2. How Do Family and Control Structures Affect Group Affiliated Firms’
2.3 Data And Methods
國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
20
2.3 D
ATAA
NDM
ETHODS2.3.1 Business groups in Taiwan
Compared to prior studies on conglomerate, business groups in Taiwan are particularly informative in the study of internal market because of two reasons. First, the existing evidence suggests that the effect of internal capital market on firm performance is not universal but depends on the institutional context where the firm is embedded. For instance, Hubbard and Palia (1999) found that given the under-developed external capital markets, the conglomerate merger wave occurred in 1960s and 70s in US was efficient, even though later it became inefficient as external capital markets became more developed over time. Taiwan as an emerging economy, in which external capital market is relatively under-developed, provides us different institutional settings from US. Second, group-affiliated firms in Taiwan offer a rich variety of ownership structure and clear definitions of group membership. Common ownership under a conglomerate means that shareholders have identical proportion of shares in all divisions within the firm. In contrast, shareholders of a business group can have different combinations of shares across different member firms within a group. Such flexibility in the ownership structure of business groups allows the controlling shareholders to increase their private benefits at the expenses of the minority shareholders, which is the key to understand the dark side of internal market (Chang, 2003; Claessens et al., 2000; Morck & Yeung, 2003). In addition, business groups also allow its ultimate controllers to use such control-enhancing mechanism as pyramid structure to increase the separation of ownership and control, thereby affecting firm value.
‧ 國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
21
2.3.2 Data Source and Sample
Taiwan’s top 2,000 listed manufacturing firms from 2000 to 2005 are used to test our hypotheses. We obtain ownership information and the financial statement data of these firms through the Taiwan Economic Journal database and then distinguish 4,355 group-affiliated firms. The Taiwan Economic Journal database subscribed by many authoritative international database providers, including DATASTREAM, REUTERS, Thomson Financial, etc., is the most credible database in Taiwan. Masulis, Pham, and Zein (2011) also utilize Taiwan Economic Journal’s ownership information to identify business groups in Asia and make a comparison between family and non-family business groups around the world. Due to the information of ownership and market value is available only for the publicly traded firms, there are 826 groups be identified and 967 firms, including 356 non-family group firms and 611 family group firms, be recognized as group-affiliated firms. Lastly, we excluded the firms controlled by government from non-family group firms and have 3,752 firm-year observations in our sample during 2000-2005.
2.3.3 The identification of group-affiliated firms in Taiwan
For identifying group-affiliated firms, Taiwan Economic Journal mainly refers to Taiwan’s Company Act Chapter VI-I and Statements of Financial Accounting Standards No.6. Taiwan’s Company Act Chapter VI-I, including Article 369-1 to 369-12, are formulated in 1997 to regulate group-affiliated firms’ business behavior. In this Act, enterprises which are independent in existence but have controlling-subsidiary or
‧ 國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
22
investment relations between each other are referred to group-affiliated firms.
Control-subsidiary relations are identified when a controlling firm has more 50%
ownership or control 50% seats on a board of directors of its subsidiary firms or when a controlling firm can directly or indirectly influence the administrative or financial systems of its subsidiaries. And the investment relations are established when cross equity shareholdings are up to 30%.
However, this Act has a loophole that let ultimate controllers can helm some subsidiary firms for their private benefits through kin relations which can let them eschew the monitoring from the market and the government. Take the ownership structure of Formosa Plastics Group in 2008 for example, Formosa Petrochemical Corp. is a joint venture established by Formosa Plastics Corporation, Nan Ya Plastics Corporation, and Formosa Chemicals & Fibre Corporation and each of these three parent companies hold less than 50% ownership of Formosa Plastics Corporation. Furthermore, each cross-holding between these three parent companies and Formosa Plastics Corporation is less than 30%. Accordingly, although all these firms are actually controlled by Wang Yung-Ching and his family, they cannot be recognized as a group in terms of Taiwan’s Company Act.
The Statements of Financial Accounting Standards No.6 issued by the Financial Accounting Standards Committee in Taiwan is a remedy for the loophole discussed above.
These Statements ask group-affiliated firms to disclose their top management team member’s kin relationship, their investment relations when their holdings of the other firm was between 20% and 50%, the foundation they donate more than one-third of its
‧ 國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
23
total capital, and their actual controllers. Again, take Formosa Plastics Group as an example. After the founder of Formosa Plastics Group, Wang Yung-Ching, resigned as chairman of Formosa Petrochemical Corp., Formosa Plastics Corporation, Nan Ya Plastics Corporation, and Formosa Chemicals & Fibre Corporation between 2006 and 2008, there are no marriage bonds between new chairmen of these firms. Nevertheless these four firms are actually controlled by the administration department of Formosa Plastics Group. Thus these four firms are still recognized as under the control of Wang Yung-Ching and his family.
Utilizing the ownership and kinship information that Taiwan’s financial authority ask listed firms to disclose, Taiwan Economic Journal can establish the chain of control and distinguish group-affiliated firms from non-group-affiliated firms. However, for the cases of 40/60 joint venture between two groups, the disclosure of investment relations that holdings are more than 20% may cause the boundary overlap between two business groups. Thus for this kind of cases, Taiwan Economic Journal will further refer to these firms’ board composition and their officers, directors, or blockholders to determine their actual controllers and to identify which group they are affiliated by. Lastly, in order to exclude these firms that ultimate owners are not actually engaged in (under financial control rather than administrative or strategic control), we investigate whether the ultimate owner, the foundation they control, or his/her private firm holds any board seat of a group-affiliated. When the ultimate do not control any board seat at group-affiliated firm, we will recognize that this firm is acquired just for financial purposes and then eliminate it.
‧ 國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
24
This process will continue until reaching the ultimate controller at the top of the chain of control. As at least two firms are controlled by the same family ultimate controller, these firms are recognized as a family business group (Hoshi, Kashyap, & Scharfstein, 1991;
Masulis et al., 2011). In contrast, when a set of firms are under the control of government, official bank, professional manager or two ultimate controllers (no one can dominate without the other’s support), it is identified as a non-family groups.
2.3.4 Dependent variable
Following Villalonga & Amit’s (2006, 2009) studies of family control and performance, we use Tobin’s q as our dependent variable. We use the ratio of the firm’s market value to total assets as a proxy for Tobin’s q, and we estimate the firm’s market value as the market value of common stock plus the book value of preferred stock and long-term and short-term loans payable. To control for industry effects, we also calculate industry-adjusted q, measured as the difference between the firm’s q and the asset-weighted average of the imputed qs of its segments, as a proxy for firm value. We use TEJ industry code, which is based on the industry code provided by Taiwan Stock Exchange, to compute industry averages.
2.3.5 Independent Variables
Family control measures. We use a dummy variable and a continuous variable to
measure family control. To ensure that a group-affiliated firm is actually controlled by the ultimate controllers, we exclude the firms that their ultimate controllers do not involve in the board. The family control dummy variable (family) equals 1 when a group-affiliated
‧ 國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
25
firm’s ultimate controllers are from a single family and they hold at least one seat on the board of this firm. When a group-affiliated firm’s founder is the only controlling shareholder or the ultimate controllers of a group-affiliated firm are from different families or institutions, such firm are considered non-family controlled firms. For the continuous measure of family control we use the family controllers’ voting rights as a proxy. And all non-family controlled firms are coded as zero.
Control rights in excess of control threshold. We construct excess control rights (excess
control) as the difference between the ultimate controllers’ control rights (La Porta et al., 1999) and the controlling shareholdings (Cubbin & Leech, 1983). La Porta et al. (1999) define control rights as the sum of controllers’ direct shareholdings and the cumulative holdings of the weakest link in a business group. Moreover, Cubbin and Leech (1983) define minority control holdings as values of P. According to their theory, we assume the voting probability equal to 99% and the degree of control equal to 1 (Z-value is 3.32) to calculate P. When the excess control rights in a firm is greater than zero, it is supposed that this firm is minority controlled; while when the excess control rights is less than zero, the ultimate controllers can’t dominate the shareholders’ meeting and need to seek other shareholders’ support for proxy fight to wrest control rights. Since the costs of large ownership will surpass its benefits when ownership separate from control, we use the inside owners’ excess control rights as a proxy for entrenched management in group-affiliated firms.
Divergence between voting rights and cash flow rights. We use the difference between voting rights and cash flow rights (divergence) as a proxy for ultimate controllers’ power
‧ 國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
26
to use other people’s money. Voting rights is the seats that the ultimate controllers hold on the board, and the cash flow rights are the sum of controlling shareholders’ direct shareholdings and indirect shareholdings. Indirect shareholdings are the sum of the cash flow rights controlled by other non-financial affiliates in a business group (La Porta et al., 1999). For example, when a controlling shareholder A holds 5 % shares of firm 1 and 10% of firm 2 (direct shareholdings), and then firm1 holds 20% shares of firm 2, for A, whose indirect shareholdings of firm 2 is 1 percent (20% multiplying 5%), therefore, A’s cash flow rights is equal to 11% (10% + 1%).
Controlling owners serve as CEO. We use a dummy variable to represent active or
passive control (Anderson & Reeb, 2003). Controlling owners as CEO is identified as active control in this study. We use a dummy variable is coded as ‘1’ if the ultimate controller or his family serve as both the chairman and the CEO, and ‘0’ otherwise.
2.3.6 Controls
Following Chang’s (2003) study, we control for both firm and group effects that may affect ownership structure and firm value. We take the log of the sales of a firm in billions Taiwan dollar as a control for firm size (sales). We also control for firm equity in billions of Taiwan dollar. Firm age is calculated as the number of years since founding and. Years of listing (listing) is the elapsed years calculated by calendar year since the last time a firm was listed. Volatility is measured as the standard deviation of profitability during the past five years. Sales growth is computed as the annual growth of sales and market share is calculated by dividing a firm’s sales by industry shipments at the TEJ industry code.
‧ 國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
27
RD intensity is measured by dividing a firm’s RD expenditure by annual sales. We use debt-to-equity ratio as a proxy for leverage. This study also includes year and industry dummies to control for time and industry effects. Furthermore, in order to identify high-tech and low-tech industries, we use TEJ industry code, which is based on the industry code provided by Taiwan Stock Exchange, rather than Standard Industry Classification. All electronic firms are coded as high-tech industry (1) and the others are coded as low-tech industry (0).
In terms of group effects, we incorporate group equity in billions Taiwan dollar and internal transactions. Internal transactions include internal trade and financial transaction (Chang & Hong, 2000). We use sales to affiliates and purchases from affiliates as proxies for internal trade and use loan to affiliates and debts from affiliates to measure internal financial transaction. These variables are defined as the ratio of each action divided by a firm’s total sales.
2.3.7 Data Analysis
We use Hausman-Taylor estimator to exam the relationship between control mechanisms and firm value for two main reasons. First, fixed effects estimators can control for unobserved firm heterogeneity that affects both ownership and performance in panel data (Himmelberg, Hubbard, & Palia, 1999), while this methodology require within variation in ownership (Zhou, 2001). Using the sample form Standard and Poor’s ExecuComp data set, Zhou (2001) found that the within variations in managerial ownership are very small comparing to between variations. Three control-enhancing mechanisms in our study have
‧ 國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
28
the same characteristics (Table 2-1). Second, the relationship between family control and firm value can be endogenous and Hausman-Taylor estimator allows for including multiple endogenous variables. Furthermore, no matter in Ordinary Least Squares, Fixed effects, or Random effects world, Hausman-Taylor estimator is a viable estimator and is always second best to the efficient estimator (Baltagi, Bresson, & Pirotte, 2003). In this study, family control measures (dummy or continuous) , divergence between voting rights and cash flow rights, control rights in excess of control threshold, controlling owners serve as CEO, and their interaction terms are treated as endogenous variables.
Table 2-2 presents the descriptive statistics and correlations between variables, and Table 2-3 presents the regression results for the full sample. In order to further identify the contingency effect of environmental uncertainty that may affect the relationship between family control mechanisms and firm value, all firms are divided into high-tech or low-tech industry group and Table 2-4 and Table 2-5 report the regression results for these two subgroups. Moreover, the results of the interaction effects between cross-shareholdings and different control-enhancing mechanisms on firm value are reported in Table 2-6 and Table 2-7. Time effects and industry effects are also controlled.
When we use Tobin’s q as our dependent variable, the industry and year dummies are included. In addition, we try to control for time effects and industry effects by using industry-adjusted q as our dependent variable, while in this case the industry and year dummies are dropped (Villalonga & Amit, 2006).
‧ 國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
29