Influence of familiarity
bias on investment
decisions
A.M. Wieland
June 5, 2020
Despite globalization of financial markets and the benefits of international portfolio diversification, a strong bias in favor of domestic assets is often observed in literature on international investment portfolios. In this paper I explore the relation between invest-ment decisions and the familiarity bias, by considering a sample of 198 companies with an investor base of 203,711 investors from 75 countries for the period 2016-2019. The ten-dency to invest in geographically proximate firms or familiarity bias seems to be driven by knowledge asymmetries. The benefits from knowledge asymmetries are largest for small domestic firms. Hence, the familiarity bias becomes less eminent as firms increase in size and diversify internationally.
Study Programme: MSc IFM Student number: S3182967 Supervisor: Dr. Adri de Ridder Co-assessor: Dr. Halit Gonenc Key words:
1
Introduction
Despite the fact that financial markets are increasingly integrated, asset portfolios are still clustered at home. As a result investors are not maximizing their returns, and firms do not fully reap the benefits of the global financial markets. In this paper I highlight some of the potential explanations for this phenomenon.
Traditionally, financial and economic models posit that investors make assumptions about future payoff of assets. The assumption following this is that investors strive for utility maximization via their portfolio composition. A high return combined with the lowest possible risk is desirable. Especially the latter can be achieved through portfolio diversification, in particular via international diversification. Since, in case of a non-global financial crisis one’s investments are diversified, and therefore, only a small percentage of the total investment portfolio is affected by the financial downturn.
However, during the global crisis at the beginning of the century the interconnected financial markets proved unable to handle or stop the impact of a financial downturn in one country. This led to a trend of deglobalization and dedollarization post financial crisis (Goldberg, 2019 ; Shavshukov and Zhuravleva, 2020). Furthermore, the increasing complexity of international financial markets, facilitated by technological advancements, perhaps made the international trade more complex than it was in the previous century. Thus, it is important to continuously update prior studies on how this development of international financial markets may alter investment decisions.
As a result of globalisation a decrease in imposed trade barriers is observed. Even though this trend was slightly reversed after the financial crisis (Shavshukov and Zhu-ravleva, 2020). In general, this implies that over the past decades global portfolio di-versification is increasingly attractive to investors, as the additional costs associated with international investment decreased. Moreover, the nature of internationalization changed, for instance the arrival of born global firms in comparison to the gradual expansion in-ternationally (Knight and Liesch, 2016). This is for instance reflected in the continuous updating of Vahlne’s Uppsala model (Vahlne, 2020).
to payout policies. The double taxation in some countries might result in market fric-tion. For instance the dividend withholding tax on payments to foreign shareholders and additional transaction costs, results in a preference toward lower dividend payments or reinvestment for foreign investors (Ferreira, Massa, and Matos, 2010). However, some investors prefer an alternative diversification strategy, namely diversification per sector. Here an investor spreads the risk as well, however not necessarily internationally. This strategy reduces the negative effect of overweighing domestic stocks in portfolio compo-sition.
Both the home bias, tendency to overweigh domestic stocks in one’s portfolio, and the familiarity bias tilt the investment decision of investors towards domestic stocks. Prior studies observe that in contrast to the expected global diversification of invest-ments, global investment patterns are mostly clustered in domestic or similar markets (Huberman, 2001 ; Dziuda and Mondria, 2012). This phenomenon may be described as a familiarity bias. This bias is founded on knowledge asymmetry, individuals presume to possess a knowledge advantage within the domestic market. Consequently, the as-sumption of superior information results in a tendency to predominantly invest close to home (Huberman, 2001 ; Dziuda and Mondria, 2012). In my study familiarity is defined as: the geographic distance in kilometers between an investor and an investment object, which may result in the belief of a knowledge advantage for domestic investors over their foreign counterparts. Based on prior studies one may argue that investors are currently not maximizing their portfolio returns, despite the reduction of international investment barriers. Within this paper I examine the influence of familiarity as a potential cause for the lack of diversification in portfolio composition. Moreover, potential moderating variables that could weaken or strengthen this effect are considered. Thereby providing an insight to management on how to attract foreign investors. Hence, within this paper I aim to provide an answer to the following research question:
Prior literature mentions several factors that may influence the relationship between familiarity and investment decisions. Firstly, higher leverage implies a higher risk. Thus, highly levered firms may be less attractive investment objects, unless one beliefs to possess superior information about the firm. Hence, it strengthens the positive effect of familiarity on the tendency to invest, whereas firm size and multinationality result in the opposite moderating effect. To clarify, larger and internationally diversified firms can be regarded as less risky or too big to fail. Moreover, these firms are better equipped to communicate internal knowledge to outside parties on an international scale. Thereby reducing the influence of familiarity on investment decisions.
Following the expectation of the paper that greater familiarity leads to more invest-ment, this relationship is empirically tested using an ordered probit model. Investor data is collected from firms with headquarters in one of these five countries, namely the US, the UK, Germany, France, and the Netherlands. Investors are divided into one of three categories based on their holdings: weak, moderate, and strong ownership. Based on prior research I expect the assumption that most investors are clustered around the firm’s headquarters to hold. Or to put it differently, the majority of investors is clustered in geographic proximity of the investment object.
This research contributes to prior research on international business and international finance by outlining the potential effect of a familiarity bias on investment patterns from a European perspective. Whilst outlining the influence of US investors and prior colonial ties. Furthermore, it discusses the impact certain firm characteristics might have on the relationship between familiarity and investment patterns, which provides useful manage-rial implications.
respectively). One may argue that the US and the UK have more in common than the UK and other European countries, due to the prior colonial ties, which resulted in simi-larities till the present day, such as the legal system and language. Both of which reduce the influence of familiarity on investment decisions (Bergin and Pyun, 2016 ; Lundholm, Rahman, and Rogo, 2018).
Figure 1: Geographic distance to an investment object per country
Figure 1 provides an overview of the geographic distance in kilometers between an in-vestor and an asset. The blue part of the bars represent the 49% of inin-vestors closest to their investment object, the orange part depicts the mean, and the grey part of the part shows the kilometer distance for the investors furthest away from their investment object (51-100%).
of American, Canadian, and Australian investors. For the Netherlands the mean value can be explained by the low number of domestic investors in combination with a majority of investors from Europe.
To continue with the empirical section of this research. Prior to the ordered probit regression a univariate analysis is conducted, and based on the results several conclu-sions can be drawn, namely that domestic investors seem to hold more in terms of value compared to foreign investors. Moreover, domestic investors also hold a larger share of the firm relative to foreign investors (1.61% and 0.28% respectively). Lastly, domestic in-vestors seem to prefer smaller firms, potentially because local knowledge may have greater benefits on their ability to estimate future returns.
The empirical findings based on the ordered probit model indicate there is a negative and statistically significant relationship between an decrease in familiarity and investment. Indicating that an increase in distance lowers holdings. However, the effect is rather small. For the full sample firm size strengthens this relationship, and multinationality reduces the influence of the familiarity bias. Hence, more internationalized firms suffer less from a familiarity bias. This second finding is in line with prior research.
To ensure the robustness of the findings several batteries of robustness tests are per-formed. In the second run of tests, the US is omitted from the sample which yields partially different results. The effect of familiarity on investment becomes positive, and the moderating effect of firm size becomes insignificant. The difference in results indicates that the preferences of US investors drive the initial results.
Although this research uses geographic proximity as a proxy of familiarity, the psychic distance may impact investors sense of familiarity. Since, lower physiological distance may reduce information costs (Bergin and Pyun, 2016). Therefore, the ordered probit analysis is run once solely for countries with prior colonial ties. Via the latter potential biases due to former colonial or historical ties can be identified. Similar to the original model, familiarity negatively influences holdings. The only significant interaction term is firm size which reduces the negative effect of familiarity on the likelihood of investment, as expected based on prior studies (i.e. Coval and Moskowitz, 1999).
method identifies if investors preference or the familiarity bias differs per group. This analysis is preformed twice, once with and once without the US. For both runs, the nega-tive effect of familiarity on the likelihood to invest becomes posinega-tive for the third quartile. The moderating effect of both multinationality and firm size remains negative.
To ensure that the zero values amongst the observation do not cause issues in the ordered probit model, a Tobit regression is run. The coefficient for familiarity remains small. Moreover, if the dependent variable ownership is treated as a continuous variable, the effect familiarity bias is negative however small.
Prior studies identified a preference for larger firms amongst foreign investors (Coval and Moskowitz, 1999 ; Dahlquist and Robertsson, 2001). Especially large institutional investors prefer larger firms as investments. To ensure that the preference towards large firms does not bias the results the sample is split in two, based on firm size. Thereafter I run a two sample t-test and the ordered probit model is re-run. The breadth of ownership is significantly larger for the bigger firms. For the sample of large firms, familiarity consis-tently negatively effects ownership, whereas for small firms these results are inconsistent. The coefficient for leverage is positive and significant solely in model (3) for the larger firms. Firm size negatively impacts ownership for both groups, the moderating effect is positive and significant for the smaller firms, indicating it strengthens the familiarity bias. The coefficients for the R&D and cash ratio differ in both tables. The investors for smaller firms prefer less innovative firms that hold more cash, the opposite holds for the investors of the larger firms.
This rest of this paper is organized as follows, first prior literature is discussed and hypotheses are formed in Section 2. In Section 3 the research method and data collection is described, whereas the empirical results are presented in Section 4. The final Section discusses the main conclusions.
2
Literature review and hypothesis development
as well as Foerster and Karolyi (1999) document the positive effect of listing on the US exchanges for non-US firms, such as an increase in the number of shareholders and the evidence of abnormal returns around the listing period. Coval and Moskowitz (1999) find that portfolio fund managers show a preference for local firms. Similar conclusion are drawn by Dziuda and Mondria (2012) who find that investment through mutual funds are still biased towards the domestic market. Multiple authors observed a preference to-wards the familiar (i.e. Grinblatt and Keloharju, 2001 ; Huberman, 2001). Huberman (2001) found that this familiarity bias even occurs within a country, as he identified that investors tend to invest in the local Regional Bell Operating Company in comparison to the Regional Bell Operating Company of other states. Moreover, Bergin and Pyun (2016) identified a home bias in investment choices, and that if investors decide to invest abroad they tend to prefer countries with returns correlated to home assets.
Furthermore, Benartzi (2001) identified the tendency of employees to invest a large fraction of their retirement money in company stocks, hence the firm most familiar to them. The 401K pension schemes in the US and UK are a good example of this. Since, research observes that employees investment portfolios are mostly unbalanced, and heavily tilted towards the employer’s stock (i.e. Pendleton and Robinson, 2018 ; Tang, Mitchell, Mottola, and Utkus, 2010 ; VanDerhei, Holden, Alsonso, and Bass, 2016 ; Walter and Corley, 2015). Even though this may have devastating results, as seen in the case of Enron. The importance of these 401K pension schemes increases the number of investors as well. Contrasting to big pension funds in Europe where individuals do not decide in which stocks their pensions are invested, individuals decide in which package they invest (Benartzi, 2001 ; Pendleton and Robinson, 2018).
in the country of investment are, the stronger the preference of foreign investors for the firm to reinvestment instead of pay dividends. Hence, based on this I expect that foreign investors prefer firms with higher R&D ratios or capital expenditures relative to domestic investors. Thus, these firms presumably suffer less from a familiarity bias.
The tendency to invest in the familiar may be driven by presumed asymmetric knowl-edge. Prior studies extensively discussed the role of information asymmetry on portfolio composition (i.e. Bergin and Pyun, 2016 ; Kang and Stulz, 1997 ; Van Nieuwerburgh and Veldkamp, 2009). The importance of knowledge asymmetry is amplified once more due to its introduction in the calculation of the variance of returns by Okawa and Van Wincoop (2012) and Bergin and Pyun (2016). One of the main aspects fuelling the assumption of asymmetric knowledge is proximity. Hence, this implies that living nearby a specific firm increases the likelihood of investing in it. This, not because of a focus on the ex-ploitation of knowledge asymmetry, but instead it is fuelled by optimism and charitable feelings toward things one feels affinity with (Huberman, 2001). People may choose to locate somewhere based on their affinity with it and their economic prospects in the area (Branikas, Hong, and Xu, 2020), since this is correlated with the demand for local stock it is important to take into account when drawing conclusions on the importance of fa-miliarity. However, even after controlling for non random location choice, distance seems to play a role in portfolio composition (Branikas, Hong, and Xu, 2020).
to foreign investment. Instead Tesar and Werner (1995) propose that geographic prox-imity is an important factor for portfolio composition. The assumption of a preference towards geographically close investment objects can be observed in investment patterns of countries as well. For instance, the Netherlands receives and sends the largest portion of foreign investment to its neighboring countries (Boutorat and Van den Berg, 2017). Bergin and Pyun (2016) also suggest that a common border may increase the chance of investing. Especially in combination with other variables that make a country more fa-miliar, or reduce the influence of asymmetric knowledge, for instance a common currency, language, or economic union (Bergin and Pyun, 2016). Moreover, Keloharju, Kn¨upfer, and Linnainmaa (2012) find that individual’s product market choices impact their invest-ment choices. They find a strong positive relationship between customer relationship and company ownership, also with regards to the size of the ownership stake. Hence, based on the previous we may assume that the familiarity bias adds another dimension to the traditional models on the risk-return trade-off for portfolio composition.
capital more efficiently amongst domestic funds (Dziuda and Mondria, 2012). Based on the previous the following is hypothesized:
H1: Investors are more likely to invest in the familiar, hence there is a positive relationship between likelihood of investing in a firm and geographic proximity to the investment object.
To continue, there are certain costs associated with conveying information from a firm to an investor (Coval and Moskowitz, 1999). Firms undertake several steps to commu-nicate internal knowledge to the public. One channel of communication are the annual reports. The costs associated with this channel are higher for the international market than for the domestic market. Since the firms need to adhere to international accounting standards and it should be written in the lingua franca. Hence, although access to foreign capital markets reduces financing costs, it may increase the disclosure costs for the firm. Under the assumption that these costs are equal, irrespective of firm size, it is relatively more expensive for smaller firms to attract foreign investors. Hence, larger firms seem to be at an advantage when attracting foreign investors, which reduces the influence of familiarity on investment decisions. Moreover, investors need to gather this information and be equipped to process it. The costs associated with this are relatively lower for local investors according to Coval and Moskowitz (1999).
this relation is not empirically tested within this paper, it does imply firms can gain from reducing the familiarity bias and thereby increasing their breadth of ownership. More-over, it has been argued in prior literature that it is a reflection of investor confidence in the firm (Badrinath and Wahal, 2002 ; Sias, Starks, and Titman, 2006). Dahlquist and Robertsson (2001) identified the preference of institutional investors to invest in firms with a larger breadth of ownership, a characteristic more likely to occur in larger firms. They argue the firm size may outweigh familiarity for investment decisions.
To continue, similar to in debt financing international diversification might alter the effects of the familiarity bias. Jang (2017) finds a positive relation between internation-alization and financing via foreign lenders. Moreover, international bond markets are more actively used by multinational companies, especially during crisis periods (Jang, 2017). These multinational firms have a physical presence in multiple countries, as well as a presence on the international equity markets. Especially the physical presence may reduce investor’s preference for domestic firms, since it increases the familiarity between the investor and the company. As mentioned before, customer relations and firm visibility may lead to this higher sense of familiarity (Keloharju, Kn¨upfer, and Linnainmaa, 2012 ; Grullon, Kanatas, and Weston, 2004).
relationship between familiarity and investment.
H2a: Multinationality negatively moderates the relationship between familiarity and investment, thereby increasing the likelihood of international portfolio diversification of investors.
H2b: Firm size negatively moderates the relationship between familiarity and investment, thereby in-creasing the likelihood of international portfolio diversification of investors.
Since the wealth of shareholders is largely concentrated in the firms they own, they consequently prefer to invest in less risky firms as this increases their expected utility (Faccio, Marchica, and Muro, 2011). This because of several disadvantages associated with debt, such as the cost of bankruptcy (Weiss, 1990). This is nicely highlighted in the first stage of the privatization scheme in the Czech republic, after the fall of communism, the former state owned firms were privatized. Authorities set a uniform share price for all firms, this allowed researchers to investigate what drives share price and purchase de-cisions, as the face value is equal for all firms and becomes irrelevant as a decision tool for investors. Hence, investors base their decision on past performance, agency costs, and the expected costs of financial distress. Hingorani, Lehn, and Makhija (1997) finds that share demand is inversely related to leverage. They opt one of the reasons for this is that higher leverage increases the chances of financial distress for the firm (Hingorani, Lehn, and Makhija, 1997).
idiosyncratic payoffs on domestic assets (Bergin and Pyun, 2016). Which may explain their preference for small, highly levered, domestic firms. Kang and Stulz (1997) found similar results on the Japanese market, foreign investors underweigh small, highly levered firms. A potential explanation for this is knowledge asymmetry, or the familiarity bias. In general, highly levered firms have greater future return uncertainty, this in combina-tion with a lack of knowledge about the firm can result in severe adverse seleccombina-tion. To clarify, uninformed investors probably face more adverse selection, whereas the informed investors (locals) are more capable to estimate the future returns of a firm. Thus, we may assume that in order to maximize utility, investors prefer to invest in firms with lower levels of leverage, unless they presume to have superior knowledge about the investment object. As this could compensate for the higher risk. Hence, the tendency to invest in the familiar is strengthened by higher levels leverage within a firm. Therefore, I propose the following hypothesis:
H3: Higher firm leverage strengthens the effect of familiarity when investing.
3
Methodology and data collection
3.1
Data collection
Data is gathered via DataStream on a large sample of firms with headquarters in five different countries, namely the US, the UK, the Netherlands, Germany, and France. The familiarity variable is based on geographic proximity, for the construction of this variable the distance in kilometers from the capital of headquarter country to the capital of the investor’s home country is used. Due to the differences between the countries, robustness can be ensured by considering country characteristics, such as GDP.
of 2016-2019. To correct for outliers, total assets, foreign sales, investor holding value, market capitalization, R&D expenditures, total sales, cash and equivalents, and long-term debt are winsorized at the 0.99 level.
As moderating variables on the relationship between familiarity and investment, sev-eral variables are considered. Firstly, multinationality is expected to reduce the effect of familiarity on investment. Multinationality is measured by the ratio of foreign sales to the total sales. Secondly, firm leverage, the higher risk associated with leverage may strengthen the relationship between familiarity and investment decisions. Following prior research, leverage is measured as the long-term debt divided by total assets (i.e. Cho, El Ghoul, Guedhami, and Suh, 2014 ; Booth, Aivazian, Demirguc-Kunt, and Maksimovic, 2001). Lastly, an increase in firm size is expected to reduce the effect of familiarity on investments. As a proxy for firm size the natural logarithm of total assets is used. All monetary variables are presented in the US$ amount. Moreover, following prior research, I control for both firm and country level characteristics (i.e. Fauver, Mingyi, Xi, and Taboada, 2017 ; Jang, 2017 ; Park, Suh, and Yeung, 2013). The following firm level con-trols are included in the model: research and development (R&D) ratio, measured by the R&D expenses divided by the total sales; Capex, capital expenditures divided by the total assets; blockholder, dummy assuming 1 if an investor owns more than 10%of the company zero otherwise; cash ratio, cash and equivalents dividend by current liabilities. The latter is included since Jang (2017) found a difference in cash holdings between multinational and domestic firms. These firm level variables are collected via Compustat IQ. To con-tinue, via the country level control variables, the influence of country characteristics on sample firms is isolated. To begin with the annual inflation rate as reported by the World Bank (Cho, El Ghoul, Guedhami, and Suh, 2014). The second country level control is GDP growth, this is measured as a percentage of GDP and collected from the World Bank (Fauver, Mingyi, Xi, and Taboada, 2017). The operalization of all variables can be found in Appendix Table A19, and the location of investors per headquarter country are depicted in Appendix Table A20.
2). Firstly, the majority of investors originate from the United States.
Figure 2: World map of investor origin
Figure 2 provides an overview of the origins of the investors included in the sample and in which of the five HQ countries they hold assets. Per investor country a pie chart shows the fraction of their total investment invested in one or more of the HQ countries.
Australian investors, almost 40% invest in US firms and 24% in firms from the UK. A similar pattern is found amongst Canadian investors, who invest most in the US, UK, and France (42%, 20.5%, and 14.5% respectively). Spanish investors mostly invested in French firms (28%) and the only investor from Curacao invested in a Dutch firm. Lastly, besides for the Dutch the majority of investors from one of the five headquarter countries invested in their domestic market.
3.2
Ordered probit model
An ordered probit model will be used to confirm or reject the hypotheses stated in Section 2. The advantage of this model is the use of a categorical dependent variable, which allows for comparison between the different categories of investors, whilst main-taining the normality assumption. The dependent variable with m = 3 categories, weak (<0.35%), moderate (0.35-7.98%), and strong (>7.98%) ownership. These thresholds are based on the 33rd and 67th percentile. First, solely focusing on the relationship between familiarity and investment, and thereafter including interaction terms to test for the po-tential moderating effect. Equation 1 shows the regression equation for the full model, the three subscripts stand for firm, country, and year (i, c, t respectively). The year fixed effects are denoted byφt. The error term is normally distributed. The full regression can be written as:
y∗i = α + β1f amiliarityi+ β2leverageit+ β3f amiliarityi∗ leverageit+ β4f irmsizeit+
β5f amiliarityi∗ f irmsizeit+ β6multinationalityit+ β7f amiliarityi∗ multinationalityit+
β8R&Dratioit+ β9cashratioit+ β10blockholderc+ β11GDP growthct+ β12inf lationct+ φt+ it (1)
assets. The interaction term familiarity*firm size shows the moderating effect of firm size on the familiarity bias. I define firm multinationality as the ratio of foreign sales to the total sales. The moderating effect of multinationality of the familiarity bias is written as the interaction term familiarity*multinationality. The model includes several firm and country level controls: R&D ratio, the R&D expenses to total sales; Capex, capital ex-penditures divided by total assets; blockholder, dummy variable equal to 1 if an investor owns more than 10% of the company, and zero otherwise; cash ratio, cash and equivalents to current liabilities; the annual inflation rate; the GDP growth, this is measured as a percentage of GDP.
Here a multivariate model is considered. This can be rewritten in vector notation, see equation 2.
yi∗= α + Xi0β + φt+ it, ∀i ∈ (1, . . . , m) and ∀t ∈ (1, . . . , k), (2)
with m = 203, 711 and k = 4. Both vectors Xi0 and β are described below. Note that as a short-hand notation, the ith observation of the variables (familiarityi, . . . , inflationi), is denoted by (X1,i, . . . , X12,i).
Xi0 = X1,i, . . . , X12,i β = β1 .. . β12
Variations on equation 2 exist. Variables may be left out of the equation, if beneficial for the relevant analysis. Furthermore, I attempt to confirm the hypotheses stated in Section 2 following the different variations of equation 2.
yi= 0 if − ∞ < y∗i ≤ T1 1 if T1< y∗i ≤ T2 2 if T2< y∗i ≤ ∞ (3)
There is no constant term, since adding something to β0, T1, T2 yields exactly the same probabilities. Then β0, T1, T2 would not be identified. The number of params to be esti-mated is equal to 2(T1, T2) + number of β.
4
Empirical results
This chapter presents the empirical findings of this research. Prior to the empiri-cal results of the ordered probit model a table of descriptive statistics and a correlation matrix are presented. Moreover, a univariate analysis is performed to identify the initial differences between domestic and foreign investors. Thereafter, the findings of the ordered probit model are discussed and elaborated upon. Lastly, several batteries of robustness tests are performed.
4.1
Summary statistics and correlation
Within this section the distribution of variables per headquarter country, as well as the descriptive statistics for the different variables are presented. Thereafter, the correlation matrix is depicted and discussed.
countries. The low value for the US is consistent with the literature, and can be explained by the 401K schemes and limited presences of institutional investors. As for the familiar-ity variable in row 4 that the distance between an investor and the investment object is the smallest in the US and the largest for the UK (1831 and 4056 km, respectively). This can be explained by the large percentage of domestic investors investing in the American firms. Whereas the further distance between British firms and their investors, is mostly based on the large percentage of US investors investing in the British firms (42%), the prior colonial ties between the two countries resulted in similarities between the two that lower the information costs, and hence, make investment more attractive despite the large distance between the two countries (5900 km). The mean foreign sales is lowest for the US and highest for the UK, see row 6. The current liabilities and relative R&D expenses in relation to sales are significantly higher in the US than for the other four countries. This may imply that the American firms are more innovative. The large amount of born global firms and importance of Silicon Valley could contribute to this result. The market capitalization and firm size is the biggest in the US ($312 billion and 18.246, respectively) and the smallest in the Netherlands ($8.6 billion and 12.642). A potential reason for this are the sizes of the two domestic markets. In case a firm is purely serving the domestic market, the total market size of an American company is far larger than for a Dutch company. The low cash ratio in the Netherlands could be related to the lower R&D ratio of the country and the importance of the agriculture industry. To clarify, a higher R&D ratio implies higher asset intangibility, which according to Leia, Qiub, and Wan (2018) may result in higher cash holdings. The mean value for multinationality is only 0.1% for the US, France and the Netherlands. It is 0.2% for the UK and Germany. However, if we look at the maximum values for multinationality the UK and US clearly have the least internationalized firms (3.3% and 1.1% of total sales is foreign). For the Netherlands and Germany the foreign sales comprises of a larger part of total sales (57.3% and 41.4%). Hence, the range between purely domestic and internationalization of firms is larger for the latter two countries. This is in line with their export focused economies.
Table 1: Descriptive statistics and distribution of the variables
The table below, Panel A, presents the different variables organized per country. Thereby, showing the differences in observation between the five headquarter countries in the sample. Row 1 shows the country abbreviations, row 2 shows the number of companies per country. Row 3-19 show the mean value observed for the different variables for a given country. 3 is ownership, investor holding value divided by the company’s market capitalization. 4 is the familiarity variable, measured as the geographic distance in km between the investor and HQ countries. 5 is the investor holding value, and 6 is foreign sales. 7 is revenue, and 8 current liabilities. 9 is R&D ratio, R&D expenses divided sales, and 10 is Capex, the capital expenditures divided by total assets. 11 depicts market capitalization. 12 depicts cash ratio, computed by dividing cash and equivalents by current liabilities. 13 is firm size, for which the natural logarithm of total assets is used as a proxy. 14 is total long-term debt, 15 depicts leverage measured as total LT debt divided by total assets. 17 and 18 show the GDP growth and inflation, respectively. Panel B shows the descriptive statistics of the variables used in the analysis using the sample of 261,547 observation from 2016-2019. Familiarity, firm size, leverage, R&D ratio, Capex, and cash ratio are computed as mentioned above. Blockholders is a dummy variable assuming a value of 1 if an investor hold more than 10% of the total company’s value. MNC, is foreign sales divided by the total sales. All monetary values in both panels are in US$ (x1000).
Panel A: Distribution of variables
Variable US UK FR DE NL Total
Number of companies 35 40 40 38 42 195
Ownership (%) 0.072 1.975 1.211 1.167 4.244 1.016
Familiarity (km) 1830.668 4063.736 2710.230 2784.851 3557.792 2551.414
Investor hold value ($) 106,096 32,096 23,884 21,367 19,082 202,525
Foreign sales($) 36,813.14 85,525.22 51,903.35 45,155.39 54,496.22 52,879.54 Revenue ($) 89,400,000 42,800,000 35,700,000 44,400,000 9,706,666 51,600,000 Current Liabilities ($) 23,600,000 13,600,000 13,700,000 13,800,000 8,461,808 14,500,000 R&D ratio (%) 0.068 0.034 0.039 0.054 0.032 0.069 Capex (%) 0.041 0.026 0.027 0.039 0.032 0.035 Market capitalization ($) 312,000,000 40,400,000 40,700,000 30,800,000 8,613,856 141,000,000 Cash ratio (%) 0.401 0.339 0.405 0.422 0.179 0.330
Firm size (lnAssets) 18.246 17.212 14.472 17.290 12.642 16.318
Total LT-debt ($) 25,100,000 12,600,000 7,655,562 12,200,000 6,730,089 12,600,000
Leverage (%) 0.271 0.268 0.204 0.204 0.183 0.233
Multinationality (%) 0.001 0.002 0.001 0.002 0.001 0.001
GDP growth (%) 0.023 0.014 0.013 0.006 0.018 0.018
Inflation (%) 0.018 0.017 0.011 0.014 0.026 0.019
Panel B: Descriptive statistics
Variable N Mean SD Minimum Maximum
Ownership (%) 202,525 1.017 6.523 0 645.064
Familiarity (km) 261,547 2,503.801 3,053.744 0 18,819
Firm size (lnAssets) 261,545 16.318 2.919 9.140 19.692
Leverage (%) 261,547 0.233 0.148 0 0.987 Multinational (%) 261,547 0.001 0.004 0 0.573 Blockholder 261,547 0.468 0.499 0 1 R&D ratio (%) 261,547 0.069 0.492 0.0001 13.011 Capex ($) 261,547 0.035 0.022 0 0.233 Cash ratio (%) 261,547 0.330 0.594 0 16.043
value of 16.318. As shown in Panel A, this can be explained by the difference in size of the Dutch and American firms. The mean value for leverage is 0.232, with the maximum of 0.743 indicating there are some highly levered firms present in the sample. This spread of observations indicates a difference in investor preference with regards to the level of leverage of investment objects. Looking at blockholders, 46.8% of investors holds more
than 10%of the firm they invested in and can be seen as a block holder. The high number of blockholders is mostly due to the importance of institutional investors in Europe and as a consequence a lower number of total investors per firm. Another potential expla-nation is the quality of the stocks of blue chip companies, which makes it a trustworthy investment. Surprisingly the relative R&D expenditures are low compared to the relative cash holdings of firms, since prior research mentioned a potential relationship between the two. Although a mean investment in R&D of a value of 7% of sales results in a significant research budget.
Table 2: Correlation Matrix
Table 3 shows the correlation among variables. (1) Ownership; (2) blockholder; (3)
fa-miliarity; (4) multinationality; (5) leverage; (6) Firm size; (7) Capex; (8) R&D
ra-tio; (9) cash ratio; (10) GDP growth; (11) Inflation. * indicates that the correlation
is significant at least at the 10% level, ** at the 5% level, and *** at the 1% level.
Variable 1 2 3 4 5 6 7 8 9 10 11 1 Ownership 1,000 ”” ”” ”” ”” ”” ”” ”” ”” ”” ”” 2 Blockholder 0.228*** 1.000 ”” ”” ”” ”” ”” ”” ”” ”” ”” 3 Familiarity 0.048*** 0.092*** 1.000 ”” ”” ”” ”” ”” ”” ”” ”” 4 Multinationality 0.035*** 0.002 0.048*** 1.000 ”” ”” ”” ”” ”” ”” ”” 5 Leverage 0.040*** 0.161*** 0.016*** 0.024*** 1.000 ”” ”” ”” ”” ”” ”” 6 Firm size 0.235*** 0.652*** 0.050*** 0.068*** 0.233*** 1.000 ”” ”” ”” ”” ”” 7 Capex 0.036*** 0.089*** 0.067*** 0.084*** 0.144*** 0.112*** 1.000 ”” ”” ”” ”” 8 R&D ratio 0.027*** 0.008*** 0.031*** 0.399*** 0.045*** 0.027*** 0.076*** 1.000 ”” ”” ”” 9 Cash ratio 0.051*** 0.105*** 0.041*** 0.375*** 0.029*** 0.148*** 0.022*** 0.765*** 1.000 ”” ”” 10 GDP growth 0.088*** 0.313*** 0.163*** 0.102*** 0.132*** 0.129*** 0.140*** 0.018*** 0.007*** 1.000 ”” 11 Inflation 0.091*** 0.406*** 0.015*** 0.045*** 0.136*** 0.711*** 0.029*** 0.043*** -0.151*** 0.307*** 1.000
projects, which reduce their value as collateral for loans, it makes economical sense that these innovative firms hold more cash. Another high coefficient is observed for firm size and blockholders (0.652 at the 1% level). Which may indicate some multicollinearity. A potential explanation for this could be that the larger firms are a safer investment for for instance pension funds and other institutional investors.
4.2
Univariate analysis
In this section a two sample mean test is performed to test potential difference between domestic and foreign investors. This serves as a first indication if domestic investors follow different investment patterns than foreign investors.
Considering Table 3 depicted below, Panel A, we observe that the mean holding value of domestic and foreign investors differs, with domestic investor’s mean holding value exceeding that of foreign investors. Since p=0.000 at a 95% confidence interval the null hypothesis, that both means are equal, is rejected. Economically this implies that the average domestic investors holdings are higher than for foreign investors. Potentially due to a knowledge advantage, as mentioned by Dziuda and Mondria (2012).
Continuing with Panel B, ownership concentration is equal to the holding value of an investor divided by the total market capitalization of the firm (investment object). A clear difference between the domestic and foreign mean is observed (1.61 and 0.280 re-spectively). Meaning that on average, domestic investors hold a larger percentage of the firm. The null hypothesis that the mean ownership concentration is similar for domestic and foreign investors is rejected (p=0.000 at 95% confidence). This is consistent with the
findings in Panel A and the preference of foreign investors for firms with a larger breadth of ownership (Dahlquist and Robertsson, 2001).
preference as well. They contribute this to the higher reward associated with knowledge asymmetry on investment made in smaller firms.
The results of the two sample mean test indicate a difference between foreign and domestic investors exists with regards to holding value, ownership concentration, and size of the investment object. These preferences can be used by firms to attract foreign in-vestors so to reduce the familiarity bias. For example, a larger breadth of ownership or an increase in firm size might attract foreign investors.
Table 3: Two sample mean t-test
This table depicts the results of the two sample mean tests. This table depicts if a
mean difference between group 0 (domestic) and group 1 (foreign) can be identified,
with regards to the holding value, the fraction of ownership, and firm size. The
frac-tion of ownership is equal to investor holding value divided by the firms market value.
Panel A: Investor holding value
Group Observations Mean Standard error SD 95% confidence interval P r(|T | > |t|) 0 112,033 10,800,000 49,069.07 16,400,000 10,700,000 10,900,000
1 90,492 6,022,313 41,500.47 12,500,000 5,940,973 6,103,654
difference 4,771,126 64,265.57 4,645,167 4,897,085 0.000
Panel B: Ownership (%)
Group Observations Mean Standard error SD 95% confidence interval P r(|T | > |t|)
0 112,033 1.61 0.024 8.135 1.564 1.660
1 90,492 0.280 0.012 3.512 0.257 0.303
difference 1.332 0.0270 1.280 1.385 0.000
Panel C: Firm size
Group Observations Mean Standard error SD 95% confidence interval P r(|T | > |t|)
0 112,033 15.348 0.007 3.091 15.334 15.363
1 90,492 18.121 0.004 1.265 18.112 18.129
difference -2.772 0.009 -2.789 -2.756 0.000
4.3
Ordered probit model
In this section the ordered probit regression models based on equation 2 are presented. Following the results of these models the hypotheses proposed in Section 2 will be con-firmed or rejected. In this run the full data sample is used, later several variations are made to ensure robustness.
interpretation of the economic effects.
Looking at the first model, a positive coefficient is observed at a 95% confidence in-terval, indication that an increase in distance results in a higher likelihood of investing. However, in the final and complete model the coefficient for familiarity is negative, as expected based on the literature. Hence, I fail to reject the null hypothesis and confirm hypothesis 1 that investors are more likely to invest in the geographically proximate in-vestment objects. Moreover, if we look at the marginal effects per group (weak, moderate, and strong ownership) an increase in geographic distance reduces the likelihood that an investor is in the moderate or strong ownership group, and thus, an investor is more likely to be in the weak ownership group. Although the effect is statistically significant, economically it is not. Therefore, I believe it is important to combine conclusions from the marginal effects with conclusions drawn in prior research. Hence, familiarity increases the chance of investing more in one firm (moderate or strong ownership), which is con-sistent with the tilted portfolio composition towards the familiar (i.e. Bergin and Pyun, 2016 ; Huberman, 2001 ; Dziuda and Mondria, 2012 ; Coval and Moskowitz, 1999), and the preference of foreign investors to invest in firms with a larger breadth of ownership (Dahlquist and Robertsson, 2001).
Next, the three moderating variables are added to the model. All three are statisti-cally significant at the 1% level, leverage is the only variable with a positive coefficient. Furthermore, higher leverage seems to attract more clustered ownership based on the marginal effect per group (moderate and strong ownership). An increase in either firm size or multinationality decreases the chance of an investor being in the moderate or strong ownership category.
Considering the different interaction terms, the interaction term between familiarity and leverage is insignificant. Hence, hypothesis 3 that “higher firm leverage strengthens the effect of familiarity when investing” is rejected. Firm size in itself has a negative effect on investment (-0.219 at 1%), the interaction term with familiarity is positive and
signifi-cant. Multinationality (MNC) seems to have a strong negative effect on investment and is significant (1%level). Additionally, the interaction term familiarity*MNC is negative and significant too (95% confidence). Hence, the familiarity bias is weaker for internationally diversified firms which confirms hypothesis 2a.
Table 4: Ordered probit model
This table presents the ordered probit regression results of familiarity on the likelihood of investing using a sample of 261,547 observations from between 2016-2019. The observed firms are from one of five countries: United States, United Kingdom, Germany, France, the Netherlands. The investors originate from one of 75 countries. The dependent variable is ownership, investor holding value divided by the company’s market capitalization, divided in three categories: weak (<0.35% of firm owned), moderate (0.35-7.98% of firm owned), and strong (>7.98% of firm owned).The main independent variable is familiarity, measured as the geographic distance in km between the investor and HQ countries. Leverage is equal to long-term debt divided by the total assets of a firm. For firm size the natural logarithm of total assets is used as a proxy. MNC, is foreign sales divided by the total sales. R&D ratio, R&D expenses divided by sales. The cash ratio is equal to cash and equivalents divided by current liabilities. Block holders is a dummy variable assuming a value of 1 if an investor hold more than 10% of the total company’s value. Model I includes only the main independent variable and the control variables. In model II the moderator variables are added to the model. Models III-V look at the three different interaction terms one by one, and model VI is the full model. Reported beneath each coefficient estimate is the standard deviation. Moreover the *, **, *** refers to 1%, 5%, 10% significance levels respectively.
Variable (1) (2) (3) (4) (5) (6) Familiarity 0.0000199** 0.0000005 0.00000266 -0.00003** 0.000002** -0.00003** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Leverage 0.088*** 0.248*** 0.082*** (0.020) (0.022) (0.022) Familiarity x leverage -0.000002 0.000003 (0.000) (0.000) Firm size -0.213*** -0.219*** -0.217*** (0.003) (0.004) (0.004)
Familiarity x firm size 0.000002** 0.000002**
(0.000) (0.000) MNC -4.511*** -4.198*** -4.710*** (1.513) (1.524) (1.546) Familiarity x MNC -0.001** 0.00004 (0.000) (0.000) R&D ratio -0.055*** -0.066*** -0.112*** -0.062*** -0.100*** -0.054*** (0.016) (0.016) (0.017) (0.016) (0.017) (0.016) Cash ratio 0.189*** 0.013 0.202*** 0.003 0.193*** 0.013 (0.009) (0.010) (0.009) (0.010) (0.009) (0.010) Blockholder -8.815 -9.890 -8.842 -9.900 -8.809 -9.881 (83.566) (85.420) (83.483) (85.216) (83.517) (85.401) GDP growth -41.124*** -25.890*** -41.864 -24.953*** -41.965*** -25.438*** (0.730) (0.792) (0.733) (0.769) (0.747) (0.780) Inflation 18.968*** 3.314** 18.502*** 2.892** 19.872*** 3.524** (1.344) (1.387) (1.344) (1.380) (1.356) (1.392) Number of observations 261,547 261,545 261,547 261,545 261,547 261,545 Prob > chi2 0.000 0.000 0.000 0.000 0.000 0.000 Pseudo R2 0.5993 0.6085 0.5996 0.6084 0.5993 0.6085
Time fixed effect Yes Yes Yes Yes Yes Yes
Notes: The table shows the ordered probit regression coefficients of all the variables used in this study: familiarity; leverage; firm size; MNC; R&D ratio; cash ratio; blockholder; GDP growth; inflation. As well as the interaction effects: familiarity*firm size; familiarity*leverage; familiarity*MNC.
to expand internationally. Via this the company most probably increases international visibility and lowers potential barriers formed by knowledge asymmetry. Both contribute to a larger breadth of ownership, which is beneficial for the firm.
For all models the prob > chi2 is equal to 0.000, indicating that as a whole the or-dered probit model is statistically significant and a good fit. The control variables have the expected coefficients based on prior studies. The blockholder variable is insignificant, perhaps due to the characteristics of an ordered probit model. However, due to the large percentage of blockholders in the sample (see Table 1) I expect that the preferences of these investors influence the other variables.
4.4
Alternative models for robustness
In this section several batteries of robustness tests are performed. Firstly, to control for the differences between the US and the other countries, due to the 401K schemes in the US and the larger number of institutional investors in Europe, the US is excluded from the second run of tests. Thereafter, the model is run only for firms with prior colo-nial ties. Although this research focuses on physical distance as a measure of familiarity, psychological distance influence the sense of familiarity as well. Since, a common lan-guage or legal system reduce information costs, and hence, lower the barriers to invest in foreign assets. Thereafter, the familiarity variable is split into four quartiles to allow for comparison between different investor groups. Again the ordered probit analysis is run once with and once excluding the US. Then a Tobit model test is performed using the same sample as in the original ordered probit model. Lastly, the sample is split based on firm size to ensure that investor preference towards a specific firm size does not cause a bias in the results.
4.4.1 Ordered probit model excluding the US
not bias the results. The relatively high number of investors located in the US may be explained by the higher number of institutional investors in Europe compared to in the US. After the deletion of US firms from the sample it comprises of 154,791 observations.
Table 5: Ordered probit excluding the US
This table presents the ordered probit regression results of familiarity on the likelihood of investing using a sample of 154,791 observations from between 2016-2019. The observed firms are from one of four countries: United Kingdom, Germany, France, the Netherlands. The investors originate from one of 71 countries. The dependent variable is ownership, investor holding value divided by the company’s market capitalization, divided in three categories: weak (<0.35% of firm owned), moderate (0.35-7.98% of firm owned), strong (>7.98% of firm owned).The main independent variable is familiarity measured as the geographic distance in km between the investor and HQ countries. Leverage is equal to long-term debt divided by the total assets of a firm. For firm size the natural logarithm of total assets is used as a proxy. MNC, is foreign sales divided by the total sales. R&D ratio, R&D expenses divided by sales. The cash ratio is equal to cash and equivalents divided by current liabilities. Block holder is a dummy variable assuming a value of 1 if an investor holds more than 10% of the total company’s value.
Variable (1) (2) (3) (4) (5) (6) Familiarity 0.000004** 0.00004** 0.0001 0.0003 0.00007*** 0.00002 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Leverage 0.097* -0.011 0.0629 (0.058) (0.082) (0.083) Familiarity x leverage 0.0001 0.00001 (0.000) (0.000) Firm size -0.088*** -0.085*** -0.085*** (0.005) (0.007) (0.007)
Familiarity x firm size -0.0001 -0.000007
(0.000) (0.000) MNC -6.570*** -4.926*** -4.335**** (1.194) (1.633) (1.644) Familiarity x MNC -0.001*** -0.001** (0.000) (0.000) R&D ratio -0.078*** -0.030 -0.080*** -0.043** -0.055*** -0.030 (0.019) (0.020) (0.019) (0.019) (0.020) (0.020) Cash ratio 0.060*** 0.001 0.063*** -0.010 0.064*** 0.001 (0.016) (0.017) (0.017) (0.017) (0.016) (0.017) Blockholder -8.048 -8.528 -8.047 -8.524 -8.062 -8.521 (59.574) (64.471) (59.574) (64.497) (59.546) (64.480) GDP growth 6.429*** 7.240*** 6.371*** 6.934*** 6.874*** 7.207*** (1.831) (1.837) (1.833) (1.836) (1.833) (1.838) Inflation 2.051 -2.329 1.962 -2.646* 2.566* -2.330 (1.498) (1.537) (1.504) (1.525) (1.502) (1.537) Number of observations 154,791 154,789 154,791 154,789 154,791 154,789 Prob > chi2 0.000 0.000 0.000 0.000 0.000 0.000 Pseudo R2 0.694 0.6954 0.694 0.695 0.694 0.6954
Time fixed effect Yes Yes Yes Yes Yes Yes
Notes: The table shows the ordered probit regression coefficients of all the variables used in this study: familiarity; leverage; firm size; MNC; R&D ratio; cash ratio; blockholder; GDP growth; inflation. As well as the interaction effects: familiarity*firm size; familiarity*leverage; familiarity*MNC. Moreover the *, **, *** refer to 1%, 5%, 10% significance levels respectively.
are similar in both models with the exemption of GDP growth, positive and significant in the model excluding the US. The positive coefficient for familiarity is unexpected based on prior research. Considering the second model of both tables, although the coefficients for familiarity are similar it is only significant in Table 5. The removal of the US significantly impacts the results for model (4) and (6), familiarity is insignificant in contrast to the negative and significant coefficient observed in Table 4. Leverage has a positive coefficient in both tables, whereas firm size and multinationality have a negative coefficient. Hence, for the moderating variables similar results are observed with and without the American firms. Lastly, looking at the interaction terms in the third model, the interaction term with leverage remains insignificant. Moreover, familiarity*MNC is significant and neg-ative similar to Table 4. With regards to the interaction term familiarity*firm size the effect is insignificant in the model without US firms contrasting to the results of Table 4. The observation is that the R&D ratio, cash ratio, block holder, and GDP growth are comparable in both tables, although the strength of the effect reduces especially for the cash ratio and GDP growth. The strong positive effect of inflation in Table 4 becomes smaller and negative in Table 5. The coefficients for the blockholder variable remains insignificant.
4.4.2 Ordered probit model for colonial ties
Although this research uses geographic proximity as a proxy of familiarity, psychic distance may impact an investor’s sense of familiarity as well. According to Bergin and Pyun (2016), similar colonial experience could result in familiar financial institutions, and a common language or legal system. All of which lower the information costs. To account for the potential influence of prior colonial ties, this section focuses on the UK and the Netherlands, and their former colonies. The US is excluded as a prior colony, since the large number of investors and the specific characteristics of the country might outweigh the other former colonies. By comparing these results with the models presented in Table 4, I attempt to ensure the robustness of the results.
Since the information costs potentially decrease as a result of prior colonial ties, one may expect that the influence of familiarity reduces as well. Comparing the two tables, the coefficients for familiarity are similar, although the effect of familiarity on investments appears slightly stronger in Table 6, contrasting the earlier formulated expectation. More-over, leverage appears to have a positive effect regardless of the model, and the interaction term with familiarity is insignificant in both tables. With regards to firm size, the effects remained significant and negative. Whereas the interaction term with familiarity is now negative and more economically significant. This would be in line with prior research (i.e. Coval and Moskowitz, 1999 ; Dahlquist and Robertsson, 2001). As well as, with the expectation that an increase in firm size reduces the effect of the familiarity bias on investment decision, as formulated in hypothesis 2b. The effects of multinationality on investments remains significant and negative, however, the interaction term with famil-iarity is insignificant in contrast to the results of Table 4 and the expectations formed in Section 2. This might be due to the lower information costs, which reduce the benefit of being a multinational company for conveying information. Since legal systems influence the accounting profession, the financial statements are expected to be more similar. In combination with a common language the barriers to international portfolio diversifica-tion reduce regardless of firm internadiversifica-tionalizadiversifica-tion.
blockholder variables display similar results for the prior colonial countries as for the full sample, blockholder being insignificant in both cases. A similar observation can be drawn for GDP growth and inflation, which both remain significant. Moreover, the pseudoR2 is higher for the model on colonial ties.
Concluding, the familiarity bias remains present amongst countries with prior colonial ties. However, the relevance of firm internationalization to reduce this bias disappears, and instead firm size becomes an influential factor in the reduction of this bias.
Table 6: Ordered probit model for colonial ties
This table presents the ordered probit regression results of familiarity on the likelihood of investing, using a sample of 137,102 observations from between 2016-2019. The observed firms are from the United Kingdom and the Netherlands. The investors originate from one of their former colonies or the country itself. The dependent variable is ownership, investor holding value divided by the company’s market capitalization, divided in three categories: weak (<0.35% of firm owned), moderate (0.35-7.98% of firm owned), strong (>7.98% of firm owned).The main independent variable is familiarity measured as the geographic distance in km between the investor and HQ countries. Leverage is equal to long-term debt divided by the total assets of a firm. For firm size the natural logarithm of total assets is used as a proxy. MNC, is foreign sales divided by the total sales. R&D ratio, R&D expenses divided by sales. The cash ratio is equal to cash and equivalents divided by current liabilities. Block holder is a dummy variable assuming a value of 1 if an investor holds more than 10% of the total company’s value. Moreover the *, **, *** refer to 1%, 5%, 10% significance levels respectively.
Variable (1) (2) (3) (4) (5) (6) Familiarity -0.0003* -0.0003* -0.0007** -0.0001*** -0.0002 0.0001 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Leverage 0.074* 0.055 0.008 (0.043) (0.070) (0.070) Familiarity x leverage 0.00002 0.0001 (0.000) (0.000) Firm size -0.136*** -0.103*** -0.101*** (0.005) (0.008) (0.008)
Familiarity x firm size -0.0008*** -0.0008***
(0.000) (0.000) MNC -4.353*** -5.543*** -3.279* (1.254) (1.702) (1.721) Familiarity x MNC -0.0004 -0.0003 (0.000) (0.000) R&D ratio -0.108*** -0.061*** -0.111*** -0.067*** -0.0887*** -0.058*** (0.020) (0.021) (0.20) (0.020) (0.021) (0.021) Cash ratio 0.108*** 0.021 0.0115*** 0.010 0.113*** 0.018 (0.017) (0.018) (0.018) (0.018) (0.017) (0.018) Blockholder -8.446 -9.245 -8.442 -8.997 -8.459 -8.989 (107.443) (132.651) (107.436) (78.001) (107.394) (77.940) GDP growth -41.582*** -31.522*** -41.944*** -30.044*** -42.927*** -31.220*** (1.369) (1.467) (1.375) (1.438) (1.391) (1.475) Inflation 25.338*** 12.616*** 25.330*** 12.115*** 26.431*** 13.020*** (1.616) (1.718) (1.616) (1.703) (1.628) (1.721) Number of observations 137,102 137,100 137,102 137,100 137,102 137,100 Prob > chi2 0.000 0.000 0.000 0.000 0.000 0.000 Pseudo R2 0.0690 0.693 0.690 0.693 0.690 0.693
4.4.3 Ordered probit model per quartile of familiarity
In this section the models of Table 4 are recreated per quartile of familiarity. This allows for comparison between the different sub-groups, for instance an increase in geo-graphic distance or firm size might have a bigger effect on portfolio composition within a certain quartile. The first quartile is equal to zero kilometers or domestic investors. The second quartile ranges from 1-430 kilometers, the third group is 435-5,900 kilometers re-moved from the investment object. The last quartile ranges from 6,090-18,819 kilometers between the investor and the asset.
In Table 7 the mean value of the different variables per quartile are presented, as well as the number of observations or investors per quartile. Comparing the mean values of the four quartiles several conclusions can be drawn. Firstly, in the second quartile the familiarity variable seems positively skewed, whereas it is negatively skewed for the fourth quartile. The first indicating that if an investor is not domestic there might be a mini-mum distance between the investor and the investment object. As for the latter, the fact that it is negatively skewed might indicate that there are only a few outliers that reach the maximum distance of 18,819 kilometers. As well as that the benefits of international portfolio diversification are outweighed by the costs after a certain amount of kilometers. To continue with the dependent variable ownership, the investors in the first quartile have the lowest mean value or concentration of ownership with 0.280. The mean value increases significantly in the second quartile after which it decreases again for the latter two quartiles (2.421, 1.596, and 1.430 respectively). Indicating that investors located within the 1-430 kilometer range of an asset tend to own a larger share of the investment object relative to the other three quartiles. This may be because the balance between the risk reduction by portfolio differentiation and higher information costs are optimal in the perception of the investor. As may be expected, a similar pattern is observed for the blockholder variable.
firms. Potentially this is caused by large institutional investors investing in domestic as-sets, since these investors favor larger firms (Dahlquist and Robertsson, 2001). The fact that leverage is highest and multinationality smallest for this same group is in line with Coval and Moskowitz (1999) findings.
Table 7: Distribution of variables per quartile
This table displays the number of observations and their mean value per variable for the different quartiles. The quartiles of familiarity are: Q1, zero kilometers; Q2, 1-430 kilometers; Q3, 435-5,900 kilometers; Q4, 6,090-18,819 kilometers. The variables for which the mean values are displayed are leverage, the long-term debt divided by the total assets. Firm size, the natural logarithm of total assets. MNC, foreign sales divided by total sales. Blockholders, dummy variable assigned a value of one if an investor holds more than 10% of a firm’s value. Ownership, the investor holder value divided by firm market capitalization. R&D ratio, R&D expenses divided by sales. The cash ratio, cash and equivalents divided by the current liabilities.
Variable Q1 Q2 Q3 Q4 Observations 91,506 10,850 57,868 43,496 Familiarity 0.000 346.025 2832.740 8094.873 Leverage 0.267 0.229 0.243 0.244 Firm size 18.121 17.107 17.374 17.516 MNC 0.001 0.002 0.002 0.002 Ownership 0.280 2.421 1.596 1.43 Block holder 0.172 0.525 0.464 0.373 R&D ratio 0.063 0.098 0.057 0.146 Cash ratio 0.392 0.458 0.398 0.515
To conclude, the mean value for the R&D ratio is significantly higher for the second and especially the fourth quartile. A similar observation can be made about the cash ratio which is in line with the relationship established by Leia, Qiub, and Wan (2018) between the two variables.
The rest of this section is devoted to discussing the ordered probit models for the different quartiles of the full sample. In the next section Table 5 is recreated for the four quartiles.
similar to Table 4. Although the effect of multinationality is still negative and significant the strength of the effect decreased.
As for the control variables, the effect of the R&D ratio changed from negative to pos-itive but remains significant. Indicating that domestic investors prefer innovative firms. The cash ratio has a positive and significant effect for most models. The effect of GDP growth is comparable for both tables, however, the effect of inflation changed to negative for the first quartile model. The pseudo R2 decreased significantly (0.416) although this is still sufficient to trust the results.
Table 8: Ordered probit model for the first quartile of familiarity
This table presents the ordered probit regression results of familiarity on the likelihood of investing, using a sample of 91,506 observations from between 2016-2019. The first quartile of familiarity consists solely of domestic investors. The dependent variable is ownership, investor holding value divided by the company’s market capitalization, divided in three categories: weak (<0.35% of firm owned), moderate (0.35-7.98% of firm owned), and strong (>7.98% of firm owned).The main independent variable is familiarity, measured as the geographic distance in km between the investor and HQ countries. Leverage is equal to long-term debt divided by the total assets of a firm. For firm size the natural logarithm of total assets is used as a proxy. MNC, is foreign sales divided by the total sales. R&D ratio, R&D expenses divided by sales. The cash ratio is equal to cash and equivalents divided by current liabilities. Blockholder is a dummy variable assuming a value of 1 if an investor hold more than 10% of the total company’s value. Moreover the *, **, *** refer to 1%, 5%, 10% significance levels respectively.
Variable (1) (2) (3) (4) (5) (6) Familiarity 0.000 0.000 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Leverage 0.104*** 0.404*** 0.104*** (0.025) (0.024) (0.025) Familiarity x leverage 0.000 0.000 (0.000) (0.000) Firm size -0.255*** -0.258*** -0.255*** (0.004) (0.004) (0.005)
Familiarity x firm size 0.000 0.000
(0.000) (0.000) MNC -3.001 3.785 -3.001 (2.844) (2.862) (2.845) Familiarity x MNC 0.000 0.000 (0.000) (0.000) R&D ratio 0.912*** 0.283*** 1.139*** 0.230*** 0.929*** 0.283*** (0.065) (0.069) (0.067) (0.066) (0.066) (0.069) Cash ratio 0.147*** -0.019 0.149*** -0.026* 0.142*** -0.019 (0.013) (0.014) (0.014) (0.014) (0.014) (0.014) Blockholder -20.575 -11.846 -23.637 -11.158 -20.834 -11.846 (103.449) (176.768) (148.004) (151.881) (93.671) (176.768) GDP growth -38.780*** -22.326*** -39.855*** -21.271*** -38.127*** -22.326*** (2.074) (2.184) (2.074) (2.118) (2.135) (2.184) Inflation -2.085 -9.125* -4.191 -9.342* -2.903 -9.125* (5.395) (5.535) (5.391) (5.497) (0.593) (5.534) Number of observations 91,506 91,504 91,506 91,504 91,506 91,504 Prob > chi2 0.000 0.000 0.000 0.000 0.000 0.000 Pseudo R2 0.397 0.416 0.399 0.416 0.387 0.416
Table 9 depicts the models shown in Table 4 for 10,850 investors from the second quartile. Hence, foreign investors who are located within a 430 kilometer range from the asset they invested in. In contrast to the previous table, familiarity has a significant effect here, namely negative as formulated in hypothesis 1. The effect is more pronounced than in Table 4, hence, the importance of geographic proximity on investment decision might be stronger for this group than for the complete sample.
Table 9: Ordered probit model for the second quartile of familiarity
This table presents the ordered probit regression results of familiarity on the likelihood of investing, using a sample of 10,850 observations from between 2016-2019. The second quartile of familiarity does not contain any domestic investors, however, investors are located relatively close to the investment object (1-430 kilometers). The dependent variable is ownership, investor holding value divided by the company’s market capitalization, divided in three categories: weak (<0.35% of firm owned), moderate (0.35-7.98% of firm owned), strong (>7.98% of firm owned).The main independent variable is familiarity, measured as the geographic distance in km between the investor and HQ countries. Leverage is equal to long-term debt divided by the total assets of a firm. For firm size the natural logarithm of total assets is used as a proxy. MNC, is foreign sales divided by the total sales. R&D ratio, R&D expenses divided by sales. The cash ratio is equal to cash and equivalents divided by current liabilities. Blockholder is a dummy variable assuming a value of 1 if an investor hold more than 10% of the total company’s value. Moreover the *, **, *** refer to 1%, 5%, 10% significance levels respectively.
Variable (1) (2) (3) (4) (5) (6) Familiarity -0.001** -0.001** -0.005 0.002 -0.001** 0.002 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Leverage 0.362** 0.894 0.484 (0.170) (1.116) (1.1147) Familiarity x leverage -0.001 -0.003 (0.003) (0.003) Firm size -0.104*** -0.047 -0.048 (0.016) (0.097) (0.099) Familiarity x firm size -0.001 -0.002
Leverage remains positive and significant in model (2), however it is insignificant in model (3) and (6). The same loss of significance is observed for firm size, although the coefficient is negative. Multinationality does not have a significant impact on investment decisions for this group, the same can be concluded for the three interaction terms. For this group of investors an increase in the cash ratio seems to reduce the likelihood of investing, in contrast to the previous group and the full sample. Contrary to the prior group inflation has a significant and positive effect. The pseudoR2 is now comparable to the one observed in Table 4 indicating its predictive power is better than for Table 8.
The 57,868 investors of the third quartile live between 435 and 5,900 kilometers from the company they invested in. Table 10 depicts the ordered probit regressions for this quartile. The familiarity variable is positive and significant for all six models. Thus, within this group an increase in geographic distance seems to have a slight positive effect on the likelihood of investing, unlike the expectations formulated in hypothesis 1, and perhaps caused by the openness of the European Union. Leverage and the interaction term of leverage and familiarity do not have a significant effect of investment. Similar to the results of Table 4, firm size has a negative effect on holdings. As formulated in hypothesis 2b, the interaction term with firm size has a negative and significant effect. The latter implies that the effect of familiarity on investment is reduced by the moderator firm size. The effect of multinationality is only significant in model (2) for this group. The interaction term remains negative and significant, once more indicating it reduces the impact familiarity has on investment decisions, see hypothesis 2a. The effect of the control variables for this group are consistent with the effects observed in Table 4, and the pseudoR2 is similar for both tables as well.
