Corporate Diversification, Relatedness, and Firm Value: Evidence from Korean Firms *

Transcription

1 Asia-Pacific Journal of Financial Studies (2008) v37 n6 pp Corporate Diversification, Relatedness, and Firm Value: Evidence from Korean Firms * Sung C. Bae Bowling Green State University, Bowling Green, USA Taek Ho Kwon Chonnam National University, Yeosu, Korea Jang Woo Lee Dongeui University, Busan, Korea Received 02 March 2007; Accepted 11 July Abstract We examine the valuation effects of diversification activities for Korean firms by diversification type and the occasion of the Korean financial crisis. Employing a unique dataset of 2,894 firm-years for the entire manufacturing industries, we find that diversification by Korean firms on average decreases firm value but its effect varies by the type of diversification. While unrelated diversification erodes firm value, related diversification is associated with a non-negative effect on firm value. These valuation effects are more pronounced before the crisis than after the crisis. Our results also show an important role of a firm s affiliation to a large business group, known as chaebols, that related diversification by chaebol-affiliated firms comes with a significant value gain. We further find that the different valuation effects of unrelated and related diversification are closely related to a firm s ownership concentration and financial leverage. Our results are robust to different samples and regression model specifications. Keywords: Related Diversification; Unrelated Diversification; Korean Firms; Korean Financial Crisis; Chaebol Affiliation * We are grateful for many helpful comments from Sridhar Gogineni, Jun-Koo Kang, Young Seok Park, seminar participants at Hallym University, session participants at the 2006 Financial Management Association meetings in Salt Lake City, Utah, USA and 2006 Korean Finance Association meetings. We thank Jinwoo Park (editor) and two anonymous referees for additional valuable comments and suggestions. All remaining errors are our own. ** Corresponding Author. Address: Professor of Finance at the Department of Finance and Real Estate, College of Commerce and Economics, Dongeui University, Busan, Korea, ; jangwoo@deu.ac.kr; Tel: ; Fax:

2 Corporate Diversification, Relatedness, and Firm Value 1. Introduction Extensive literature has examined the consequences of corporate diversification. On the one hand, earlier empirical studies document a diversification discount that diversified firms trade at an average discount relative to single-segment firms (Lang and Stulz, 1994; Berger and Ofek, 1995; Servaes, 1996). On the other hand, correcting for sample selection bias, later studies find a diversification premium (Villalonga, 1999; Whited, 2001; Campa and Kedia, 2002; Graham, Lemmon, and Wolf, 2002; Villalonga, 2004a, 2004b). 1) While the existing literature has actively focused on the corporate diversification of developed countries including the U.S., the international evidence regarding the valuation effect of corporate diversification in emerging markets is limited and inconclusive. Furthermore, the inquiry into the valuation effect of firm diversification has paid little attention to the supposedly different valuation effect of two diversification types of related and unrelated diversification, while focusing primarily on the marginal valuation effect of diversification relative to non-diversification. This issue should not be taken lightly since a firm s targeted diversification can have a vastly different effect on firm value depending on the relatedness of the new business to the current business (Rumelt, 1974; Chatterjee and Wernerfelt, 1991; Fan and Lang, 2003). We intend to fill this gap by examining how related and unrelated corporate diversifications affect the market values of Korean firms. Our study makes important contributions to the existing literature on corporate diversification in several aspects. First, we provide new empirical evidence on corporate diversification for Korean firms in one of the world s leading emerging markets. Unlike developed markets, emerging markets are characterized with a lack of reliable information and a high degree of information asymmetry, leading to a high degree of market imperfection. This suggests that there may be great room to enhance firm value if a firm creates an internal production market through diversification. On the contrary, diversification by firms in emerging markets may decrease firm value because the greater information asymmetry and lower monitoring of firms by the market may increase the agency costs associated with diversification. Due to these 1) See Martin and Sayrak (2003) for a survey of recent literature on the effect of corporate diversification on shareholder value. 1026

3 Asia-Pacific Journal of Financial Studies (2008) v37 n6 differences, the empirical evidence documented for firms in the developed markets would not be directly applicable to firms in emerging markets. A number of prior studies offer unsettled evidence on this issue for emerging market firms. Lins and Servaes (2002) argue that the severe market imperfections found in emerging markets increase potential agency costs associated with diversification and that greater asymmetric information allows management and large shareholders to exploit minority shareholders more easily. Consistent with these arguments, they find that diversified firms from East Asian markets trade at a discount of approximately 7% compared to single-segment firms. Baek, Kang, and Park (2004) also find that highly diversified Korean firms have significantly lower share returns than undiversified counterparts. In contrast, Khanna and Palepu (2000) argue that diversification may be more valuable in emerging markets than in more developed economies. Fauver, Houston, and Naranjo (2003) find a diversification premium or no discount for firms in less developed economies. The majority of prior studies, however, use only a limited number of large firms in emerging markets available in the Worldscope database. Second, we shed new empirical lights on the relation between firm value and two types of diversification: related and unrelated diversification. Our study is the first of its kind that examines this issue for Korean firms. Related diversification refers to expansion into the category of existing lines of business, and unrelated diversification refers to expansion into an industry different from existing lines of business. As Chatterjee and Wernerfelt (1991) note, optimal diversification would be different according to firm resources, suggesting that the diversification type is directly linked to firm value. Related diversification may offer both positive and negative valuation effects relative to unrelated diversification, and existing literature offers mixed evidence on this issue (see Lewellen, 1971; Rumelt, 1974; Amihud and Lev, 1981; Nayyar, 1993; Markides and Williamson, 1994; Claessens et al., 2003; Doukas and Kan, 2004). Third, Korean firms offer a unique opportunity to investigate the effect of macroeconomic factors on corporate diversification strategies. We take the Korean financial crisis that broke out in late 1997 as a special event of the macroeconomic factors and examine this issue for Korean firms over the pre- and post-crisis periods. During the past three decades of Korea s economic growth, Korean firms were often criticized for their supposed involvement in reckless expansion schemes in the name of diversification and for their catalytic role for the 1997 Korean financial crisis. Following the cri- 1027

4 Corporate Diversification, Relatedness, and Firm Value sis, Korean firms have been driven to make fundamental changes in corporate governance and corporate diversification strategies by both internal and external forces. In spite of the strong need of changes in corporate financing and diversification strategies, however, there still appears to be lack of efforts by Korean firms, consequently only with a limited success. In this regard, we test whether the Korean financial crisis has brought a significant change in Korean firms diversification strategies and their effects on firm value. Finally, we use alternative measures of both firm value and the degree of a firm s diversification activities to avoid methodological problems associated with these two key measures in the Berger and Ofek s (1995) excess value method to assess gain or loss from the diversification. 2) Other studies suggest that the Berger and Ofek s method causes a possible downward bias in computing firm value (see, e.g., Villalonga, 1999; Campa and Kedia, 2002; Graham, Lemmon, and Wolf, 2002). Along with the methodological problems, there is a practical limitation in securing the median ratios from the representative firms in the same lines of business since there are often only few firms available in the same industry in the emerging markets. Unlike previous studies, we measure the degree of a firm s related and unrelated diversification separately using Caves diversification index method (Caves et al., 1980, pp ). We develop our unique dataset by compiling all necessary data information for a total of 2,894 firm-years covering the entire Korean manufacturing industries during the periods, three years before and after the 1997 Korean financial crisis. We find that the diversification activities by Korean firms are, on average, associated with a decline in firm value. When the diversification is classified into two types of diversification, we find strikingly different valuation effects of related and unrelated diversification on firm value. Specifically, related diversification is associated with a non-negative, though not positive, effect on firm value, while unrelated diversification is associated with a significantly negative effect on firm value. Another key find- 2) Excess value of a diversified firm is obtained from the natural logarithm of the ratio of a firm's actual value to its imputed value. The imputed value of each segment is measured by the median ratio of a firm s total capital to its total assets (or sales or earnings) within the industry the single segment belongs to, multiplied by the segment s total assets (or sales or earnings). Hence, the sum of the imputed values of a firm s segments represents the value of the firm as if all of its segments are operated as stand-alone businesses. Berger and Ofek (1995) document that comparing the sum of these stand-alone values to the firm s actual value shows a 13% to 15% average value loss from diversification during

5 Asia-Pacific Journal of Financial Studies (2008) v37 n6 ing of our study is the significantly different valuation effect of related diversification by Korean firms between the pre-and post-crisis period. An expansion into related business by Korean firms adds firm value during the pre-crisis period, but decreases firm value during the post-crisis period. These results suggest that it became more difficult for Korean firms to generate gains even from related diversification activities during the post-crisis period. Our findings are generally consistent with those from previous studies that document evidence supporting that the economic crisis in Korea has a significant and negative effect on the value of firms. We also find that chaebol-affiliation erodes firm value, but we find no evidence that diversification by chaebol-affiliated firms destroys firm value. Our results show that while unrelated diversification by chaebol-affiliated firms has a non-negative, though not positive, effect on firm value, related diversification by these firms has a positive and significant valuation effect before the financial crisis. Our results further show that a firm s corporate governance (proxied by ownership concentration) and financial leverage have significantly opposite effects on the values of firms which engage in related diversification before the financial crisis. To be more specific, firms engaging in related diversification are associated with higher firm value when they have less concentrated ownership and/or have higher financial leverage. For chaebol-affiliated firms, the effects of ownership concentration and financial leverage are more pronounced and different depending on the diversification type. While a firm s ownership concentration has a positive effect on chaebol firms unrelated diversification, it has a negative effect on chaebol firms related diversification; hence, unrelated diversification by a chaebol-affiliated firm is associated with a higher firm value when the firm has higher ownership concentration and/or lower financial leverage. In contrast, related diversification by a chaebol-affiliated firm is associated with a higher firm value when the firm has lower ownership concentration and/or higher financial leverage. 2. Diversification, Firm Value, and Korean Financial Crisis 2.1 The Effect of Diversification on Firm Value Existing literature documents both costs and benefits associated with corporate di- 1029

6 Corporate Diversification, Relatedness, and Firm Value versification. On the one hand, corporate diversification may enhance firm value in several respects. First, multidivisional firms can operate efficiently by coordinating specialized divisions (Chandler, 1977). Second, diversification can mitigate the underinvestment problem through an efficient allocation of assets by creating an internal capital market (Weston, 1970; Stulz, 1990). Third, diversification can help lower capital costs and increase capital raising capacity since earnings streams from diversified divisions have low correlations to make corporate cash flows stabilized (Lewellen, 1971). Fourth, related diversification is advantageous in building up economies of scale and utilizing strategic assets (Markides and Williamson, 1994). On the other hand, corporate diversification may destroy firm value for several reasons. First, diversified firms may engage in investment projects with a negative NPV by making a bad use of increased capital raising capacity and free cash flow (Jensen, 1986; Stultz, 1990). Second, multidivisional firms may delay withdrawal of failed segments by allowing cross-subsidization (Meyer, Milgrom, and Roberts, 1992). Third, multidivisional firms may face greater information asymmetry costs than independent firms because of increased information asymmetry between headquarters and divisions (Harris, Kriebel, and Raviv, 1982; Myerson, 1982). Methodologically, a key measure in the analysis of valuation effects of diversification is firm value. Berger and Ofek (1995) measure excess value as the percentage difference between a firm s total value and the sum of imputed values for its segments as stand-alone entities to assess gain or loss in value from diversification. The imputed value of each segment is measured by the median ratio of a firm s total capital to its total assets (or sales or earnings) within the industry the single segment belongs to, multiplied by the segment s total assets (or sales or earnings). Hence, the sum of the imputed values of a firm s segments represents the value of the firm as if all of its segments were operated as stand-alone businesses. Berger and Ofek (1995) document that comparing the sum of these stand-alone values to the firm s actual value shows a 13% to 15% average value loss from diversification during Numerous subsequent studies apply the Berger and Ofek s methodology and offer evidence that diversification firms on average trade at a discount relative to singlesegment firms. More recent studies show, however, that the discount is only the product of sample selection bias. Villalonga (1999) and Campa and Kedia (2002) find that diversified firms trade at a discount prior to diversifying, suggesting that firms diversify to en- 1030

7 Asia-Pacific Journal of Financial Studies (2008) v37 n6 hance firm value. Controlling for the endogeneity of the decision to diversify, both studies find that the diversification discount disappears or even turns into a premium. Graham, Lemmon, and Wolf (2002) note that the Berger and Ofek s method is likely to overestimate diversification discounts since it assumes that diversified segments can attain the industrial average firm value, hence underestimating gain from the diversification. 3) They assert that Berger and Ofek s (1995) results stem from their not controlling sampling errors when computing imputed values. 4) They show that half or more of the discount appears because the segments acquired by diversifying firms were also discounted prior to their acquisition. Given that both diversifying firms and their targets trade at a discount prior to diversification, it is not surprising that diversified firms exhibit a discount. Overall, the findings of the more recent studies suggest that diversification, in itself, does not destroy the value. Gomes and Livdan (2004) develop a general dynamic model of the optimal behavior of a firm, predicting that diversified firms have, on average, lower values of Tobin s q than focused firms do. In their model, firms diversify only when they become relatively unproductive in their current industries. They argue that this endogenous selection mechanism accounts for the lower valuation of diversified firms. Using three different treatment effects estimators, Villalonga (2004b) finds that diversification on average does not destroy the value, consistent with the evidence in Villalonga (1999), Campa and Kedia (2002), and Graham, Lemmon, and Wolf (2002). In this study, we test two different views of the effect of diversification on firm value for Korean firms before and after the Korean financial crisis. Focusing on a single country in this way allows us to examine diversification index measures at a level of detail that would be hard to aggregate across countries. 2.2 The Effect of Related and Unrelated Divesification on Firm Value Existing literature documents mixed evidence on the valuation effects of related 3) For instance, when managerial problems drive a firm to take over another firm whose value is below the industrial average, the Berger and Ofek s method that uses industrial averages is likely to underestimate excess value of the diversified firm. 4) Mansi and Reeb (2002) argue that diversification reduces shareholder value but enhances bondholder value. They report that Berger and Ofek s (1995) results are obtained because debt value is estimated as book value. When both equity and debt are estimated as market value, they find no evidence of diversification discount. 1031

8 Corporate Diversification, Relatedness, and Firm Value diversification relative to unrelated diversification. On the one hand, Markides and Williamson (1994) point out several merits associated with related diversification including economies of scope in the short run, utilizing accumulated core competence and the potential to build strategic assets by using such core competence in the long run. Rumelt (1974) asserts that related diversification can increase firm value because it can allow firms to jointly use both skill and resources. Nayyar (1993) argues that benefits from a positive relation in an existing business and from economies of scope are available from related, but not from unrelated, diversification. Doukas and Kan (2004) find that bidders engaging in unrelated acquisitions experience larger declines in excess cash flow and greater valuation discounts than do bidders engaging in related acquisitions. Using the Business Information Tracking Series (BITS) as an alternative data source, Villalonga (2004a) finds that diversified firms on average trade at a significant premium relative to comparable portfolios of single-business firms. He interprets these findings as evidence that there is a discount to unrelated (conglomerate) diversification, but a premium to related diversification. 5) On the other hand, Lewellen (1971) argues that if diversification is carried out within similar industries, the insurance effect from unrelated diversification would be so insignificant that an increase in firm value by way of debt capacity augmentation will not take shape. Amihud and Lev (1981) also show that unrelated diversification is associated with lower firm risk due to the existence of multiple lines of business with imperfectly correlated returns. As a whole, existing studies suggest that related diversification can generate a positive effect on firm value if adequately carried out. Existing literature also suggests that the motive for related diversification may be different from that for unrelated diversification and that the effect of diversification on firm value may be different depending on the diversification type, irrespective of the way the degree of diversification is measured. 2.3 The Effect of the Korean Financial Crisis Several studies investigate various issues related to the Korean financial crisis and 5) Villalonga (2004a) shows that since related diversification is likely to predominate over conglomeration, when all diversification types are pooled together as they are in BITS, the net effect on firm value becomes a premium. 1032

9 Asia-Pacific Journal of Financial Studies (2008) v37 n6 document evidence supporting significant changes and reforms across the financial sector, corporate sector, and public sector. Chang, Kang, and Shin (2004) describe two major changes following the Korean financial crisis. First, business rules of the financial sector gradually became more aligned with global standards. Under the reform programs, commercial banks with a capital adequacy (BIS) ratio of below 2% were to be given management improvement orders from the government, such as the complete write-off equity capital, suspension of operation, and merger with healthier financial institutions. In addition, financial institutions with a certain asset size were required to appoint outside directors, set up an auditing committee, and appoint a compliance officer. All these reforms resulted in a significant improvement in the soundness and profitability of the financial sector. The average BIS ratio of commercial banks increased from around 7.0% at the end of 1998 to 10.5% at the end of The average debt to equity ratio of manufacturing companies fell to below 140% in 2002, a significant decline from a level near 400% in Second, at the urge of the International Monetary Fund and the Korean government, the Korean corporate sector has undergone massive restructuring reforms on shareholder rights, corporate governance, management transparencies, and financial structures, among others. For example, cross-debt guarantees declined to almost zero level by 2000 among affiliates of the top five chaebols, and by 2002 among those of the remaining thirty largest chaebols. In addition, legislative reforms on corporate governance made it obligatory for the companies listed on the Korea Stock Exchange (KSE) to appoint outside directors (effective in 1998), and for large KSE-listed companies to establish an audit committee (effective in 1999). Hence, by the end of 2001, the average number of outside directors per KSE-listed companies increased to 2.3, accounting for 34.8% of the total number of directors. Furthermore, by the end of 2001, more than 22% of KSE-listed companies introduced auditing committees. A fair disclosure system was also installed to enhance management transparency in accounting practices and increase firms responsibilities for their public information announcements. Reforms in the corporate sector along with the financial sector s gradual adoption of global standards are believed to help restoring the principle of shareholder value maximization among Korean companies. Despite these reforms, Korean chaebols have continued to operate as conglomerate business groups. De facto CEO and owner 1033

10 Corporate Diversification, Relatedness, and Firm Value of chaebols are still members of founding family. Furthermore, the affiliates of chaebols are linked to each other through circular shareholdings, and share a same brand, a same business philosophy, and a same pool of manpower. In this regard, it is a critical empirical issue whether the reforms in the corporate and financial sectors following the Korean financial crisis brought in fundamental changes in the diversification trend of Korean firms, especially chaebol firms, and their impact on the relation between the diversification and firm value. 3. Data and Measurement of Diversification Indexes 3.1 Data Our sample represents publicly traded manufacturing firms listed on the KSE from covering three years before and after the 1997 Korean financial crisis. The base industry (or main line of industry) of our sample firms belongs to Korea Standard Industrial Classification (KSIC) 15 through 36, spanning twenty-one manufacturing industries from food and beverage (KSIC 15) to furniture manufacturer (KSIC 36), except for the tobacco industry (KSIC 16) which is not included due to the lack of sufficient sample firms. Hence, we exclude firms in the financial services and utility industries. Sales volume is used as a basis for determining the base industry and computing the degree of diversification. We collect sales data from the Korean Association of Listed Companies database. Each sales item is classified with reference to KSIC three-digit level, excluding those noted as others or those with an ambiguous classification. We also exclude firms without necessary accounting entries, stock return data or sales data. Our final sample consists of 2,894 firm-years in twenty-one industries over the periods. 3.2 Measurement of Diversification Indexes We employ three measures of the degree of a firm s diversification drawn from the Caves weighted index of diversification (Caves et al., 1980, pp ). The first measure of diversification index, CDX, is computed as: 1034

11 Asia-Pacific Journal of Financial Studies (2008) v37 n6 J CDX = p d (1) j= 1 j jh where J = the total number of products in a firm; pj = sales of product j as a percentage of total sales; H = the base product with largest sales; 0 if product j belongs to the same three-digit KSIC as the base product H, djh = 1 if product j belongs to a different three-digit KSIC than the base product H. 2 if product j belongs to a different two-digit KSIC than the base product H. As used by extensive studies in the economics and finance literature, CDX measures the degree of a firm s diversification into both related and unrelated lines of business. A higher value of CDX indicates a greater diversification. The next two measures of diversification index assess the degree of relatedness of a firm s new diversification activity relative to the current lines of business. Berger and Ofek (1995) consider firms operating in less than two-digit SIC codes as practicing related diversification. They measure the degree of relatedness as the difference between the total number of segments reported by a diversified firm and the number of segments with a different main two-digit SIC code. 6) The Berger and Ofek s definition of relatedness, however, may be misleading. For example, according to Berger and Ofek (1995), if a firm currently engages in unrelated diversification at the two-digit SIC level and subsequently diversifies into a three-digit SIC business within the previous unrelated two-digit SIC level, this latter diversification is regarded as related diversification. This classification is controversial, and hence, we use an alternative way to classify related and unrelated diversification. 7) The second diversification index, CDXU, considers only two-digit KSIC codes and thus measures the degree of a firm s unrelated inter-industry diversification. Similar to CDX, CDXU is computed as: 6) Their regression results show a positive and significant relation between firm value and related diversification, which they interpret as suggesting that relatedness mitigates the value loss from diversification. 7) Rumelt (1974) uses a combination of objective and subjective criteria to classify relatedness. Fan and Lang (2000) use commodity flow data in U.S. input-output (IO) tables and construct IO-based measures, so as to capture interindustry and intersegment (within a diversified firm) vertical relatedness and complementarity. 1035

12 Corporate Diversification, Relatedness, and Firm Value J CDXU = p d (2) j= 1 j where J = the total number of products in a firm; pj = sales of product j as a percentage of total sales; H = the base product with largest sales; djh = jh 1 if product j belongs to a different two-digit KSIC than the base product H, 0 if product j belongs to the same two-digit KSIC as the base product H. The third diversification index, CDXR, is computed by first selecting a base industry with the largest sales volume within two-digit KSIC industries for each firm and then estimating the diversification index based on the three-digit KSIC codes within the two-digit KSIC base industries. Hence, it measures the degree of a firm s related intra-industry diversification. Similar to CDX and CDXU, CDXR is computed as: J CDXR = p d (3) j= 1 j jh where J = the total number of products in a firm; pj = sales of product j as a percentage of total sales; H = the base product with largest sales; 1 if product j belongs to a different three-digit KSIC than the djh = base product H within two-digit KSIC base industry, 0 otherwise. 3.3 Regression Model for the Effect of Diversification on Firm Value We conjecture that if diversification increases firm value, the diversification measure of CDX would be positively related to the measure of firm value in a regression with the effects of other variables being controlled. Since a firm s value can be affected by factors other than the firm s diversification activities, we use seven control variables in the regression model. The first three control variables are based on Tobin s q theory and include natural log of total assets (LASST) as proxy for firm size, R&D expenditures divided by total sales (RNDPS) as proxy for firm s growth, and 1036

13 Asia-Pacific Journal of Financial Studies (2008) v37 n6 standard deviation of daily stock returns over the past year (SRSTD) as proxy for firm s risk (Chauvin and Hirschey, 1993; Bhagat and Welch, 1995). The next two control variables are the proportion of shares owned by a firm s largest shareholder (OWN) relative to total shares as proxy for corporate governance structure, and leverage ratio (LEV; measured by total debt divided by total assets) to control for corporate financial risk factor. These two variables are included as they are expected to affect firm value differently before and after the Korean financial crisis other than the firm s diversification activity. The last two conrol variables are twenty industry dummy variables (ID) and six year dummy variables (YD) to control for industry and year differences, respectively, in the regression. The industry dummy variables span a total of twenty-one industries from food and beverage (KSIC 15) to furniture manufacturer (KSIC 36), except for the tobacco industry (KSIC 16) which is not included due to the lack of sufficient sample firms. Data for all accounting items and daily stock returns are collected from the Korea Investors Service database. Putting these variables together, we examine the relation between diversification and firm value in the following regression model: FV a a LASST a RNDPS a SRSTD a OWN a LEV i = i 2 i 3 i 4 i 5 i 20 6 (4) + acdx + a ID + a YD + ζ 6 i 6+ j j 27+ k k i j= 1 k= 1 where FV is firm value, LASST is natural log of total assets, RNDPS is the ratio of R&D expenditures to total sales, SRSTD is the standard deviation of daily stock returns. OWN is the relative ownership of largest shareholder, LEV is total debt over total assets, CDX is the Caves weighted index of diversification measured by equation (1), ID is industry dummy variable, and YD is year dummy variable standing for each year. In order to examine the valuation effects of two different types of diversificationrelated versus unrelated diversification, we estimate the regression equation (5) by substituting CDX in equation (4) with CDXU and CDXR in the following manner: FV b b LASST b RNDPS b SRSTD b OWN b LEV i = i 2 i 3 i 4 i 5 i 20 6 (5) + b CDXU + b CDXR + b ID + b YD + η 6 i 7 i 7+ j j 28+ k k i j= 1 k=

14 Corporate Diversification, Relatedness, and Firm Value In equation (5), CDXU represents KSIC two-digit diversification index, measuring the degree of a firm s unrelated (inter-industry) diversification. CDXR represents threedigit KSIC diversification index within two-digit KSIC base industry, measuring the degree of a firm s related intra-industry diversification. 3.4 Measurement of Firm Value Due to the methodological problems associated with computing imputed values of diversified firms and lack of sufficient single-segment firms in the same industry in the Korean market, we do not employ the excess value method developed by Berger and Ofek (1995). Instead, we measure firm value as the sum of market value of common stock, book value of preferred stock, and book value of debt, and standardize it by total assets, as used in the existing literature on diversification. A preliminary examination of the standardized firm values reveals that their values vary widely, ranging from 0.35 to Jarque-Bera test statistics of these standardized firm values reject the null hypothesis of normal distribution. Hence, using the standardized firm values as dependent variable in a regression may cause a serious heteroskedasticity problem. One way to solve this problem would be to exclude these peculiar values from the data set. If the peculiar values are the outcome of firms diversification activities, however, this approach would cause a selection bias and yield a biased estimation of the relation between diversification and firm value. Moreover, since the peculiar values are rather evenly distributed, it is difficult to sort out peculiar ones from the data set. Due to the potential problem with the above approach, we employ an alternative one in which we first rank the standardized firm values according to their magnitude each year, and then assign a numeric value of two to the top 40% of firm values, one to the middle 20%, and zero to the bottom 40%. Hence, according to this classification, firms with a numeric value of two (zero) represent firms with high (low) value in a particular year. Because of the categorical and ordinal nature of the dependent variable, we employ the ordered dependent variable model for estimation using the maximum likelihood method. The ordered dependent variable model can be either probit model or logit model based on the distribution assumptions on residuals, and the estimation results from both models are known to have little difference (see Greene (2000)). We employ 1038

15 Asia-Pacific Journal of Financial Studies (2008) v37 n6 ordered logit model for our estimation. Although the ordered logit analysis using ordered dependent variables may not provide exact estimates of the relation between diversification and firm value, it would generate probabilistic directions of the effect of the diversification on firm value. For example, a positive (negative) coefficient for an independent variable would indicate that an increase in the independent variable is more likely to be associated with an increase (decrease) in firm value. The ordered dependent variable model used in our study has the merit of preserving peculiar values for the analysis, which are discarded as outliers in prior studies. 8) 4. Summary Statistics of Key Variables and Diversification Indexes 4.1 Summary Statistics of Key Variables Table 1 presents summary statistics of key variables and results from comparing key variables by two subperiods of pre- and post-korean financial crisis. Firm values standardized by total assets (FV) vary in a wide range of to , whose finding supports our rationale of using the ordered dependent variable model. A typical Korean firm in our sample spends on average less than 1% of its revenue on R&D (RNDPS), has an average debt ratio (LEV) of 70.3%. The largest shareholder in a Korean firm owns an average ownership (OWN) of 26.5% during our sample period. Comparing variables between pre- and post-crisis periods, we find that both mean and median values of FV, RNDPS, and LEV for sample Korean firms are significantly (at least at the 10% level in case of t-test for RNDPS) higher for the pre-crisis period than for the post-crisis period. This reflects that Korean firms experience significant declines in their values, spend significantly less in R&D investments, and have significantly lower leverage ratios following the Korean financial crisis. On the contrary, both mean and median values of SRSTD, LASST, and OWN are significantly (at least at the 5% level in case of chi-square test for LASST) higher for the post-crisis period than for the pre-crisis period, indicating that the asset size, level of risk, and owner- 8) We also perform ordinary least squres regression analysis using the raw values of standardized firm values for the purpose of robustness check. The results are reported in Table 10, which are qualitatively the same as those from the ordered logit models. 1039

16 Corporate Diversification, Relatedness, and Firm Value ship concentration of Korean firms increase significantly following the financial crisis. Table 1. Summary Statistics of Key Varibles by Periods The sample consists of 2,894 firm-years for the period Full period is from 1994 to 2000, pre-crisis period is from 1994 to 1996, and post-crisis period is from 1998 to FV stands for firm value measured as the sum of book value of debt, book value of preferred stock, and market value of equity standardized by total assets. LASST is natural log of total assets, RNDPS is R&D expenditures divided by sales, SRSTD is standard deviation of daily stock returns, OWN is the ownership of largest shareholder, and LEV is leverage ratio measured by total debt divided by total assets. Means are given with medians below in parentheses except for minimum and maximum values for full period and t-statistics and chi-square statistics given for differences. t- and χ 2 -statistics are to test for differences in means and medians, respectively, between pre- and post-crisis periods. Full Peirod Pre-Crisis Period Post-Crisis Period Difference (Pre-Post) Minimum (Maximum) Mean (Median) Mean (Median) Mean (Median) t-statistic (χ 2 -statistic) FV (16.414) (0.981) (1.037) (0.859) *** (0.178 *** ) LASST (10.809) (8.143) (8.108) (8.273) *** ( ** ) RNDPS (0.611) (0.000) (0.0003) (0.0) * (0.005 *** ) SRSTD (0.190) (0.039) (0.026) (0.053) *** ( *** ) OWN (0.890) (0.256) (0.242) (0.277) *** ( *** ) LEV (15.271) (0.640) (0.723) (0.675) *** (0.048 *** ) No. of Obs. 2,894 1,326 1,146 Note) ***, **, * denote significance at the 1%, 5%, 10% level, respectively. Our results are in general consistent with the implications of previous studies that the Korean financial crisis has affected negatively the value of Korean firms as well as other firm characteristics. 4.2 Sample Distributions and Characteristics of Diversification Index Measures Table 2 reports the distribution of sample firms by year and descriptive statistics of 1040

17 Asia-Pacific Journal of Financial Studies (2008) v37 n6 three diversification index measures used in our paper. CDX follows the Caves index (1980, pp ) and considers up to three-digit KSIC codes. CDXU computes the index within two-digit KSIC codes only, and CDXR computes the index by selecting an industry with the biggest sales volume within two-digit KSIC industries and then estimating the diversification index of diversification within the three-digit KSIC codes among them. Table 2. Descriptive Statistics of Diversification Indices CDX follows Caves, Porter, Spence, and Scott (1980, pp ) which considers KSIC threedigits. CDXU computes the index within KSIC two-digits only, and CDXR computes the index by selecting an industry with the biggest sales volume within two-digit industries and then estimating index of diversification within three-digits among them. Years Total No. of Firms No. of Firms CDX CDXU CDXR Mean Median No. of Firms Mean Median No. of Firms Mean Median ,146 1,326 2, , , As shown in the CDX measure, for the entire sample period of , 59.4% (1,718/2,894) of sample firm-years engage in diversification activities, encompassing an average of 25.7% of total annual sales. Among firm-years that are diversified during the same period, about 73.5% (1,263/1,718) engage in unrelated diversification across different industries by diversifying 9.9% of total annual sales within the twodigit KSlC level. In contrast, about 40.7% (699/1,718) of firm-years that are diversified engage in related, intra-industry diversification by diversifying 4.5% of total annual sales, as evidenced by the results on CDXR. Thus, it is confirmed that Korean firms usually depend on unrelated diversification for growth instead of related diversification. Although not reported explicitly in Table 2, 244 firms, or 13% of firm-years that are diversified, engage in both unrelated and related diversification during the periods. 1041

18 Corporate Diversification, Relatedness, and Firm Value Table 2 also shows contrasting evidence on the diversification activities of Korean manufacturing firms before and after the Korean financial crisis that broke out in late In terms of average sales involved in diversification, both CDX and CDXU diversification measures increase steadily over the pre-crisis period ( ), reach the highest values of 27.5% and 11.0% for CDX and CDXU, respectively, in 1997, and decline steadily during the post-crisis period ( ). A similar trend is observed for median values of sales for CDX with a peak at 8.9% in These findings are further supported by the evidence that the median value of CDX during the post-crisis period is 0.00%, substantially lower than 6.5% during the pre-crisis period. We see that Korean firms become less inclined to diversification activities after the financial crisis. Overall, the results in Table 2 indicate that a substantially larger number of Korean firms engage in unrelated diversification into different industries in the twodigit KSIC level and that the diversification activities of Korean firms change (decrease) considerably following the Korean financial crisis in late In addition, we find that 40.6% (1,176/2,894) of sample firm-years do not engage in diversification activities. 4.3 Summary Statistics of Standardized Firm Value Table 3 shows summary statistics of the standardized firm value by diversification activity (diversified vs. non-diversified firms) and period (pre-vs. post-crisis period) for the three measures of diversification index. As shown in Panel A where diversification is measured by CDX, diversified firms have significantly (at the 1% level both by t-statistic and z-statistic) lower firm values than non-diversified firms in terms of both mean and median values for the full period, the pre-crisis period and the postcrisis period. Similar results are obtained from the diversification measures of CDXU (panel B) and CDXR (panel C). As shown in Panels B and C, means and medians of the standardized firm value of diversified firms are significantly less than those of non-diversified firms over the full period, regardless of the diversification type. It is also evident that both diversified firms and non-diversified firms suffer a significant value loss following the Korean financial crisis. Hence, the Korean financial crisis affected significantly and negatively the value of both diversified and nondiversified Korean firms. 1042

19 Asia-Pacific Journal of Financial Studies (2008) v37 n6 Table 3. Characteristics of Standardized Firm Values of Korean Firms by Diversification Indices and Periods The sample consists of 2,894 firm-years for the period Full period is from 1994 to 2000, pre-crisis period is from 1994 to 1996, and post-crisis period is from 1998 to CDX follows Caves, Porter, Spence, and Scott (1980, pp ) which considers KSIC three-digits. CDXU is computed within KSIC two digits only, and CDXR is computed by selecting an industry with the biggest sales volume within two-digit industries and then estimating index of diversification within three-digits among them. Firm value is measured as the sum of market value of common stock, book value of preferred stock, and book value of debt, divided by total assets. Means are given with medians below in parentheses. Tests for difference are in means and medians between pre-crisis period and post-crisis period. Full Period ( ) Mean (Median) Pre-Crisis Period ( ) Mean (Median) Post-Crisis Period ( ) Mean (Median) Difference (Pre-Post) t-statistic (z-statistic) Panel A: Diversification vs. Non-diversification Based on CDX Diversified Firms Non-Diversified Firms t-statistic z-statistic (0.971) (1.005) *** *** (1.028) (1.055) *** *** (0.851) (0.870) ** ** (13.45 *** ) 2.20 ** (10.42 *** ) Panel B: Unrelated Diversification (Diversification in KSIC Two Digits) vs. Non-Diversification Based on CDXU Diversified Firms Non-Diversified Firms t-statistic z-statistic (0.973) (0.991) ** (1.020) (1.052) *** *** (0.859) (0.859) (11.24 *** ) 2.30 ** (12.80 *** ) Panel C: Related Diversification (Diversification within Base Industry) vs. Non-diversification Based on CDXR Diversified Firms Non-Diversified Firms t-statistic z-statistic (0.950) (0.995) *** *** (1.027) (1.040) *** *** Note) ***, **, * denote significance at the 1%, 5%, 10% level, respectively (0.825) (0.862) *** *** 3.66 *** (10.07 *** ) 1.51 (13.87 *** ) 1043

20 Corporate Diversification, Relatedness, and Firm Value 5. Regression Analysis and Results 5.1 Pearson Correlation Coefficients Before we perform ordered logit regression analysis of firm value on diversification, we first examine correlation coefficients of variables used in the regression. Table 4 shows Pearson correlation coefficients among FV, three diversification index measures, and five control variables in the regression. The correlation coefficient between CDX and CDXU (measure of unrelated inter-industry diversification) is and significant at the 1% level, indicating that the diversification measure widely used in the previous is indeed the inter-industry, unrelated diversification in substance according to our classification. The correlation coefficient between CDX and CDXR (measure of related intra-industry diversification) is and significant at the 1% level. In contrast, the correlation coefficient between CDXU and CDXR is and not significant at the 10% level, indicating that a firm s unrelated diversification and related diversification are substitute. Table 4. Pearson Correlation Coefficients The sample consists of 2,894 firm-years for the period FV stands firm value (sum of book value of debt, book value of preferred stock, and market value of equity) stadardized by total assets. CDX follows Caves weighted diversification index, which considers KSIC three digits. CDXU measures inter-industry diversification and considers KSIC two digits only. CDXR measures intra-industry diversification and is computed by selecting an industry with the biggest sales volume within two-digit industries and then estimating index of diversification within KSIC three digits among them. LASST is natural logarithm of total assets, RNDPS is R&D expenditures divided by sales, SRSTD is standard deviation of daily stock returns, OWN is the ownership of largest shareholder, LEV is leverage ratio measured by total debt divided by total assets. FV CDX CDXU CDXR LASST RNDPS SRSTD OWN CDX CDXU *** CDXR *** LASST *** *** *** *** RNDPS *** ** *** SRSTD *** *** OWN *** *** *** *** LEV *** * *** *** Note) ***, **, * denote significance at the 1%, 5%, 10% level, respectively. 1044

21 Asia-Pacific Journal of Financial Studies (2008) v37 n6 Firm value standardized by total assets (FV) is not significantly correlated with any of the three diversification index measures. However, its categorized values, though not reported here, have significant and negative (at the 1% level) correlations with all three diversification index measures, implying that firms diversification activities have significant and negative impact on firm value. Firm value is significantly and negatively correlated with firm size (LASST) at the 1% level and ownership concentration (OWN) at the 1% level, but is significantly and positively correlated with R&D activities (RNDPS) at the 1% level, firm risk (SRSTD) at the 1% level, and leverage (LEV) at the 1% level. LASST, a measure of firm size, has a positive and significant (at the 1% level) correlation with the three diversification index measures, implying that firms diversification activities are closely related to their asset size. LASST is also significantly and positively correlated with RNDPS (at the 1% level), but significantly and negatively with firm value (FV) (at the 1% level), SRSTD (at the 1% level), OWN (at the 1% level), and LEV (at the 10% level). Hence, a larger firm tends to show lower standardized firm value, invest more in R&D, have lower risk in terms of stock return volatility, have less concentrated (or more diluted) ownership, and use less leverage. OWN (ownership concentration) is negatively and significantly (at the 1% level) correlated with FV, CDX, CDXR, and LASST. Ownership concentration, however, is not significantly correlated with CDXR (related diversification). The level of leverage (LEV) shows no significant correlation with any of the three measures of diversification index (CDX, CDXU, CDXR). No pairs of variables have correlation coefficients large enough for us to worry about multicollinearity problem. 5.2 Ordered Logit Regression Analysis of Firm Value on Diversification Table 5 presents results from the ordered logit analysis of Korean firms diversification activities on firm value, as estimated by equation (4), for the full and two subperiods with several control variables. The estimated regression coefficients of control variables are similar regardless of the period examined and in general as expected. Firm value (FV) is negatively related to firm size (LASST) but positively to R&D expenditures (RNDPS), ownership concentration (OWN), and financial leverage (LEV), all significant at least at the 5% 1045