MSC INTERNATIONAL ECONOMICS AND BUSINESS
Relationship between Foreign Investment and Key
Determinants of Corporate Governance —— Case
Study of China & Germany
By:
Lu Zhou s1655744
October 2007
Thesis supervisor: Dr. Gjalt de Jong
Relationship between Foreign Investment and
Key Determinants of Corporate Governance
—— Case Study of China & Germany
Lu Zhou
ABSTRACT
When financial sources are limited within the domestic country, it is important to understand the factors triggering foreign investors to invest in a certain company. According to an opinion survey done by McKinsey in 2002, the level of corporate governance is a crucial factor for attracting foreign investment. This study examines the relationship between corporate governance and foreign investment within two countries: China as an emerging market and Germany as a developed market economy.
TABLE OF CONTENTS
Page
1. Introduction ……….…… 4
2. Theoretical Background and Hypotheses……… 6
1. Introduction
While the term ‘Corporate Governance’ has no universally agreed definition, in one sense it refers to the combined legal and regulatory system that establishes the structure and role of the corporation’s board and defines the rights and obligations of directors, management, shareholders and other stakeholders (Kimber et al., 2005). In other words, corporate governance deals with the procedures for directing, monitoring and controlling the activities of the corporation by defining the context of decision-making within the enterprise. Clear corporate governance is especially important for investors, because they want to be informed what companies do with their invested money. If this information is not clear, investors have the possibility to invest in companies all over the world. Arthur Levitt, former chairman of the US Securities and Exchange Commission, said: “If a country does not have a reputation for strong corporate governance practices, capital will flow elsewhere. If investors are not confident of the level of disclosure, capital will flow elsewhere. If a country opts for lax accounting and reporting standards, capital will flow elsewhere.” (Sunday Times Business Times, SA, 24 February 2002). In 2002, McKinsey & Company carried out a global investor opinion survey. Their sample consisted of more than 200 institutional investors. The survey result showed that 80% of the respondents would pay a premium for investing in a well-governed company. The size of the premium increases together with the country-level risk. Institutional investors frequently claim that they avoid foreign firms that are poorly governed, particularly when it comes to investments in emerging markets (McKinsey, 2002). From the view of companies, foreign investors can play an important role in funding corporations, especially in countries where domestic sources of outside financing are limited. Moreover, according to Cooper et al. (2007), well-governed firms have better investment opportunities; particularly, higher market-to-book ratio and output-to-capital ratio.
evidence concerning this question. Therefore, I have the motivation to investigate the following two research questions in my paper. Do investors really avoid foreign firms with a low level of corporate governance? And is this possible trend even stronger for firms within emerging market economies? To answer these questions, I selected two countries to conduct my research, one with an emerging market economy (China) and one with a developed market economy (Germany).
2. Theoretical Background and Hypotheses
2.1 Theoretical Background
Prior empirical studies on corporate governance and foreign investment can be divided into those based on country-level data and on firm-level data. In both cases, the results of these studies are rather vague; for example, both Dahlquist et al. (2003) and Chan et al. (2004) research a combination of emerging and developed market countries and find that of a battery of country-level governance variables, only a proxy for government expropriation risk matters. Thus, a better understanding of corporate governance and more research of foreign investment on both levels is needed (Leuz et al., 2006). Based on the literature discussed below, I construct my model using independent variables, which are discussed and used in other related studies.
the poorly governed firms attract less foreign investment; this finding is directly related to country’s information rules and legislation rather than economic development.
Stijn Claessens (2002) argues that a greater access to financing by firms is stimulated by better property rights, which indicates higher development of financial and capital markets. In his article, Claessens selected creditor and shareholder rights as the indicators of property rights. Better creditor rights are strongly and positively associated with the depth of the financial system, leading to an increasing number of lenders and extension of financing. Moreover, a better quality of shareholder protection increases the country's stock markets. These issues lead to an increase in amounts invested by firms, as well as the growth of firms.
Since corporate governance cannot be measured directly, it is necessary to determine factors which can be used to identify the level of corporate governance. A neat and explicit summary of such factors can be found in Stijn Claessens’ paper (Claessens et al., 2004). He addressed the positive link between high-level corporate governance and growth, development and welfare in general. Specifically, Claessens mentions five aspects of high-level corporate governance. Those are increased access to external financing by firms, higher firm valuation, better operational performance, reduced risks of financial crises and better relationships with stakeholders. I will use these factors to construct my hypotheses later on.
The reasons why the percentage owned by largest shareholder, leverage, number of shareholders, dividend yield, ratio of book to market and total assets could be determinants of corporate governance can be explained as follows.
corporate governance level is lower. Moreover, as mentioned in the article ‘Corporate Governance and Development’, Claessens believes that good corporate governance should have the ability to reduce risks of financial crises. Since leverage is the ratio of total debt to total assets, it can indicate a company’s potential financial risk. Consistent with Claessens, Lins, Leuz and Warnock (2006) also use leverage as a proxy for expected firm-level corporate governance. Concerning the number of shareholders, Claessens mentioned that it is possible to associate high-level corporate governance with generally better relationships with all shareholders because firms have to coordinate their relationships with banks, bondholders, labor and government, which can affect firm’s management in different ways. Dividend yield and book to market ratio are selected because, besides reduced risks of financial crises, Claessens also emphasized on firm valuation, better operational performance. He argues that there is a positive relationship between high-level corporate governance and operational performance (for instance, dividend yield, higher valuation, higher profits and growth). Furthermore, Lins, Leuz and Warnock (2006) also use book to market ratio as a proxy for growth. The authors argue that a preference for growth stocks can be reflected in a tendency to hold low book to market value stocks. The total assets variable is also derived from Claessens’ paper (Corporate Governance and Development), which states that valuation level can affect corporate governance level as well. Similarly, Lins, Leuz and Warnock (2006) also use total assets as a proxy for expected firm-level corporate governance. Based on the explanation above, these six variables are reasonable determinants of corporate governance.
2.2 Hypotheses
firms. Moreover, firms with higher accounting disclosure quality and higher outside ownership concentration in this region performed better during the crisis.
Ownership concentration has impact on foreign investment. The protection of investors has shown to be highly important for foreign investors because, in many countries, minority shareholders and creditors are being exploited by the controlling shareholders. When foreign investors finance firms, they face a risk, and many times, the possible returns on their investments are being hidden by the controlling shareholders or managers. Good corporate governance can help foreign investors to protect themselves against this behavior by controlling shareholders (La Porta et al., 2000).
Moreover, Filatochev et al. (2001), suggest that ownership concentration has a negative effect on investment due to the lack of minority shareholder protection. The data in this paper is based on Russian companies, and the authors argue that this finding is relevant for transition economies. Consequently, I expect that the blockholder impact on foreign investment in China is greater than in Germany. Authors use the percentage owned by the largest shareholder as a measure of ownership concentration. Clearly, powerful shareholders have substantial impact on foreign investment, and the hypothesis is:
Hypothesis 1: the higher the number of percentage owned by largest shareholder, the less the foreign investments.
Hypothesis 2: the higher the level of leverage, the less the foreign investments.
An important mission of corporate governance is the coordination of relationships with shareholders. However, coordination becomes more difficult when the number of shareholders increases, because a large number of shareholders can lead to difficulties in decision making. In other words, when there are many shareholders, it is more likely that decision making becomes more complicated. Strategic constituencies and power asymmetries may arise in this case, which is mentioned in Sudarat Ananchotikul’s article as one of the causes for less investment. (Ananchotikul, 2006) Since my sample mostly consists of large listed companies where ownership and management are separated, power asymmetries and decision making problems are more likely to arise when the number of shareholders is large. Therefore, the hypothesis is as follows:
Hypothesis 3: the larger the number of shareholders, the less the foreign investments.
Lins, Leuz and Warnock (2006) argue that dividend yield reflects company’s growth prospects, and hence is a good measure of corporate governance. This view is also supported by McGee (2005), who states that the direct responsibility of boards is to provide returns for shareholders; thus, if a company is satisfying shareholders’ needs, it has good corporate governance. Linz, Leus and Warnock (2006) found that lower dividend yield with a high plowback ratio reflects good growth prospects for the firm, thus making the firm more attractive for investors. Thus, the hypothesis is as follows: Hypothesis 4: the higher the dividend yield, the less the foreign investments.
higher the excess, the lower the equity level; consequently, the lower the firm valuation. (Claessens et al., 2004)
Moreover, Joh (2003) found that better corporate governance is linked to improved rates of return on equity, higher valuation, higher profits and sales growth. However, the causality relationship is questionable, because firms could operate better due to other reasons. Chidambaran et al. (2006) found no statistically significant causality effect between performance and corporate governance, while suggesting that firms might adjust their behavior in response to performance changes.
There is significant evidence that well-performing firms attract more foreign investment. Bilsen and van Maldegem (1999) found that, in transition economies, foreign-owned firms perform better on average. However, it is not clear whether the foreigners targeted above-average firms, which means that causality between foreign investment and firm performance is a questionable issue. Nevertheless, Freund and Djankov (2000) found that foreign investors seek for profitable companies. An appropriate measure of firm performance is book-to-market value, which reflects how firm’s value has changed over time due to its operations. Thus, the hypothesis for this variable is as follows:
Hypothesis 5: the higher the ratio of book to market, the less the foreign investments.
are not well protected, foreigners prefer to acquire the controlling share of the company. Based on these findings, I state the hypothesis:
Hypothesis 6: a higher amount of total assets will attract more foreign investments.
2.3 Control Variable
Eiichi Tomiura’s (2003) paper reexamines the relationship between R&D and FDI. The author argues that excluding firms which do not engage in FDI and/or R&D overstates the impact of R&D on FDI significantly. This paper uses firm-level data of 118,300 Japanese manufacturers. The data consists of firm characteristics and FDI in 1998, which implies that the study is cross-sectional. Firm characteristics include sales, capital, industry classification and R&D spending, whereas the FDI measure is the number of foreign affiliates or subsidiaries of Japanese companies (data on the FDI size not available). The companies are classified according to the participation in R&D and/or FDI activities, which allows comparing results from two samples: restricted to firms engaging in both FDI and R&D and unrestricted sample. For these two samples, the OLS regression model is used, and the result proves that the impact of R&D on FDI is overestimated in the restricted sample.
According to ADBI Institute’s1 paper, there is also some evidence that suggests that the foreign investment approval process was skewed in favor of key high-tech industries such as metallurgy, basic chemicals, machinery, pharmaceuticals, fertilizer, electronics, and motor vehicles. Based on this finding, I choose metallurgy, chemicals, machinery & equipment, medication & biomedicine, fertilizer, electronics, information technology and telecommunication industries as the high-tech industries in my research.
Next to that, Freund and Djankov (2000) have examined foreign investment issues in the Republic of Korea. The research suggests that foreign companies seek for
1
profitable and relatively large firms with low debt ratios. Moreover, foreign investors target high-value-added sectors, such as electronics, chemical products, motor vehicles, industrial machinery, telecommunications, etc. Based on the support of aforementioned studies, I think this variable is quite important and can be considered as a control variable which accounts for industry. Thus, I state the hypothesis of this dummy variable – high-tech industry variable as follows:
Hypothesis 7: the high-tech industries will have more foreign investments.
3. Research Methodology
3.1 Economic Model
Based on the theories described in sections 1 and 2, I construct a model for the two countries’ research. For both countries, I randomly select 100 listed companies. The reason why I choose 100 observations as the sample size for each country is explained below.
"It is sometimes presumed that a sample should be based on some agreed percentage of the population from which it is taken. The view that there is a constant percentage, often thought to be around 10 per cent, which can be applied when sampling populations of all kinds and sizes is quite wrong". (Chisnall, Peter M., Marketing Research, Maidenhead, UK, McGraw-Hill, 1986)
In my paper, I use “Sample Size Calculator”2 to calculate my sample size. Before using the sample size calculator, there are two terms that need to be defined. They are confidence interval and confidence level. The confidence interval is the plus-or-minus figure. Usually, it appears for example in newspapers or television opinion poll results. The confidence level tells that how sure the researcher can be. Confidence level is expressed as a percentage and represents how often the true percentage of the population who would pick an answer lies within the confidence interval. For instance, the 95% confidence level means you can be 95% certain and similarly the 99% confidence level means you can be 99% certain. The wider the confidence interval that a researcher is willing to accept, the more certain the researcher can be that the whole population result would be within that range. This could be easily understood by taking an example. Suppose a researcher asked a sample of 2000 people in a city which brand of bread they preferred, and 60% said brand A. If 95% confidence level and +/- 20 confidence interval adopted, then the researcher can be 95% sure that between 40 and 80% of all the people in the city actually do prefer that brand, but the researcher cannot be so sure that between 59 and 61% (if +/- 1 confidence interval adopted) of the people in the city prefer the brand.
In order to operate sample size calculator, the populations of the observed objects for both China and Germany have to be known. China listed companies in this paper including all those listed companies in Shanghai Stock Exchange and Shenzhen Stock Exchange in 2006. In 2006, there are 819 listed companies3 in Shanghai Stock Exchange and 634 listed companies4 in Shenzhen Stock Exchange; totally 1453 companies. And in Germany, there are 648 listed companies5 in the same year.
After collecting all the information necessary to calculate my sample size by using Sample Size Calculator, I choose the 95% confidence level as most researchers use
2
Sample Size Calculator: for details please go to http://www.surveysystem.com/sscalc.htm
3
Source from Shanghai Stock Exchange website.
4
Source from Shenzhen Stock Exchange website.
5
and input the confidence interval which I prefer (+/- 10). For calculating the minimum needed sample size of China case, I input 1453 in population box and the calculator gives the sample size needed at 95% confidence level together with a +/- 10 confidence interval is 90. Similarly, Germany sample size can be obtained in the same way. If input population is 648, the calculator gives the sample size needed at 95% confidence level together with a +/- 10% confidence interval is 84. Actually, if I adopt +/- 9 confidence interval, the calculated needed sample size of Germany is exactly 100. These calculated numbers indicate that if the sample size of China listed companies is not less than 90 and the sample size of Germany listed companies is not less than 84, it can be considered large enough. Based on calculation above, as well as considering the convenience of calculation and comparison of two-country cases in the latter sector, I decide to use a 100-country sample in both China and Germany cases. Besides, as the large sample theory suggests, the larger the sample size is, the more accurate result that a sample could offer. Therefore, sample size of 100 firms for both China and Germany listed companies would be appropriate and sufficient.
I selected six variables based on previous researches: percentage owned by largest shareholder, leverage, number of shareholders, dividend yield, book-to-market value and logarithm of total assets. There are two reasons that why use logarithm of total assets. First, log form allows for direct estimation of elasticity. Second, the use of logarithm of total assets makes it possible to avoid the hard currency transformation task. This is simply because China and Germany use different currency, and if the direct number of total assets is used, then it is necessary to use the exchange rate to switch all observations in the same currency. However, once log form is adopted, it reflects the percentage change of total assets. Therefore, since these two countries use different local currency on measuring total assets, transforming this variable into logarithm form is more comparable and convenient.
whether the firms belongs to high-tech industry or not. Therefore, my model is as follows: FI = β0 + β1 * POLS + β2 * LVR + β3 * NS + β4 * DY + β5 * RATIO_BTM + β6 * LOG_TA + δ * DUM + et Where: FI = foreign investment
POLS = percentage owned by largest shareholder LVR = leverage
NS = number of shareholders DY = dividend yield
RATIO_BTM = book to market ratio LOG_TA = logarithm of total assets
DUM = dummy variable (1 if high-tech industry; 0 if otherwise) β0 = intercept, the value of Y when all Xi values are 0
β1, 2, 3, 4, 5, 6 = the coefficients of those six independent variables.
δ = the coefficients of dummy variable et = error term
My seven hypotheses are:
Hypothesis 1: the higher the number of percentage owned by largest shareholder, the less the foreign investments.
Hypothesis 2: the higher the level of leverage, the less the foreign investments. Hypothesis 3: the larger the number of shareholders, the less the foreign investments. Hypothesis 4: the higher the dividend yield, the less the foreign investments.
As presented above, my research has 200 observations in two countries. There are three main categories of research methodology. First is time series analysis, second is cross section analysis, and the last one is panel data. Theoretically, Time series method easily suffers such problem like autocorrelation of error term. Besides, it may also have a problem of heteroskedasticity of error term, and this will lead to the error term showing in non-stochastic process in the regression estimation. Cross section study, unlike the time series, observes population subset at one point in time. The most significant advantages of this method are the enhanced possibility to use random sample, conduction within short period of time, cost advantage. The most significant disadvantage is the difficulty to prove the causality of a relationship between variables, which could make the regression estimation biased. Similarly, cross section analysis may also face the issue of heteroskedasticity in the error term. Furthermore, this method may have other problems like measurement error or omitted variable bias. The panel data method is able to overcome the disadvantage of cross-sectional data when comparing to the above two methods. However, in this thesis, my 200 observations from the two countries are only for one year, year 2006. Therefore, it is impossible to use time series data analysis. Besides, it is also not suitable to use panel data analysis. The reason is that panel data method usually refers to the pooling of observations on a cross section of firms over several time periods, but in this paper, there is nothing related to time periods problem. Since there are some zero foreign investment values in both China and Germany sample data, more suitable methodology for my research model is censored regression analysis (Tobit model). Furthermore, censored regression method even allows for heteroskedasticity. In following sectors I will present the implementation of the analysis for this economic model.
3.2 Data Collection and Description
All the data that I obtained for China and Germany listed companies are from second hand resources. The source of the China data is the Yahoo Finance website, while the source of Germany data is AMADEUS database.
Definitions of variables are explained below. First, dependent variable – foreign investment (FI) is defined as percentage owned by foreign shareholders of total capital of the listed company. Percentage owned by largest shareholder (POLS) is the largest shareholder that controlled the percentage of total listed company shares. Leverage (LVR) is calculated by total debt over total assets, while ratio of book to market (RATIO_BTM) is measured as book value per share divided by market share price in the end of 2006. Number of shareholders (NS) represents the total number of shareholders of a listed company in the end of 2006. Dividend yield (DY) is defined as yearly yields of share divided by year-end market price of the share (2006). Log of total assets (LOG_TA) is the natural logarithm of total assets.
As mentioned in the Introduction sector, my research is based on the firm-level data. I randomly select6 100 listed companies in both China and Germany. In total, there are 200 observations; for each observation, there are six independent variables, one (or two if in the two-country joint data case) dummy variable(s) and a dependent variable (for details please see sector 3.1 Economic Model).
Obviously, foreign investment data appears in percentage form. Now move on to the dummy variable. By adding a dummy variable to the economic model, I expect that this additional, firm-specific dummy will make my research and analysis more accurate. As mentioned Section 3.1 Economic Model, I choose metallurgy, chemicals, machinery & equipment, medication & biomedicine, fertilizer, electronics, information technology, and telecommunication as the high-tech industries in my research. Therefore, when a listed company belongs to one of the eight industries, this
6
chosen listed company is categorized as belonging to high-tech industry. Accordingly, its dummy variable’s value is 1; otherwise, the value is 0.
3.2.2 Summary Statistics
Table 1 and Table 2 provide summary statistics for China and Germany samples respectively. Both tables contain mean, median, maximum, minimum, standard deviation and observations numbers etc. statistics. From Table 1 and Table 2, it is obvious that Germany sample has more foreign investments according to mean, median and maximum statistics (i.e. 16.5553%, 11.125% and 86.94% respectively). They are close to 5, 8 and 3 times of relative Chinese statistics respectively. The minimum percentage of foreign investments is 0 for both China and Germany.
3.3 Econometric Techniques and Procedures
The statistical technique analysis in this research consists of two main steps. First is the correlation test. This not only includes testing the correlations between the dependent variable and independent variables, but also tests the correlations between independent variables. Then, I will use regression to fulfill my analysis. Because my data include some zero foreign investment values, the least squares method is not suitable. In this case, using least squares method, for example, ordinary least squares (OLS) will be biased and will no longer be the best linear unbiased estimator (BLUE)7. Based on the characteristics of my data, the censored regression method will be adopted instead. Besides, one of the most common ways of doing censored regression - the maximal likelihood method (ML) will be used for the regression. More specifically, the Tobit model will be used to analyze the data. All the technical research will be carried out in EViews software. In order to get better and more reliable result, two subsamples will be tested both separately and jointly. This means that analysis of China and Germany sample data, and two-country joint data will follow these steps.
4. Empirical Results
In this section, empirical results and discussions will be introduced. First, the correlations of data will be tested. Then, there is one more thing that has to be noticed before using the censored regression. When testing censored regression in EViews software, the error term distribution is requested. Hence, there is a need to test the distribution of the error term of my sample. Finally, censored regression test with one of three error term distributions (i.e. normal distribution, logistic distribution and extreme value distribution) will be suggested to adopt.
7
Gauss-Markov Theorem: under the assumptions SR1-SR5 of the linear regression model, the estimator b1 and b2
have the smallest variance of all linear and unbiased estimators of β1 and β2. They are the Best Linear Unbiased
4.1 Correlations Test
4.1.1 Correlations of China Data
The correlations among the independent variables and between dependent variable and independent variables for China case are shown in Table 3.
First, I consider the correlation between dependent variable and independent variables. Table 3 contains different correlations between foreign investment and independent variables, and dummy variable. Comparing to the other correlations, it seems that the percentage owned by largest shareholder has a higher possibility of having correlation with foreign investment (0.582173), while number of shareholders, dividend yield and dummy variable have least possibility of correlation with foreign investment (-0.047166), (0.051471) and (-0.068683). The rest three independent variables seem to have a less significant relationship with foreign investment (all close to 0.15). Among all the six independent variables and one dummy variable, two independent variables and the dummy variable, specifically, the number of shareholders, book to market ratio and the high-tech dummy variable has negative correlation to foreign investment. The negative signs of these two independent variables are consistent with their corresponding hypotheses, i.e. H 3 and H 5, which expect negative relationships between number of shareholders and book to market ratio to foreign investment respectively.
variables. Even when it tends to have correlation with logarithm of total assets, they are negatively correlated and the value is around 0.28. The rest correlation values among independent variables, including dummy variable, are all no more than 0.23, whether positive or negative. Therefore, the dummy variable can be considered to have no significant correlation with all independent variables.
To summarize, according to the rule of thumb in ‘Undergraduate Econometrics’8 multicollinearity occurs when correlation between independent variables is higher than 0.8. There is no correlation in Table 3 that is higher than 0.8, which means all of the absolute correlation values between every two independent variables are no higher than 0.8. Therefore, the multicollinearity problem may not occur if this sample is examined with the multi-regression model.
4.1.2 Correlations of Germany Data
The correlations between the dependent variable and independent variables and among the independent variables for Germany case are shown in Table 4.
Just as I did in the analysis of China data, I first consider the correlation between dependent variable and independent variables. The results from the Table 4 show that the highest correlations appear between foreign investment and leverage (0.314424) and between foreign investment and number of shareholders (0.258490), while the lowest correlation is -0.074890, which is between foreign investment and dummy variable. Meanwhile, the rest correlations between dependent variable and independent variables are all around 0.1, no matter negatively or positively correlated.
Table 4 also shows the results of correlations among the independent variables. The most obvious significant correlations are between the logarithm of total assets and number of shareholders (0.674674), and between logarithm of total assets and
8
leverage (0.614650). Except that, results also show that there is some correlation between number of shareholders and book to market ratio (0.307151), as well as between number of shareholders and leverage (0.306631). The rest correlations among the independent variables are all close to 0.1 or even much lower. Especially the dummy variable tends to have less correlation with all the other variables.
Finally, according to the rule of thumb in ‘Undergraduate Econometrics’ there is no correlation in Table 4 which is higher than 0.8; this is the same as I found in China case. Since all of the absolute correlation values between every two independent variables are not higher than 0.8, the multicollinearity problem may not occur if this sample is examined with the multi-regression model.
4.2 Censored Regression Test
Aforementioned paragraphs provide a rough idea of this multivariate linear model. According to the specific characteristics of the sample data in this paper, all the dependent variables are successive, but they have certain restrictions. This type of model refers to limited dependent variable models. In this paper, the most significant restriction of this multivariate linear model is that some dependent variable values are zero. This characteristic of sample data requires a censored regression model – a type of the limited dependent variable models9. Because the left censored (i.e. the lowest value of foreign investment) of both China and Germany’s sample data are zero, this multivariate linear model could be a normative censored regression model, i.e. Tobit model.
Censored regression model refers to the condition when the dependent variable is only partly observed. This model usually considered as follows:
9
Where yi* is latent variable and σ is a scale parameter10.
In the canonical censored regression model, known as the Tobit model, the errors are normally distributed, and the observed data y is given by11:
Actually, Tobit regression model is a special case of censored regression model. It was introduced by James Tobin in 1958. In his model, James assumed the error term of the demand of durable commodity is normally distributed; in other words, ei ~ N (0,
σ2), i = 1, 2, …, n. The multivariate linear model in this paper could be a normative censored regression model, i.e. Tobit model, because the left censored of both China and Germany’s sample data are zero. However, in the Tobit model, error term is assumed to be normally distributed. Thus, it is necessary to test the error term distribution of all the data that will be used in this paper. All the test process will be conducted in EViews software.
EViews provides tools to perform maximum likelihood estimation of censored regression model and to use the results for the further analysis. Maximum likelihood is useful for my model because it is used to fit mathematical models also to censored data and offers a way of tuning the free parameters12 of the model to provide an optimum fit. While executing censored regression model in EViews, there is an option about error term distribution which has to be selected. EViews gives three options of error term distribution: standard normal distribution, logistic distribution and extreme value distribution13.
10
Plese see EViews 6 Users Guide II, p232 equation (30.24). 11
Plese see EViews 6 Users Guide II, p233 equation (30.25). 12
A free parameter is a number used to define a theory thoroughly enough so as to make useful predictions. This number should be determined by experiment but some theories include parameters that have not been verified by observation.
13
In other words, the distribution of error term has to be tested before operating Tobit regression. Two common ways can be used to check the assumption of normally distributed errors. One method is histogram of the residuals, and the other method gives statistics that can be used to formally test a null hypothesis that the residuals come from a normal distribution. This formal test called Jarque-Bera test. Jarque-Bera statistic is given by
where S is skewness and k is kurtosis. EViews software can calculate Jarque-Bera statistic and show it in the histogram. Jarque and Bera found that when the residuals are normally distributed, the Jarque-Bera statistic has a chi-squared distribution with 2 degrees of freedom. Therefore, the null hypothesis of normally distributed errors can be rejected if a calculated value of the statistic exceeds a critical value selected from the chi-squared distribution with 2 degrees of freedom.
4.2.1 Test of the error term distribution of China data
The histogram of the ordinary least squares residuals of China data is presented in Table 5.
The 5% critical value from a χ2 distribution with 2 degrees of freedom is 5.99 and the Jarque-Bera statistic from Table 5 is 57.50116. Because 57.50116 > 5.99, null hypothesis of normally distributed errors should be rejected. That is to say, the error term of China data is not normally distributed. Therefore, the Tobit model (normative censored regression model) would not be the best choice for China data. Consequently, I expect some other censored regression models be used for testing the multivariate linear model in this paper.
standard deviation of China data’s error term, which are 4.49E-16 (very close to 0) and 4.360786 respectively. From these, the variance of error term of China data equates to 4.360786 * 4.360786, which is close to 19.016455. Recall the previous paragraphs; footnote 13 gives the exact criteria for error term distribution. Unluckily, it seems that the calculated expected value and variance of China data’s error term fit none of those three types. However, comparing to the other distributions, logistic distribution error term has the highest variance with a zero expected value. Thus, it seems that the logistic distribution of error term is the only possible choice for China case when executing the censored regression model.
4.2.2 Test of the error term distribution of Germany data
The histogram of the ordinary least square residuals of Germany data is presented in Table 6.
Again, the 5% critical value from a χ2 distribution with 2 degrees of freedom is 5.99 and now the Jarque-Bera statistic from Table 6 is 31.65769. Because 31.65769 > 5.99, the null hypothesis of normally distributed errors should be rejected. That is to say, the error term of Germany data is not normally distributed. Therefore, the Tobit model (normative censored regression model) would not be the best choice for Germany data. Thus, I expect some other censored regression models could be used for testing the multivariate linear model in this paper.
4.2.3 Censored regression test of China data
The censored regression result for China is presented in Table 7. Here, the distribution of error term of China data is assumed to be logistic distribution.
According to the results of Table 7, percentage owned by largest shareholder is the only statistic that is significant at the 5% significance level and two-tailed test. The p-value of the percentage owned by largest shareholder is 0.0000. Hence, the coefficient of the percentage owned by largest shareholder (0.756848) is significant and positively related to foreign investment. Unfortunately, this conclusion is opposite to the original hypothesis (hypothesis 1) which expects that the higher number of percentage owned by largest shareholder will lead to less foreign investments. The finding that percentage owned by largest shareholder is positively related to foreign investment can be explained by the fact that foreign investors prefer to invest in firms with greater ownership concentration under risky conditions. (Tóth and Zemčík, 2006) Moreover, there is evidence that minority shareholders in China are typically exploited by blockholders (Hovey, 2004); thus, foreign investors prefer to become blockholders themselves and control larger percentage of the shares.
For the rest test results, no evidence in Table 7 provides clear relationships between independent variables (including intercept coefficient and the dummy variable) and the dependent variable. Since all these p-values are insignificant, all the relevant hypotheses (hypotheses 2, 3, 4, 5, 6 and 7) are not supported. The relationships between all these independent variables (including intercept coefficient and the dummy variable) and the dependent variable are vague.
meaning that official support is available for the firm. (Hovey, 2004) Therefore, foreign investors may avoid firms with high leverage due to lower profitability and higher risk, or favor such firms due to the political support they have. Because of this reason, my analysis failed to produce clear results of the relationship between leverage and the foreign investments. Second, since the minority shareholder rights in China are not well-protected, the “number of shareholders” variable has a much weaker impact on the quality of corporate governance; it mostly depends on ownership concentration, as mentioned before. This could be the reason of the insignificant effect of number of shareholders on foreign investment. Third, in contrast to Lins, Leuz and Warnock (2006), Brown and Caylor (2004) found that high dividend yield means better performance of the firms and higher returns to shareholders. However, this finding does not reflect the possibility that low dividend yield stocks can reflect good prospects for firm growth and thus better returns in the long run. This might explain the ambiguous results of my analysis. Fourth, the insignificant relationship between book to market ratio and the foreign investments can be explained by the fact that foreign investors might see profitable opportunities in recently transformed state-owned enterprises14, which are typically less efficient and have poor corporate governance. (Hovey, 2004) Consequently, since some large companies are recently transformed state-owned enterprises, they can still be favored by foreign investors despite the high book to market ratio, because Chinese government provides additional benefits for these investors. (“Overview of FDI in China”, 2004) This leads to vague results of the relationship between book to market ratio and foreign investment. Fifth, since my sample of China data mostly has large companies (mean logarithm of total assets for China data is greater than for Germany data), the effect of the logarithm of total assets on foreign investment becomes less significant. The reason is that the scale of measurement is too narrow. Typically, foreigners tend to invest in large companies due to the availability and accessibility of information on these companies, and in my sample, only large companies are visible.
14
However, it could be much more significant on a broader scale, if we take into account medium size enterprises as well. Sixth, in China case, high tech companies are mostly state-owned, and there are some restrictions for foreign investors to enter; therefore, despite the fact that previous researches show a positive link between high tech industries and foreign investment, there are limitations for this possibility in China.
4.2.4 Censored regression test of Germany data
The censored regression result for Germany is presented in Table 8. Here, the distribution of error term of Germany data is assumed to be logistic distribution.
It is clear from Table 8 that three independent variables are significant at the 5% significance level and two-tailed test. Those are leverage, number of shareholders and the logarithm of total assets. In the censored regression model, the p-value of the leverage now is 0.0001. In contrast, the p-values of the number of shareholders and the logarithm of total assets are 0.0017 and 0.0196 respectively. However, due to the positive coefficients of leverage and the number of shareholders (38.83551 and 0.415567 respectively), and a negative coefficient of logarithm of total assets (-2.410009), the conclusions from this censored regression model are opposite to the original hypotheses (hypothesis 2, 3 and 6). Nevertheless, one more point should be noticed from Table 8. The p-value of the dividend yield is 0.0422, which is significant at 10% significance level and two-tailed test or at 5% significance level and one-tail test. Furthermore, with a huge negative coefficient of dividend yield (-240.3425), this result perfectly supports the hypothesis 4, which states that the more the dividend yield the less the foreign investment will be expected.
possibilities for privileged consumption. For high-growth firms, the optimal amount of debt is lower, because the interests of shareholders and managers are more aligned. Thus, if the firm operates in a stable environment and faces low growth, high leverage can be a good way to minimize agency costs. Under these circumstances, foreigners would prefer to invest in firms with high leverage. Moreover, firms are more likely to reduce leverage if their credit rating is downgraded (Kisgen, 2006); consequently, firms with high leverage can have high credit ratings, which are an important factor for choosing the investment target. This could explain the positive relationship between leverage and foreign investment. Second, in contrast to China case, minority investors in Germany are better protected against exploitation by blockholders. Moreover, large number of shareholders can be an indicator of a large, well-known firm, which is possibly better known abroad. Consequently, this firm can attract more foreign investment. Third, according to Table 4, the logarithm of total assets has a correlation of 0.615 and 0.675 with leverage and number of shareholders respectively. Although these correlations are lower than the break even value of the rule of thumb (0.8), these correlations could have impact on the coefficient of the logarithm of total assets.
they need to achieve rapid growth, which means good prospects for investors. These two issues could be an explanation for the ambiguous result of relationship between percentage owned by the largest shareholder and the foreign investments. Furthermore, according to Lins, Leuz and Warnock (2006), book to market value is a good proxy for growth stocks. Thus, the effect of book to market value on foreign investments may depend on investor’s preference of the stock types. Typically, high-tech firms have growth stocks, since a large part of earnings is retained in order to engage in R&D projects. (Investopedia) For the Germany sample, nearly half of the listed companies are high-tech firms; consequently, the result of the regression is ambiguous.
4.3 Two-Country Joint Data Test
It makes sense that the bigger the sample size is, the more accurate the empirical testing will be. This idea is based on the big sample theorem. Therefore, it is useful to operate a two-country joint data test, since the results of the two-country joint data are expected to be better than any of all the previous relative individual testing results. There is one change in the economic model which is presented in the sector 3.1. Since now two countries data are united, it is very useful to introduce a new dummy variable (DUM_2) to the previous equation. The value 1 of DUM_2 refers to China (i.e. Chinese listed company) while the value 0 of DUM_2 refers to Germany (i.e. German listed company). Therefore, the new equation is:
FI = β0 + β1 * POLS + β2 * LVR + β3 * NS + β4 * DY + β5 * RATIO_BTM + β6 *
LOG_TA + δ * DUM + ζ * DUM_2 + et
Where:
FI = foreign investment
LVR = leverage
NS = number of shareholders DY = dividend yield
RATIO_BTM = book to market ratio LOG_TA = logarithm of total assets
DUM = high-tech dummy variable (1 if high-tech industry; 0 if otherwise) DUM_2 = country dummy variable (1 if China; 0 if Germany)
β0 = intercept, the value of Y when all Xi values are 0
β1, 2, 3, 4, 5, 6 = the coefficients of those six independent variables.
δ, ζ = the coefficients of dummy variables et = error term
4.3.1 Correlations test
The correlations among the independent variables and between dependent variable and independent variables for two-country joint data are shown in Table 9.
According to Table 9, leverage and the country dummy (DUM_2) seem to have a quite high correlation. Besides, the dividend yield and country dummy (DUM_2) also has quite high correlation (0.618850). Except these two correlations, the other correlations are quite low. Then, according to the rule of thumb in ‘Undergraduate Econometrics’ , since most correlations in Table 9 are much lower than 0.8 in absolute values, the multicollinearity problem may not occur if this sample is used to run the multi-regression model.
4.3.2 Censored regression test of two-country joint data
4.3.2(a) Test of the error term distribution of two-country joint data The histogram of the residuals of two-country joint data is presented in Table 10.
441.5050. Because 441.5050 > 5.99, the null hypothesis of normally distributed errors should be rejected. That is to say, the error term of two-country joint data is not normally distributed. Therefore, the Tobit model (normative censored regression model) would not be the best choice for two-country joint data. Thus, other censored regression models should be used for testing the multivariate linear model in this paper.
Just as in the China and Germany cases, error term of two-country joint data is also not normally distributed. Hence, the distribution of error term is more likely to belong to logistic distribution or extreme value distribution. Table 10 gives the mean and the standard deviation of two-country joint data’s error term, which are 1.63E-15 (very close to 0) and 11.73241 respectively. The variance of error term of two-country joint data equals to 11.73241 * 11.73241, which is close to137.64944. Given the same reason that I explained in previous cases, the logistic distribution of error term seems to be the only possible choice for two-country joint case when executing the censored regression model.
4.3.2(b) Censored regression test of two-country joint data
The censored regression result for two-country joint data is presented in Table 11. Here, the distribution of error term of two-country joint data is assumed to be logistic distribution.
0.6263 to 0.1273 (Germany censored regression analysis). However, now the coefficient of the percentage owned by largest shareholder is not significant compared to the China censored regression analysis (0.0000). This leads to ambiguous result for hypothesis 1 which expects a negative relationship between the percentage owned by largest shareholder and the foreign investment. This result can be explained by different conditions of shareholder right protection in Germany and China. As mentioned before, under the conditions of uncertainty foreign investors target firms with large ownership concentration in order to avoid exploitation by domestic blockholders (Tóth and Zemčík, 2006). Third, the ratio of book to market is significant at 5% significance level (two-tailed test) with coefficient -7.351249. This result fully supports the hypothesis 5 which expects a negative relationship between the ratio of book to market and the foreign investment. Fourth, the logarithm of total assets is significant at 5% significance level (one-tail test) with coefficient 0.932811. Therefore, this result supports the hypothesis 6 which expects a positive relationship between the logarithm of total assets and the foreign investments. Fifth, the new dummy variable (DUM_2) with coefficient -17.00242, which is treated as the countries distinguish variable, is significant at the 5% significance level and two-tailed test. Thus, if the dummy variable is one (for China), the predicted value of foreign investments is 17 percent less, holding all other factors constant. Consequently, Germany has more foreign investment than China.
comparing to Germany one. The p-value of the number of shareholders is 0.0017 (which is significant at 5% significance level and two-tailed test) together with a positive coefficient in Germany censored regression analysis. However, now the p-value of leverage is 0.3819. This result may have occurred due to adding China data (insignificant relationship) to Germany data (significant relationship), leading to a decrease in significance for the joint sample. Third, dividend yield becomes more insignificant (p-value = 0.8642) as well. Due to the fact that the hypothesis 4 was supported for Germany censored regression and not supported for China censored regression, the ambiguous result for the joint country sample is to be expected. Besides, the high-tech dummy variable is also worth noticing because in all the regression tests, this dummy variable is always insignificant. Hence, the original hypothesis 7 was not supported in all regression tests. Since all tests of three different samples show that the control variable (high tech dummy) is insignificant, it seems that this control variable is not closely related to the foreign investment.
5. Conclusion
The thesis investigates the relationship between foreign investment and determinants of corporate governance. Since corporate governance cannot be measured directly, it is necessary to determine factors which can be used to identify the level of corporate governance. In this paper, corporate governance determinants are percentage owned by largest shareholder, leverage, number of shareholders, dividend yield, book to market ratio and total assets. Here, percentage owned by largest shareholder (POLS) is the percentage of total shares held by the largest shareholder. Leverage (LVR) is calculated as total debt over total assets, while ratio of book to market (RATIO_BTM) is measured as book value per share divided by market share price in the end of 2006. Number of shareholders (NS) represents the total shareholders number in the end of 2006 of a listed company. Dividend yield (DY) is defined as yearly yields of share divided by year-end market price of the share (2006). Log of total assets (LOG_TA) is the natural logarithm of total assets.
and two-country joint sample, I found that for China sample only percentage owned by largest shareholder has significant positive relationship with foreign investment; however, the original hypothesis is supposed to have a negative relationship. For Germany sample, hypothesis 4 which states that the higher the dividend yield the less the foreign investments, is fully supported by the empirical result. Besides, in contrast to the original hypotheses which expected negative relationship on foreign investment, leverage, number of shareholder and logarithm of total assets are found to have positive relationships with foreign investment. For the two-country joint sample, both hypotheses of book to market ratio and logarithm of total assets are supported at 5% significance level. More specifically, book to market ratio seems to have a negative effect on foreign investment, while logarithm of total assets is found to have a positive effect on foreign investment. The other aspects of corporate governance do not have clear relationship with foreign investment. Furthermore, the country dummy variable (DUM_2) is also significant at 5% significance level with two-tailed test. From this result, it can be concluded that Germany (DUM_2 = 0) seems to have more foreign investment than China (DUM_2 = 1), and this conclusion is consistent with sample statistics summaries. The results of this study also have interesting implications for listed companies in both developed market economy and emerging market economy to attract the foreign investment. According to the findings, listed companies should focus on reducing their stocks’ book to market ratio and increasing their total assets to show that they have quite big firm sizes. Moreover, though the empirical results show that developed market economy (Germany) has more foreign investment than emerging market economy (China), it would also be helpful for attracting foreign investment if listed companies in China pay more attention on percentage owned by largest shareholder while listed companies in Germany focus on reducing their dividend yield by investing in successful projects.
REFERENCES
Aggarwal, Raj and Kyaw, NyoNyo, “Leverage, Investment Opportunities, and Firm Value: A Global Perspective on the Influence of Financial Development”, 2006 Ananchotickul, Sudarat, “Does Foreign Investment Really Improve Corporate
Governance? Evidence from Thailand”, University of California, Berkeley, 2006 Arikawa, Miyajima, Saito, “Investment and Corporate Governance: Evidence from
Japan in the 1990s”, 2002
Bilsen, Valentijn and Van Maldegem, Peter, “Foreign Direct Investment and Enterprise Performance in Transition Countries: Evidence From Russia and Ukraine”, 1999
Brown, Lawrence and Caylor, Marcus, “Corporate Governance and Firm Performance”, Georgia State University, 2004
By The Shanghai Stock Exchange and Organization for Economic Co-operation and Development (OECD)
Chan, Kalok, Vicentiu M. Covrig, and Lilian K. Ng, “What Determines the Domestic Bias and Foreign Bias? Evidence from Mutual Fund Equity Allocations Worldwide”, Journal of Finance, 60, 1495-1534, 2004
Chidambaran, Palia, Zheng, “Does Better Corporate Governance “Cause” Better Firm Performance?”, 2006
Chisnall, Peter M., Marketing Research, Maidenhead, UK, McGraw-Hill, 1986
Claessens, Stijn and Berglöf, Erik, “Enforcement and Corporate Governance” World Bank Policy Research Working Paper No. 3409, 2004
Claessens, Stijn, and Joseph P. H. Fan, “Corporate Governance in Asia: A survey”, International Review of Finance 3:71-103, 2002
Comparison of New York Stock Exchange Corporate Governance Rules and China Corporate Governance Rules for Listed Companies
Cooper, Bohren and Priestley, “Corporate Governance and Real Investment Decisions”, SSRN working paper, 2007
Corporate Governance Reform in the EU by 2nd Meeting of the Task Force, 30 September 2004 CEPS, 1 Place du Congrès, 1000 Brussels
Dahlquist, Magnus, Pinkowitz, Lee, René Stulz, and Rohan Williamson, “Corporate Governance and the Home Bias”, Journal of Financial and Quantitative Analysis 38, 87-110, 2003
Erica Weintraub Austin, Bruce E. Pinkleton, 2000, “Strategic Public Relations Management: Planning and Managing Effective Communication Programs”, Lawrence Erlbaum Associates (ISBN 0805831606)
Fausten, Dietrich K., “Asymmetric Information and the Composition of Foreign Investment”, Monash University, 2003
Filatochev, Kapelyushnikov, Dyomina, Aukutsionek, “The Effects of Ownership Concentration on Investment and Performance in Privatized Firms in Russia”, Managerial and Decision Economics, 22:299-313, 2001
“Foreign Direct Investment and Enterprise Performance in Transition Countries: Evidence from Russia and Ukraine” by ESRC Centre for Business Research, University of Cambridge, Working Paper No. 144
Fowler, Jr., Floyd J. Survey Research Methods - Applied Social Research Methods Series - Volume 1. California: 1984
Freund, Caroline and Djankov, Simeon, “Which Firms Do Foreigners Buy? Evidence from the Republic of Korea”, Policy Research Working Paper No. 2450, 2000 Hanna, Michael, “Corporate Governance Issues at CarBoulevard.com A Case Study”
Haas School of Business University of California at Berkeley, 2004
Harrison, Ann E., Margaret S. McMillan, “Does Direct Foreign Investment Affect Domestic Firms Credit Constraints?”, 2002
Hovey, Martin, “Corporate Governance in China: an Empirical Study of Listed Firms”, Griffith Business School, August 2004
Investopedia, A Forbes Media Company,
Jia-gui, Chen, Qun-hui Huang, “Comparison of Governance Structures of Chinese Enterprises with Different Types of Ownership” Senior Fellows, Chinese Academy of Social Sciences, 2001
Joh, Sung Wook, “Corporate governance and firm profitability: evidence from South Korea before the economic crisis”, Journal of Financial Economics, 68:287-322, 2006
Karuna, Christo, “CEO reputation and corporate governance”, The Paul Merage School of Business, University of California
Kimber, D. P. Linton and G. O’Neill, “Corporate governance in the Asian Pacific region: A selective review of developments in Australia, China, India and Singapore”, Asia Pacific Journal of Human Resources, 43(2): 180-97, 2005 Kisgen, Darren, “Do Firms Target Credit Ratings or Leverage Levels?”, Boston
College, 2006
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny, “Investor Protection and Corporate Governance”, Journal of Financial Economics 58:3-27, 2000
Lele, Priya and Siems, Mathias, “Shareholder Protection: A Leximetric Approach”, University of Cambridge, 2006
Lemmon, Michael and Lins, Karl, “Ownership Structure, Corporate Governance, and Firm Value: Evidence from the East Asian Financial Crisis”, The Journal of Finance, 2003
Leung, Erika, Liu Lily, Shen, Lu, Taback, Kevin, Wang, Leo, “Financial Reform and Corporate Governance in China”, MIT Sloan School of Management Cambridge, Massachusetts, 2002
Lins, Leuz, Warnock, “Do Foreigners Invest Less in Poorly Governed Firms?”, Finance Working Paper No. 43/2004, revised 2006
McGee, John; Thomas, Howard; Wilson, David, “Strategy: Analysis and Practice”, McGraw-Hill, 2003
McKinsey and Company, 2002, “Global Investor Opinion Survey” <http://www.mckinsey.com/governance>
Muradzikwa, Samson, “Foreign Investment in SADC”, Working Paper 02/67, School of Economics University of Cape Town, 2002
Myers, “The Capital Structure Puzzle”, Journal of Finance, 39(3):575-592, 1984 “Overview of FDI in China”, Ministry of Commerce of People’s Republic of China,
June 14, 2004, <http://english.mofcom.gov.cn/aarticle/topic/bizchina/fdiinchina /200406/20040600233643.html >
Policy Dialogue on Corporate Governance in China
Schipani, Cindy A., Junhai, Liu, “Corporate Governance in China: Then and Now”, Columbia Business Law Review, 2002:1-69
Structuring for Success --- The first 10 years of capital markets in China Sunday Times Business Times, SA, 24 February 2002, p12
“The High Dividend Yield Return Advantage: An Examination of Empirical Data Associating Investment in High Dividend Yield Securities with Attractive Returns Over Long Measurement Periods.”, Tweedy, Browne Fund Inc.
Tomiura, Eiichi, “FDI into Asia and domestic R&D intensity of Japanese manufacturers: firm-level relationship”, 2003
Tóth, Peter and Zemčík, Petr, “What Makes Firms in Emerging Markets Attractive to Foreign Investors? Micro-Evidence from the Czech Republic”, Charies University, Center for Economic Research and Graduate Education, Academy of Sciences of the Czech Republic, Economics Institute, Working Paper Series (ISSN 1211-3298), 2006
Two Foreign Investment Policies Will Be Adjusted, <http://en.epochtimes.com/news/6-11-28/48644.html>
