times the beginning of month price per share, p, divided by the total beginning of the quarter value of the portfolio, MV, invested by investor i in market J, or in all non-home markets, J ≠ I, in USD. The resulting turnover measure, TO, for investor i in each target market J, relative to holdings in that market and relative to holdings in all non-home markets, is formally presented in Eqs. (1a) and (1b), respectively:

TOi; J;t

1 ¼  2

TOi; J;t ¼

1  2

Si;t   X p j;t  S j;t j¼1

MV i; J;t

1 þ  2

Si;t   X p j;t  S j;t j¼1

X J≠I

MV i; J;t

þ

1  2

Si;t   X p j;t  B j;t j¼1

ð1aÞ

MV i; J;t Si;t   X p j;t  B j;t j¼1

X

MV i; J;t

:

ð1bÞ

J≠I

Throughout the analysis, we classify each investor's home market holdings as those holdings where the country of security's exchange is the same as the investor's reported domicile. With this classification we treat each investor's turnover in cross-listed securities that are listed in the investor's home country as turnover in the home market. Similarly, turnover in securities that are domiciled in the investor's home market, but listed in foreign exchanges is classified as non-home market turnover. We choose this treatment of cross-listed securities (as opposed to treating cross-listed securities abroad that are domiciled in the home country as home market turnover) because we are more interested in investor's turnover in different market environments rather than in individual securities. Moreover, the number of cross-listed securities as a share of investors' portfolio is trivial. 3.2.2. Cultural distance Cultural distance from the investor's i home country I to the target market J is computed based on primary dimensions of culture from

Hofstede.7 Our cultural distance index is based on four Hofstede's primary dimensions of culture: Individualism, Masculinity, Power Distance, and Uncertainty Avoidance.8 Following Kogut and Singh (1988),9 we compute the cultural distance, CD, for each investor country I from each target market J as follows:

CDI; J ¼

 2 4 H n; J −H n;I X n¼1

Vn

=4

ð2Þ

Where Hn,I and Hn,J are the nth cultural dimension of an investor from countries I and J, respectively, and Vn is the variance of the nth cultural dimension. 3.2.3. Longitudinal distance In order to test H3,and the effect of latitudinal and longitudinal distances on trading frequency and portfolio allocation, we break down geographical distance from the investor's i home country I to the target country J into latitudinal and longitudinal distances. Because each degree of longitude is different, so that the distance between degrees is the highest at the equator, we round to the nearest 5° of latitude and use coordinates of capital cities from CEPII and the Latitude/Longitude Distance Calculator to compute the latitudinal distance between countries in kilometers.10 3.2.4. Target market liquidity When we examine institutional investors' turnover in their home markets we control for the cultural characteristics of the home market, while also conditioning on home market liquidity. To proxy for liquidity we create a volume variable by aggregating all available home market securities' daily trading volumes from Compustat Global's Security Daily into a quarterly trading volume for each country. The trading volume is the total value of shares traded each quarter of each stock scaled by the stock's average market value based on its daily closing price. The quarterly trading volume for all stocks is computed based on the aggregate quarterly trading volume of all publicly traded securities scaled by the market's total capitalization at the end of each quarter. 3.2.5. Under/overweighting of target markets Previous research has documented a link between cultural similarity and portfolio allocation (Aggarwal et al., 2012; Anderson et al., 2011). Before we explore this link in more detail, we first confirm previous findings using our dataset of institutional investors' cross-border portfolio holdings. Then, we examine whether both geographical distances across longitudes and latitudes affect portfolio investment. For this purpose, we generate a bias variable, which is defined as the difference between each investor's actual investment in a particular target market as a share of the investor's non-home market portfolio and the expected allocation to that market. The expected weight allocation of each market is the investable market value of market J relative to the total market value of the investable world according to 7 We also repeat all the analyses using cultural distances from GLOBE study, whose primary dimensions of culture are similar to Hofstede's, except that the study is newer and instead of five dimensions, GLOBE classifies national culture into nine primary dimensions. These results are reported in robustness checks in Tables 4. 8 Consistent with previous research, we omit the fifth cultural dimension, long-term orientation, from the calculation of cultural distance because it is not available for a large number of countries. 9 Kandogan (2012) offers an improvement to Kogut and Singh (1988) method. We use the method from Kandogan (2012) also as a robustness check and report results in Table 4. We however, report the main results for Kogut and Singh (1988) method, given that it is the standard method used in the literature. 10 We also examine the distance effect on trading for countries that do and do not use daylight saving time. However, the fact that our data is only available with quarterly frequency creates limitation in exploring the daylight saving time phenomenon in more detail since it is usually applied in the middle on the quarter. Therefore, we recommend that future research should address this issue using a more appropriate dataset.

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

FTSE. More formally, the under/overweight bias variable is computed as follows: MV i; J MV J biasi; J ¼ X −X MVi;I≠ J MV J J;I≠ J

ð3Þ

J;I≠ J

where bias is calculated for each investor i in each available market J, including those markets where investor i has zero investment. MVi,J is the market value of all securities investor i holds in market J, and MVJ is the market capitalization value of the corresponding market. A visual presentation of the expected investment weight for the countries included in our sample is displayed in panels A and B of Fig. 1 for the years 2000 and 2010, respectively.

39

3.2.6. Regression specifications In our main analysis, to test for the determinants of trading frequency, we employ different specifications of multivariate regressions with investor country and target country fixed effects, similar to the methodology used by Lane and Milesi-Ferretti (2008). The specification for the double fixed effect regression is the following: TOi; J≠ ¼ ϕI þ ϕ J þ β1 Z I; J þ β2 X i≠ þ εi; J≠

ð4Þ

where TOi,J,t is the quarterly turnover by investor i in the target market J in quarter t. ΦI and ΦJ are the indicator variables for each investor's home country and target countries I and J. The investor country fixed effects control for country characteristics that may explain generally higher or lower levels of turnover. The target country fixed effects control for

Expected Weights, 2000 HONG KONG 1.3%

OTHER 8.9%

ITALY 2.2% UNITED STATES 51.8% CANADA 2.3% NETHERLANDS 2.4% SWITZERLAND 2.8% GERMANY 3.3% FRANCE 4.5%

UNITED KINGDOM 9.8%

JAPAN 10.6%

Expected Weights, 2010 OTHER 19.37%

UNITED STATES 41.88%

SOUTH KOREA 2.14% BRAZIL 2.44% SWITZERLAND 3.05% GERMANY 3.20% AUSTRALIA 3.40%

FRANCE 3.98% CANADA 4.12% UNITED KINGDOM 8.15%

JAPAN 8.26%

Fig. 1. Expected weights based on float market capitalization weights in 2000 and 2010. Expected weights are computed based on FTSE investable indexes. The weight of each market is defined as the total market capitalization of each market in U.S. dollars divided by the total market capitalization of all investable markets in U.S. dollars.

40

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

features of the target country that may explain generally higher or lower turnovers in the target market. ZI,J is a vector of explanatory variables, which includes the main variable of interest for this study — the cultural distance between the investor country and the target market. ZI,J also includes other bilateral control variables for the investor and target country pairs. These variables include Trust in others, Latitudinal distance, Longitudinal distance, Time zone differential, Common legal standard and Common language. Xi,t is a vector of explanatory variables that are investor i specific, or investor i-target market J specific, and vary over time. These variables include the size of the investor's total portfolio (investor size), size of the assets in country J (Investor size in target) and the investor's experience in market J. All the specifications used in the analysis also include year indicators and the standard errors are twoway clustered errors based on investor home country–target country pairs. To ensure that multicollinearity is not a concern, we test the variance inflation factor (VIF) for the bilateral variables and never simultaneously include two variables in the regression that produce high VIF score. The VIFs as well as a correlation matrix of the control variables included in this paper are reported in Appendix C. 4. Results 4.1. Summary statistics Table 1 shows the summary statistics of the sample's institutional investors' turnover. Panel A presents the average turnovers by investor type, panel B by style, and panel C by the investors' self-reported turnover volume. We display the turnovers for each category of institutional investors at home and abroad as well as the difference between the two. Additionally, we report the number of investors that belong to each category. Panel A of Table 1 shows that the most common investor type in the sample is Investment Adviser, followed by Hedge Funds and Mutual Funds. For all investor types, except for Broker, home market turnover is, on average, larger (or statistically indifferent) compared with the nonhome market turnover. The largest differences, that are also statistically significant between home and non-home turnovers, are in Investment Bank Asset Management, Investment Adviser, and Hedge Funds. The summary statistics in panel B reveal that the home market turnover exceeds the non-home market turnover in every investor self-reported style category, but the unreported one. The largest number of investors in the sample that report their style belong to Value, GARP, Growth, and Yield investors. The largest differences in home markets and non-home market turnovers are observed in Growth and Aggressive Growth investors by a significant margin. In panel C, not surprisingly, the most frequent traders are the investors that self-report themselves as high turnover investors. Interestingly, for “Very high” turnover investors, the difference between the home and non-home market turnovers is by far the greatest. The turnover difference between home market and non-home market almost monotonically decreases from “Very high” to “Very low” turnover investors (“Very low” turnover investors actually have lower, but statistically insignificant, turnover at home). Overall, the results presented in Table 1 generally suggest that investors, regardless of type, style or classification, trade much more frequently in their home market than they do abroad. Table 2 shows the average turnovers of institutional investors based on their country of domicile. The reported turnovers are the time series equally weighted averages for each country. We report turnovers for home markets as a share of domestic portfolio (HTO) as well as for foreign markets as share of the foreign portfolio (FTO). In addition, we report cultural characteristics of the countries used in the study. These include: Individualism (IDV), Uncertainty Avoidance (UAI), TRUST, Assertiveness (ASS), and Institutional Collectivism (COLL). The table reveals that home market turnovers exceed turnovers abroad in most investor countries. The only exception where the foreign turnover is

higher and statistically significant is in Thai investors' portfolios. The difference in home market and foreign market turnover is the greatest in US investors' portfolios followed by Austrian, Taiwanese, Chinese, and Spanish investors. 4.2. Determinants of turnover In Table 3 we present the results from our first regression analyses. We first test for determinants of institutional investors' turnover in each of the target markets. As per equation (1b), we measure turnover for each of the institutional investor's target markets as a share of the investor's total portfolio market value in all non-home markets. The main variables of interest are cultural distance along with other bilateral control variables. In order for the results to be consistent with H1, we expect Cultural distance to be negatively related to institutions' turnover. Common language and common Legal origin may also proxy for cultural differences, and we expect their signs to be positive. In addition to cultural distance from Hofstede, we also include Trust in others variable in the analysis for a subset of European investors. We expect Trust in others to take on a positive sign, implying that investors trade with higher frequencies in markets they trust more. Similar to Cultural distance, Longitudinal distance, Time zone differential, and Latitudinal distance are also expected to be negatively related to turnover, consistent with H3. We also include simple geographical Distance in the regression, and expect it to be negative. In all regressions, we also control for institutions' characteristics, which include investors' Experience, measured as quarters of presence in the target market, with Table 1 Summary statistics of institutional investors. Table 1 shows turnovers at home and abroad for institutional investors in the sample based on their self-reported type (in panel A), selfreported style (in panel B), and self-reported turnover (in panel C). The table reports the number of investors from each category (#) as well as the average turnovers at home (HTO) and foreign markets (FTO) as a share of investors' home market portfolio and foreign market portfolio, respectively. The reported turnovers are time series averages. The last two columns show the difference (Diff.) between home market and foreign market average turnover and the t-statistic of the differential. #

HTO

FTO

Diff.

T-stat

Panel A: Investor type Investment Adviser Hedge Fund Company Mutual Fund Manager Bank Management Division Broker Insurance Management Division Pension Fund Insurance Company Broker/Inv Bank Asset Mgmt Private Banking Portfolio Foundation/Endowment Market Maker Arbitrage

4229 1370 966 813 231 217 206 153 148 111 40 18 2

11.31 21.62 12.05 9.8 14.05 9.98 7.08 5.3 11.67 11.63 7.97 25.51 48.53

5.99 15.75 5.65 5 20.18 4.99 7.17 4.04 5.15 10.8 8.81 29.33 19.4

5.32 5.86 6.4 4.8 −6.13 4.99 −0.09 1.26 6.52 0.83 −0.85 −3.83 29.13

[22.51]⁎⁎⁎ [4.18]⁎⁎⁎ [14.64]⁎⁎⁎ [12.61]⁎⁎⁎ [−2.81]⁎⁎⁎ [6.42]⁎⁎⁎

[0.57] [0.24] n/a n/a

Panel B: Investor style Value GARP Growth Yield Deep Value N/A Aggressive Growth Index

2484 2126 1208 1153 912 309 274 38

15.38 11.29 16.91 9.91 13.86 8.09 14.69 8.5

5.87 7.24 6.6 6.97 8.14 17.45 6.2 2.44

9.5 4.06 10.31 2.94 5.72 −9.36 8.49 6.06

[22.94]⁎⁎⁎ [9.07]⁎⁎⁎ [19.04]⁎⁎⁎ [6.27]⁎⁎⁎ [7.48]⁎⁎⁎ [−2.75]⁎⁎⁎ [9.25]⁎⁎⁎ [3.5]⁎⁎⁎

Panel C: Investor turnover N/A Medium Low Very Low High Very High

5060 915 887 681 639 322

10.503 16.02 10.35 6.47 25.91 35.35

5.732249 7.18 5.1 7.34 15.91 27.24

4.77 8.84 5.25 −0.87 10 8.12

[25.52]⁎⁎⁎ [15.11]⁎⁎⁎ [11.42]⁎⁎⁎

⁎ Significant at 10% level. ⁎⁎ Significant at 5% level. ⁎⁎⁎ Significant at 1% level.

[−0.12] [1.11] [6.49]⁎⁎⁎

[−0.95] [6.13]⁎⁎⁎ [2.31]⁎⁎⁎

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

41

Table 2 Turnovers of institutional investors at home and in non-home markets. Table 2 shows the cultural characteristics of the investors' home markets used in the analysis. These include Individualism (IDV), Uncertainty avoidance (UAI), Trust, Assertiveness (ASS), and Institutional collectivism (COLL). The table also reports the number of investors from each country (#) as well as the average turnovers at home (HTO) and foreign markets (FTO) as a share of home market portfolio and foreign market portfolio, respectively. The reported turnovers are time series averages. The last two columns show the difference between home market and foreign market average turnover and the t-statistic of the differential. COUNTRY

IDV

UAI

TRUST

ASS

COLL

#

HTO

FTO

Diff.

T-stat

Argentina Australia Austria Belgium Brazil Canada China Czech Rep. Denmark Finland France Germany Greece Hong Kong Hungary India Ireland Israel Italy Japan Malaysia Mexico Netherlands New Zealand Norway Poland Portugal Singapore South Africa South Korea Spain Sweden Switzerland Taiwan Thailand UK USA

46 90 55 75 38 80 20 58 74 63 71 67 35 25 80 48 70 54 76 46 26 30 80 79 69 60 27 20 65 18 51 71 68 17 20 89 91

86 51 70 94 76 48 30 74 23 59 86 65 112 29 82 40 35 81 75 92 36 82 53 49 50 93 104 8 49 85 86 29 58 69 64 35 46

0.176 0.461

4.22 4.28 4.62

3.66 4.29 4.3

4.2 4.05 3.76

3.83 4.38 4.77

3.8 3.81 4.13 4.55 4.58 4.67 4.79 3.73 3.92 4.23 4.07 3.59 3.87 4.45 4.32 3.42

4.8 4.63 3.93 3.79 3.25 4.13 3.53 4.38 4.63 4.46 3.68 5.19 4.61 4.06 4.46 4.81

4.06 3.65 4.17 4.6 4.4 4.42 3.38 4.51 3.92 3.64 4.15 4.55

4.53 3.92 4.9 4.62 5.2 3.85 5.22 4.06 4.59 4.03 4.27 4.2

3.73 8.49 13.39 6.95 7.73 9.20 15.02 8.79 7.68 7.48 9.05 11.25 11.71 11.17 6.86 13.93 6.95 8.58 11.33 11.10 9.90 11.31 7.92 5.71 8.37 7.56 9.50 10.31 10.69 8.34 12.73 9.23 9.51 16.61 6.79 13.15 15.36

7.51 8.56 4.66 3.79 17.22 10.38 6.67 2.07 3.68 2.91 5.20 5.87 3.94 5.21 2.16 7.69 7.00 3.38 4.75 7.91 5.30 5.54 8.41 8.25 5.74 4.76 5.90 4.45 6.63 11.52 4.58 4.44 6.47 8.22 8.10 6.89 3.73

−3.78 −0.06 8.74 3.17 −9.50 −1.18 8.35 6.73 4.00 4.57 3.85 5.39 7.77 5.95 4.70 6.24 −0.05 5.20 6.58 3.19 4.59 5.77 −0.49 −2.54 2.63 2.80 3.60 5.86 4.06 −3.18 8.15 4.79 3.04 8.39 −1.31 6.26 11.64

[−1.29] [−0.06] [8.58]⁎⁎,⁎⁎⁎ [2.34]⁎⁎⁎

0.094 0.428 0.523

6 194 95 54 59 392 100 14 71 66 317 389 30 137 12 58 46 61 129 162 65 7 63 18 54 53 59 112 108 65 234 137 400 68 52 884 3728

0.589 0.188 0.368 0.411 0.233

0.292 0.391 0.088 0.156 0.45 0.512 0.742 0.19

0.188 0.282 0.2 0.68 0.539 0.242 0.415 0.305 0.393

[−1.46] [−1.62] [6.46]⁎⁎⁎ [5.11]⁎⁎⁎ [3.55]⁎⁎⁎ [5.20]⁎⁎⁎ [4.68]⁎⁎⁎ [8.57]⁎⁎⁎ [5.90]⁎⁎⁎ [6.39]⁎⁎⁎ [3.56]⁎⁎⁎ [2.44]⁎⁎⁎ [−0.03] [2.78]⁎⁎⁎ [7.43]⁎⁎⁎ [3.06]⁎⁎⁎ [4.31]⁎⁎⁎ [2.44]⁎⁎⁎ [−0.32] [−0.83] [1.26] [1.86]⁎ [1.60] [5.76]⁎⁎⁎ [2.37]⁎⁎⁎ [−1.32] [11.2]⁎⁎⁎ [5.17]⁎⁎⁎ [4.54]⁎⁎⁎ [3.77]⁎⁎⁎ [−2.11]⁎⁎⁎ [7.37]⁎⁎⁎ [20.13]⁎⁎⁎

⁎ Significant at 10% level. ⁎⁎ Significant at 5% level. ⁎⁎⁎ Significant at 1% level.

a negative expected sign, investors' total value of portfolio (Investor size), and investors' total market value in each target market (Investor size in target). We expect this latter market value to be positively related to turnover, as larger holdings are naturally traded more in dollar terms. Investor country, target country and year indicators are included in all of the regressions. Consistent with our hypothesis, both panels suggest a negative and statistically significant relation between the cultural distance between the investor and the target market. Cultural distance is negatively related to turnover with statistical significance. Of the other cultural variables, common Legal origin is positive and significant while Common Language and Trust in others are not significant. Interestingly, of the geographic bilateral controls, only the Longitudinal distance, Time zone differential, and simple Distance are negative and significant. Latitudinal distance, however, is insignificant in both specifications. These results suggest that the distance effect on investors' turnover is directionally depended. The turnover declines substantially as investors move farther from their home market, but more so when they move away from their home market across time zones. Overall, the results from the bilateral controls provide support to H1 and H3.11 11 As a robustness test we also independently examine the relation between longitudinal and latitudinal distances with trading frequency in four different regions of the world (Asia, Europe, North America and Pacific). While the result of the sign of longitudinal distance coefficient is mostly stable, the sign of the latitudinal distance coefficient is less robust.

As for the investor control variables, Table 3 reveals that experience is negatively related to trading frequency. A possible explanation is that after the initial accumulation of a position, investors decrease the frequency of their trading activity. While investors' total portfolio size is weakly positively related to turnover, the coefficient of investors' portfolio size in the target market is positive and significant, consistent with expectation. Table 4 repeats the analyses from Table 3, but with several different cultural proxies. The dependent variable is investors' turnover in target market I as a share of the investors' total portfolio market value in nonhome markets. The variables of interest are the investor-target country pair variables that vary across country pairs and include several measures of cultural differences between the investor and the target market. These independent variables include the Cultural distance between the investor and the target market, measured based on Hofstede's primary dimensions of culture (Eq. (2)) from Table 3. The alternative cultural variables include: Globe, which is the cultural distance between investor and target market from GLOBE study's nine primary dimensions of culture. CD, SD, and MD are the Correlation Distance, Standardized Euclidean Distance, and Mahalanobis Distance respectively , all computed using Hofstede's four dimensions of culture based on Kandogan (2012). Kandogan suggests these measures to control for some cross-correlations in Hofstede's primary dimensions of culture as an improvement to the Kogut and Singh's (1988) method. Investor level controls are the same from Table 3. The results reported in

42

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

Table 3 Determinants of investors' turnover relative to all holdings abroad. Table 3 shows results from OLS regressions, where the dependent variable is the investors' turnover in target market J as a share of the investors' total portfolio market value in non-home markets (Eq. (1b)). All the specifications include investor and target country fixed effects. The variables of interest are the investor-target country pair variables. These bilateral variables include the Cultural distance between the investor and the target market, measured based on Hofstede's primary dimensions of culture (Eq. (2)), Trust in others for the European subset of the sample, geographical Distance in the log of kilometers, the log of distance in latitude and longitude in kilometers (Distance, latitude and Distance, longitude), Time zone differential in hours, common Legal origin indicator, and a Common language indicator between the investor and the target market. Investor level controls include investors' Experience in each target market J, measured in quarters present. Investor size and Investor size in target in logs, which are the quarterly total market value of the investor and the quarterly market value invested in each target market J, respectively. All regressions also include year indicators. The standard errors are two-way clustered errors based on home country–target country pairs. The t-statistics values are reported under the coefficients.15 15 We repeat the analysis of this table only with countries that use daylight savings time with respect to time zone differential and longitudinal distance. The results are similar in magnitude and significance to the results reported above and available upon request. (1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

−0.4067⁎⁎⁎ [6.01]

Cultural Distance Trust in others

−0.2907⁎⁎⁎ [4.64] 0.4172 [1.07]

Distance

−0.7783⁎⁎⁎ [5.75] −0.1395 [1.32]

Distance, latitude Distance, longitude

−0.5709⁎⁎⁎ [5.38]

Time zone differential

−0.2502⁎⁎⁎ [5.68]

Legal origin

0.0453 [0.70] −0.5318⁎⁎⁎ [5.17]

0.7115⁎⁎⁎ [3.78]

Common language −8.0165⁎⁎⁎ [5.69] 0.1737⁎⁎⁎

Experience Investor size in target Investor size Observations Adjusted R2

[4.93] 1.0697⁎ [1.84] 231,620 0.1853

−5.7281⁎⁎⁎ [6.23] 0.0504 [1.27] 3.7371⁎⁎ [2.31] 58,718 0.1829

−8.0028⁎⁎⁎ [5.52] 0.1430⁎⁎⁎ [3.55] 1.1065⁎ [1.84] 231,620 0.1910

−8.1050⁎⁎⁎ [5.61] 0.1803⁎⁎⁎ [4.91] 0.9703 [1.61] 231,620 0.1811

−7.9902⁎⁎⁎ [5.54] 0.1458⁎⁎⁎ [3.65] 1.1363⁎ [1.91] 231,620 0.1902

−8.0567⁎⁎⁎ [5.69] 0.1493⁎⁎⁎ [3.85] 1.0364⁎ [1.73] 231,620 0.1914

−8.0257⁎⁎⁎ [5.68] 0.1748⁎⁎⁎ [4.88] 1.1039⁎ [1.87] 231,620 0.1836

0.2566⁎ [1.67] −0.1207 [0.70] −7.9151⁎⁎⁎ [5.62] 0.1402⁎⁎⁎ [3.57] 1.2008⁎⁎ [2.08] 231,620 0.1936

⁎ Significant at 10% level. ⁎⁎ Significant at 5% level. ⁎⁎⁎ Significant at 1% level.

Table 4 mainly confirm the results from Table 3 that support H1. All cultural distance measures are negative and statistically significant with the exception of the Globe cultural distance. The result is still negative, but statistically insignificant. In Table 5, we report results from regressions that allow for investors' home country characteristics to vary while fixing the characteristics of the target country. This type of regression allows us to test whether investors' home country specific cultural characteristics influence trading behavior in addition to the bilateral cultural and geographic variables. The investor country specific cultural characteristics include Uncertainty avoidance, Individualism and Trust of the investor country. We also include GDP per capita and GDP in logarithm of the investor country to ensure that omitted macroeconomic characteristics are not driving the result of the cultural characteristics. The results presented in Table 5 continue to confirm H1 and H3. Both cultural distance and distance across longitudes are negative and significant across specifications. Distance across latitudes is not significant. Common language and legal origin are positive as expected, but statistically insignificant. The country specific cultural dimensions provide support for H2a. As the Uncertainty avoidance of the investor country increases, the trading frequency in foreign markets decreases. In other words, more ambiguity averse investors are less willing to transact in equity markets when information asymmetry is higher. While Trust has the expected positive sign, it is not statistically significant. As for Individualism, it carries the expected positive sign, as per H2c, when it is included without GDP per capita control variable. However, the sign turns negative and significant when the GDP per capita is included in the regression. It should be noted, that Individualism is highly correlated with the wealth of the nation (GDP per capita), so that richer countries also tend to be more individualistic. Previous

literature documents a positive relationship between trading activity and individualism, but only without the wealth control, which may explain the difference in our results to the results in Chui et al. (2010). 4.3. Under-diversification and turnover It is possible that the relationship between trading frequency and cultural distance and longitudinal distance is driven partially by investors' asset allocation to the target markets. In this section, we first closely examine the relationship between asset allocation and cultural distance. We then repeat the analysis presented in Table 3, while defining target market turnover as a share of target market investment, so that the turnover measure captures the possible bias in asset allocation. In Table 6 we test for determinants of investors' asset allocation to target market J. The purpose of this table is not only to mainly confirm previous literature's findings, but also to examine the effect of latitudinal and longitudinal geographical distances on portfolio allocation. Investors' allocation Bias to target market J is the dependent variable and the bilateral control variables, which include Cultural distance, Longitudinal distance, and Latitudinal distance are the main variables of interest. These three control variables are expected to take on negative signs. We also include the Trust in others variable for the small subset of European investors, which is expected to be positive (consistent with Guiso et al., 2004). Additionally, we include bilateral controls for Time zone differential, simple geographic Distance, common Language and Legal origin. The Bias in investors' allocation to target market J is calculated as per Eq. (3) and effectively measures the under-and overweighting of each target market J in the investor's portfolio as s share of all the investor's foreign portfolio allocation.

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

43

Table 4 Alternative cultural variables and investors' turnover relative to all holdings abroad. Table 4 repeats the analyses from Table 3, but with several different cultural proxies. The dependent variable is the investors' turnover in target market J as a share of the investors' total portfolio market value in non-home markets. All the specifications include investor and target country fixed effects. The variables of interest are the investor-target country pair bilateral variables. They include several measures of cultural differences between the investor and the target market. These independent variables include the Cultural distance between the investor and the target market, measured based on Hofstede's primary dimensions of culture (Eq. (2)). Globe is the cultural distance between investor and target from GLOBE study's nine primary dimensions of culture. CD, SD, and MD are the Correlation Distance, Standardized Euclidean Distance, and Mahalanobis Distance, all computed using Hofstede's 4 dimensions of culture based on Kandogan (2012). Investor-level controls include investors' Experience in each target market J, measured in quarters of presence. Investor size and Investor size in target are in logs. All regressions also include year indicators. The standard errors are two-way clustered errors based on home country–target country pairs. The t-statistics values are reported under the coefficients. (1) Cultural distance

(2)

(3)

(4)

(5)

−0.4067⁎⁎⁎ [6.01] −0.0705 [1.37]

Globe CD

−0.8575⁎⁎⁎ [5.08]

SD

−0.5939⁎⁎⁎ [6.66]

MD Experience Investor size in target Investor size Observations Adjusted R2

−8.0165⁎⁎⁎ [5.69] 0.1737⁎⁎⁎ [4.93] 1.0697⁎ [1.84] 231,620 0.1853

−5.8729⁎⁎⁎ [6.99] 0.1631⁎⁎⁎ [6.04] −0.4370 [0.67] 167,466 0.1600

−8.0220⁎⁎⁎ [5.72] 0.1720⁎⁎⁎ [4.86] 1.1302⁎⁎ [1.97] 231,620 0.1859

−7.9272⁎⁎⁎ [5.71] 0.1693⁎⁎⁎ [4.81] 1.1032⁎ [1.91] 231,620 0.1875

−0.5906⁎⁎⁎ [6.16] −7.9510⁎⁎⁎ [5.70] 0.1704⁎⁎⁎ [4.83] 1.0837⁎ [1.86] 231,620 0.1865

⁎ Significant at 10% level. ⁎⁎ Significant at 5% level. ⁎⁎⁎ Significant at 1% level.

The results reported in Table 6 support the idea that the bilateral variables that are related to investors' trading frequency are also related to portfolio weights of different target markets. Cultural distance and all of the geographic distance variables are negatively related to the portfolio weights of the target markets. This implies that culturally and geographically distant target markets are relatively underweighted in investment portfolios while geographically and culturally close target markets are relatively overweighed.12 Although the results show that both Latitudinal distance and Longitudinal distance are related to Bias in portfolio allocation, Longitudinal distance is more significant and larger in magnitude. This result provides new evidence to what determines portfolio allocations across the globe. While geographical distance is a known determinant of portfolio allocation, we document that the geographical distance matters much more, when investor moves further away from the target market across time zones.13 Of the other bilateral controls, common Legal origin is positive and significant as expected and Common language is not statistically significant. Home bias and international under-diversification in asset allocation have been documented before (Beugelsdijk & Frijns, 2010; Chan, Covrig, & Ng, 2005; Coval & Moskowitz, 1999, 2001; French & Poterba, 1991; Grinblatt & Keloharju, 2001 among others). The portfolio allocation analysis presented in Table 6 examines the international underdiversification behavior differently. Most previous studies focus on portfolio level under-diversification by computing a home bias and foreign bias measures for each investor. Our analysis contributes to the literature by focusing on the determinants of portfolio allocation 12

For robustness, we repeat the analysis reported in Table 6 but with several different cultural proxies. These independent variables include the cultural distance between investor and target from GLOBE study's nine primary dimensions of culture, Correlation Distance, Standardized Euclidean Distance, and Mahalanobis Distance. The results are consistent with the results presented in Table 6 and omitted from this version of the paper for brevity. 13 As a robustness test we also independently examine the relation between longitudinal and latitudinal distances with Bias in four different regions of the world (Asia, Europe, North America and Pacific). While the result of the sign of longitudinal distance coefficient is mostly stable, the sign of the latitudinal distance coefficient is less robust.

at the country level. We use country-pair specific bilateral control variables in order to explain portfolio allocation to target countries, similar to Aggarwal et al. (2012). Our dataset also allows us to investigate determinants of portfolio allocation at the institutional level, unlike previous studies that focus on aggregate country level portfolio flows. As Table 6 indicates, Cultural distance and Longitudinal distance relate negatively to portfolio allocation. These findings suggest that we are more likely to observe higher turnovers in the markets where higher shares of the portfolio are allocated (as documented in Table 3), when we compute turnover as a share of all foreign portfolio values for the investors (as per Eq. (1b)). Although in Table 3 we control for total investment in each target market, as a robustness check, in Table 7 we examine whether Cultural distance is related to trading volumes, when turnover is scaled by each target market J's assets in investors' portfolio (Eq. (1a)). Overall, the results presented in Table 7 are similar to the results presented in Table 3, but the coefficients of the Cultural distance are somewhat smaller in magnitude. Also, when Geographical Distance is separated into Longitudinal distance and Latitudinal distance, the Longitudinal distance is more significant and larger in magnitude. This highlights the robustness of our initial results subject to a stricter turnover definition. It also appears that, unlike in Table 3, total Investor size is negatively related to turnover. This suggests that large investors trade less, possibly due to the price impact that they are likely to impose.

4.4. Culture and home market turnover To further investigate the relationship between cultural characteristics of investors' home markets and the observed trading frequencies, we also test whether the country-specific cultural characteristics relate to turnover in home market securities. Table 8 shows the results from OLS regressions, where the dependent variable is investors' turnover in home market as a share of the investors' portfolio in the home market. The measures of culture to test

44

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

Table 5 Investor country's culture and investors' turnover relative to all holdings abroad. Table 5 shows the results from analyses that repeat Table 3's analyses, but in addition to investor country– target country specific variables, we also include investor country specific cultural variables. The dependent variable is the turnover from Eq. (1b). All the specifications include target country fixed effects and the main variables of interest are the bilateral variables and the investor country cultural variables. These bilateral variables include the Cultural distance between the investor and the target (Eq. (2)), distance in latitude and longitude in logs of kilometers (Distance, latitude and Distance, longitude), common Legal origin indicator, and a Common language indicator between the investor and the target market. Investor country controls include cultural Individualism, Uncertainty avoidance, and Trust. Additionally, we include the log of GDP and GDP per capita of the investor. Investor level controls include investors' Experience in each target market J measured in quarters of presence. Investor size and Investor size in target are in logarithm. All regressions also include year indicators. The standard errors are two-way clustered errors based on home country–target country pairs. The t-statistics values are reported under the coefficients.16 16 The results of this table were also replicated using the alternative turnover measure of Eq. (1a) as the dependent variable. Results are similar in magnitude and significance. We discuss the result in the text, but omit the table for brevity.

Cultural distance Distance, latitude Distance, longitude Legal origin Common language Experience Investor size in target Investor size Individualism

(1)

(2)

(3)

(4)

(5)

−0.3101⁎⁎⁎ [5.66] −0.0162 [0.20] −0.2639⁎⁎⁎

−0.2876⁎⁎⁎ [5.50] 0.0515 [0.70] −0.4304⁎⁎⁎

−0.2941⁎⁎⁎ [5.50] 0.0300 [0.43] −0.4255⁎⁎⁎

−0.3227⁎⁎⁎ [5.15] 0.0509 [0.63] −0.4483⁎⁎⁎

−0.3129⁎⁎⁎ [4.99] 0.0535 [0.68] −0.4401⁎⁎⁎

[3.04] 0.2280 [1.37] 0.3237⁎ [1.66] −7.7635⁎⁎⁎ [5.95] 0.1612⁎⁎⁎

[4.51] 0.2352 [1.47] 0.2785 [1.52] −8.0647⁎⁎⁎ [6.21] 0.1475⁎⁎⁎

[4.65] 0.2565 [1.58] 0.2101 [1.20] −8.2158⁎⁎⁎ [6.46] 0.1306⁎⁎⁎

[4.48] 0.2255 [1.25] 0.2344 [1.13] −8.7487⁎⁎⁎ [6.12] 0.1405⁎⁎⁎

[4.26] 0.2659 [1.47] 0.1689 [0.83] −8.7398⁎⁎⁎ [5.93] 0.1385⁎⁎⁎

[4.33] 1.2522⁎⁎ [2.18] 0.0283 [0.06]

[4.03] 1.2674⁎⁎ [2.21] −1.2625⁎⁎⁎ [2.78]

[3.58] 1.2658⁎⁎ [2.20]

[3.33] 1.5660⁎⁎ [2.47]

[3.38] 1.5734⁎⁎ [2.49] −1.3847⁎⁎ [2.19] −0.9311⁎⁎

−0.3058 [1.04]

Uncertainty avoidance Trust GDP per capita GDP Observations Adjusted R2

231,620 0.1841

0.1573 [0.82] 0.3143⁎⁎⁎ [5.94] 231,620 0.1883

0.0109 [0.06] 0.2367⁎⁎⁎ [4.30] 231,620 0.1873

0.2242 [0.28] 0.1852 [0.65] 0.2581⁎⁎⁎ [2.85] 201,931 0.1970

[2.06] 0.0362 [0.05] 0.1436 [0.56] 0.3484⁎⁎⁎ [4.22] 201,931 0.1980

⁎ Significant at 10% level. ⁎⁎ Significant at 5% level. ⁎⁎⁎ Significant at 1% level).

hypotheses H2a, H2b, and H2c include Uncertainty avoidance, Trust, and Individualism. As a robustness check for Individualism we also include two possible overconfidence measures from the GLOBE study: Institutional Collectivism (opposite interpretation to individualism) and Assertiveness. We also control for financial market and macroeconomic characteristics. The financial market controls are market Volume, market Capitalization and total Trading cost. The total Trading cost control is also broken into Commissions, Fees, and Market impact in specification (1). The macroeconomic controls comprise of the logarithms of GDP and GDP per capita. Finally, we include the following investor level controls: Investors' Experience in home market measured in quarters and the Investor size, which is the logarithm of total market value of the institution. The results reported in Table 8 continue to support H2a. It appears that higher ambiguity aversion of the investors' home market is related to lower trading frequency in the home market securities. The sign of Individualism is negative, which is inconsistent with our overconfidence hypothesis, but statistically insignificant. On the other hand, cultural Assertiveness is positively related to trading frequency. Cultural Trust is also positive, but without any statistical significance. The other control variables mostly carry their expected signs. Overall, trading volume in the home market and capitalization of the home market are both positively related to trading frequency. Trading costs are negatively related to trading frequency, but only without controls for market capitalization.

5. Conclusion In this study we examine the effect of culture in investors' home market and cultural distance between investors' home markets and target markets on institutional investors' trading frequency. We make several contributions to the existing literature. First, we find strong evidence that institutions trade at much higher frequency in their home market and in markets that are culturally closer to their home market compared with more distant markets. Our findings support the idea that as cultural distance increases, information about target markets' securities becomes more difficult to access and interpret, and as a result, investors choose more passive positions and are less willing to bet on their knowledge. Previous literature documents that an investor's competence influences trading frequency. Here we show that the same investor's competence can vary from market to market, and as a result we observe different magnitudes of turnover within the investor's own portfolio. Second, we show that investors from more ambiguity averse counties trade less frequently at home and abroad. Third, we find that geographical distance has an impact on trading frequency, so that investors' turnover is lower in geographically distant markets, but only when the distance is measured in longitude. Latitudinal distance is not a significant determinant of trading frequency. These results provide new evidence on possible reasons for observed turnovers across the international markets and merit a new avenue for future research.

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

45

Table 6 Determinants of investor portfolio allocation to target. Table 6 shows the results from OLS regressions, where the dependent variable is the investors' portfolio allocation bias in target market J. Investors' bias is calculated as per Eq. (3). All the specifications include investor and target country fixed effects. The variables of interest are the investor-target country bilateral variables. These independent variables include the Cultural distance between the investor and the target market, measured based on Hofstede's primary dimensions of culture (Eq. (2)), Trust in others for the European subset of the sample, geographical Distance in the log of kilometers, distance in latitude and longitude in logs of kilometers (Distance, latitude and Distance, longitude), Time zone differential in hours, common Legal origin indicator, and a Common language indicator between the investor and the target market. Investor level controls include investors' total market value in logarithm (Investor size). All regressions also include year indicators. The standard errors are two-way clustered errors based on home country–target country pairs. The t-statistics values are reported under the coefficients.17 17 We repeat the analysis of this table only with countries that use daylight savings time with respect to time zone differential and longitudinal distance. Results are similar in magnitude and significance to this table and available upon request. (1)

(2) −0.1642⁎⁎⁎ [2.89]

Cultural distance Trust in others

(3)

(4)

4.0463⁎⁎⁎ [4.34]

−0.7255⁎⁎⁎ [3.85]

−0.2113⁎⁎⁎ [3.59]

Distance, latitude Distance, longitude Time zone differential Legal origin

Observations Adjusted R2

−0.0407 [1.38] 1,976,908 0.3296

−0.0003 [0.00] 182,338 0.0202

(6)

(7)

(8) −0.1366⁎⁎⁎ [2.72]

Distance

Common language Investor size

(5)

−0.0354 [1.24] 1,976,908 0.3346

−0.0447 [1.53] 1,976,908 0.3300

−0.3651⁎⁎⁎ [3.99]

−0.0331 [1.17] 1,976,908 0.3329

−0.1376⁎⁎⁎ [2.96] −0.3528⁎⁎⁎ [3.53]

−0.1244⁎⁎⁎ [3.57]

−0.0366 [1.26] 1,976,908 0.3320

0.4029⁎⁎ [2.36]

−0.0399 [1.35] 1,976,908 0.3297

0.2752⁎ [1.65] −0.0072 [0.04] −0.0370 [1.29] 1,976,908 0.3339

⁎ Significant at 10% level. ⁎⁎ Significant at 5% level. ⁎⁎⁎ Significant at 1% level.

Table 7 Determinants of investor turnover relative to target holdings. Table 7 shows results from OLS regressions, where the dependent variable is the investors' turnover in target market J as a share of the investors' portfolio in market J (Eq. (1a)). All the specifications include investor's home country and target country fixed effects. The variables of interest are the investor's home country–target country pair variables. These bilateral variables include the Cultural distance between the investor and the target market, measured based on Hofstede's primary dimensions of culture (Eq. (2)), Trust in others for the European subset of the sample, geographical Distance in the log of kilometers, distance in latitude and longitude in logs of kilometers (Distance, latitude and Distance, longitude), Time zone differential in hours, common Legal origin indicator, and a Common language indicator between the investor and the target market. Investor level controls include investors' Experience in each target market J, measured in quarters of presence. Investor size and Investor size in target are in logarithm. All regressions also include year indicators. The standard errors are two-way clustered errors based on home country–target country pairs. The t-statistics values are reported under the coefficients.18,19 18 We repeat the analysis of this table only with countries that use daylight savings time with respect to time zone differential and longitudinal distance. Results are similar in magnitude and significance to this table and available upon request. 19 We repeat the analysis with several cultural variables, similar to Table 4. Results are similar to this. Cultural distance is negative, but smaller in magnitude compared with the result in Table 4. (1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

−0.1101⁎⁎ [1.99]

Cultural distance

−0.1127⁎ [1.89] −0.0779 [0.14]

Trust in others Distance

−0.3213⁎⁎⁎ [3.84]

Distance, latitude

−0.0970⁎ [1.65]

Distance, longitude

−0.2271⁎⁎⁎ [3.70]

Time zone differential

−0.0088 [0.17] −0.2223⁎⁎⁎ [3.49]

−0.0792⁎⁎⁎ [3.46]

Legal origin

0.0267 [0.21]

Common language Experience Investor size Observations Adjusted R2

1.9748⁎ [1.79] −3.0418⁎⁎⁎ [2.67] 223,674 0.0421

⁎ Significant at 10% level. ⁎⁎ Significant at 5% level. ⁎⁎⁎ Significant at 1% level.

9.3510⁎⁎⁎ [4.17] −9.3786⁎⁎⁎ [3.80] 56,528 0.0563

1.8581⁎ [1.68] −3.0328⁎⁎⁎ [2.67] 223,674 0.0423

1.9509⁎ [1.76] −3.0852⁎⁎⁎ [2.70] 223,674 0.0421

1.8796⁎ [1.70] −3.0209⁎⁎⁎ [2.67] 223,674 0.0423

1.8894⁎ [1.71] −3.0600⁎⁎⁎ [2.70] 223,674 0.0422

1.9771⁎ [1.78] −3.0515⁎⁎⁎ [2.69] 223,674 0.0420

−0.1917 [1.38] 0.0416 [0.24] 1.8780⁎ [1.70] −3.0273⁎⁎⁎ [2.68] 223,674 0.0423

46

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

Table 8 Determinants of home market turnover with cultural proxies. Table 8 shows the results from OLS regressions, where the dependent variable is the investors' turnover in home market as a share of the investors' portfolio in the home market. The dependent variable is computed based on quarterly buys and sells for home market, similar to Eq. (1a). The variables of interest are the investor country variables that measure the culture of the investors' home market. They include the Individualism, Uncertainty avoidance, Trust in others, and two measures of culture from GLOBE:Institutional Collectivism (opposite interpretation to individualism) and Assertiveness. We also control for financial market and macroeconomic characteristics. We include market Volume, market Capitalization, total Trading cost that is also broken into Commissions, Fees, and Market impact. The log of GDP and GDP per capita is included as macroeconomic controls. In addition, we include investor level controls: investors' Experience in home market measured in quarters and the Investor size which is the log of total market value of the institution. All regressions also include year indicators. The standard errors are clustered errors based on investor country. The t-statistic values are reported under the coefficients. (1)

(2)

Volume

0.1284 [1.38]

Capitalization Trading cost Commissions Fees Market impact

−0.2350⁎⁎⁎ [2.96] −0.0003 [0.00] −0.2248⁎⁎ [2.13]

(3)

(4)

−0.0075 [0.17] 2.2880⁎⁎⁎

0.0176 [0.42] 1.2833⁎

[8.45] 0.0985⁎⁎⁎ [3.41]

[1.88] 0.0852⁎ [1.94]

−0.0290 [1.26]

Individualism Uncertainty

(5) 0.0156 [0.43] −0.1681 [0.22] 0.0770⁎⁎ [2.43]

(6) 4.3598⁎⁎⁎ [6.38] 0.1618 [0.24] 0.1002⁎⁎ [2.19]

(7) 4.3005⁎⁎⁎ [7.06] −0.3084 [0.41] 0.0757⁎ [1.83]

−0.8676 [0.87]

Assertiveness

2.7739⁎⁎⁎ [3.13]

Trust

Experience

16.3402⁎ [1.93] −0.1461⁎⁎ [2.58]

−6.7996 [1.60] −0.1701⁎⁎⁎ [3.84]

−5.3583 [0.84] −0.1735⁎⁎⁎ [4.13] 0.3660 [0.58] 1.0916⁎⁎

GDP per capita GDP Observations Adjusted R2

234,440 0.0235

234,440 0.0318

[2.28] 234,440 0.0329

[2.69] 0.1017⁎⁎ [2.13]

−0.0437⁎ [1.80]

Collectivism

Investor size

0.0113 [0.28] 1.3625⁎⁎

−3.2151 [0.77] −0.1752⁎⁎⁎ [4.29] 0.4860 [0.96] 1.9565⁎⁎⁎ [3.35] 234,440 0.0335

1.0784 [0.11] −0.1777⁎⁎⁎ [4.39] 0.5901 [0.83] 1.0087⁎ [1.77] 219,888 0.0369

1.6499 [0.18] −0.1779⁎⁎⁎ [4.42] 0.3445 [0.52] 1.1399⁎⁎ [2.64] 219,888 0.0378

1.1878 [0.28] −2.8879 [0.58] −0.1785⁎⁎⁎ [4.53] −0.0565 [0.08] 1.0228⁎⁎ [2.67] 227,626 0.0321

⁎ Significant at 10% level. ⁎⁎ Significant at 5% level. ⁎⁎⁎ Significant at 1% level.

Appendix A. Hofstede's primary dimensions of culture14 1. Uncertainty avoidance index (UAI) deals with a society's tolerance for uncertainty and ambiguity. It indicates to what extent a culture programs its members to feel either uncomfortable or comfortable in unstructured situations. Unstructured situations are novel, unknown, surprising, or different from usual. Uncertainty avoiding cultures try to minimize the possibility of such situations by strict laws and rules, safety and security measures. Uncertainty avoiding countries are also more emotional and are motivated by inner nervous energy. 2. Individualism (IDV) as opposed to collectivism, is the degree to which individuals are integrated into groups. On the individualist side we find societies in which the ties between individuals are loose: everyone is expected to look after herself and her immediate family. In collectivist societies people from birth onwards are integrated into strong, cohesive groups. 3. Power distance index (PDI) measures the extent to which the less powerful members of organizations and institutions accept and expect that power is distributed unequally. It suggests that a society's level of inequality is endorsed by the followers as much as by the

leaders. Power and inequality are extremely fundamental facts of any society and while all societies are unequal, some are more unequal than others. 4. Masculinity (MAS) versus femininity refers to the distribution of roles between the genders. The survey studies reveal that (a) women's values differ less among societies than men's values; (b) men's values from one country to another contain a dimension from very assertive and competitive and maximally different from women's values on the one side, to modest and caring and similar to women's values on the other. The assertive pole has been called ‘masculine’ and the modest, caring pole ‘feminine’. The women in feminine countries have the same modest, caring values as the men; in the masculine countries they are somewhat more assertive and competitive, but not as much as the men, so that these countries show a gap between men's values and women's values. 5. Long-Term Orientation (LTO) versus short-term orientation: this fifth dimension was found in a study among students in 23 countries around the world. Values associated with Long-Term Orientation are thrift and perseverance.

Appendix B. Correlation matrix, Variance Inflation Scores (VIFs). 14 From Geert Hofstede's website: http://www.Geert-Hofstede.com and from Culture Consequences, 2001, 2nd edition, pages xix–xx).

The table below shows the correlation matrix of the bilateral independent variables used in the analyses. Time zone differential,

E. Beracha et al. / International Review of Financial Analysis 31 (2014) 34–47

longitudinal distance and simple distance are highly correlated and are not included in our regressions simultaneously (correlations N 0.85).

Cultural distance

Common Common language legal

Cultural 1 distance Common −0.2397 1 language Common −0.3423 0.5441 legal Distance 0.2567 0.0828 Time zone 0.2865 0.1005 diff. Distance, 0.2016 0.055 latitude Distance, 0.2257 0.0772 longitude

Distance Time zone diff.

Distance, Distance, latitude longitude

1 −0.0026 1 0.0208 0.8798 0.1693 0.5895 −0.0141 0.9374

1 0.3703 1 0.9049 0.4069

1

Appendix C. Variance Inflation Factors (VIFs). The table below reports the Variance Inflation Factors (VIF) associated with a regression that includes all the bilateral controls, trading frequency and the other selected controls as independent variables and turnover as the dependent variable. Time zone differential, longitudinal distance and simple distance are highly correlated and never included simultaneously in any of the regressions (VIF scores N 10).

Variable

VIF

1/VIF

Panel A: All bilateral variables Cultural distance Common language Common legal Distance Time zone diff. Distance, latitude Distance, longitude

1.37 1.49 1.72 16.30 6.55 2.44 13.11

0.731 0.672 0.580 0.061 0.153 0.410 0.076

1.28 1.47 1.63 1.31 1.39

0.783 0.680 0.612 0.762 0.718

Panel B: Bilateral variables included simultaneously Cultural distance Common language Common legal Distance, latitude Distance, longitude

References Aggarwal, R., & Goodell, J. W. (2009). Markets and institutions in financial intermediation: National characteristics as determinants. Journal of Banking and Finance, 33, 1770–1780. Aggarwal, R., Kearney, C., & Lucey, B. (2012). Gravity and culture in foreign portfolio investment. Journal of Banking and Finance, 36, 525–538. Ahern, K. R., Daminelli, D., & Fracassi, C. (in press). Lost in translation? The effect of cultural values on mergers around the world. Journal of Financial Economics (in press). Akerlof, G. A. (1997). Social distance and social decisions. Econometrica, 65, 1005–1027. Anderson, C. W., Fedenia, M., Hirschey, M., & Skiba, H. (2011). Cultural influences on home bias and international diversification by institutional investors. Journal of Banking and Finance, 35, 916–934. Arrow, K. J. (1974). The limits of organization. New York: W. W. Norton & CO, Inc. Barber, B.M., & Odean, T. (2000). Trading is hazardous to your wealth: The common stock investment performance of individual investors. Journal of Finance, 55, 773–806.

47

Barber, B.M., & Odean, T. (2001). Boys will be boys: Gender, overconfidence, and common stock investment. Quarterly Journal of Economics, 116, 261–292. Beck, T., Demirgüç-Kunt, A., & Levine, R. (2003). Law, endowments, and finance. Journal of Financial Economics, 70, 137–181. Benos, A. V. (1998). Overconfident speculators in call markets: Trade patterns and survival. Journal of Financial Markets, 1, 353–383. Beugelsdijk, S., & Frijns, B. (2010). A cultural explanation of the foreign bias in international asset allocation. Journal of Banking and Finance, 34, 2121–2131. Caballé, J., & Sákovics, J. (2003). Speculating against an overconfident market. Journal of Financial Markets, 6, 199–225. Chakrabarti, R., Gupta-Mukherjee, S., & Jayaraman, N. (2009). Mars–Venus marriages: Culture and cross-border M&A. Journal of International Business Studies, 40, 216–236. Chan, K., & Covrig, V. (2012). What determines mutual fund trading in foreign stocks? Journal of International Money and Finance, 31, 793–817. Chan, K., Covrig, V., & Ng, L. (2005). What determines the domestic bias and foreign bias? Evidence from mutual fund equity allocations worldwide. Journal of Finance, 60, 1495–1534. Choi, N., Fedenia, M., Skiba, H., & Sokolyk, T. (2013). Informed trading or familiarity bias? Evidence from the performance of international institutional investors. Working paper. Christelis, D., & Georgarakos, D. (2013). Investing at home and abroad: Different costs, different people? Journal of Banking and Finance, 37, 2069–2086. Chui, A.C. W., Titman, S., & Wei, K. C. J. (2010). Individualism and momentum around the world. Journal of Finance, 65, 361–392. Coval, J.D., & Moskowitz, T. J. (1999). Home bias at home: Local equity preference in domestic portfolios. Journal of Finance, 54, 2045–2073. Coval, J.D., & Moskowitz, T. J. (2001). The geography of investment: Informed trading and asset prices. Journal of Political Economy, 109, 811–841. Djankov, S., La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (2008). The law and economics of self-dealing. Journal of Financial Economics, 88, 430–465. Dorn, D., & Huberman, G. (2005). Talk and action: What individual investors say and what they do. Review of Finance, 9, 437–481. Felbermayr, G. J., & Toubal, F. (2010). Cultural proximity and trade. European Economic Review, 54, 279–293. Franks, J., Mayer, C., & Wagner, H. (2006). The origins of the German corporation—finance, ownership and control. Review of Finance, 10, 537–585. French, K. R., & Poterba, J. M. (1991). Investor diversification and international equity markets. American Economic Review, 81, 222–226. Glaser, M., & Weber, M. (2007). Overconfidence and trading volume. Geneva Risk Insurance Review, 32, 1–36. Graham, J. R., Harvey, C. R., & Huang, H. (2009). Investor competence, trading frequency, and home bias. Management Science, 55, 1094–1106. Grinblatt, M., & Keloharju, M. (2001). How distance, language, and culture influence stockholdings and trades. Journal of Finance, 56, 1053–1073. Grinblatt, M., Keloharju, M., & Linnainmaa, J. (2012). IQ, trading behavior, and performance. Journal of Financial Economics, 104, 339–362. Grossman, S. J., & Stiglitz, J. E. (1980). On the impossibility of informationally efficient markets. American Economic Review, 70, 393–408. Guiso, L., Sapienza, P., & Zingales, L. (2004). The role of social capital in financial development. American Economic Review, 94, 526–556. Guiso, L., Sapienza, P., & Zingales, L. (2006). Does culture affect economic outcome? Journal of Economic Perspectives, 20, 23–48. Guiso, L., Sapienza, P., & Zingales, L. (2008). Social capital as good culture. Journal of the European Economic Association, 6, 295–320. Guiso, L., Sapienza, P., & Zingales, L. (2009). Cultural biases in economic exchange? Quarterly Journal of Economics, 124, 1095–1131. Hart, O. (1995). Firms, contracts, and financial structure. New York: Oxford University Press. Heath, C., & Trevsky, A. (1991). Preferences and beliefs: Ambiguity and the competence in choice under uncertainty. Journal of Risk Uncertainty, 4, 5–28. Hofstede, G. (1980). Culture's consequences: International differences in work-related values. Beverly Hills, CA: Sage. Hofstede, G. (2001). Culture's consequences: Comparing values, behaviors, institutions and organizations across nations (2nd ed.)Thousand Oaks, CA: Sage. House, R. J., Hanges, P. J., Javidan, M., Dorfman, P. W., & Gupta, V. (2004). Culture, leadership, and organizations: The GLOBE study of 62 societies. Thousand Oaks, CA: Sage. Huang, R. R. (2008). Tolerance for uncertainty and the growth of ambiguous industries. Journal of Development Economics, 87, 333–353. Kandogan, Y. (2012). An Improvement to Kogut and Singh measure of cultural distance considering the relationship among different dimensions of culture. Research in International Business and Finance, 26(2), 196–203. Kogut, B., & Singh, H. (1988). The effect of national culture on the choice of entry mode. Journal of International Business Studies, 19, 411–432. Lai, S., Ng, L., Zhang, B., & Zhang, Z. (2013). Do mutual funds trade differently at home and abroad? Working paper. Lane, P. R., & Milesi-Ferretti, G. M. (2008). International investment patterns. Review of Economics and Statistics, 90, 538–549. Portes, R., & Rey, H. (2005). The determinants of cross-border equity flows. Journal of International Economics, 65, 269–296. Stulz, R. M., & Williamson, R. (2003). Culture, openness, and finance. Journal of Financial Economics, 70, 313–349. Van Nieuwerburgh, S., & Veldkamp, L. (2010). Information acquisition and underdiversification. Review of Economic Studies, 77, 779–805.