SEARCH

SEARCH BY CITATION

Keywords:

  • G32;
  • G34

Abstract

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Data and Variables
  5. III. Empirical Evidence
  6. IV. Conclusion
  7. Appendix
  8. References

We examine the governance role of multiple large shareholder structures (MLSS) to determine their valuation effects in a sample of 1,252 publicly traded firms from nine East Asian economies. We find that the presence, number, and size of multiple large shareholders are associated with a significant valuation premium. Our results also show that the identity of MLSS influences corporate value and that the valuation effects of MLSS are more pronounced in firms with greater agency costs. Our results imply that MLSS play a valuable monitoring role in curbing the diversion of corporate resources.


I. Introduction

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Data and Variables
  5. III. Empirical Evidence
  6. IV. Conclusion
  7. Appendix
  8. References

Corporate governance research, starting with La Porta, Lopez-de-Silanes, and Shleifer's (1999) seminal study, shows that ownership in public firms outside the United States and the United Kingdom is concentrated in the hands of very few major shareholders. These are typically members of a wealthy family who tend to use controlling devices, such as top-down chain of control pyramids and multiple class shares, to secure control rights in excess of their ownership rights. This separation of ownership and control enables controlling shareholders to extract private benefits at the expense of minority shareholders (Johnson et al. 2000; Volpin 2002). Although a growing body of research examines the implications of excess control by the largest shareholder (e.g., Faccio, Lang, and Young 2001; Claessens et al. 2002; La Porta et al. 2002; Attig et al. 2006; El Ghoul et al. 2009), scant attention has been given to the governance role of multiple large shareholder structures (MLSS), despite their pervasiveness around the world. Recent studies profiling corporate ownership structures reveal that a significant number of firms are controlled through MLSS. For instance, Claessens, Djankov, and Lang (2000) and Faccio and Lang (2002) document that MLSS exist in 32.2% of East Asian firms and 45.26% of Western European firms. We investigate the governance role of MLSS by examining its effect on corporate value in East Asia.

Two theoretical views on the benefits and costs of MLSS motivate us to investigate their potential governance role. One view is that MLSS may play a valuable monitoring role in curbing the extraction of private benefits. Bennedsen and Wolfenzon (2000) suggest that control structures with multiple large shareholders may be efficient when equity holdings are evenly distributed among the controlling shareholders. Similarly, Bloch and Hege (2001) show that two large shareholders refrain from extracting private benefits because they compete for corporate control by attracting minority shareholders. The alternative viewpoint casts doubt on the effectiveness of shared control in producing better corporate governance. Zwiebel's (1995) model suggests that moderate-sized blockholders are prone to be in cahoots with each other to appropriate divisible private benefits. Kahn and Winton (1998) identify occurrences where large shareholders prefer to opportunistically trade on private information rather than monitor management. Another strand of literature considers both views of the role of MLSS in corporate governance. For instance, Gomes and Novaes (2005) develop a theoretical decision model that suggests that concentrated control in the hands of one large investor provides better protection to minority shareholders than shared control among blockholders. However, their model also shows that sharing control increases efficiency in less protective economies (e.g., weak legal system, less transparent firms).

The preceding discussion suggests that, theoretically, MLSS can engage in efficient monitoring or in the extraction of divisible private benefits of control. We investigate these two competing MLSS governance hypotheses to determine which one dominates and in which situations. To test these hypotheses we relate corporate valuation to the presence and attributes of MLSS in a sample of 1,252 publicly traded firms from nine East Asian economies.

Our work contributes to the literature in at least two ways. First, we augment empirical research on the governance impact of MLSS by examining the effect of the largest shareholder's control contestability in East Asian economies. Laeven and Levine (2008) address the effects of MLSS in Western European countries and Maury and Pajuste (2005) in Finland. The weak legal environment and the different ultimate ownership patterns in East Asia, compared to other regions examined, provide an opportune setting in which to study the effects of MLSS on agency costs (Faccio and Lang 2002). Our research also relates to other studies on agency problems embedded in ultimate ownership structures of East Asian firms. Claessens et al. (2002) document significant firm value discounts associated with higher excess control rights by the largest shareholder, whereas Fan and Wong's (2005) evidence implies that high-quality auditing plays a corporate governance role by reducing the extent of these discounts. We augment these studies by showing that MLSS also play a valuable monitoring role in curbing the diversion of corporate resources in East Asia. Second, we investigate the situations under which MLSS may play a more efficient governance role. Because most of the recent empirical evidence suggests that agency problems around the globe lie in the misalignment between the control and ownership stakes of ultimate owners, we assess the effects of MLSS in shaping the outcome of the largest shareholder's excess control. We also examine whether the presence of excess free cash flows affects the governance role of MLSS.

Our empirical analysis yields several results consistent with the efficient-monitoring hypothesis of MLSS in East Asia. We find that the presence and number of multiple large shareholders, other than the largest shareholder, are associated with a significant valuation premium. After controlling for firm-level characteristics and industry effects, we estimate that MLSS firms trade at a 9.44% average premium over firms with a single large shareholder. Furthermore, our results suggest that adding one large shareholder to the firm's ownership structure enhances corporate valuation by 4.96% on average. We also find that higher contestability of the largest shareholder's voting power by the second-largest shareholder increases firm value.

Three other important findings emerge from our analysis. First, our measures of control dispersion provide empirical support to Bennedsen and Wolfenzon's (2000) theoretical predictions and stress the relevance of control contestability in determining the governance role of MLSS. We find that shareholders of similar size (i.e., low dispersion) are more efficient in their governance role. Second, we document a more pronounced valuation effect of MLSS when the second-largest shareholder is either a family or the state, which shows that blockholder identity plays a role in shaping MLSS valuation effects. Third, we find that the valuation effects of MLSS are more pronounced in firms with severe agency problems, particularly those driven by a firm's free cash flows. This evidence implies that MLSS play a stronger governance role in firms where the likelihood of corporate diversion is greater.

II. Data and Variables

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Data and Variables
  5. III. Empirical Evidence
  6. IV. Conclusion
  7. Appendix
  8. References

Sample

The main source of ownership data is Claessens, Djankov, and Lang (2000), covering firms from nine East Asian countries: Hong Kong, Indonesia, Japan, Korea, Malaysia, Philippines, Singapore, Taiwan, and Thailand. The data set identifies the ultimate controlling shareholders of 2,980 East Asian firms as well as their ultimate cash flow (ownership) and voting rights (control) as of December 1996 or the end of the 1996 fiscal year. The data set also includes information on the presence of multiple large shareholders as well as their control stakes, which enables us to test their role in corporate governance. We exclude firms that do not have a controlling shareholder who owns more than 10% of the voting rights from the analysis. To obtain industry affiliations and financial data related to firm-level variables in 1996, we hand-match the ownership data set with the 1997 Worldscope database. We drop firms with insufficient financial data to measure corporate valuation and other firm-level variables. Following previous research, we also eliminate financial firms (Standard Industrial Classification [SIC] codes between 60 and 69) from our analysis. These filters yield data on ownership structure, corporate valuation, industry affiliation, and firm-level characteristics for 1,252 East Asian firms in 1996.

Variables

Firm Valuation Following prior research (e.g., Morck, Shleifer, and Vishny 1988; La Porta et al. 2002; Claessens et al. 2002; Lins 2003), we use Tobin's Q as our proxy for firm valuation (TOBQ). We define TOBQ as the ratio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equity.

Ultimate Ownership and Control of the Largest Shareholder We follow Claessens et al. (2002) to gauge the incentive and entrenchment effects of the largest shareholder on firm value. We use the ultimate share of ownership rights to capture the degree of incentive alignment with minority shareholders; we label this variable CASH1. To capture the entrenchment effect, we construct the continuous variable CONTMCASH, which equals the largest ultimate owner's control rights in excess of ownership rights.

MLSS We employ variables reflecting various attributes of MLSS. We start by identifying whether the firm has more than one large shareholder by constructing a dummy variable (MOWNERS) set to 1 if at least one large shareholder other than the largest shareholder controls more than 10% of the firm, and 0 otherwise. Based on the efficient-monitoring (entrenchment) hypothesis, we expect this variable to have a positive (negative) effect on corporate value.

Our second variable measures the number of blockholders other than the largest, up to the fifth, that control more than 10% of the firm (NOWNERS). This construct allows us to separate the coalition formation effects (Bennedsen and Wolfenzon 2000) and the bargaining effects (Gomes and Novaes 2005) of MLSS. The coalition-formation hypothesis predicts that, all else equal, the larger the number of shareholders, the greater the likelihood that the winning coalition will hold a small equity stake, thereby externalizing the outcome of its actions and resulting in a decline in firm value. In this case, a negative relation between NOWNERS and firm value is expected.1 Alternatively, the bargaining effects hypothesis suggests that disagreement among a large number of shareholders implies that projects diluting minority shareholders' interests will be rejected, which in turn translates into value premium. Consequently, a positive relation between NOWNERS and firm value is expected. Because it is unclear which hypothesis, a priori, should dominate, the relation between corporate value and the number of large shareholders is an empirical question.

In a second step, we refine our analysis to investigate the importance of control contestability and dispersion in shaping the governance role of MLSS. According to the efficient-monitoring hypothesis, greater control contestability enhances other large shareholders' monitoring incentives.2 However, under the entrenchment hypothesis (Zwiebel 1995), large shareholders extract private benefits that are proportional to their control stakes; therefore, greater contestability is associated with higher dilution of minority shareholders' interests. To empirically examine these two competing arguments we consider the voting rights of the second-largest shareholder (CONT2). We then measure the second-largest shareholder's relative power vis-à-vis the largest shareholder using the ratio of the voting rights of the second-largest shareholder to the voting rights of the largest shareholder (VOTING21). Similarly, we consider a dispersion ratio (DISPERSION1), defined as the difference in the control stakes of the largest and second-largest shareholders over their sum (CONT1 − CONT2)/(CONT1 +−CONT2). The higher this ratio, the lower is the contestability of the control of the largest shareholder by another shareholder. We also consider a second proxy for MLSS, control dispersion (DISPERSION2), measured as the Herfindahl index of the differences between the voting rights of the five largest shareholders (CONT1 − CONT2)2+ (CONT2 − CONT3)2+ (CONT3 − CONT4)2+ (CONT4 − CONT5)2. All else equal, higher DISPERSION2 implies lower contestability of the control of the largest shareholder by the coalition of large shareholders. Based on the efficient-monitoring (entrenchment) hypothesis, we expect a negative (positive) effect of variables measuring the dispersion of voting rights (i.e., DISPERSION1,DISPERSION2) on corporate value.

Control Variables Our selection and specification of control variables closely follow recent international corporate governance research. We proxy for firm size with the natural logarithm of total assets in U.S. dollars (SIZE), we measure leverage with the ratio of long-term debt to total assets (LEVERAGE), we capture investment with the ratio of capital expenditure to total assets (CAPEX), and we control for growth opportunities using lagged sales growth (SALESGR). Claessens et al. (2002) contend that small firms are less diversified in East Asia, leading to lower diversification discounts. Thus, we expect SIZE to bear negatively with TOBQ. Faccio, Lang, and Young (2005) show that controlling shareholders in East Asia use debt financing, mainly obtained from related parties, to acquire more resources to expropriate. Consequently, we expect LEVERAGE to be negatively related to TOBQ in our sample. Finally, La Porta et al. (2002) argue that firms with better growth opportunities should exhibit higher performance. Therefore, we expect SALESGR and CAPEX to be positively associated with TOBQ. To limit the influence of outliers, we winsorize all continuous variables at the 1st and 99th percentiles.

Descriptive Statistics

Table 1 reports the number of observations and provides descriptive statistics by country for all variables used in the analyses. The Appendix summarizes the definitions and data sources for all variables. There is wide variation in the number of firms in each country. Japan is the most representative, totaling 477 firms, followed by Korea and Hong Kong, which account for 155 and 133 firms, respectively. The Philippines is the least representative with only 54 firms. Consistent with Claessens et al. (2002), we find that Malaysian firms exhibit the highest performance with a mean (median) TOBQ of 2.18 (1.90). In contrast, we find that Thai firms have the lowest performance with a mean (median) TOBQ of 1.23 (1.00). This result is likely driven by the fact that we exclude widely held firms from our sample whereas Claessens et al. do not.

Table 1. Descriptive Statistics by Country.
Variable Hong KongIndonesiaJapanKoreaMalaysiaPhilippinesSingaporeTaiwanThailandTotal
  1. Note: This table reports descriptive statistics on Tobin's Q, multiple large shareholder structures, and control variables for 1,252 nonfinancial firms from nine East Asian countries in 1996. The variables are: TOBQ, ratio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equity; MOWNERS, dummy variable set to 1 if at least one large shareholder other than the largest shareholder controls more than 10% of the firm, and 0 otherwise; NOWNERS, number of other large shareholders (up to the fifth) controlling more than 10% of the firm; CONT2, ultimate voting rights of the second-largest shareholder; VOTING21, ratio of voting rights of the second-largest shareholder to voting rights of the largest shareholder; DISPERSION1, difference between the largest and the second-largest shareholders' voting rights to their sum; DISPERSION2, Herfindahl index of the differences between the voting rights of the five largest shareholders; CASH1, ultimate cash flow rights of the largest shareholder; CONTMCASH, ultimate cash flow rights minus ultimate voting rights of the largest shareholder; SIZE, natural logarithm of total assets in millions of U.S. dollars; SALESGR, growth rate in sales over the previous year; LEVERAGE, ratio of long-term debt to total assets; and CAPEX, ratio of capital expenditures to total assets.

TOBQMean1.3901.3421.4321.0142.1831.6511.5661.8371.2301.475
Median1.0751.0241.2960.9511.8891.2491.3011.6541.0001.263
MOWNERSMean0.2860.5440.0960.1800.5820.6300.6330.4810.8860.332
Median0.0001.0000.0000.0001.0001.0001.0000.0001.0000.000
NOWNERSMean0.3530.6470.1070.1800.8090.6850.8260.6671.5860.439
Median0.0001.0000.0000.0001.0001.0001.0000.0002.0000.000
CONT2Mean3.7597.6471.0492.2667.5458.0008.7165.30914.7214.419
Median0.00010.0000.0000.00010.00010.00010.0000.00020.0000.000
VOTING21Mean0.1310.2730.0770.1090.2790.3270.3490.2730.4210.181
Median0.0000.2820.0000.0000.2500.3130.3230.0000.4300.000
DISPERSION1Mean0.8270.6490.9160.8680.6380.5870.5680.6620.4550.775
Median1.0000.5601.0001.0000.6000.5240.5121.0000.3981.000
DISPERSION2Mean0.1050.1330.0290.0550.0910.0740.0850.0590.1020.064
Median0.0780.0860.0120.0480.0630.0490.0540.0360.0670.040
CASH1Mean28.03028.72110.20619.50726.83625.01923.56919.87736.60319.584
Median26.00026.0008.00018.00024.00024.00022.00020.00036.00018.000
CONTMCASHMean4.3468.9855.1193.0335.1453.3897.4774.8272.3514.956
Median0.0007.0006.0000.0000.0000.0008.0003.0000.0000.000
SIZEMean12.72612.15113.34213.51912.55312.23512.28112.89812.44312.945
Median12.52611.98813.07813.48812.57712.17912.24912.81212.38312.784
SALESGRMean0.1110.1360.0450.1410.2900.2500.1330.0290.1650.112
Median0.0910.1200.0250.1240.1430.1950.0770.0160.0950.058
LEVERAGEMean0.1040.1890.1350.1990.1010.1170.1110.1190.1940.139
Median0.0870.1500.1060.1930.0570.0860.0580.1040.1640.113
CAPEXMean0.0610.0930.0280.0840.0780.1260.0750.0590.0810.060
Median0.0440.0690.0170.0650.0620.1150.0540.0370.0480.040
N  133 68 477 150 110 54 109 81 70  1,252

MLSS are present in approximately one-third of the firms in our East Asian sample. This is smaller than the proportion of 46% reported in Western European firms (Faccio and Lang 2002). This figure also shows considerable cross-country variation in our sample. MLSS are most common in Thailand (88.6%) and Singapore (63.3%), and least frequent in Japan (9.6%). The variables measuring the power of other large shareholders, their number, and the dispersion of voting rights among them are generally consistent with this sorting.

The largest shareholder owns 19.58% of the cash flow rights, on average, and voting rights exceed cash flow rights by an average of 4.96%. However, the patterns of ownership and excess control show systematic differences across the East Asian countries. Firms from Thailand and Hong Kong exhibit the most concentrated cash flow rights at 36.60% and 28.03%, respectively, and Japanese firms exhibit the least concentrated cash flow rights at 10.21%. For excess control, firms from Indonesia and Singapore lead the East Asian countries with an average separation of 8.99% and 7.48%, respectively, and Thai firms show the lowest average separation of 2.35%. Finally, descriptive statistics on firm-level characteristics show that Korean and Japanese firms are the largest, those in Malaysia exhibit the highest sales growth, firms in Korea are the most leveraged, and Indonesian firms invest the greatest amounts.

III. Empirical Evidence

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Data and Variables
  5. III. Empirical Evidence
  6. IV. Conclusion
  7. Appendix
  8. References

Univariate Tests

We present initial insights on the relation between firm valuation and MLSS from correlations and univariate tests. Table 2 reports Pearson's correlation coefficients for all regression variables. The results indicate that TOBQ is significantly and positively correlated with MOWNERS,NOWNERS,CONT2, and VOTING21, and significantly and negatively correlated with DISPERSION1. These pairwise correlations, all significant at the 1% level, lend preliminary support to the efficient-monitoring hypothesis of MLSS. The pairwise correlation coefficients among the other explanatory variables are generally low (consistently below 40%), providing some assurance that multicollinearity is not affecting our multivariate results.

Table 2. Correlation Between the Regression Variables.
 TOBQ1234567891011
  1. Note: This table reports Pearson correlations between all regression variables for a sample of 1,252 nonfinancial firms from nine East Asian countries in 1996. The variables are: TOBQ, ratio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equity; MOWNERS, dummy variable set to 1 if at least one large shareholder other than the largest shareholder controls more than 10% of the firm, and 0 otherwise; NOWNERS, number of other large shareholders (up to the fifth) controlling more than 10% of the firm; CONT2, ultimate voting rights of the second-largest shareholder; VOTING21, ratio of voting rights of the second-largest shareholder to voting rights of the largest shareholder; DISPERSION1, difference between the largest and the second-largest shareholders' voting rights to their sum; DISPERSION2, Herfindahl index of the differences between the voting rights of the five largest shareholders; CASH1, ultimate cash flow rights of the largest shareholder; CONTMCASH, ultimate cash flow rights minus ultimate voting rights of the largest shareholder; SIZE, natural logarithm of total assets in millions of U.S. dollars; SALESGR, growth rate in sales over the previous year; LEVERAGE, ratio of long-term debt to total assets; and CAPEX, ratio of capital expenditures to total assets. The p-values are in parentheses. Ownership data are from Claessens, Djankov, and Lang (2000).

1. MOWNERS0.129           
(.000)           
2. NOWNERS0.1110.885          
(.000)(.000)          
3. CONT20.1270.9010.818         
(.000)(.000)(.000)         
4. VOTING210.1100.8820.7990.885        
(.000)(.000)(.000)(.000)        
5. DISPERSION1−0.119−0.942−0.847−0.916−0.988       
(.000)(.000)(.000)(.000)(.000)       
6. DISPERSION20.034−0.140−0.139−0.111−0.2180.207      
(.236)(.000)(.000)(.000)(.000)(.000)      
7. CASH10.0870.1830.1670.2100.022−0.0710.774     
(.002)(.000)(.000)(.000)(.428)(.012)(.000)     
8. CONTMCASH−0.037−0.056−0.064−0.044−0.0630.0660.163−0.362    
(.189)(.049)(.023)(.123)(.025)(.020)(.000)(.000)    
9. SIZE−0.169−0.195−0.201−0.205−0.1660.179−0.155−0.224−0.008   
(.000)(.000)(.000)(.000)(.000)(.000)(.000)(.000)(.775)   
10. SALESGR0.0750.0870.0830.1040.077−0.0820.0730.109−0.0140.017  
(.008)(.002)(.003)(.000)(.006)(.004)(.010)(.000)(.622)(.553)  
11. LEVERAGE−0.1440.0050.0100.0060.002−0.007−0.063−0.039−0.0380.3390.068 
(.000)(.858)(.715)(.835)(.932)(.819)(.026)(.167)(.183)(.000)(.017) 
12. CAPEX0.0330.1730.1230.1630.119−0.1400.1650.224−0.0300.0250.1510.200
(.244)(.000)(.000)(.000)(.000)(.000)(.000)(.000)(.294)(.369)(.000)(.000)

We shed further light on the relation between TOBQ and MLSS using graphical evidence. In Figure I, we plot mean and median TOBQ for firms with MLSS and those with a single large shareholder. Consistent with the correlation analysis, firms with MLSS are worth more: the average MLSS premium is 15.1%. This result does not seem to be driven by outliers, as we obtain the same pattern for median values, although the difference is more conservative at 6.4%. In unreported results, we also find that the MLSS premium is positive in seven out of nine countries in our sample.

image

Figure I. The Univariate Relation between Tobin's Q and the Presence of Multiple Owners.

Download figure to PowerPoint

Table 3 compares firm valuation and other firm characteristics after bisecting the sample according to the presence or absence of MLSS. The table shows significant differences between the two subsamples of firms, and the MLSS premium is statistically significant. The largest shareholder achieves higher cash flow rights and a lower separation between ownership and control in firms with MLSS. Also, firms with MLSS are significantly smaller, exhibit greater growth, and invest more.

Table 3. Univariate Tests.
Variable MOWNERS = 1 (A)MOWNERS = 0 (B)Difference (A) – (B)t-test (Wilcoxon test)
  1. Note: This table presents mean and median difference tests of Tobin's Q and control variables for 1,252 nonfinancial firms from nine East Asian countries in 1996. The variables are: TOBQ, ratio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equity; CASH1, ultimate cash flow rights of the largest shareholder; CONTMCASH, ultimate cash flow rights minus ultimate voting rights of the largest shareholder; SIZE, natural logarithm of total assets in millions of U.S. dollars; SALESGR, growth rate in sales over the previous year; LEVERAGE, ratio of long-term debt to total assets; and CAPEX, ratio of capital expenditures to total assets. Ownership data are from Claessens, Djankov, and Lang (2000).

  2. ***Significant at the 1% level.

  3. **Significant at the 5% level.

TOBQMean1.6161.4040.2124.59**
Median1.3211.2420.079(2.44**)
CASH1Mean23.09017.8405.2506.57***
Median22.00015.0007.000(8.05***)
CONTMCASHMean4.4345.216−0.782−1.97**
Median0.0003.000−3.000(−2.98***)
SIZEMean12.55113.141−0.589−7.01***
Median12.55012.951−0.401(−6.28***)
SALESGRMean0.1480.0940.0543.10***
Median0.0760.0510.025(2.15**)
LEVERAGEMean0.1400.1380.0010.18
Median0.1080.116−0.008(−0.87)
CAPEXMean0.0760.0520.0246.20***
Median0.0530.0350.018(5.79***)
N 416836  

Regression Analysis

Although the correlation and univariate analyses offer preliminary evidence supporting the efficient-monitoring hypothesis, we perform a multivariate analysis to more rigorously examine the governance role of MLSS. Table 4 reports our main results from regressing firm value on MLSS proxies while controlling for the effects of other factors. Although most empirical governance studies use Tobin's Q as a proxy for firm value, we recognize that Tobin's Q may also measure growth opportunities. To further reduce bias from potential cross-sectional heterogeneity in growth opportunities, we control for industry fixed effects in all regressions, based on Campbell's (1996) industry classification.3

Table 4. The Effects of Multiple Large Shareholder Structures on Firm Valuation: Main Evidence.
Panel A. Presence and Number of Multiple Shareholders
 
Variable (pred. sign)MOWNERS (+)NOWNERS (+)CONT2 (+)CASH1 (×100) (+)CONTMCASH (−)SIZE (−)SALESGR (+)LEVERAGE (−)CAPEX (+)Intercept (?)NR2
  1. Note: This table presents regressions of firm value on multiple large shareholder structures and control variables for 1,252 nonfinancial firms from nine East Asian countries in 1996. The dependent variable is TOBQ, ratio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equity. The independent variables are: MOWNERS, dummy variable set to 1 if at least one large shareholder other than the largest shareholder controls more than 10% of the firm, and 0 otherwise; NOWNERS, number of other large shareholders (up to the fifth) controlling more than 10% of the firm; CONT2, ultimate voting rights of the second-largest shareholder; VOTING21, ratio of voting rights of the second-largest shareholder to voting rights of the largest shareholder; DISPERSION1, difference between the largest and the second-largest shareholders' voting rights to their sum; DISPERSION2, Herfindahl index of the differences between the voting rights of the five largest shareholders; CASH1, ultimate cash flow rights of the largest shareholder; CONTMCASH, ultimate cash flow rights minus ultimate voting rights of the largest shareholder; SIZE, natural logarithm of total assets in millions of U.S. dollars; SALESGR, growth rate in sales over the previous year; LEVERAGE, ratio of long-term debt to total assets; and CAPEX, ratio of capital expenditures to total assets. All regressions control for industry effects (not reported). Ownership data are from Claessens, Djankov, and Lang (2000). Robust t-statistics are in parentheses.

  2. ***Significant at the 1% level (one-tailed when directional predictions are made, and two-tailed otherwise).

  3. **Significant at the 5% level (one-tailed when directional predictions are made, and two-tailed otherwise).

(1)0.135***  0.010−0.005−0.073***0.172**−0.705***0.1002.404***1,2520.090
(2.642)  (0.053)(−1.156)(−3.437)(2.135)(−3.683)(0.271)(7.999)  
(2) 0.072** 0.021−0.005−0.075***0.173**−0.712***0.1652.432***1,2520.087
 (1.891) (0.116)(−1.150)(−3.517)(2.160)(−3.685)(0.446)(8.106)  
(3)  0.009**−0.008−0.005−0.073***0.168**−0.711***0.1212.414***1,2520.089
  (2.191)(−0.045)(−1.222)(−3.500)(2.089)(−3.704)(0.326)(8.165)  
 
Panel B. Control Contestability
 
Variable (pred. sign)VOTING21 (+)DISPERSION1 (−)DISPERSION2 (−)CASH1 (×100) (+)CONTMCASH (−)SIZE (−)SALESGR (+)LEVERAGE (−)CAPEX (+)Intercept (?)NR2
 
(1)0.188**  0.092−0.004−0.074***0.171**−0.701***0.1282.405***1,2520.088
(2.249)  (0.516)(−1.006)(−3.494)(2.119)(−3.669)(0.343)(8.011)  
(2) −0.181*** 0.160 −0.071***0.168**−0.692***0.0842.497***1,2520.088
 (−2.565) (0.962) (−3.384)(2.082)(−3.591)(0.226)(8.482)  
(3)  −1.160**0.616** −0.079***0.177**−0.709***0.1892.451***1,2520.086
  (−2.004)(2.184) (−3.691)(2.212)(−3.681)(0.520)(8.447)  

Main Evidence on the Role of MLSS

Reflecting the various proxies capturing the effects of MLSS, several specifications are reported in Table 4.4 Following the theoretical predictions on the effects of MLSS previously outlined, and for expositional convenience, we group our core results into two panels. In Panel A, we report the valuation effects of MLSS variables that reflect the existence of (at least) a second large shareholder (MOWNERS), the number of blockholders beyond the largest shareholder (NOWNERS), and the control size of the second-largest shareholder (CONT2). In Panel B, we report the valuation effects of MLSS variables, measured in relative terms, to capture control contestability. We examine the effects of the ratio of the voting rights of the second-largest shareholder to the voting rights of the first-largest shareholder (VOTING21), the ratio of the dispersion in control stakes between the largest and second-largest shareholders (DISPERSION1), and the dispersion of the voting rights of the five largest shareholders (DISPERSION2).

Presence of Multiple Shareholders and Firm Valuation The estimated coefficient on the presence of multiple large shareholders, MOWNERS (specification (1) in Panel A of Table 4), is positive and statistically significant, suggesting that the mere presence of multiple large shareholders (other than the very largest controlling shareholder) is associated with higher firm valuation, consistent with the univariate tests. This suggests that when all other variables are set to their mean values, MLSS show a 9.44% value premium over firms with a single large shareholder. The results of specification (2) suggest that increasing the number of large shareholders improves corporate governance and thus enhances corporate value in East Asian firms. The estimated coefficient of the number of blockholders (NOWNERS) implies that adding a second large shareholder to a firm with a single large shareholder yields a 4.96% increase in firm valuation.5 This result is consistent with the bargaining effect of MLSS: the presence of a large number of controlling shareholders reduces the diversion of corporate resources because corporate decisions, especially those involving private benefits, require mutual consent among large shareholders. Specification (3) indicates that the control rights of the second-largest shareholder (CONT2) are positively and significantly associated with firm value (at the 1% level). In sum, the evidence in Panel A supports the premise that the presence of large shareholders, beyond the largest, is likely to play a decisive role in curbing the extraction of private benefits by the controlling shareholder in East Asian corporations.

Control Contestability and Firm Valuation The control contestability variables reported in Panel B of Table 4, specification (1) suggest that VOTING21, the proxy for the size of the second-largest shareholder relative to the controlling owner, is significantly and positively associated with higher firm value. Reinforcing this finding, the estimated coefficient on DISPERSION1, displayed in specification (2), is negative and statistically significant at the 1% level. Reflecting its first-order economic importance, this coefficient estimate implies that a 1 standard deviation (0.338) increase in DISPERSION1 translates into a 6.12% (0.181 × 0.338 = 6.12%) decrease in Tobin's Q, representing 4.84% of the median Tobin's Q. This proxy tests whether unequal distribution of voting rights reduces the incentives and monitoring power of the second-largest shareholder. That is, it captures the control contestability of the largest owner by the second-largest shareholder, with higher values implying lower contestability. Accordingly, the significant, negative coefficient on DISPERSION1 indicates that unequal distribution of voting rights between the largest and second-largest shareholders compromises the monitoring power of the latter, resulting in lower firm value. Further evidence of the effects of control contestability on corporate value is reported in specification (3) of Panel B. Here, the estimated coefficient on DISPERSION2 is negative and statistically significant, suggesting that an uneven distribution of control rights among the five largest shareholders reduces the effectiveness of MLSS monitoring. This may increase the diversion of corporate resources, again resulting in lower firm valuation. Our findings on control dispersion provide empirical support to Bennedsen and Wolfenzon's (2000) theoretical predictions that dilution of power leads to low levels of diversion. The estimated coefficients on DISPERSION1 and DISPERSION2 highlight the importance of having shareholders of relatively similar size to achieve more efficiency in their governance role.

As for the firm-specific controls, we find that firm size (SIZE) and leverage (LEVERAGE) are significantly and negatively associated with firm value, suggesting that smaller and less leveraged firms have higher valuation, consistent with Lins (2003), among others. In addition, we find that the variables SALESGR and CAPEX enter the regressions with significant and positive coefficients across all specifications. These findings suggest that firms with better future growth opportunities and higher investment enjoy higher valuations.6

Summary and Comparison with Related Work Our results support the efficient-monitoring hypothesis of MLSS. The presence, number, and voting power of controlling shareholders beyond the largest shareholder translate into improved monitoring that helps restrain the incentives and ability of the largest shareholder to expropriate firm value. We also find that contestability of the largest shareholder's control (i.e., dispersion of shareholdings) influences the governance role of MLSS. Uneven distribution of blockholders' shareholdings (e.g., control or ownership rights) increases the likelihood of a winning coalition with small cash flow rights and reduces the contestability of the largest shareholder's control, which translates into a negative valuation effect. Our evidence on the dispersion of control rights corroborates that of Maury and Pajuste (2005) for a sample of Finnish firms. It is also consistent with Laeven and Levine (2008), who find a negative relation between corporate valuation and the dispersion of cash flow rights across multiple large shareholders in a sample of publicly traded Western European firms. For comparison with Laeven and Levine, we use the difference between the control rights of the two largest shareholders (CONT1 – CONT2) as an alternative measure of shareholdings dispersion.7 We find that (CONT1 – CONT2) loads negatively, in line with Laeven and Levine's findings for Western Europe. Reflecting its economic significance, the coefficient estimate on our shareholding dispersion proxy (CONT1 – CONT2) implies that a 1 standard deviation (13.24) increase in dispersion translates into a 10% decrease in valuation, that is, around 8% in the median Tobin's Q, compared to an 18% decrease in Laeven and Levine's sample of Western European firms. Our result of a positive and significant effect on corporate value with the mere existence of multiple blockholders is contrary to this variable's nonsignificant valuation effect reported in Maury and Pajuste. The discrepancy might be attributable to differences in the nature and severity of agency problems and the institutional environment, especially the protection of minority interests, between our East Asian sample and Western European firms (e.g., Claessens et al. 2002; Faccio and Lang 2002). We conjecture that, all else equal, the mere presence of multiple controlling shareholders in East Asian firms is valuable in reducing corporate diversion activities, perhaps because they sidestep deficiencies in the external institutional environment.

Evidence on the Role of MLSS When Expected Agency Costs Are Severe

Prior corporate governance research suggests that excess control rights and excess free cash flows are responsible for severe agency costs. In Table 5 we investigate whether and how the extent and significance of agency problems in Asian firms shape the monitoring role of MLSS. We split the sample into firms with and without separation between ownership and control rights of the controlling shareholder (specifications (2) and (1), respectively). As in Table 4, for expositional convenience we present the results in two panels according to the type of MLSS proxy. The results support the premise that the governance role of MLSS is more pronounced in firms with more serious agency problems. We report distinguishable estimated coefficients for NOWNERS,CONT2,VOTING21, and DISPERSION1 between the two subsamples, although the differences are not statistically significant. In particular, the coefficient estimates on NOWNERS and VOTING21 are statistically insignificant in the subsample of firms without excess control (specification (1)). However, the coefficients on both variables are positive and highly significant in the subsample of firms with excess control (specification (2)). The estimated coefficients on DISPERSION1 provide further evidence of the valuable effects of MLSS in firms with a higher likelihood of expropriation. DISPERSION1 displays a negative and significant coefficient at the 1% level in firms with excess control, suggesting that an unequal distribution of voting rights between the largest and second-largest shareholders compromises the monitoring incentives of the latter, resulting in lower firm value.

Table 5. The Effects of Multiple Large Shareholder Structures on Firm Valuation: Subsample Analysis.
Variable (pred. sign)CONTMCASH = 0 (1)CONTMCASH > 0 (2)Difference (z-stat) (2) – (1)FCF ≤ Median (3)FCF > Median (4)Difference (z-stat) (4) – (3)
  1. Note: This table presents regressions of firm value on multiple large shareholder structures and control variables for 1,252 nonfinancial firms from nine East Asian countries in 1996. Each cell refers to a regression including all control variables used in Table 4. The dependent variable is TOBQ, ratio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equity. The independent variables are: MOWNERS, dummy variable set to 1 if at least one large shareholder other than the largest shareholder controls more than 10% of the firm, and 0 otherwise; NOWNERS, number of other large shareholders (up to the fifth) controlling more than 10% of the firm; CONT2, ultimate voting rights of the second-largest shareholder; VOTING21, ratio of voting rights of the second-largest shareholder to voting rights of the largest shareholder; DISPERSION1, difference between the largest and the second-largest shareholders' voting rights to their sum; and DISPERSION2, Herfindahl index of the differences between the voting rights of the five largest shareholders. CONTMCASH is the ultimate cash flow rights minus ultimate voting rights of the largest shareholder. FCF is free cash flow, defined as earnings before interest, taxes, depreciation, and amortization, less interest, taxes, and common dividends. All regressions control for industry effects. Ownership data are from Claessens, Djankov, and Lang (2000). Robust t-statistics are in parentheses.

  2. ***Significant at the 1% level (one-tailed when directional predictions are made, and two-tailed otherwise).

  3. **Significant at the 5% level (one-tailed when directional predictions are made, and two-tailed otherwise).

  4. *Significant at the 10% level (one-tailed when directional predictions are made, and two-tailed otherwise).

Panel A. Presence and Number of Multiple Shareholders
 
MOWNERS (+)0.138**0.162**0.0240.0290.214***0.185*
(1.937)(2.153)(0.232)(0.423)(2.677)(1.773)
NOWNERS (+)0.0490.134**0.085−0.0270.148***0.175**
(0.989)(2.271)(1.099)(0.618)(2.467)(2.371)
 
Panel B. Control Contestability
 
CONT2 (+)0.007*0.014**0.0070.0010.015***0.014*
(1.384)(1.986)(0.814)(0.259)(2.396)(1.793)
VOTING21 (+)0.1340.292***0.1580.0300.322***0.292*
(1.162)(2.385)(0.942)(0.293)(2.397)(1.713)
DISPERSION1 (−)−0.143*−0.258***−0.115−0.029−0.294***−0.265*
(−1.462)(−2.414)(−0.793)(−0.324)(−2.572)(−1.825)
DISPERSION2 (−)−1.700*−1.715**−0.0150.146−2.785**−2.931*
(−1.612)(−1.687)(−0.010)(−0.167)(−2.109)(−1.850)
N  643  609   534  587 

Jensen (1986) argues that free cash flows are associated with opportunistic managerial behavior because they can be used in value-destroying activities rather than being distributed to shareholders. In closely held firms, managers are usually related to controlling shareholders (e.g., La Porta, Lopez-de-Silanes, and Shleifer 1999) and the potential misuse of free cash flows is likely to benefit the ultimate owners. Accordingly, we test whether the monitoring role of MLSS is more valuable in firms with excess free cash flows.8 We distinguish firms with excess free cash flows as those with above-sample-median free cash flows, defined as earnings before interest, taxes, depreciation, and amortization, less interest, taxes, and common dividends. The results, reported in specifications (3) and (4) in Table 5 for the two subsamples, lend strong support to the premise that the monitoring role of MLSS is more prominent in firms with excess free cash flows. We find that none of the MLSS proxies is significant in the subsample of firms with low free cash flows. In contrast, the results indicate significant relations between MLSS proxies and firm valuation in the subsample of firms with excess free cash flows, suggesting that MLSS monitoring is more effective in these firms. For example, the coefficient on MOWNERS is positive and statistically significant (at the 1% level) with its point estimate, when all other variables in specification (4) are held at their mean values, indicating that the MLSS value premium is 13.7%. Reinforcing this evidence, we report economically and statistically significant (at the 1% level) positive relations between CONT2, VOTING21, and firm value.

Although the coefficients on DISPERSION1 and DISPERSION2 are statistically indistinguishable from zero in the subsample of firms with low free cash flows, the corresponding coefficients are negative and highly statistically significant in the subsample of firms with substantial free cash flows. In contrast to Panel A, we generally report statistically significant differences in the coefficients on MLSS variables between the two subsamples, based on the extent of free cash flows. In line with Bennedsen and Wolfenzon's (2000) prediction, this evidence emphasizes the importance of shareholding dispersion in determining the governance role of MLSS.

The findings in Table 5 suggest that the MLSS monitoring role is more valuable in firms with expected agency problems, in particular, those resulting from high free cash flows. A plausible interpretation is that large shareholders tend to exercise a more efficient monitoring role when they fear corporate diversion by the largest controlling shareholder.

Does Ownership Identity Matter?

Prior research suggests that the identity of shareholders is important in understanding the link between ownership structure and firm value. Claessens, Djankov, and Lang (2000) classify shareholders into three main categories: family held, widely held, and state owned. They find that minority shareholder expropriation is more likely when the largest controlling shareholder is a family or a state. Accordingly, we examine the effects of the interaction between the identity and size of the second-largest shareholder on firm valuation. The underlying assumption is that different types of owners may have distinct incentives and abilities to monitor the controlling shareholder.

Specification (1) of Table 6 indicates that the presence of a family (FAMILY2) or a state (STATE2) as the second-largest shareholder is significantly and positively associated with firm value, suggesting superior monitoring by these two types of investors. In contrast, we find no evidence that control by a widely held firm (WH2) is associated with higher valuation. This implies lower incentives for widely held firms to monitor the controlling owner and reduce the extraction of private benefits. This explanation is consistent with Villalonga and Amit (2006), who stress that large institutional shareholders—whether a bank, an investment fund, or a widely held corporation—are less inclined to engage in monitoring because the resulting gains will be diluted among a large number of owners. In specification (2), we examine the effects of the interaction between identity and the size of voting rights held by the second-largest shareholder (CONT2) on firm valuation. Except when a widely held firm is the second-largest shareholder, we find positive and statistically significant coefficients on both interaction terms, suggesting that family and state play an important role in curbing the extraction of private benefits by the largest shareholder. This is reinforced by the results in specification (3), indicating positive and statistically significant coefficients on the interactions of family and state ownership with VOTING21.

Table 6. The Effects of the Interaction between the Identity and Size of the Second-Largest Shareholder on Firm Valuation.
Variable (pred. sign)Full SampleCONTMCASH = 0CONTMCASH > 0FCF ≤ MedianFCF > Median
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)   (15)
  1. Note: This table presents regressions of firm value on multiple large shareholder structures and control variables for 1,252 nonfinancial firms from nine East Asian countries in 1996. The dependent variable is TOBQ, ratio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equity. The independent variables are: FAMILY2, dummy variable set to 1 if the second-largest shareholder is a family, and 0 otherwise; WH2, dummy variable set to 1 if the second-largest shareholder is a widely held firm, and 0 otherwise; STATE2, dummy variable set to 1 if the second-largest shareholder is a state, and 0 otherwise; CONT2, ultimate voting rights of the second-largest shareholder; VOTING21, ratio of voting rights of the second-largest shareholder to voting rights of the largest shareholder; CASH1, ultimate cash flow rights of the largest shareholder; CONTMCASH, ultimate cash flow rights minus ultimate voting rights of the largest shareholder; SIZE, natural logarithm of total assets in millions of U.S. dollars; SALESGR, growth rate in sales over the previous year; LEVERAGE, ratio of long-term debt to total assets; and CAPEX, ratio of capital expenditures to total assets. CONTMCASH is the ultimate cash flow rights minus ultimate voting rights of the largest shareholder. FCF is free cash flow, defined as earnings before interest, taxes, depreciation, and amortization, less interest, taxes, and common dividends. Robust t-statistics are in parentheses. All regressions control for industry effects. Ownership data are from Claessens, Djankov, and Lang (2000).

  2. aDenotes that the differences between the subsample coefficients are significant at the 5% level.

  3. ***Significant at the 1% level (one-tailed when directional predictions are made, and two-tailed otherwise).

  4. **Significant at the 5% level (one-tailed when directional predictions are made, and two-tailed otherwise).

  5. *Significant at the 10% level (one-tailed when directional predictions are made, and two-tailed otherwise).

FAMILY2 (+)0.188**  0.206**  0.199*  0.180*  0.197*  
(2.116)  (1.875)  (1.326)  (1.409)  (1.495)  
WH2 (+)0.045  0.085  0.017  −0.097  0.143*a  
(0.732)  (0.900)  (0.225)  (−1.237)  (1.551)  
STATE2 (+)0.411***  0.209*  0.703***  0.126  0.666***a  
(2.739)  (1.453)  (2.660)  (1.240)  (2.632)  
FAMILY2 * CONT2(+) 0.010*  0.009  0.014  0.011  0.010 
 (1.578)  (1.194)  (1.227)  (1.268)  (1.020) 
WH2 * CONT2(+) 0.002  0.005  −0.002  −0.009*  0.010*a 
 (0.357)  (0.717)  (−0.370)  (−1.587)  (1.410) 
STATE2 * CONT2(+) 0.038***  0.015*  0.059***a  0.011  0.052***a 
 (3.028)  (1.511)  (3.347)  (1.282)  (2.998) 
FAMILY2 * VOTING21(+)  0.234**  0.196  0.347*  0.300*  0.207
  (1.737)  (1.161)  (1.573)  (1.444)  (1.056)
WH2 * VOTING21(+)  0.012  0.050  0.009  −0.196**  0.185
  (0.131)  (0.313)  (0.100)  (−1.938)  (1.234)
STATE2 * VOTING21(+)  0.959***  0.367**  1.529***a  0.315*  1.270***a
  (2.907)  (1.695)  (3.016)  (1.528)  (2.808)
N      1,252      1,252      1,252       643       643       643       609       609       609       534       534       534       587       587   587
R20.0960.1010.1010.0930.0890.0880.1420.1650.1670.1540.1560.1570.1200.1270.125

We also examine the effects of the type of second-largest shareholder in firms with higher potential of corporate diversion. The results are reported in specifications (4) through (15) in Table 6. The estimated coefficients on the presence of the state as the second-largest shareholder, as well as its control stake and voting power (relative to the largest shareholder), are positive and significant across the different specifications. Both the significance and magnitude of the effects are generally more important in the subsample of firms with severe agency problems (specifications (7)–(9) and (13)–(15)). We test whether the estimated coefficients in the subsample analysis are significantly different from each other and find that the state consistently plays a more significant role in firms with excess control than in firms without separation between ownership and control. A similar result is obtained when we use a firm's free cash flows as a proxy for agency costs. These findings, together with the evidence for the full sample (specifications (1)–(3)), suggest that the state plays an efficient monitoring role as a second shareholder. This is seen in the higher valuation of our East Asian firms, particularly those with more pronounced agency costs.

The results suggest that having a family or the state as a second-largest shareholder helps alleviate the risk of corporate diversion and enhances firm value. This may seem counterintuitive, as a family or state may be expected to be associated with weaker governance. The competition for corporate control between controlling families and the desire of the state to maximize proceeds from future privatizations (see, e.g., Megginson and Netter 2001) plausibly explain these findings.

Robustness Checks

We conduct sensitivity tests to gauge the robustness of our evidence on the effects of MLSS on firm value. In Table 7 we reproduce our core results reported in Tables 4 through 6, using industry-adjusted Tobin's Q as an alternative proxy for firm valuation to control for industry effects (Villalonga and Amit 2006). Our previous findings are not affected by using this alternative proxy, and we continue to estimate significant and positive relations between the various MLSS proxies and firm value. The effects are generally more economically and statistically important in firms with high agency costs (specifications (3) and (5)). Similarly, we continue to find support for more efficient monitoring by a family or state as second-largest shareholders.

Table 7. The Effects of Multiple Large Shareholder Structures on Firm Valuation: Robustness Tests, Set 1.
Variable (pred. sign)Full Sample (1)CONTMCASH = 0 (2)CONTMCASH > 0 (3)FCF ≤ Median (4)FCF > Median (5)Full Sample (6)Full Sample (7)Full Sample (8)
  1. Note: This table presents regressions of firm value on multiple large shareholder structures and control variables for 1,252 nonfinancial firms from nine East Asian countries in 1996. Each cell in columns (1) to (5) refers to a regression including all control variables used in Table 4. Regressions reported in columns (6) and (8) also include all control variables used in Table 4. The dependent variable is industry-adjusted TOBQ, ratio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equity. The independent variables are: MOWNERS, dummy variable set to 1 if at least one large shareholder other than the largest shareholder controls more than 10% of the firm, and 0 otherwise; NOWNERS, number of other large shareholders (up to the fifth) controlling more than 10% of the firm; CONT2, ultimate voting rights of the second-largest shareholder; VOTING21, ratio of voting rights of the second-largest shareholder to voting rights of the largest shareholder; DISPERSION1, difference between the largest and the second-largest shareholders' voting rights to their sum; DISPERSION2, Herfindahl index of the differences between the voting rights of the five largest shareholders; FAMILY2, dummy variable set to 1 if the second-largest shareholder is a family, and 0 otherwise; WH2, dummy variable set to 1 if the second-largest shareholder is a widely held firm, and 0 otherwise; and STATE2, dummy variable set to 1 if the second-largest shareholder is a state, and 0 otherwise. CONTMCASH is the ultimate cash flow rights minus ultimate voting rights of the largest shareholder. FCF is free cash flow, defined as earnings before interest, taxes, depreciation, and amortization, less interest, taxes, and common dividends. Ownership data are from Claessens, Djankov, and Lang (2000). Robust t-statistics are in parentheses.

  2. ***Significant at the 1% level (one-tailed when directional predictions are made, and two-tailed otherwise).

  3. **Significant at the 5% level (one-tailed when directional predictions are made, and two-tailed otherwise).

  4. *Significant at the 10% level (one-tailed when directional predictions are made, and two-tailed otherwise).

MOWNERS (+)0.148***0.145**0.160**0.0340.220***   
(2.808)(2.020)(2.034)(0.497)(2.715)   
NOWNERS (+)0.079**0.0560.126**−0.0190.149***   
(2.066)(1.124)(2.072)(−0.450)(2.496)   
CONT2 (+)0.010***0.008*0.013**0.0020.015***   
(2.357)(1.473)(1.897)(0.330)(2.436)   
VOTING21 (+)0.209***0.157*0.285**0.0390.333***   
(2.473)(1.341)(2.301)(0.369)(2.511)   
DISPERSION1 (–)−0.197***−0.158*−0.252**−0.037−0.304***   
(−2.738)(−1.600)(−2.311)(−0.402)(−2.675)   
DISPERSION2 (–)−0.991**−1.605*−1.481*0.190−2.721**   
(−1.709)(−1.519)(−1.520)(0.213)(−2.064)   
FAMILY2 (+)     0.195**  
     (2.166)  
WH2 (+)     0.061  
     (0.974)  
STATE2 (+)     0.432***  
     (2.857)  
FAMILY2 * CONT2 (+)      0.011* 
      (1.630) 
WH2 * CONT2 (+)      0.003 
      (0.631) 
STATE2 * CONT2 (+)      0.039*** 
      (3.048) 
FAMILY2 * VOTING21 (+)       0.252**
       (1.839)
WH * VOTING21 (+)       0.040
       (0.442)
STATE2 * VOTING21 (+)       0.987***
       (2.944)
N1,2526436095345871,2521,2521,252
R20.0600.0690.070

In a second set of robustness checks, reported in Table 8, we examine whether the effects of MLSS on firm performance extend to other periods and alternative proxies. In specifications (1) through (9), we report results from annual regressions for the eight years around 1996—the year when ownership data were collected—while assuming that a firm's ownership structure is constant (La Porta, Lopez-de-Silanes, and Shleifer 1999). The results generally indicate significant relations between our key MLSS proxies and firm value, suggesting that our previous conclusions are not driven by the specific year. This is further supported by the highly significant coefficients on all MLSS proxies when we estimate pooled regressions for 1992–2000 (reported in column (10)) to control for year effects. In particular, the valuation effects of MLSS generally hold during the Asian financial crisis of 1997 and 1998.9 Our unreported results remain practically unchanged when we replicate Table 5 for 1997 and the pooled years 1997 and 1998. These findings suggest that the role of MLSS in corporate governance was also important during the Asian financial crisis, when the incentives for expropriation were strong (Mitton 2002).

Table 8. The Effects of Multiple Large Shareholder Structures on Firm Valuation: Robustness Tests, Set 2.
Variable (pred. sign)TOBQROA 1992-2000 (11)Net Margin 1992-2000 (12)Sales Growth 1992-2000 (13)
1992 (1)1993 (2)1994 (3)1995 (4)1996 (5)1997 (6)1998 (7)1999 (8)2000 (9)1992-2000 (10)
  1. Note: This table presents regressions of firm value, return on assets, net margin, and sales growth on multiple large shareholder structures and control variables for a panel of 8,168 nonfinancial firm-year observations from nine East Asian countries over 1992–2000. Each cell refers to a regression including all control variables used in Table 4. The dependent variables are TOBQ, ratio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equity; ROA, ratio of earnings before interest, taxes, depreciation, and amortization to total assets; Net Margin, ratio of net income to sales; and Sales Growth, growth of sales over the previous year. The independent variables are: MOWNERS, dummy variable set to 1 if at least one large shareholder other than the largest shareholder controls more than 10% of the firm, and 0 otherwise; NOWNERS, number of other large shareholders (up to the fifth) controlling more than 10% of the firm; CONT2, ultimate voting rights of the second-largest shareholder; VOTING21, ratio of voting rights of the second-largest shareholder to voting rights of the largest shareholder; DISPERSION1, difference between the largest and the second-largest shareholders' voting rights to their sum; and DISPERSION2, Herfindahl index of the differences between the voting rights of the five largest shareholders. All regressions control for industry effects (not reported). Regressions (10), (11), (12), and (13) also control for year effects (not reported). Ownership data are from Claessens, Djankov, and Lang (2000). Robust t-statistics are in parentheses.

  2. ***Significant at the 1% level (one-tailed when directional predictions are made, and two-tailed otherwise).

  3. **Significant at the 5% level (one-tailed when directional predictions are made, and two-tailed otherwise).

  4. *Significant at the 10% level (one-tailed when directional predictions are made, and two-tailed otherwise).

Panel A. Presence and Number of Multiple Shareholders
 
MOWNERS (+)0.204**0.177**0.135**0.196***0.135***0.102**0.064**0.121**0.0440.118***0.006***0.028***0.041***
(2.147)(1.803)(1.772)(3.307)(2.642)(2.220)(1.711)(2.438)(0.814)(6.331)(2.734)(2.554)(5.135)
NOWNERS (+)0.172***0.092*0.091**0.113***0.072**0.072**0.034*0.067**0.0240.071***0.002*0.018**0.026***
(2.796)(1.384)(1.675)(2.941)(1.891)(2.153)(1.432)(1.933)(0.702)(5.454)(1.470)(1.920)(4.729)
 
Panel B. Control Contestability
 
CONT2 (+)0.009*0.012*0.0030.011***0.009**0.004*0.0020.006**0.0040.006***0.000***0.001**0.003***
(1.322)(1.606)(0.547)(2.478)(2.191)(1.432)(0.859)(1.806)(1.05)(4.53)(2.732)(2.266)(5.095)
VOTING21 (+)0.245*0.282*0.1260.252***0.188**0.107*0.0370.185**0.0570.147***0.011***0.041**0.072***
(1.555)(1.593)(0.908)(2.403)(2.249)(1.490)(0.702)(2.233)(0.632)(4.768)(3.001)(2.144)(5.562)
DISPERSION1 (−)−0.244**−0.252**−0.140−0.245***−0.181**−0.123**−0.055−0.163***−0.055−0.146***−0.009***−0.039**−0.058***
(−1.862)(−1.719)(−1.241)(−2.873)(−2.565)(−1.971)(−1.177)(−2.349)(−0.738)(−5.620)(−3.033)(−2.208)(−5.214)
DISPERSION2 (−)−1.199*−0.914−0.800−1.347***−1.160**−1.174***−0.671**−0.243−0.203−0.798***−0.022−0.103−0.068
(−1.471)(−0.975)(−1.184)(−2.382)(−2.004)(−2.494)(−1.725)(−0.506)(−0.402)(−4.284)(−0.921)(−1.137)(−0.787)
N  308  492  627  751  1,252  1,229  1,243  1,196  1,070  8,168  7,674  8,153  8,168

To address whether our findings hold when using non-market-based performance proxies, we report regression results in specifications (11), (12), and (13) that replace Tobin's Q with return on assets (ROA), net margin, and sales growth, respectively. Our core evidence on the role of MLSS in East Asia remains essentially the same. For instance, the presence (MOWNERS) and number (NOWNERS) of blockholders, the voting rights of the second-largest shareholder (CONT2 and VOTING21), and our first measure of shareholding dispersion (DISPERSION1) are consistently statistically and economically significant across the three performance models from 1992 to 2000, lending further support to the governance role of MLSS in East Asia.

Finally, in unreported tests that replicate Table 4, all coefficients on the MLSS variables remain unaffected in both level and statistical significance when we include the square of the cash flow rights of the largest shareholder to control for the nonmonotonic relation between ownership structure and corporate value (Morck, Shleifer, and Vishny 1988). However, the estimated coefficient on the square of the cash flow rights of the largest shareholder is negative but statistically insignificant.

IV. Conclusion

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Data and Variables
  5. III. Empirical Evidence
  6. IV. Conclusion
  7. Appendix
  8. References

Corporate governance studies generally focus on the shareholding attributes of the largest shareholder, giving little attention to the role of multiple large shareholders. We therefore investigate whether the presence, number, size, and identity of multiple large shareholders play a significant monitoring role in curbing the extraction of private benefits by the largest controlling shareholder in a sample of East Asian firms. Our results provide robust evidence that MLSS play a genuine corporate governance role. First, we find that firms with MLSS trade at a premium, relative to firms with a single large shareholder, suggesting that multiple large shareholders provide valuable monitoring. Second, increasing the number of large shareholders is associated with higher corporate valuation, consistent with the idea that a large number of shareholders creates more disagreement about projects that harm minority shareholders. Third, we document that greater contestability of the largest shareholder's voting power by the second-largest shareholder, as well as a more equal distribution of control stakes between these two largest shareholders (and among a firm's five largest blockholders), increases firm value. This suggests that the incentives and abilities of multiple large shareholders to monitor the controlling owner depend on the dispersion of the voting rights among the controlling shareholders. Fourth, we find that the monitoring role of MLSS is more pronounced in firms with severe agency problems. Finally, we find that the presence and voting power of families or the state as second-largest shareholders are associated with higher firm valuation.

Our results imply that MLSS constitute an effective corporate governance mechanism for firms located in East Asia. These findings can be used by regulators and policy makers at various levels to promote MLSS by, for instance, relaxing the restrictions on foreign ownership and facilitating the issuance of private equity.

Appendix

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Data and Variables
  5. III. Empirical Evidence
  6. IV. Conclusion
  7. Appendix
  8. References

Appendix: Variables, Definitions, and Sources

VariableDefinitionSource
Panel A. Firm-Specific Variables
 
TOBQRatio of the market value of assets to their book value, where the market value of assets is the market value of common stock plus the book value of assets minus the book value of equityAuthors' calculations based on Worldscope data
SIZENatural logarithm of total assets in millions of U.S. dollarsAs above
SALESGRGrowth rate in sales over the previous yearAs above
LEVERAGERatio of long-term debt to total assetsAs above
CAPEXRatio of capital expenditures to total assetsAs above
FCFFree cash flow, defined as earnings before interest, taxes, depreciation, and amortization, less interest, taxes, and common dividendsAs above
ROARatio of earnings before interest, taxes, depreciation, and amortization to total assetsAs above
Net MarginRatio of net income to salesAs above
 
Panel B. Ownership and Control Variables
 
MOWNERSDummy variable set to 1 if at least one large shareholder other than the largest shareholder controls more than 10% of the firm, and 0 otherwiseAuthors' calculations based on Claessens, Djankov, and Lang (2000) data
NOWNERSNumber of other large shareholders (up to the fifth) controlling more than 10% of the firmAs above
CONT2Ultimate voting rights of the second-largest shareholderAs above
VOTING21Ratio of voting rights of the second-largest shareholder to voting rights of the largest shareholder, CONT2/CONT1As above
DISPERSION1Difference between the largest and the second-largest shareholders' voting rights to their sum, (CONT1 – CONT2)/(CONT1 + CONT2)As above
DISPERSION2Herfindahl index of the differences between the voting rights of the five largest shareholders,(CONT1 – CONT2)2+ (CONT2 – CONT3)2+ (CONT3 – CONT4)2+ (CONT4 – CONT5)2As above
CASH1Ultimate cash flow rights of the largest shareholderAs above
CONTMCASHUltimate voting rights minus ultimate cash flow rights of the largest shareholderAs above
Footnotes
  • 1

    This is in line with Rozeff's (1982) prediction that shareholding concentration may not be the only relevant proxy of agency costs. Rozeff suggests that a small number of blockholders, a proxy for ownership dispersion, may play a more valuable monitoring role than a large number of blockholders in determining optimal firm dividend policy.

  • 2

    In Bennedsen and Wolfenzon (2000), increasing contestability while keeping the number of shareholders constant increases the equity stake of the winning coalition, which translates into actions more aligned with the interests of minority shareholders. In Bloch and Hege (2001), low contestability decreases the intensity of the competition for corporate control among large shareholders, resulting in less commitment to refrain from consuming private benefits.

  • 3

    For comparison, especially with Claessens et al. (2002) and Laeven and Levine (2008), we capture firm value with Tobin's Q. For robustness, we report tests that examine whether our findings are sensitive to using alternative performance measures.

  • 4

    Although we aim to capture various dimensions of MLSS (e.g., size, contestability, and dispersion), we test for the effect of each proxy separately given their high correlation, as shown in Table 2.

  • 5

    We also estimate that adding four large shareholders to a firm with a single large shareholder translates into a 19.85% increase in firm value.

  • 6

    All regressions in the subsequent tables include the same controls as in Table 4, but for brevity we report the coefficient estimates of our key test variables (MLSS-related proxies). We generally continue to find similar effects of these controls on firm value.

  • 7

    Although Laeven and Levine (2008) consider the dispersion of cash flow rights of the two largest shareholders, we use the dispersion of voting rights because cash flow rights data, covering major shareholders beyond the controlling owner, are not available for the East Asian sample. The control and ownership stakes of the blockholders are usually highly correlated.

  • 8

    We thank an anonymous referee for insightfully suggesting this analysis.

  • 9

    Mitton (2002) defines the crisis period from July 1997 through August 1998.

References

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Data and Variables
  5. III. Empirical Evidence
  6. IV. Conclusion
  7. Appendix
  8. References