Keywords: Corporate governance, Investment opportunity JEL classification: G34

ACADEMIA ECONOMIC PAPERS 31 : 3 (September 2003), 301 331 When Will the Controlling Shareholder Expropriate Investors? Cash Flow Right and Investment Opportunity Perspectives Konan Chan Department of Finance National Taiwan University Shing-yang Hu Department of Finance National Taiwan University Yan-zhi Wang Department of Finance National Taiwan University Keywords: Corporate governance, Investment opportunity JEL classification: G34 Correspondence: Yan-zhi Wang, Ph. D. Student of the Department of Finance, National Taiwan University, Taipei 106, Taiwan. Tel: (02) 2368-5846; Fax: (02) 2366-0764; E-mail: D89723007@ms89.ntu. edu.tw. We acknowledge Yeh-ning Chen and two anonymous referees for their insightful comments to this paper.

09:18 AM page:302 ABSTRACT In this paper, we examine how investment opportunities influence the impact of cash flow rights on a firm value. Previous papers argue that cash flow rights serve as the incentives of a controlling shareholder to expropriate outside investors, and document that cash flow rights increase firm value. We find that when firms have opportunities to invest in positive NPV projects, cash flow rights do not increase firm value. However, when there exist investment opportunities and cash flow rights are relatively low, cash flow rights still increase firm value. Only when cash flow rights are relatively high, will firm value not be affected by cash flow rights, given the potential investment gains. Hence, in the presence of profitable projects, whether cash flow rights increase firm value depends on the level of cash flow rights. These results suggest a non-linear relationship between cash flow right of the controlling shareholder and firm valuation.

09:18 AM page:303 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) 1. INTRODUCTION Corporate governance is an important issue in corporate finance, and modern corporate governance theories focus on the effect of cash flow rights held by the ultimate controlling shareholder. 1 Past studies argue that the cash flow rights reflect incentives of the controlling shareholders to expropriate outside investors. Large cash flow rights reduce incentives of the controlling shareholder to transfer firm resources to private benefits. For example, La Porta et al. (2002) and Claessens et al. (2002) find that there exists a cash flow right effect where firm value are positively associated with cash flow rights. The literature presents two important factors which determine how cash flow rights affect firm value. One is investor protection mechanisms, and the other is investment opportunity. Investor protection mechanisms were first documented by La Porta et al. (2002), which follows their path-breaking papers about the impact of investor protection on financial development (La Porta et al. (1999, 2000)). They argue that when laws and government policy are advantageous to outside investors and well enforced, the controlling shareholder has huge costs to expropriate firm resources, and the cash flow right is not important to firm value. As a result, the cash flow right effect decreases with the level of law protection to investors. This result is further supported in recent studies (e.g., Volpin (2002), Lemmon and Lins (2003), and Nenova (2003)). As for the other factor, Lemmon and Lins (2003) argue that when there exist investment opportunities, cash flow rights do not increase firm value. Treating the Asian financial crisis as a negative impact on investment opportunities, they examine the relationship between stock returns and cash flow rights during and before the crisis. They find that cash flow rights did not affect stock returns before the crisis, but did have a strong positive impact during the crisis for East Asian countries. However, how do investment opportunities influence the positive relationship between cash flow rights and firm value in general? Although Lemmon and Lins (2003) argue that this positive relationship will vanish when there exist profitable projects, it is not evident whether this phenomenon exists outside of the financial crisis. Moreover, whether the impact of investment opportunities on this positive relationship is 1 Shleifer and Vishny (1997) survey the literature on corporate governance. Since then researchers have focused on the issue of the effect of cash flow rights (see a further survey of Becht et al. (2002)). 303

09:18 AM page:304 Academia Economic Papers 31:3 (2003) related to the cash flow right level is unclear. Thus, the purpose of this paper is to give a thorough examination of the relationship among firm value, cash flow rights and investment opportunities. In this paper, we argue that as investment opportunities exist, cash flow rights do not increase firm value. When investing resources into positive NPV projects rather than transferring resources, the controlling shareholder has chances to increase future wealth in proportion to future cash flows. Thus, the controlling shareholder will wait for future benefits and expropriate less now. Moreover, the investment opportunity can be treated as an opportunity cost of expropriation. When an investment opportunity is present, the cash flow right does not matter so much because the controlling shareholder has more opportunity costs inexpropriation. Therefore, the positive relationship between cash flow rights and firm value exists only when investment opportunities are not available. In addition, we propose that when firms have positive NPV projects and cash flow rights are relatively low, cash flow rights still increase firm value. This argument is different from Lemmon and Lins (2003). In particular, given the presence of investment opportunities, if a controlling shareholder has tiny cash flow ownership, she will get nearly no profits from future investment gains due to her small cash flow right. 2 As a result, she tends to expropriate rather than invest resources into positive NPV projects even when there exist investment opportunities. On the contrary, if a controlling shareholder has more cash flow ownership, she will wait for the future investment gains, and expropriate less now. Accordingly, under the presence of investment opportunities, whether cash flow rights raise firm value is dependent on the level of cash flow rights. In this paper, we employ the long-run industry sales growth as a measure of investment opportunity. This measure is designed to assess how investment opportunities alter the positive relationship between cash flow rights and firm value over time. In particular, Lemmon and Lins (2003) emphasize that there exists a severe endogenous problem between the firm s investment decisions and firm value. A number of past studies also document that the firm value and investment opportunity are jointly determined (e.g., McConnell and Muscarella (1985), Fazzari and Petersen (1993), and Cho (1998)). In our paper, since a single firm s decision will hardly affect the industry equilibrium in a competitive market, industry characteristics like concentration and 2 Theoretically, a controlling shareholder can controlafirmwithzerocashflowownershipthrough pyramid ownership structure and cross holding (Bebchuk et al. (1999)). 304

09:18 AM page:305 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) industry growth will not be changed by a firm. Therefore, the industry sales growth is almost an exogenous variable to individual firms, and can avoid the endogeneity problem mentioned in Lemmon and Lins (2003). We choose the third quartile as the critical value to divide firms into high and low growth industries. This methodology makes sure that investment opportunity shocks are significant enough. When the investment opportunity shock is large, the impacts of investment gains on reducing incentives to expropriate will be significant as well. In addition, we use capital expenditure and R&D expenditure as alternative measures of investment opportunities to check the robustness of our results. Our empirical results are consistent with the notion that cash flow rights are independent of firm value with the presence of investment opportunities. However, when firms have positive NPV projects and cash flow rights are low, cash flow rights still increase firm value. For example, for industries with high sales growth, the Tobin s q is not associated with cash flow rights. On the contrary, for industries with low sales growth, the Tobin s q increases in cash flow rights. Besides, for industries with high sales growth, the positive relationship between Tobin s q and cash flow rights exists when cash flow rights are relatively low, but disappears when cash flow rights are relatively high. These results suggest a non-linear relationship between the cash flow right and firm value. This paper contributes to the corporate governance literature in two ways. First, we extend La Porta et al. (2002) and Lemmon and Lins (2003) to give an in-depth examination of the relationship among cash flow rights, investment opportunities and firm value. By our methodology, we can investigate the impact of investment opportunities on the cash flow right effect in cross-sectional analyses. Second, our research relates to the literature examining the impact of corporate ownership structure on firm valuation. 3 Previous studies suggest a non-linear relationship between ownership and firm valuation. We also find a non-linear relationship between the cash flow right and firm value depending on investment opportunities. To our knowledge, this is the first study to document this non-linear relationship. The paper is organized as follows. Section 1 is the introduction. Section 2 3 Early studies focus on the non-linear relationship between firm value and ownership, rather than the cash flow right by the ultimate controlling shareholder. The ownership in these papers is usually defined as the shareholding of the five largest shareholders, or by management board (e.g., Demsetz and Lehn (1985), Morck et al. (1988), McConnell and Servaes (1990), Cho (1998), and Demsetz and Villalonga (2001)). Instead, subsequent papers after Shleifer and Vishny (1997) examine the relationship between cash flow rights and firm performance (e.g., La Porta et al. (2002), Claessens et al. (2002), and Lemmon and Lins (2003)). 305

09:18 AM page:306 Academia Economic Papers 31:3 (2003) presents a simple model. Section 3 describes the data. Section 4 presents the empirical findings. Section 5 shows the industry effect. Section 6 discusses the robustness of results and Section 7 concludes. 2. A SIMPLE MODEL In this section, we introduce a simple model, which is similar to Johnson et al. (2000) and La Porta et al. (2002). The hypothesis settings in our model follow assumptions of these two papers. Although we follow some of their assumptions, it is emphasized that we present a non-linear relationship between cash flow rights and firm value, which has never been documented in the literature. Here we describe our main hypotheses as follows. We assume that the unique controlling shareholder is an entrepreneur who has the cash flow right α of the firm, and α is exogenously determined. The firm has the amountofcashi, which is invested in a project with the gross rate of return R, and the firm has no cost in the investment. Besides, as a private benefit of controlling the firm, the entrepreneur can divert a share of the firm resources to herself. The diversion can take the form of salary, transfer pricing, or subsidized personal loans, etc. We further assume s 0 to avoid the problem of negative expropriation. For the case of negative s, it means that the entrepreneur will invest in the project with her own money due to good timing. If she would like to finance the project with additional money, then the firm needs to issue SEO. However, the entrepreneur could not buy all of the shares newly issued, since the other shareholders also have the right to take the offering. Thus, it is less meaningful and more complicated for negative s. Moreover, when the controlling shareholder diverts share s from firm resources, she faces a cost function c(s), c(s) = s2 2, (1) where c is the share of resources that a controlling shareholder wastes when s is diverted. 4 We assume the cost c to be borne by the entrepreneur rather than by all the 4 La Porta et al. (2002) considers c as a function of s and k, which is the level of law protection. However, we do not discuss law protection in our paper and omit it. 306

09:18 AM page:307 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) shareholders. Using these assumptions, the entrepreneur maximizes: α(1 s)ri + si c(s)i, (2) where the first term is the gain from investments after diverting, and the remaining terms are the benefit from expropriation. Notice that the entrepreneur only can divert the original resources of the firm, so R is absent in the last two terms. Since the solution for optimal s is independent of I, the objective function becomes: α(1 s)r + s s2 2. (3) The first order condition is: s = 1 αr if αr < 1 = 0 if αr 1. (4) By optimization of s to the entrepreneur, called s, an entrepreneur can maximize the utility. Besides, the firm value is defined by (1 s )RI q, sofirmvalueisnondecreasing in α from equation (4). We call the positive relationship between cash flow rights and firm value, the cash flow right effect. From equation (4), we find there exists a non-linear relation, and it leads a non-linear relationship between cash flow rights and firm value. Such a non-linear relation is due to the assumption of non-negative s, which has been mentioned above. Based on the equation (4), s would be larger than zero if and only if R<1/α. This means that if R is lower than 1/α, the controlling shareholder will expropriate the firm value. Since future investment opportunities can be treated as opportunity costs of expropriation to the controlling shareholder, when opportunity costs are higher, the controlling shareholder will not divert firm resources and profits. Johnson et al. (2000) also mention this pattern in their model, and suggest that the controlling shareholder will not expropriate firm value if αr 1. 5 Furthermore, when firms have gainful 5 Although they also argue the same pattern, we discuss different phenomena. They focus on how the investor protection mechanism affects the firm value with investment opportunity shocks, while we focus 307

09:18 AM page:308 Academia Economic Papers 31:3 (2003) projects in which to invest, the cash flow right does not matter so much. Hence, the positive relationship between cash flow rights and firm value will vanish due to gains from investments. In contrast, the firm is increasing with the cash flow right, ie., the cash flow right effect is present. Therefore, we have the following testing hypothesis. H1 For firms with enough investment opportunity shocks, cash flow rights of the controlling shareholder do not influence the firm value in general. In this paper, we also argue that there is a non-linear relationship between cash flow rights and firm value. When α is greater than 1/R, the controlling shareholder will not divert the firm value from equation (4). Given R greater than one, the entrepreneur need not own 100% cash flow right to prevent expropriation. Thus, given investment opportunities, we separate firms into two groups, one with larger cash flow rights (α 1/R), the other with lower cash flow right (α < 1/R). For the first group, investment opportunities eliminate the cash flow right effect. Firm value are independent of cash flow rights if firms have positive NPV projects. However, for the second group, the impact from investment opportunities is so small that firm value still increase with cash flow rights. Thus, whether investment opportunities influence the positive relationship between cash flow rights and firm value depends on the level of cash flow rights. The theoretical model is presented in the following figures. For Figures 1 and 2, we draw the relationship between firm value and expropriation given investment opportunities in a bold line (the line with R>1). The other notations are the same as the model. When the firm has profitable projects, incentives of the controlling shareholder to expropriate become lower. For the cash flow right higher than 1/R, the impact of cash flow rights on firm value is eliminated by investment opportunities, so the bold line with high α is horizontal. On the other hand, when cash flow rights are low, the cash flow right still positively influences the firm value. Thus, the slope is positive in the left-hand side of the bold line in Figure 2. This implies that the firm value has a non-linear relationship with the cash flow right. Therefore, we propose our second testing hypothesis. H2 For firms with investment opportunities but low (high) cash flow rights, firm value are (not) increasing along with cash flow rights. on the impact of investment opportunities on the cash flow right effect. 308

09:18 AM page:309 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) s R = 1 R > 1 1/R 1 Å low Å high Å Figure 1 Relationship between α and s q R = 1 R > 1 1/R 1 Å low Å high Å Figure 2 Relationship between α and q 309

09:18 AM page:310 Academia Economic Papers 31:3 (2003) 3. DATA Our initial sample contains listed Taiwanese companies during the period of 1987 to 2000 from the Taiwan Economic Journal (TEJ) database. We obtainedthe information of family groups from China Credit Information Service from 1987 to 1998 in order to calculate the cross-holding ownership in each of the family groups. For firms not owned by family groups, we calculate the cash flow right and voting right of the ultimate controlling shareholder based on La Porta et al. (1999, 2000 and 2002). Since we don t have family group data for 1999 and 2000, we collect only non-family firms in these two years. Moreover, we exclude financial firms, observations within one year of IPO, and sample firms without ownership information. 6 We also follow La Porta et al. (2002) to require that the controlling shareholder who has the largest shareholding through pyramid or cross-ownership controls have at least 10% of the voting rights. We delete firms with dispersed ownership structure because even the largest shareholder in these firms may have only limited ability to expropriate. Finally, in order to reduce the impact of outliers, we delete the top and bottom 2% observations based on Tobin s q, cash flow rights and R&D expenditure ratios. As a result, our final sample contains 918 firm-year observations for 178 firms. The variables we use in the paper are defined as follows: Tobin s q = the book value of debt plus the market value of common equity divided by the book value of assets. CF right = The cash flow right held by the controlling shareholder. R&D expenditure = R&D and advertisement expenditures divided by annual sales minus the industry median. Sales growth = the geometric average of sales growth over the past three years. We follow previous literature to compute the cash flow right and voting right through pyramid or cross-holding ownership (e.g., Bebchuk (1999), La Porta et al. 6 The reason for excluding data within one year of IPO is that the firm value of IPO tends to be high, and the ownership tends to be concentrated. Thus, we delete these observations to ensure that results are not due to the features of IPOs. 310

09:18 AM page:311 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) (year 1995) Ritek Corp CDIB 10.54% Cho-kai VC 4.37% Cho-ou VC 7.03% China VC 3.51% Cho-ou VC 62.5% China VC 37.13% Note: CDIB is China Development Industrial Bank; VC means venture capital company. Figure 3 Ownership Structure of Ritek Corp (1999, 2000, 2002)). We first identify the ultimate controlling shareholder of a firm, and then compute his/her cash flow right and voting right. 7 For example, we illustrate the ownership structure of Ritek Corp in Figure 3. In this case, we have a two-tier pyramid ownership structure, and the cash flow right and the voting right are 13.62% (0.625 10.54% + 7.03%) and 17.57% (10.54% + 7.03%), respectively, for the ultimate controlling shareholder, Cho-ou VC. As for the calculation of the cash flow right and voting right, we have more detailed information in the Appendix. In this paper, we use the portfolio test to check our hypotheses. We categorize the sample based on cash flow rights and industry sales growth, respectively. First, high and low cash flow right groups are separated by the median cash flow right of the controlling shareholder. Second, high and low industry sales growth groups are classified by the 3 rd quartile of the long-run industry sales growth. Long-run industry sales growth is the geometric average of annual industry sales growth rates during the past three years (called industry sales growth hereafter). Annual industry sales growth is the simple average of individual firms sales growth in each industry. Over our sample period, we have 335 industry-year observations for three-year industry sales growth, andthe thirdquartile of these industry-year observations is 17.3%. We use the 17.3% to separate the sample into high and low growth groups. If an industry sales growth rate is higher than 17.3%, all firms in the industry in that year will be classified in the high growth group. 8 7 The controlling shareholder can be a family, state, financial institution or a corporation. 8 For example, the electronic industry sales growth of 31.4% in the year 2000 is computed based on the following steps. First, we obtain sales growth rates for individual firms (e.g., TSMC s annual sales growth rates are 14.3%, 45.6% and 127.3% from 1998 to 2000; UMC s annual sales growth rates are 26.5%, 58.1% and 260.5% from 1998 to 2000). Second, we compute the simple average of annual sales growth for three years, which are 31.8%, 27.2% and 35.5% during 1998 to 2000, respectively. Then 311

09:18 AM page:312 Academia Economic Papers 31:3 (2003) One may have two questions regarding our industry sales growth grouping. The first one is why we use past three-year industry sales growth as a proxy for future growth. We assume past sales growth is a reasonable proxy for future growth because using industry level characteristic avoids the problem of endogeneity. In particular, Lemmon and Lins (2003) emphasize that there exists a severe endogenous problem between the firm s investment decisions and firm value. A number of past studies also document that the firm value and investment opportunity are jointly determined (e.g., McConnell and Muscarella (1985), Fazzari and Petersen (1993), and Cho (1998)). In our opinion, since a single firm s decision will hardly affect the industry equilibrium in a competitive market, industry characteristics like concentration and industry growth will not be changed by a firm. Therefore, the industry sales growth is almost an exogenous variable to individual firms, and its use as proxy can avoid the endogeneity problem mentioned in Lemmon and Lins (2003). The other question is why we choose the 3 rd quartile to be the critical value in the proxy of the investment opportunity. In the case of Lemmon and Lins (2003), they select the Asian financial crisis as an investment opportunity shock. Since the shock is so large that controlling shareholders lose their incentive to run the business, firm value would increase with cash flow rights due to the cash flow right effect. Therefore, we choose the 3 rd quartile of the investment opportunity proxy in classifying groups to make sure that investment opportunity shocks are large enough to obtain meaningful and significant results. 4. EMPIRICAL RESULTS Our primary valuation measure is the Tobin s q. Panel A in Table 1 presents the summary statistics. Panel B shows the difference between groups. In panel B of Table 1, we find that Tobin s q for high and low cash flow rights are 1.947 and 1.772, respectively. The difference is 0.175, which is significant. This result supports the cash flow right effect of La Porta et al. (2002), where cash flow rights increase firm value. we get the (long-run) industry sales growth of 31.4% ( 3 (1 + 31.8%)(1 + 27.2%)(1 + 35.5%) 1). All firms in the electronics industry in year the 2000 are classified as high growth since 31.4% is greater than 17.3%. 312

09:18 AM page:313 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) Table 1 Summary Statistics Panel A: Summary statistics Mean S.D. Min. Max. Median Q1 Q3 Skew Kurtosis Tobin s q 1.8595 0.9407 0.6699 7.7748 1.6392 1.2402 2.1874 2.1669 6.9201 CF right 0.1942 0.1085 0.0288 0.5100 0.1719 0.1146 0.2478 1.0506 0.7167 R&D exp. 0.0029 0.0130 0.0223 0.0763 0.0000 0.0031 0.0051 2.0654 6.6292 Sales growth 0.0997 0.2647 0.5005 3.6941 0.0545 0.0112 0.1335 5.2421 49.6485 Panel B: Tobin s q by grouping High Low Diff. CF right 1.947 1.772 0.175** (1.05) (0.81) t = 3.06 N = 459 N = 459 Industry sales growth 2.039 1.810 0.237** (0.99) (0.92) t = 3.22 N = 200 N = 718 Note: Panel A in the Table shows the summary statistics. Panel B presents the average Tobin s q by cash flow rights and industry sales growth. Tobin s q is the book value of debt plus the market value of common equity divided by the book value of assets. CF right is the cash flow right held by the controlling shareholder. R&D expenditure is the R&D and advertisement expenditures divided by annual sales minus the industry median. Sales growth is the geometric average of the past three years sales growth. Industry sales growth is the geometric average of annual industry sales growth for the past three years. Annual industry sales growth is the simple average of individual firms sales growth in each industry. CF right groups are divided by the median CF right. Industry sales groups are divided by the 75 th percentile of industry-year observations. Numbers in the parentheses are standard deviations, and N is the number of observations in each cell. * and ** represent the significance levels of 5% and 1%, respectively, based on a two-tailed t-test. The difference of Tobin s q between high and low industry sales growth groups is 0.237, which is significant. Although we find a positive relationship here, there are two ways of investment opportunities can increase firm value. One is the direct effect of investment opportunities creating profits from projects. The other is the indirect effect in which investment opportunities eliminate expropriation incentives of the controlling shareholder (Johnson et al., 2000). Therefore, we need the following empirical analyses for the indirect effect. 313

09:18 AM page:314 Academia Economic Papers 31:3 (2003) 4.1 Impact of investment opportunities on the cash flow right effect Our first hypothesis is that for firms with enough investment opportunities, cash flow rights of the controlling shareholder do not influence firm value in general. We divide observations by a two-way independent sort for cash flow rights and investment opportunities, and present the results in Table 2. In panel A of Table 2, when the industry sales growth is higher then the 3 rd quartile, the difference across high and low cash flow rights is only 0.06, which is not significant. In contrast, as the industry sales growth is lower, the difference between high and low cash flow rights is 0.192, which is significantly different from zero. Therefore, the cash flow right of the controlling shareholder positively influences the firm value if investment opportunities are absent. In addition, the cash flow rights do not affect firm value in the presence of investment opportunities. 9 Moreover, our paper uses regression to check our findings by controlling time effect. Because our samples are from 1987 to 2000, we need to control time effects to ensure that the findings in Table 2 are reliable. Pooling 918 observations, we regress the Tobin s q on the cash flow right with year random effect. The regression results are presented in Table 3. First, models I and II in Table 3 show the simple analyses similar to La Porta et al. (2002), which present a positive relationship between cash flow rights and firm value. Second, model III and IV regress the Tobin s q on the cash flow right in high and low industry sales growth groups, respectively. As a result, the coefficient of the cash flow right in model III is significant, but it is insignificant in model IV. This finding suggests that investment opportunities eliminate the effect of cash flow rights on firm value. When a firm has profitable projects in which to invest, whether cash flow rights are high or low is not vital. The firm value will be independent of the cash flow right. Therefore, the coefficient of cash flow rights is insignificant in model IV. 10 9 We also test the hypothesis by comparing the first and the last quartile, since the use of the symmetric sub-sample can avoid estimation biases. Consequently, the differences between high and low cash flow rights are 0.06 (t = 0.42) and 0.378 (t = 2.18) for high and low industry sales growth, respectively. This result still supports our first hypothesis. We thank one referee for this insightful suggestion. 10 Although not reported in the paper, we did try various firm characteristics, such as the firm size and R&D expenditure as control variables in the regression. Coefficients of cash flow rights with and without industrial sales growths are 0.7255 (t = 1.16) and 0.9845 (t = 3.98), respectively. Accordingly, our main result still holds in this robustness check. 314

09:18 AM page:315 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) Table 2 Tobin s q Ranked by Cash Flow Rights and Investment Opportunities Industry sales growth High CF right Low CF right Diff. 2.065 2.005 0.060 High (1.02) (0.96) t = 0.42 N = 113 N = 87 1.909 1.717 0.192** Low (1.06) (0.76) t = 2.77 N = 346 N = 372 Note: This Table shows the Tobin s q under a two-way independent sort of cash flow rights and investment opportunities. The investment opportunity is measured by industry sales growth in this table. Numbers in the parentheses are standard deviations, and N is the number of observations in each cell. * and ** represent the significance levels of 5% and 1%, respectively, based on a two-tailed t-test. Intercept Sales growth CF right Table 3 Random Effect Regressions of Tobin s q Model III- Model IV- Model I Model II low industry sales high industry sales growth group growth group 1.872** 1.841** 1.856 1.874 (9.73) (9.47) (9.06) (10.03) 0.189 0.598** 0.317 (1.61) ( 2.87) (1.93) 0.832** 0.847** 1.088** 0.684 (3.38) (3.44) (4.17) (1.12) N 918 918 718 200 Note: This Table presents random effect regressions. The dependent variable is Tobin s q. Models I and II use the full sample to run regressions, where CF right is cash flow right held by the controlling shareholder, and sales growth is the geometric average of threeyear individual sales growth. Model III uses the low industry growth sub-sample, and model IV uses the high industry growth sub-sample. The correlation coefficient of CF right and sales growth is 0.025. Parameter estimations are based on regressions with year random effects. Numbers in the parentheses are t-statistics. * and ** represent the significance levels of 5% and 1%, respectively, based on a two-tailed t-test. 315

09:18 AM page:316 Academia Economic Papers 31:3 (2003) Rather than year random effect regressions, we also regress Tobin s q on the cash flow right with ordinary least square method (OLS). The empirical findings are the same as random effect regressions, and we do not present them here. As we mention above, the selection of the 3 rd quartile of industry sales growth makes the investment opportunity shock large enough. We also employ different cutoff points as the standard of the investment opportunity but do not present these in the paper. In addition to the 75% cut-off point, the median, 60%, 70%, 80% and 90% critical values are employed. As a result, the impact of investment opportunities is insignificant until the cut-off point of 70% for regression results. However, portfolio tests for all classifications support our first hypothesis. Therefore, we believe that the investment opportunity shock is large enough when the industry sales growth is greater than the third quartile. Our findings are similar to those of Lemmon and Lins (2003). However, we employ industry sales growth as investment opportunities, which can be applied to crosssectional analyses. Hence, the methodology in our paper does not rely on particular events or timing, and is more general, too. 4.2 Precise relationship among variables Our second hypothesis is that for firms with investment opportunities but low (high) cash flow rights, firm value are (not) increasing along with cash flow rights. In the other words, the impact of investment opportunities on the cash flow right effect depends on the level of cash flow rights. Table 4 presents the Tobin s q of firms in the high industry sales growth group. In order to analyze this non-linear relationship, we divide the full sample into low CF right, middle CF right and high CF right groups. Low, middle and high CF right groups are divided by the lowest, the middle, and the highest cash flow right tercile, respectively. As a result, we find that the average Tobin s q in the low and middle CF right groups are 1.852 and 2.264. The difference across these two levels is 0.412, which is significant differently from zero. In contrast, the average Tobin s q in the high CF right groups is 2.025, and the difference between the middle and high CF right groups is insignificant. Therefore, given high industry sales growth but low cash flow rights of the controlling shareholder, cash flow rights still have a positive effect on firm value. On the contrary, when firms have high industry sales growth and relatively high cash flow rights, cash flow rights are independent of firm value. This suggests that whether investment opportunities alter the positive relationship betweencash flow rights and firm value depends on the level of cash flow rights. 316

09:18 AM page:317 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) Table 4 Tobin s q Ranked by Cash Flow Rights for High Industry Sales Growth Low CF right Middle CF right Diff. 1.852 2.264 0.412* (0.89) (1.02) t = 2.27 N = 58 N = 53 Middle CF right High CF right Diff. 2.264 2.025 0.239 (1.02) (1.03) t = 1.34 N = 53 N = 89 Note: This Table shows Tobin s q sorted by cash flow rights for the high industry sales growth group. Low, middle and high CF right are the lowest, the middle, and the highest cash flow right tercile respectively. Numbers in the parentheses are standard deviations, and N is the number of observations in each cell. * and ** represent the significance levels of 5% and 1%, respectively, based on a two-tailed t-test. Furthermore, the empirical result above presents a non-linear relationship between cash flow rights and firm valuation. When the cash flow right is relatively low, there is a significant effect of the cash flow right on Tobin s q. When the cash flow right is relatively high, there is no difference between different levels of cash flow rights. Consequently, there is a non-linear relationship here, as shown in Figure 2. Because the sample in this paper contains a time trend from 1987 to 2000, we test the second hypothesis with year random effect regression. In the models, we regress the Tobin s q on cash flow rights, and the empirical results are presented in Table 5. Models I and II both use observations in the high industry sales growth group. Model I contains a sample from low CF right and middle CF right groups, and model II contains a sample with middle and high CF right groups. Corresponding to the grouping in Table 4, the sample of model I comes from the first row, and the sample of model II comes from the second row in panel A of Table 5. Thus, model I contains 111 observations with relatively low cash flow rights and high industry sales growth. Model II has 142 observations with relatively high cash flow rights and high industry sales growth. As a result, the coefficient of the cash flow right in model I is 5.245, which is significant. In contrast, the coefficient of the cash flow right in model II is statistically zero. Therefore, this result also confirms our testing hypothesis. Whencash flow rights are relatively low, cash flow rights still increase the firm value even in the presence of 317

09:18 AM page:318 Academia Economic Papers 31:3 (2003) Table 5 Regressions of Tobin s q for High Industry Sales Growth Model II- middle, high CF right groups Model I- low and middle CF right groups Intercept Sales growth CF right 1.379** 2.403** (5.48) (8.23) 0.090 0.271 (0.53) (1.46) 5.245** 0.848 (2.84) ( 0.96) N 111 142 Note: Only firms with high sales growth are included in this Table. Model I contains observations in the low and middle CF right groups. Model II has observations in the middle and high CF right groups. Parameter estimations are based on regressions with the year randomeffect. Numbers in the parenthesesare t-statistics. * and ** represent the significance levels of 5% and 1%, respectively, based on a two-tailed t-test. investment opportunities. On the contrary, this pattern will disappear if cash flow rights are relatively high. Furthermore, this finding shows strong evidence on the non-linear relationship between cash flow rights and firm valuation. Though not reported here, we also run OLS regressions by model I and II, and the results are similar. 5. INDUSTRY EFFECTS ON THE TESTING HYPOTHESES In previous section, we document the impact of investment opportunity on the positive cash flow right effect. The impact will decrease the incentive of the controlling shareholder to expropriate the firm value, and reduce the effect of cash flow rights on firm value. Moreover, the magnitude of the impact also depends on the level of the cash flow right. When the cash flow right is tiny, the impact of the investment opportunity will be absent. However, we cannot rule out the possibility that the industry effect could explain why there exists a positive relationship between cash flow rights and firm value. It is very likely that industries have higher cash flow rights as well as higher firm value, thus we can observe the positive relationship. Besides, it is possible that our results are a consequence of the industry effect. The firm might have a strong relationship 318

09:18 AM page:319 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) between cash flow rights and firm value just because the industry declines, and have an insignificant relationship when the industry grows. This suggests that the impact of investment opportunities on the cash flow right effect is due to the industry effect, rather than expropriation incentives. In this section, we address these concerns by the following three tests. 11 First, we list the Tobin s q and the cash flow right for each industry and present the result in Table 6. In the table, we could not find an obvious relationship between cash flow rights and firm value. For example, the electronics industry in Taiwan has a high Tobin s q (2.297). However, the cash flow right is not high (0.182), and is even less than the mean cash flow right of the sample (0.194). Thus, the positive relationship doesn t seem to come from the industry effect. Second, we test our hypotheses by carefully controlling the industry effect. In particular, we use the excess Tobin s q as the proxy for the firm value to eliminate any potential impact from the industry level. The excess Tobin s q is defined as the firm s Tobin s q minus its industry median. Though not reported here, our results are consistent with what we show in the previous tables. For the first hypothesis, the difference of excess q betweenhighandlowcash flow right is significant for the high industry sales growth group, but the difference is insignificant for the low growth group. This suggests that the cash flow right increases the excess q in the absence of investment opportunities, but the positive relation disappears with the appearance of investment opportunities. Moreover, the empirical results are also consistent with our second hypothesis, given investment opportunities. The difference of excess q between high and middle cash flow right groups is significantly different from zero, but insignificant between middle and high cash flow right groups. As a result, the cash flow right effect may still exist depending on the level of cash flow rights by the proxy of excess Tobin s q. Finally, we use industry fixed and random effect regressions to check our results. Even controlling the industry fixed and random effects, we still find strong evidence supporting our testing hypotheses. Again, we don t report the table here to save space. Overall, we don t find any evidence that our results are due to the industry effect. This shows that investment opportunities influence the firm s decision on expropriation. So, there exists no difference between the levels of cash flow rights with the presence of investment opportunities,whereas there is significant difference without 11 We appreciate that one referee kindly points out this concern. 319

09:18 AM page:320 Academia Economic Papers 31:3 (2003) Industry Table 6 Industry Tobin s q and the Cash Flow Right Industry code N Tobin s q CF right Mean S.D. Mean S.D. Cement 11 32 1.968 1.291 0.126 0.078 Food 12 89 1.590 0.529 0.161 0.066 Plastics 13 115 2.122 0.884 0.283 0.137 Textiles 14 176 1.695 0.778 0.173 0.084 Electric, Machinery 15 14 1.234 0.393 0.205 0.109 Appliance, Cable 16 23 1.442 0.493 0.184 0.073 Chemical 17 37 1.949 0.705 0.163 0.073 Glass, Ceramics 18 24 2.044 0.894 0.194 0.091 Paper, Pulp 19 44 2.091 1.134 0.229 0.134 Steel, Iron 20 50 1.327 0.399 0.187 0.097 Rubber 21 7 1.553 0.520 0.127 0.043 Automobile 22 21 1.692 0.352 0.124 0.031 Electronic 23 73 2.297 0.995 0.182 0.090 Construction 25 51 1.352 0.567 0.172 0.060 Transportation 26 39 1.551 0.461 0.227 0.136 Tourism 27 29 2.760 1.422 0.227 0.104 Department Stores 29 52 2.187 1.183 0.180 0.140 Other 99 37 2.000 0.987 0.230 0.112 Average 50.72 1.825 0.187 Note: This Table shows the Tobin s q and the cash flow right for each industry. Industry classification codes are from Taiwan Stock Exchange. Correlation coefficient between q and CF right for industry level is 0.315. the investment opportunity. Besides, the impact coming from the investment opportunity will vanish when the cash flow right is relatively small, suggesting a non-linear relationship between cash flow rights and firm value here. 6. ROBUSTNESS OF THE RESULTS In this section, we discuss four issues of robustness. First of all, we use industry capital expenditure and R&D expenditure ratio of a firm as investment opportunities. 12 12 We acknowledge that one referee suggests we employ the capital expenditure for the measurement 320

09:18 AM page:321 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) Second, we employ panel data regressions in empirical analyses. Third, we divide the sample into two different groups by the median of cash flow rights, and analyze them by regression to robustly check. Fourth, we try to explain what can be done about the endogeneity of ownership. The reason for treating industry capital expenditure as the investment opportunity is as follows. A firm with more positive NPV projects in which to invest will buy more plants, machines and other equipment. Thus, increases in fixed assets can serve as proxy for the investment opportunity in the future. As with the rule of industry sales growth, we use the 3 rd quartile for industry capital expenditure as the critical point of the investment opportunity. Industry capital expenditure is defined as the change in total industry fixed assets to the total industry assets. The empirical results are presented in Table 7. In panel A and B, when the industry capital expenditure is high, the difference between high and low cash flow rights is insignificant. In contrast, there exists reasonable difference between high and low cash flow right if the industry capital expenditure is low. Besides, in panel C and D, there is no significant difference when cash flow rights are relatively high, but a significant difference when cash flow rights are relatively low with high industry capital expenditure. All these results support our main arguments. In addition, we also use the R&D expenditure as the proxy of the investment opportunity. Although the R&D expenditure ratio is endogenous to the firm, it still presents some kinds of investment opportunities. Firms who have high R&D and advertising expenditures hold more profitable projects, and have more chances to increase firm valuation. Thus, we also employ the R&D expenditure and advertising expenditure ratio as the proxy of the investment opportunity. Dividing data into the high and low R&D expenditure ratio by 3 rd quartile, we show the empirical results in Table 8. Panels A and B in Tables 7 and 8 are to test the first hypothesis (H1), and the panels C and D test the second hypothesis (H2). The empirical results are consistent with the findings of using industry grouping, and accept our two hypotheses. Since we have the sample from 1987 to 2000, we use the panel regression to check empirical results. When firms have low industry sales growths, the panel regressions show that the coefficient of the cash flow right is significant at 1% level. In contrast, the coefficient is not significant when firms have high industry sales growths. In addition, for firms in the high industry sales growth group, the coefficient of the cash flow right is significant when the cash flow right is relatively low. However, the coefficient is of investment opportunity. 321

09:18 AM page:322 Academia Economic Papers 31:3 (2003) Table 7 Robustness Check of Using Capital Expenditure As Investment Opportunities Panel A: Tobin s q by capital expenditure and CF right Industry capital expenditure High CF right Low CF right Diff. 2.196 1.959 0.237 High (0.92) (0.82) t = 1.64 N = 80 N = 69 1.894 1.740 0.154* Low (1.07) (0.80) t = 2.25 N = 382 N = 387 Intercept Sales growth CF right Panel B: Regressions of Tobin s q Model I- low capital expenditure group Model II- high capital expenditure group 1.848** (8.93) 1.861** (9.56) 0.081 (0.63) 0.366 (1.32) 0.814** (2.98) 0.873 (1.55) Panel C: Tobin s q for high industry capital expenditure Low CF right Middle CF right Diff. 1.746 2.307 0.561** (0.58) (1.01) t = 3.25 N = 42 N = 47 Middle CF right High CF right Diff. 2.307 2.152 0.154 (1.01) (0.90) t = 0.84 N = 47 N = 60 Panel D: Regressions of Tobin s q for high industry capital expenditure groups Model III- Model IVlow, middle CF right group middle, high CF right group 1.312** 2.429** Intercept (5.06) (8.72) 0.131 0.322 Sales growth (0.41) (0.96) 6.12** 0.862 CF right (3.71) ( 1.02) Note: This Table presents the robustness checks by using industry capital expenditure as investment opportunities. The high and low industry capital expenditures are divided by the 3 rd quartile. The finding of panel A is similar to the result in Table 2. Panel B lists the regression results like Table 3. Panel C is similar to Table 4. Panel D is the regression results corresponding to panel C. Models I and II in panel B contain 769 and 149 observations respectively. The sample in panel D comes from the group with high industry capital expenditure. Model III in panel D has 89 observations in low and middle CF right groups and, Model IV in panel D has 107 observations in middle and high CF right groups. Regression results are estimated by year random effect. Numbers in the parentheses are t-statistics. * and ** represent the significance levels of 5% and 1%, respectively, based on a two-tailed t-test. 322

09:18 AM page:323 When Will the Controlling Shareholder Expropriate Investors? (Chan, Hu, and Wang) Table 8 Robustness Check of Using R&D Expenditure Ratios As Investment Opportunities Panel A: Tobin s q by R&D expenditure and CF right R&D expenditure High CF right Low CF right Diff. 2.002 1.928 0.074 High (1.17) (0.84) t = 0.55 N = 113 N = 117 1.929 1.718 0.211* Low (1.01) (0.79) t = 3.06 N = 346 N = 342 Intercept Sales growth CF right Panel B: Regressions of Tobin s q Model I- Model IIlow R&D expenditure ratio group high R&D expenditure ratio group 1.761** 1.933** (9.35) (7.81) 0.176 0.457 (1.49) (1.10) 0.972** 0.874 (3.74) (1.10) Panel C: Tobin s q for high R&D group Low CF right Middle CF right Diff. 1.871 2.035 0.164** (0.74) (1.01) t = 1.19 N = 74 N = 88 Middle CF right High CF right Diff. 2.035 1.976 0.059 (1.01) (1.26) t = 0.32 N = 88 N = 68 Panel D: Regressions of Tobin s q for high R&D groups Model III- Model IVlow, middle CF right group middle, high CF right group Intercept 1.521** (6.41) 2.035** (6.59) Sales growth 0.444 (1.07) 0.843 (1.58) CF right 3.086* (2.17) 0.199 (0.19) Note: This Table presents the robustness checks by using R&D expenditure ratios as investment opportunities. The high and low R&D expenditure are divided by the 3 rd quartile. The finding of panel A is similar to the result in Table 2. Panel B lists the regression results like Table 3. Panel C is similar to Table 4. Panel D is the regression results corresponding to panel C. The sample in panel D comes from the group with high R&D expenditure. Model III in panel D contains 162 observations in low and middle CF right groups, and Model IV in panel D has 156 observations in middle and high CF right groups. Regression results are estimated by year random effect. Numbers in the parentheses are t-statistics. * and ** represent the significance levels of 5% and 1%, respectively, based on a two-tailed t-test. 323