The Effect of Culture on Corporate Risk Taking

(1)

0

The Effect of Culture on Corporate Risk Taking

MSc International Financial Management Master Thesis

Final Version

Karim Salman Student number: 1842145 Email: k.salman@student.rug.nl Supervisor: Dr. M. Hernandez Tinoco

(2)

1

Abstract

This study investigates the impact of culture on corporate risk taking. Existing theory suggests that individualism has a positive relationship on risk taking, uncertainty avoidance and harmony a negative relationship on risk taking. Taking a sample of 38 countries, for the years 2010 till 2012, we report significant results presenting that individualism has a positive impact on firm risk taking and harmony a negative impact. As for uncertainty avoidance, we could not find significant results to support our hypothesis in our main tests, with the exception of the robustness tests which shows that uncertainty avoidance does in fact affect risk taking negatively. Furthermore, we provide evidence that culture indirectly affects risk taking though formal institutional channels.

(3)

2

1. Introduction

Most of the prior research done on corporate risk taking has been taken more from the perspective of formal institutions (legal origin, law enforcement, corporate governance) that have an influence on the level of risk that companies take (Li et al., 2013). Relatively less research has been done on the informal institutions (culture) that also govern and explain the way companies assess how much risk to take (Li et al., 2013). It is the interplay of both formal and informal institutions that shapes the way which countries operate, with culture having an impact on the way firms are managed and governed (corporate governance) in a specific country or region.

The overall objective of this paper is to test whether culture has a direct on risk taking by exploring two of the five cultural dimensions, individualism and uncertainty avoidance, identified by Hofstede (2001), and Schwartz’s (1994, 2004) harmony dimension. We then control for firm level (size and leverage) as well as country level (formal institutions) variables to show that culture does matter in corporate risk taking. We further test whether culture affects corporate risk taking indirectly, meaning a chain of influences from cultural values to formal institutions and then to risk taking. That being said, this paper complements the research of Li et al. (2013) by following the basic form of the methodology they employ, but instead we will estimate the models using the OLS regression whereas Li et al. (2013) use HLM regressions (Hierarchal Linear Models).

Therefore the aim of this paper is to study the effects of culture on the level of risk taking, where we intend to answer the question:

How does culture affect the level of risk taken by firms?

(4)

3 cope with uncertainty and turbulent environments and value innovation. We therefore hypothesize that individualism will be positively related to higher levels of risk taking, while uncertainty avoidance and harmony being negatively related to higher levels of risk taking.

To test our three hypotheses, we conduct an empirical study using a sample of 15,059 firms in 38 countries from the period 2010 till 2012. Using the standard deviation on return on assets and R&D as our two proxies for risk taking, we provide significant results concerning that individualism has a positive effect and harmony a negative effect on risk taking. As for uncertainty avoidance having a negative effect on risk taking, we do not find evidence initially; only after conducting robustness tests do we find that uncertainty avoidance has a negative significant effect but only when R&D is the risk taking measure, and other results show that uncertainty avoidance is significant but has a positive effect. We provide two robust tests in our paper, in one of them we include two additional cultural dimensions (power distance and masculinity) from Hofstede (1980, 2004) and find that individualism and harmony are significant when the risk taking proxy is standard deviation on the return on assets, and uncertainty avoidance and harmony are significant when R&D is the risk taking measure. As for the two additional dimensions, only power distance is significant when the risk taking measure is the standard deviation on the return on assets, and when the risk taking measure is R&D we find significant results for both masculinity and power distance. For the second robust test we include inflation and GDP as additional country level controls and find individualism and harmony are significant when standard deviation on the return on assets is the risk taking measure, while individualism, uncertainty avoidance, and harmony are all significant when R&D is the dependent variable. For Inflation and GDP growth, we find no significant results when the dependent variable is standard deviation on the return on assets, and inflation is only significant when R&D is the measure. As for the indirect effects, we find evidence that culture indirectly effects corporate risk taking through the formal institutions. We find results that are consistent in signs and reinforcing, while other results are contradicting and have an opposite sign.

(5)

4 The remainder of the paper is structured as follows. In the next section we review previous literature related to our topic and formulate hypotheses. Section 3 describes the data sample and collection and explains our models. In section 4, univariate as well as multivariate results are presented and discussed. The last section, section 5, includes the conclusion and potential avenues for further research.

2. Literature review and hypothesis development

2.1 Prior research

This paper continues to build upon the literature that exists between corporate risk taking and culture in the international arena. As mentioned earlier, most of the research done on corporate risk taking has been taken from the perspective of formal institutions. John et al. (2008) provide evidence that investor protection is positively related to corporate risk taking, meaning that countries with strong investor protection will lead to lower levels of managers extracting private benefits and hence take on riskier projects. Furthermore, Acharya et al. (2011) argue that stronger creditor rights reduce corporate risk taking as these creditor rights push firms to divert acquisitions that are not value adding, in addition to reduced corporate leverage. We will focus on informal institutions (culture) and its effect on corporate risk taking.

(6)

5 individualism. Han et al. (2010) find empirical evidence that individualism and uncertainty avoidance does have an impact on the earnings discretion of managers. They add that it is the cultural dimensions (informal institution), in addition to the formal institutional environment, are complementary factors that determine the earnings discretion and show that earnings discretion is effected by the interaction between investor protection and national culture. Kreiser et al. (2010) find a significant negative relation between uncertainty avoidance and risk taking, however they do not find a significant positive relationship between individualism and risk taking.

In addition, Kanagaretnam et al. (2011) find evidence that high individualism and low uncertainty avoidance cultures tend to have higher earnings management in the banking sector. They argue that the influence of culture in the banking industry is even more important than in other industries due to the uncertainty that exists in the complexity of operations, and risk assessment in the banking industry. Another study by Kanagaretnam et al. (2014) find evidence that countries which are individualistic, where senior management are self-oriented, self-motivated, and enjoy a high degree of autonomy, and score low in uncertainty avoidance (where senior managers prefer to cope with turbulent and uncertain environments) are more risk taking than countries that score low in individualism and high in uncertainty avoidance. Whereas the first study of Kanagaretnam et al. (2011) show significant effects on the earnings quality in the banking industry, their second study also finds evidence in the banking industry, in that there is a significant relationship between culture and risk taking, but in terms of bank reporting conservatism.

(7)

6 Harmony is another cultural variable that is proven by research to have a negative significant relationship on risk taking (Li et al., 2013; Licht et al., 2007). Unlike the two cultural dimensions, individualism and uncertainty avoidance, defined by Hofstede (2001), Schwartz (1994, 1999) has defined three sets of cultural value dimensions that also help understand the national culture of countries (Licht et al., 2007). Among these three sets is the mastery/harmony dimension, which deals with the relationship of people towards the social world (Licht et al., 2007). Li et al. (2013) provide evidence that Harmony has a significant negative relationship towards risk taking, meaning that low harmony (mastery) cultures tend to take more risks as managers are more comfortable with dynamic environments and stand up for their rights compared to their high harmony countries counterparts, who accept the world as it already is and are uncomfortable with conflicts. In addition, Licht et al. (2007) find evidence that harmony is negatively related to risk taking, by stating that countries that score low in the harmony dimension tend to take more risks as they are more individualistic and competitive in nature and stand up for their own rights.

2.2 Hypotheses development

Individualism, the first cultural dimension, refers to individuals who are self-sufficient and tend to normally rely on themselves, in addition to decision making in countries that score high on individualism are made at the individual’s level (Kanagaretnam et al., 2014). Li et al. (2013) state that countries that score high in individualism tend to be highly competitive and emphasize personal freedom and autonomy, and conclude from their analysis, that individualistic countries leads to more corporate risk taking. Furthermore, Shao et al. (2013) find that cultures that are highly individualistic invest more in long term rather than short term assets which are more risky, hence they take more risks. Licht et al. (2005) argue that because individualistic cultures emphasize autonomy, freedom, and competitiveness they require formal institutions that are able to protect them. Countries that have equity-based markets and are of common law origin are characteristics of an individualistic culture and tend to take more risks (Licht et al., 2007), compared to their collectivistic counterparts who emphasize group work and rely on informal relationships and networks instead of formal institutions (Li et al., 2013). It can be concluded that countries that emphasize individualism are more risk seeking and invest more in risky assets. Therefore, the following hypothesis can be made:

Hypothesis 1: Individualism is positively related to higher levels of risk taking.

(8)

7 opposed to countries that prefer rules and a rigid approach (Gray et al., 2013). Gray et al. (2013), found that countries that score low in uncertainty avoidance tend to take more risks while countries that score high take fewer risks. Kanagaretnam et al. (2014) also state that countries that score low in uncertainty avoidance are more risk seeking due to the fact that they are characterized as having a market based financial system and more volatile earnings. Kwok and Tadesse (2006) highlight that countries that score low in uncertainty avoidance are characterized by countries that have an equity-based financial system, as they state that equity-based markets are more ambiguous as well as require tolerance for uncertainty and turbulent environments. On the other hand, countries that score high on uncertainty avoidance prefer to go by the rules in addition to them having a bank-based financial system (Kwok and Tadesse, 2006). Furthermore, John et al. (2008) state that societies that score high in the uncertainty avoidance dimension do not really supportive of the market-based financial practices, which leads to the expectation that there is lower shareholder protection in these countries and hence lower levels of risk taking. In other words, countries that score low in uncertainty avoidance are more risk seeking.

Hypothesis 2: Uncertainty avoidance is negatively related to higher levels of risk

taking.

Our third dimension of interest, harmony, refers to fitting with the environment in a harmonious manner and to accept the world as it is rather than trying to change or exploit it (Schwartz, 1999). Li et al. (2013) state that countries that score low on the harmony dimension are assertive and exploit others if necessary, and managers are accepting to these institutions and environments that fosters these behaviors, whereas countries that score high on the harmony dimension are more accepting of the world as it is in addition to that they are uncomfortable with conflict and assertive behavior.

Hypothesis 3: Harmony is negatively related to higher levels of risk taking. 3. Methodology

3.1 Data sample and collection

(9)

8 Israel, Italy, Japan, Malaysia, Mexico, the Netherlands, New Zealand, Norway, Peru, Philippines, Poland, Portugal, Singapore, Spain, Sweden, Switzerland, Thailand, Turkey, U.K, and the U.S namely. The intention for choosing these countries in our analysis is to include samples of both developing and developed countries as opposed to focusing solely on one of them. The division between developed and developing countries is taken from the World Bank country overview index (World Bank, 2015). The extracted firms are based on the ISIN codes from the industry sector of the countries, meaning that all industrial companies such as manufacturing, retail and wholesale, information and communication, and so forth are included in the analysis. The ISIN codes of the firms from the financial and insurance sectors have been excluded from the analysis.

(10)

9 calculated the firm level variables for every firm year observation. In addition, all firm level variables have been winsorised at the 1% and 99% percentiles to remove the effect of outliers.

Table 1, on page 11, displays the country level variables and the number of firms extracted from each of the sample countries. The number of firms as seen in the table varies from 4 in Czech Republic to 2,753 in japan. Furthermore, the top three countries scoring the highest in the individualism dimension are the U.S (91), Australia (90), and the U.K (89), and the three countries with the lowest scores are Indonesia (14), Peru (16), China, Singapore, and Thailand all have a score of (20). The three countries ranking highest in uncertainty avoidance are Greece (112), Portugal (104), and Japan (92), while the three lowest ranking countries in uncertainty avoidance are Singapore (8), Denmark (23), and Sweden (29). Lastly, for the harmony dimension, the top three ranking countries are Italy (4.9), Germany (4.7), and Greece (4.69). The three countries scoring the lowest in the harmony dimension are Israel (3.35), Thailand (3.67), and Malaysia (3.68).

3.2 Measuring risk taking

The measures of risk taking we apply for our empirical research is the standard deviation of return on assets (ROA), which measures the operational risk taking of firms (Mihet, 2012), and R&D which is used normally as a measure of a firm’s risky policies (Li et al., 2013). In other words, StdDev (ROA) measures the overall risk that is taken by firms and R&D highlights risk taking of long term investments by firms. These proxies have been used in many researches (Li et al., 2013; Mihet, 2012 amongst others) and are the two of the most prominent proxies in literature. We will use both proxies in our analysis.

The standard deviation of the return on assets (from henceforth StdDev (ROA)), our first measure, is calculated by the ratio of EBITDA divided by total assets (ROA) from the years 2010 till 2012, and then calculating StdDev (ROA) by using the average ROA 3 years years for each firm.

(11)

10

3.3 Measuring culture

(12)

11 Table 1

Country level

variables

Country N IND UA HAR ASDI CPI

Argentina 38 46 86 4.27 0.34 1 Australia 979 90 51 4.13 0.76 3 Austria 41 55 70 4.62 0.21 3 Belgium 58 75 94 4.15 0.54 2 Brazil 153 38 76 4.04 0.27 1 Canada 655 80 48 4.2 0.64 1 Chile 90 23 86 4.49 0.63 2 China 1,288 20 30 3.76 0.76 2 Czech Republic 4 58 74 4.66 0.33 3 Denmark 82 74 23 4.32 0.46 3 Finland 92 63 59 4.59 0.46 1 France 307 71 86 4.5 0.38 0 Germany 372 67 65 4.7 0.28 3 Greece 103 35 112 4.69 0.22 1 Hungary 14 80 82 4.38 0.18 1 India 1,435 48 40 3.98 0.39 2 Indonesia 198 14 48 3.99 0.65 2 Ireland 31 70 35 3.9 0.79 1 Israel 70 54 81 3.35 0.73 3 Italy 147 76 75 4.9 0.42 2 Japan 2,753 46 92 4.3 0.50 2 Malaysia 633 26 36 3.68 0.95 3 Mexico 69 30 82 4.58 0.17 0 Netherlands 67 80 53 4.19 0.20 3 New Zealand 73 79 49 4.19 0.95 4 Norway 94 69 50 4.63 0.42 2 Peru 37 16 87 3.91 0.45 0 Philippines 95 32 44 4.08 0.22 1 Poland 156 60 93 4.24 0.29 1 Portugal 33 27 104 4.56 0.44 1 Singapore 413 20 8 3.99 1.00 3 Spain 75 51 86 4.65 0.37 2 Sweden 194 71 29 4.54 0.33 1 Switzerland 135 68 58 4.55 0.27 1 Thailand 381 20 64 3.67 0.81 2 Turkey 152 37 85 4.31 0.43 2 United Kingdom 808 89 35 3.69 0.95 4 United States 2,734 91 46 3.82 0.65 1

(13)

12

3.5 Measuring control variables

The control variables that we account for in the regression analysis conducted are size, leverage, market to book ratio (the firm level variables), anti-self-dealing index, creditors’ protection index, and world bank country governance index (the country level variables). Each will be explained briefly.

Firm size is calculated by taking the natural logarithm (Ln) of total assets (in U.S dollars). Li et al. (2013) show that the larger the firms are, the less likely culture will influence corporate risk taking because larger firms depend on management systems that are highly controlled. They find that firm size is negatively and significantly related in both the stdDev (ROA) and R&D. Hence, firm size can be seen as negatively related to corporate risk taking. On the other hand, Mihet (2012) finds that size is positive and significantly related when the risk measure is StdDev (ROA) and not significant when the risk measure is R&D. Furthermore, leverage is controlled by dividing total debt from total assets (total debt/total assets). Both Mihet (2012) and Li et al. (2013) report that leverage has a negative relationship to the risk measures, with the exception being that Mihet (2012) does not find a significant relationship when the risk measure is R&D. As an additional firm level control variable, we include the market-to-book ratio which is calculated as market capitalization/common stock.

Anti-self-dealing index, constructed by Djankov et al. (2008) is used as a measure to look at the legal protection of minority shareholders against the expropriation by corporate insiders where they focus on mechanisms such as litigation and disclosure. Creditors’ protection index is a collection of four provisions, ranked from 0 to 4, that are discussed by La porta et al. (1998). These four variables are: “ 1) the absence of an automatic stay in reorganization, 2) the requirement of creditors’ consent or the minimum dividend required for a debtor to file for reorganization, 3) the ranking of secured creditors first in organization, and 4) the removal of incumbent management upon filing for reorganization.” (Li et al., 2013). In addition, world bank country governance index is measured by using 6 key dimensions, which are: Rule of Law, control of corruption, government effectiveness, regulatory quality, voice and accountability, and political stability and lack of violence (World Bank, 2015). These country level variables are obtainable from the Datastream database as well.

3.6 Models

(14)

13 Model 1:

𝑅𝑅𝑅𝑅 𝑚𝑚𝑚𝑅𝑚𝑚𝑚 = 𝛽0+ 𝛽1𝐼𝐼𝐼 + 𝛽2𝑈𝑈 + 𝛽3𝐻𝑈𝑅 + 𝜀𝑖 Where:

Risk measure is StdDev (ROA) or R&D, 𝛽₀ is the constant, 𝛽₁ is the coefficient of IND, IND is Individualism, 𝛽2 is the coefficient of UA, UA is Uncertainty Avoidance, 𝛽3 is the coefficient of HAR, HAR is harmony, and 𝜀𝑖 is the error term.

This regression model is our basic model that we test in order to find out whether there is a significant effect of our independent variables on the dependent variable. The dependent variable above, risk measure, will be used once for the StdDev (ROA) measure and another regression for R&D. Based on our hypotheses, individualism (IND) is expected to have a positive sign while uncertainty avoidance (UA) and harmony (HAR) a negative sign. OLS regressions, using the statistical program E-views, are employed for our analysis in order to investigate the effects of the set of covariates on the risk taking measures.

After regressing the main effects of culture on the proxies for risk taking (StdDev (ROA) and R&D), we then account for our firm and country level control variables to determine whether the effects and power of our first model changes slightly or a lot. This brings us to Model 2:

𝑅𝑅𝑅𝑅 𝑚𝑚𝑚𝑅𝑚𝑚𝑚 = 𝛽0+ 𝛽1𝐼𝐼𝐼 + 𝛽2𝑈𝑈 + 𝛽3𝐻𝑈𝑅 + 𝛽4𝑆𝐼𝑆𝑆 + 𝛽5𝐿𝐿𝑅𝐿 + 𝛽6𝑀/𝐵 𝑚𝑚𝑟𝑅𝑟 + 𝛽7𝑈𝑆𝐼𝐼 + 𝛽8𝐶𝐶𝐼 + 𝛽9𝑊𝐶𝐿𝐼 + 𝜀𝑖

Where:

(15)

14

4. Discussion of results

4.1 Univariate results

Table 2 below displays the firm level descriptive statistics. We see that StdDev (ROA) and R&D have a mean (standard deviation) of 0.330 (4.158) and 0.026 (0.131), respectively.

StdDev

(ROA) R&D IND UA HAR SIZE LVRG

M/B

ratio ASDI CPI WCGI Mean 0.330 0.026 58.466 55.779 4.059 12.319 0.2348 1662.01 0.605 1.945 0.853

Maximum 111.187 2.573 91 112 4.9 19.168 1 372797 1 4 1.868

Minimum 0.001 0 14 8 3.35 1.946 0 0 0.1700 0 -0.557

Std. Dev. 4.158 0.131 26.370 23.874 0.300 2.271 0.220 16582.67 0.204 0.901 0.780 Observations 45,376 45,148 45,179 45,179 45,179 45,148 45,147 42,676 45,179 45,179 45,179

Table 3 shown below shows the univariate correlations between the dependent variables and the independent variables, both at the firm level and country level. As in the article by Li et al. (2013), when looking at the relationships between our independent variables, we observe a negative significant relationship between individualism and uncertainty avoidance, as well as a significant negative relationship between individualism and harmony, in addition to a significantly strong positive relationship between uncertainty avoidance and harmony. Also, when we look at both measures for risk taking and their correlations with the independent variables we see that they are all significant with the expected signs. More specifically, StdDev (ROA) and R&D have a significant positive relationship with individualism, and a significant negative relationship with both uncertainty avoidance and harmony.

Furthermore, there seems to be a pairwise correlation that is not significant, which is between the relationship between harmony and leverage. Also, we see that correlation between uncertainty avoidance and leverage and between StdDev (ROA) and leverage is only significant at the 10% significant level. Furthermore, we note that the pairwise correlation between SIZE and World bank Country Governance Index (WCGI) is significant at the 5% and 10% level. In addition, we observe that there a strong negative correlations between Anti Self-Dealing index and uncertainty avoidance, and between Anti Self-Dealing index and

Notes: StdDev(ROA) is the standard deviation on the return on assets, R&D is R&D expenditure over total assets, IND refers to individualism, UA is uncertainty avoidance, HAR is harmony, SIZE is firm size, LVRG is leverage, M/B ratio is the Market-to-Book ratio, ASDI is Anti Self-Dealing Index, CPI is Creditors’ Protection Index, and WCGI is World bank Country Governance Index.

Table 2

(16)

15 harmony. These correlations suggest that the legal standards and corporate governance (formal institutions) in a country does indeed reflect its national cultural background (Li et al., 2013; Licht et al., 2005, 2007). Also, we see that the relationship between leverage and Creditors protection index is negative and significant, which is consistent with the research of Acharya et al. (2011) who provide support that corporate leverage decreases with stronger creditor rights.

4.2 multivariate regression results

As mentioned earlier, the Ordinary Least Squares (OLS) regression method, using the E-views statistical program, will be used to analyze and interpret the results. Table 4, on the next page, presents the OLS regressions from Model 1, for both StdDev (ROA) and R&D. Most of the results are as expected except for the cultural dimension uncertainty avoidance. When looking at the StdDev (ROA) as the dependent variable, we observe that individualism and harmony are highly significant and have the desired sign; positive relationship between

StdDev

(ROA) R&D IND UA HAR SIZE LVRG

M/B

ratio ASDI CPI WCGI StdDev (ROA) 1 R&D 0.1402 (0.000) 1 IND 0.069 (0.000) 0.164 (0.000) 1 UA -0.026 (0.000) -0.046 (0.000) -0.150 (0.000) 1 HAR -0.037 (0.000) -0.063 (0.000) -0.054 (0.000) 0.604 (0.000) 1 SIZE -0.033 (0.000) -0.195 (0.000) -0.076 (0.000) 0.167 (0.000) 0.123 (0.000) 1 LVRG 0.008 (0.086) 0.037 (0.000) -0.069 (0.000) 0.008 (0.068) -0.004 (0.391) 0.058 (0.000) 1 M/B ratio 0.012 (0.000) 0.020 (0.000) 0.118 (0.000) -0.046 (0.000) -0.081 (0.000) 0.059 (0.000) -0.024 (0.000) 1 ASDI 0.024 (0.000) 0.031 (0.000) 0.064 (0.000) -0.530 (0.000) -0.680 (0.000) -0.135 (0.000) -0.116 (0.000) 0.025 (0.000) 1 CPI -0.037 (0.000) -0.086 (0.000) -0.127 (0.000) -0.229 (0.000) -0.159 (0.000) -0.143 (0.000) -0.105 (0.000) -0.098 (0.000) 0.490 (0.000) 1 WCGI 0.033 (0.000) 0.099 (0.000) 0.636 (0.000) 0.156 (0.000) 0.323 (0.000) 0.012 (0.014) -0.147 (0.000) 0.045 (0.000) 0.129 (0.000) 0.045 (0.000) 1 Table 3 Correlation table

(17)

16 StdDev (ROA) and individualism and a negative relationship between StdDev (ROA) and harmony. Thus, H1 (stating that individualism is positively related to higher levels of risk taking) and H3 (stipulating that harmony is negatively related to higher levels of risk taking) is Supported. On the other hand, uncertainty avoidance does not appear to have a significant relationship with the dependent variable StdDev (ROA) in addition to that the sign appears to be positive, which is controversial to our H2 (stating that uncertainty avoidance is negatively related to higher levels of risk taking). So, H2 is not supported.

(18)

17

N=45,117 N=45,148

Model 1 Model 1

StdDev (ROA) R&D

Coefficient t-statistic Coefficient t-statistic

C 1.660 5.672*** 0.089 9.382***

IND 0.011 14.971*** 0.001 36.711***

UA 0.001 1.068 0.0001 3.134***

HAR -0.499 -6.282*** -0.029 -11.336***

Moreover, table 5 below presents our regressions with the inclusion of firm level and country level control variables, Model 2.

The results with respect to the cultural dimensions here differ slightly from the results in table 4. We see that individualism, like in table 4, is highly significant and positive for both risk taking measures. For the harmony dimension, we see that for StdDev (ROA) the significance is now at the 5% level and not at the 1%, yet it is still significant and negatively related. With respect to R&D, we still see a highly negative relationship with harmony. Unlike in table 4, uncertainty avoidance is now not significant at any level with R&D as the

N=42,602 N=42,654

Model 2 Model 2

StdDev (ROA) R&D

C 1.725 2.583*** 0.356 6.818*** IND 0.009 6.816*** 0.0003 8.234*** UA 0.001 0.697 4.19E-05 1.194 HAR -0.315 -2.282** -0.048 -10.888*** SIZE -0.055 -6.253*** -0.012 -42.316*** LVRG 0.234 2.584*** 0.034 11.874***

M/B ratio 1.47E-06 1.234 1.74E-08 0.462

ASDI 0.523 2.559** -0.0159 -2.451** CPI -0.212 -8.096*** -0.0157 -18.931*** WCGI 0.027 0.505 0.020 11.615*** *** refers to significance at P<.01 Notes:

Dependent variables: StdDev (ROA) is the standard deviation on the return on assets; R&D is R&D expenditure over total assets.

IND refers to individualism, UA is uncertainty avoidance, HAR is harmony, and C is the constant. Table 4 OLS regression Table 5 OLS regression *** refers to significance at P< .01 ** refers to significance at P< .05

(19)

18 risk taking measure. For StdDev (ROA) uncertainty avoidance is still positive and not significant. However, we see in the univariate correlations table that both our risk taking measures are negative and significant with the uncertainty avoidance dimension.

Furthermore, regarding our control variables, we see that size is highly significant and negatively related to StdDev (ROA) and R&D. For StdDev (ROA), a suggestion of why firm size is negatively related could be due to that larger firms that have larger earnings are linked with lower operating risks (Li et al., 2013). With R&D being the risk taking measure, the negative relationship with firm size could be because larger firms, who have higher earnings, have lower R&D expenditures which is due to the economies of scale in innovation (Li et al., 2013). In both proxies of risk taking we see that they have a positive and highly significant relationship with leverage, suggesting that firms that use more leverage tend to take on more risky projects. This result is contrary to the results of Li. et al. (2013) and Mihet (2012) where they find negative significant results. Market-to-book ratio is insignificant in both StdDev (ROA) and R&D. In accordance with Acharya et al. (2011), Mihet (2012), and Li et al. (2013), we find that creditor rights is highly significantly related to both StdDev (ROA) and R&D, which implies that countries where creditor protection is high, the lower the risk taking. Furthermore, we see two different results concerning the relationship between index and our risk taking measures, where the relationship between anti-self-dealing-index and StdDev (ROA) has a positive significant relationship (at the 5% confidence level), and a significant negative relationship with R&D (also at the 5% confidence level). The result obtained from R&D as the proxy for risk taking is more in line with Mihet (2012) that shows there is a negative relationship between anti-self-dealing index, meaning the stronger the protection of minority shareholders, the less risk taking by firms. Li et al (2013) find no significance between anti-self-dealing-index and risk taking in any of their models, except for the exception that when they include leverage in the model, they find a significant negative relationship at the 10% confidence level.

Overall, we find evidence supporting H1 (individualism is positively related to higher levels of risk taking) and H3 (harmony is negatively related to higher levels of risk taking), while we find no support for H2 (uncertainty avoidance is negatively related to higher levels of risk taking). Thus, we find significant direct effects of culture.

(20)

19 We see from the table above that culture does affect corporate risk taking indirectly through the channels of formal institutions, where Li et al. (2013) have also presented results which show that culture effects corporate risk taking indirectly.

The results for individualism (IND) are contradicting to the results of Li et al. (2013) as they find insignificant results for ASDI and CPI. We see from our results that the more individualistic a country is, the stronger their corporate governance structure is and a stronger protection of minority shareholders, while countries that score high in individualism tend to have weaker creditor rights. La porta et al. (1998) state for example in the U.S, management can look for protection (called chapter 11) which gives them a great deal of power and control, which leaves the creditors relatively vulnerable as they can only collect their money or collateral with a delay.

For uncertainty avoidance (UA), we see that all the institutional variables are significant as well. Since uncertainty avoidance accentuates giving power to authorities and conformity, we show that countries that score low in uncertainty avoidance have better minority shareholder protection. Li et al. (2013) have reported the same result with respect to the ASDI variable. Contrary to their study though, we find CPI is positively related to UA while Li et al. (2013) find a negative relationship between the two variables. Furthermore, we see that WCGI has a positive significant relationship which would suggest that the higher uncertainty avoidant cultures have better governance structures.

Looking at harmony (HAR), we also find contradicting results compared to Li et al. (2013) similar to the UA dimension. Like Li et al. (2013) we find that ASDI is significant and negative to HAR, meaning that harmonious cultures are not high in anti-self-dealing measures. Contrary to them, we find that CPI is positive and significant while Li et al. (2013)

N=45,179 N=45,179 N=45,179

IND UA HAR

Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic C 44.683 148.778*** 88.099 293.063*** 4.511 1688.472*** ASDI 7.808 15.579*** -68.850 -137.240*** -1.218 -273.257*** CPI -5.385 -47.526*** 1.377 12.142*** 0.078 76.994*** WCGI 22.898 199.782*** 7.829 68.241*** 0.157 154.201*** Table 6

Indirect effects culture

*** refers to significance at P<.01

Notes: Dependent variables: IND is individualism, UA is uncertainty avoidance, and HAR is harmony

(21)

20 find a negative and significant result between the two. Also, we report a positive significant relationship between HAR and WCGI.

Overall, we can see that culture does indeed have an indirect effect on corporate risk taking though formal institutions, although few of the results are contradictory to Li et al. (2013) we nonetheless find evidence of the indirect effect of culture.

4.2 Robustness tests

We further continue our analysis by including a couple of sensitivity check to determine the robustness shown in the results above.

The first sensitivity test includes two other cultural dimensions of Hofstede’s (2001), masculinity and power distance, to our estimation model and present the results in table 7 shown below.

The results for StdDev (ROA) do not change with respect to the results in table 5 for the most, with the exception that some of the variables are only significant at the 10% level instead of 5% and some variables have become highly significant compared to the ones in

N=42,602 N=42,654

StdDev(ROA) R&D

C 1.879 2.669*** 0.369 16.541*** IND 0.007 4.622*** 7.66E-05 1.561 UA -0.0003 -0.239 -0.0002 -4.705*** MAS 0.001 1.048 0.0003 8.075*** PDI -0.004 -1.829* -0.001 -9.607*** HAR -0.271 -1.905* -0.036 -8.348*** SIZE -0.056 -6.425** -0.012 -43.378*** LVRG 0.229 2.529** 0.033 11.618***

M/B ratio 1.44E-06 1.209 1.07E-08 0.283

ASDI 0.584 2.828*** -0.004 -0.681 CPI -0.228 -8.258*** -0.019 -21.668*** WCGI -0.020 -0.344 0.011 6.119*** Table 7 OLS regression Cultural dimensions *** refers to significance at P<.01 ** refers to significance at P<.05 * refers to significance at P<.1

Notes: Dependent variables: StdDev (ROA) is the standard deviation on the return on assets; R&D is R&D expenditure over total assets.

(22)

21 table 5. Specifically, the significance of harmony, firm size, and leverage have been reduced, though they are still significant in addition to that the anti-self-dealing index is now highly significant relative to table 5. As for the two additional cultural dimensions included, only power distance has a significant negative relationship with StdDev (ROA), which suggests that firms who demonstrate an uneven distribution of power, high power distance countries, will take less risk than cultures that score low in the power distance dimension (Kreiser et al., 2010). Masculinity has no significance. An interesting point, however, is that although uncertainty avoidance is not significant, it does appear to have the expected sign postulated in our H2 (uncertainty avoidance is negatively related to higher levels of risk taking).

(23)

22 In the second sensitivity test we include inflation and GDP growth in the equation, where the results are shown in table 8.

With respect to StdDev (ROA) as the risk taking measure, the results are similar to the results of table 6, but inflation and GDP growth is included rather than the additional cultural dimensions. We see that individualism is positive and significant in addition to harmony being negative and significant to the dependent variable. Uncertainty avoidance remains to be insignificant but does have the expected negative sign. Additionally, inflation and GDP growth are not significant.

Using R&D as the risk measure we see that all three cultural values are highly significant and with the desired signs (individualism is positively related to the measure of risk taking, and uncertainty avoidance and harmony are negatively related). Using this regression with inflation and GDP growth, we see that H1, H2, and H3 are supported. Also, we see that inflation is significant while GDP growth remains insignificant. Firm size and leverage are still significant, like the previous tables, along with anti-self-dealing index and

N=42,491 N=42,543 StdDev(ROA) R&D

C 1.871 2.703*** 0.405 18.489*** IND 0.009 6.552*** 0.0005 10.587*** UA -0.0002 -0.171 -0.0003 -5.886*** HAR -0.309 -2.227** -0.047 -10.680*** SIZE -0.056 -6.375*** -0.012 -43.640*** LVRG 0.238 2.624*** 0.035 12.205***

M/B ratio 1.45E-06 1.212 1.09E-08 0.289

INF -0.018 -1.474 -0.005 -12.512*** GDPGR 0.0005 0.054 -0.0004 -1.270 ASDI 0.423 1.952* -0.044 -6.414*** CPI -0.199 -7.189*** -0.012 -14.038*** WCGI -0.023 -0.370 0.006 3.025*** *** refers to significance at P<.01 ** refers to significance at P<.05 * refers to significance at P<.1

Notes: Dependent variables: StdDev (ROA) is the standard deviation on the return on assets; R&D is R&D expenditure over total assets.

IND is individualism, UA is uncertainty avoidance, HAR is harmony, SIZE is firm size, LVRG is leverage, M/B ratio is market-to-book ratio, INF is inflation, GDPGR is GDP growth, ASDI is anti-self-dealing index, CPI is creditors’ protection index, WCGI is worldbank country governance index, and C is the constant.

Table 8

OLS regression

(24)

23 creditors’ protection index. The full tables of the regressions, 4 till 8, are provided in appendices A till K at the end of the paper.

5. Conclusion

Because the mentality and perception of people differ from country to country, studying risk taking from a cultural lens seems appropriate. Hofstede (1980, 2001) identified the four cultural dimensions in which society in different countries have their own perception and mentality where this in turn leads them to the decisions they make. As an example, we can say that two countries are highly individualistic and risk seeking, but the way they interpret and take the steps towards making these risky decisions might differ between the individuals.

In this paper, we have examined the cultural impact on risk taking by looking at firms that are from different cultures and legal origins, as well as taking representatives from the developed world and from the developing world. Our findings based on the study of 38 countries, from the years 2010 till 2012, we find that individualism and harmony have a significant (with their expected signs; positive for individualism and negative for harmony) effect on risk taking. In other words, we see that the countries that are ranked high in the individualism dimension (where societies are money and achievement oriented) and low in harmony dimension (societies that exploit their opportunities) are more risk taking countries. We did not find evidence supporting our second hypothesis concerning the effect of uncertainty avoidance on corporate risk taking, with the exception when we conducted our robustness tests we found that uncertainty avoidance is negative and significant, but only with R&D being the proxy of risk taking. Furthermore, by finding results consistent with other papers, we can relate and help apprehend that culture as an informal institution is an important factor when considering risk taking, and not only the formal institutions (Gray et al., 2013). These results are the direct effects of culture on risk taking. This could aid investors by being more aware of how the cultural system in a country works to understand the way firms operate and act upon certain situations, and adjust to these criteria among others to come to a conclusion on where to invest. Furthermore, we also provide evidence that culture does affect risk taking indirectly, by using formal institutions as channels though which it then affects risk taking.

(25)

24 investigation will allow for a more detailed view of risk taking that has occurred in countries as the crisis may have had a tremendous effect on the level of risk taking by firms, especially in countries characterized by high individualism, low uncertainty avoidance, and low harmony. Furthermore, future research could analyze the composition of the board as well because now-a-days boards are made up of individuals that are from different cultural backgrounds, and that may in turn have affect in the level of risk taking eventually.

(26)

25

References

Acharya, V. V., Amihud, Y., & Litov, L. (2011). Creditor rights and corporate risk-taking.

Journal of Financial Economics, 150-166.

Djankov, S., La Porta, R., Lopez-de-Silanes, F., & Schleifer, A. (2008). The law and economics of self-dealing. Journal of Financial Economics, 430-465.

Gray, S. J., Kang, T., & Yoo, Y. K. (2013). National Culture and International Differences in the Cost of Equity Capital. management international review, 899-916.

Han, S., Kang, T., Salter, S., & Yoo, Y. K. (2010). A cross-country study on the effects of national culture on earnings management. Journal of International Business Studies, 123-141.

Hofstede, G. H. (1980). Culture's Consequences: International Differences in Work-Related Values. Sage, Thousand Oaks, CA.

Hofstede, G. H. (2001). Culture’s Consequences: Comparing Values, Behaviors,

Institutions, and Organizations Across Nations, 2nd ed. Sage, Thousand Oaks, CA. John, K., Litov, L., & Yeung, B. (2008). Corporate Governance and Risk-Taking. The

Journal of Finance, 1679-1728.

Kanagaretnam, K., Lim, C. Y., & Lobo, G. J. (2011). Effects of national culture on earnings quality of banks. Journal of International Business Studies, 853-874.

Kanagaretnam, K., Lim, C. Y., & Lobo, G. J. (2014). Influence of National Culture on Accounting Conservatism and Risk-Taking in the Banking Industry. Accounting

Review, 1115-1149.

Kreiser, P. M., Marino, L. D., Dickson, P., & Weaver, K. M. (2010). Cultural Influences on Entrepreneurial Orientation: The Impact of National Culture on Risk Taking and Proactiveness in SMEs. Entrepreneurship Theory and Practice, 959-984.

Kwok, C. C., & Tadesse, S. (2006). National culture and financial systems. Journal of

(27)

26 La Porta, R., Lopez-De-Silanes, F., Shleifer, A., & Vishny, R. W. (1997). Legal Determinants

of External Finance. The Journal of Finance, 1113-1155.

Li, K., Griffin, D., Yue, H., & Zhao, L. (2013). How does culture influence corporate risk-taking? Journal of Corporate Finance, 1-22.

Licht, A. N., Goldschmidt, C., & Schwartz, S. H. (2005). Culture, Law, and Corporate Governance. International Review of Law and Economics, 229-255.

Licht, A. N., Goldschmidt, C., & Schwartz, S. H. (2007). Culture rules: The foundations of the rule of law and other norms of governance. Journal of Comparative Economics, 659-688.

Mihet, R. (2012). Effects of Culture on Firm Risk-Taking: A Country and Cross-Industry Analysis. IMF Working Paper, 2-51.

Schwartz, S. H. (1994). Beyond individualism-collectivism: new dimensions of values. In: Kim, U., Triandis, H.C., Kagitcibasi, C., Choi, S.C., Yoon, G. (Eds.),

Individualism and Collectivism: Theory, Method and Application. Sage, Newbury Park, CA

Schwartz, S. H. (1999). A Theory of Cultural Values and Some Implication for Work.

Applied Psychology:An International Review, 23-47.

Schwartz, S. H. (2004). Mapping and interpreting cultural differences around the world. In: Vinkins, H., Soeters, J., Esters, P. (Eds.), Comparing Cultures: Dimensions of Culture in a Comparative Perspective. Brill Academic Publishers, Boston, MA.

Shao, L., Kwok, C. C., & Zhang, R. (2013). National culture and corporate investment.

Journal of International Business Studies, 745-763.

World Bank (2015). Country and Lending Groups. (http://data.worldbank.org/about/country- and-lending-groups). Retrieved 01-01-2015.

World Bank (2015). Worldwide Governance Indicators. (http://databank.worldbank.org/data/views/variableselection/selectvariables.aspx?source=Wor ldwide-Governance-Indicators) Retrieved 24-01-2015

(28)

27

Appendices

Appendix A – OLS regression Model 1, StdDev (ROA)

Variable Coefficient Std. Error t-Statistic Prob.

IDV 0.010654 0.000712 14.97125 0.0000

UA 0.001072 0.001004 1.068235 0.2854

HAR -0.499968 0.079586 -6.282083 0.0000

C 1.659507 0.292562 5.672318 0.0000

R-squared 0.006045 Mean dependent var 0.312501

Adjusted R-squared 0.005979 S.D. dependent var 3.976983 S.E. of regression 3.965077 Akaike info criterion 5.593016

Sum squared resid 709259.2 Schwarz criterion 5.593789

Log likelihood -126166.1 Hannan-Quinn criter. 5.593260

F-statistic 91.44949 Durbin-Watson stat 0.974152

Prob(F-statistic) 0.000000

Appendix B - OLS regression Model 1, R&D

IDV 0.000850 2.31E-05 36.71138 0.0000

UA 0.000102 3.27E-05 3.13405 0.0017

HAR -0.029350 0.002589 -11.33599 0.0000

C 0.089294 0.009518 9.38184 0.0000

Adjusted R-squared 0.031800 S.D. dependent var 0.131116

S.E. of regression 0.129014 Akaike info criterion -1.257702

Sum squared resid 751.4052 Schwarz criterion -1.256930

Log likelihood 28395.36 Hannan-Quinn criter. -1.257459

Notes:

Dependent variable: StdDev (ROA) is the standard deviation on the return on assets. IND is individualism, UA is uncertainty avoidance, HAR is harmony, and C is the constant.

Notes:

Dependent variable: R&D is R&D expenditure over total assets.

(29)

28

Appendix C - OLS regression Model 2, StdDev (ROA)

IDV 0.008722 0.001280 6.816216 0.0000

UA 0.000770 0.001105 0.696569 0.4861

HAR -0.314689 0.137903 -2.281960 0.0225

SIZE -0.054630 0.008737 -6.252525 0.0000

LVRG 0.233539 0.090388 2.583739 0.0098

M/B ratio 1.47E-06 1.19E-06 1.234232 0.2171

ASDI 0.522815 0.204284 2.559263 0.0105

CPI -0.211631 0.026140 -8.096149 0.0000

WCGI 0.027034 0.053550 0.504832 0.6137

C 1.725370 0.667916 2.583213 0.0098

S.E. of regression 4.006064 Akaike info criterion 5.613730

Appendix D - OLS regression Model 2, R&D

IDV 0.000334 4.06E-05 8.234189 0.0000

UA 4.19E-05 3.51E-05 1.193760 0.2326

HAR -0.047641 0.004376 -10.88756 0.0000

SIZE -0.011719 0.000277 -42.31570 0.0000

LVRG 0.034001 0.002864 11.87365 0.0000

M/B ratio 1.74E-08 3.78E-08 0.461800 0.6442

ASDI -0.015893 0.006483 -2.451359 0.0142

CPI -0.015698 0.000829 -18.93081 0.0000

WCGI 0.019740 0.001699 11.61544 0.0000

C 0.356423 0.021193 16.81799 0.0000

Notes:

Dependent variable: StdDev (ROA) is the standard deviation on the return on assets.

IND is individualism, UA is uncertainty avoidance, HAR is harmony, SIZE is firm size, LVRG is leverage, M/B ratio is market-to-book ratio, ASDI is anti-self-dealing index, CPI is creditors’ protection index, WCGI is worldbank country governance index, and C is the constant.

Notes:

(30)

29

Appendix E - OLS regression, Cultural dimensions, StdDev (ROA)

IDV 0.007157 0.001548 4.622121 0.0000 UA -0.000329 0.001375 -0.238936 0.8112 MAS 0.001367 0.001304 1.048153 0.2946 PDI -0.003698 0.002022 -1.828827 0.0674 HAR -0.270714 0.142124 -1.904779 0.0568 SIZE -0.056437 0.008784 -6.425006 0.0000 LVRG 0.228682 0.090434 2.528700 0.0115

M/B ratio 1.44E-06 1.19E-06 1.209260 0.2266

ASDI 0.584251 0.206563 2.828442 0.0047

CPI -0.227799 0.027585 -8.258039 0.0000

WCGI -0.020160 0.058585 -0.344122 0.7308

C 1.879325 0.704154 2.668912 0.0076

Log likelihood -119566 Hannan-Quinn criter. 5.614499

Notes:

(31)

30

Appendix F - OLS regression, Cultural dimensions, R&D

IDV 7.66E-05 4.91E-05 1.561198 0.1185

UA -0.000205 4.36E-05 -4.704509 0.0000 MAS 0.000334 4.13E-05 8.074605 0.0000 PDI -0.000616 6.41E-05 -9.607436 0.0000 HAR -0.037588 0.004503 -8.347965 0.0000 SIZE -0.012058 0.000278 -43.37843 0.0000 LVRG 0.033233 0.002860 11.61791 0.0000

M/B ratio 1.07E-08 3.77E-08 0.282612 0.7775

ASDI -0.004459 0.006545 -0.681209 0.4957

CPI -0.018929 0.000874 -21.66752 0.0000

WCGI 0.011359 0.001856 6.119326 0.0000

C 0.368991 0.022308 16.54072 0.0000

Notes:

(32)

31

Appendix G - OLS regression, Inflation and GDP growth, StdDev (ROA)

IDV 0.009369 0.00143 6.552412 0.0000

UA -0.000237 0.001387 -0.170850 0.8643

HAR -0.308787 0.138683 -2.226564 0.0260

SIZE -0.056285 0.008829 -6.374817 0.0000

LVRG 0.238089 0.090746 2.623676 0.0087

M/B ratio 1.45E-06 1.20E-06 1.211746 0.2256

INF -0.017860 0.012118 -1.473844 0.1405 GDPGR 0.000475 0.008822 0.053854 0.9571 ASDI 0.423068 0.216769 1.951699 0.0510 CPI -0.198929 0.027672 -7.188916 0.0000 WCGI -0.023478 0.063466 -0.369939 0.7114 C 1.870764 0.692150 2.702832 0.0069

Sum squared resid 683495 Schwarz criterion 5.618814

Notes:

(33)

32

Appendix H - OLS regression, Inflation and GDP growth, R&D

IDV 0.000479 4.53E-05 10.58749 0.0000

UA -0.000259 4.39E-05 -5.886341 0.0000

HAR -0.046906 0.004392 -10.67990 0.0000

SIZE -0.012189 0.000279 -43.64039 0.0000

LVRG 0.035021 0.002869 12.20514 0.0000

M/B ratio 1.09E-08 3.78E-08 0.288811 0.7727

INF -0.004802 0.000384 -12.51156 0.0000 GDPGR -0.000355 0.000279 -1.270029 0.2041 ASDI -0.044038 0.006866 -6.413746 0.0000 CPI -0.012300 0.000876 -14.03791 0.0000 WCGI 0.006081 0.002010 3.024847 0.0025 C 0.405249 0.021919 18.48878 0.0000

Adjusted R-squared 0.081909 S.D. dependent var 0.132609 S.E. of regression 0.127062 Akaike info criterion -1.288000

Appendix I – OLS regression, indirect effects culture, IND

ASDI 7.808099 0.501201 15.57876 0.0000

CPI -5.384569 0.113297 -47.52595 0.0000

WCGI 22.89751 0.114612 199.7820 0.0000

C 44.68263 0.300330 148.7782 0.0000

Adjusted R-squared 0.482159 S.D. dependent var 26.37004 S.E. of regression 18.97619 Akaike info criterion 8.724335 Sum squared resid 16267322 Schwarz criterion 8.725107 Log likelihood -197074.4 Hannan-Quinn criter. 8.724578

Notes:

IND is individualism, UA is uncertainty avoidance, HAR is harmony, SIZE is firm size, LVRG is leverage, M/B ratio is market-to-book ratio, INF is inflation, GDPGR is GDP growth, ASDI is anti-self-dealing index, CPI is creditors’ protection index, WCGI is worldbank country governance index, and C is the constant.

Notes:

Dependent variable: IND is individualism.

(34)

33

Appendix J – OLS regression, indirect effects culture, UA

ASDI -68.85006 0.501677 -137.2398 0.0000

CPI 1.377017 0.113405 12.14248 0.0000

WCGI 7.828721 0.114721 68.24129 0.0000

C 88.09916 0.300615 293.0627 0.0000

Adjusted R-squared 0.367039 S.D. dependent var 23.87440 S.E. of regression 18.99419 Akaike info criterion 8.726232 Sum squared resid 16298206 Schwarz criterion 8.727004

Log likelihood -197117.2

Hannan-Quinn

criter. 8.726475

Appendix K – OLS regression, indirect effects culture, HAR

ASDI -1.218332 0.004459 -273.2565 0.0000

CPI 0.077600 0.001008 76.99435 0.0000

WCGI 0.157218 0.001020 154.2008 0.0000

C 4.511032 0.002672 1688.472 0.0000

Adjusted R-squared 0.684166 S.D. dependent var 0.300374 S.E. of regression 0.168807 Akaike info criterion -0.720027 Sum squared resid 1287.305 Schwarz criterion -0.719255 Log likelihood 16269.05 Hannan-Quinn criter. -0.719784

Notes:

Dependent variable: UA is uncertainty avoidance.

ASDI is anti-self-dealing index, CPI is creditors’ protection index, WCGI is worldbank country governance index, and C is the constant.

Notes:

Dependent variable: HAR is harmony.