Faculty Research BH' I' 1 X SCrm 10

Industry Structure and Value-motivated Conglomeration Timothy R. Burch University of Miami Vikram Nanda University of Michigan M. P. Narayanan University of Michigan September 2002 Contact author: M. P. Narayanan University of Michigan Business School Ann Arbor, MI 48109-1234 Ph: (734) 763-5936 email: mpn@umich.edu

ABSTRACT Recent literature has been largely negative in its assessment of corporate diversification, arguing that diversified firms destroy value and are prone to agency problems. In theory, however, conglomeration can be beneficial and the popular press often discusses conglomeration decisions as being driven by industry conditions. From existing value maximization theories of conglomeration, we identify two industry factors (growth opportunities and industry concentration) that relate to the degree of conglomeration in an industry. Specifically, we argue that value theories imply that both these factors are negatively related to the degree of conglomeration. Using a panel of 50 industries across 20 years, we find these factors help explain conglomeration in the manner predicted by value theories. To measure the benefits of conglomeration on an industry level, we construct "industry excess values" using the excess values of conglomerate firms (i.e., the market values of conglomerates relative to their imputed values from single-segment firms). Our empirical analysis shows that industry excess values are higher when value theories predict conglomeration should be value enhancing. Although our evidence does not rule out the role of agency theories in explaining conglomeration, we conclude that value-maximization also plays an important role.

INDUSTRY STRUCTURE AND VALUE-MOTIVATED CONGLOMERATION I. Introduction As evidenced by the last few decades, a firm's decision to modify the scope of its activities - by engaging in diversifying acquisitions, for instance - can have significant consequences for shareholder value. Decisions affecting the firm's boundaries and focus are, therefore, among the most consequential managers make. These decisions are also among the ones most scrutinized by outside investors. In keeping with its importance for shareholder value and social welfare, firm scope has been actively studied. Research in recent years has produced a substantial body of empirical and theoretical work that investigates the benefits and drawbacks of conglomeration (see Montgomery (1994) and Stein (2001) for a survey of this literature). Empirical studies in the area have tended to analyze the impact of changes in firm diversification on the firm's stock price value, its investment policies and performance. There is a large body of research that investigates whether conglomeration creates value or is driven by agency motives, resulting in inefficient allocation of capital and other resources. These studies typically focus on the individual conglomerate firm and examine the firm's organizational structure and investment policies to understand if conglomeration creates or destroys value through its effect on resource allocation. The impact of industry structure on the extent to which firms in an industry are organized under the conglomerate form, however, has received much less attention in the academic literature. The influence of industry factors becomes apparent when we consider the intertemporal variation in conglomeration in various industries. It is well known that while conglomeration in the U.S. increased during the sixties and seventies (Servaes (1996)), firms have become increasingly focused on their core activities during the eighties and nineties. Comment and Jarrell (1995) document, for instance, that about 56% of exchange-listed firms in 1988 were in only one industry segment, compared to only 38% in 1979. Consistent with this pattern, evidence in the paper indicates that, on average, industries experienced a decrease in conglomeration. The interesting observation, however, is that despite the average trend the picture at the industry level is more complex, with considerable variation in the pattern of conglomeration across industries. This is apparent from Figure 1, which plots the degree of

2 conglomeration in four selected industries over the period 1978-1997. As indicated, while the degree of conglomeration in the measuring and control devices industry follows the general pattern and steadily declines over this period, the same cannot be said about the other three industries in Figure 1. In the telephone communication industry for instance, the degree of conglomeration generally increases over much of this period, while despite some early fluctuations, there is little overall change in the computer and office equipment industry. In contrast, the motion picture production and services industry exhibits wild swings but a net decline in the degree of conglomeration. While relatively little attention has been paid to the role of industry structure in the academic literature on conglomeration, and to the degree of conglomeration in entire industries, it is commonplace for the popular press to discuss diversifying or refocusing actions by firms in terms of industry opportunities and technological and strategic considerations. For instance, several mergers between telecommunication firms and cable TV companies or between producers of entertainment products and cable and satellite companies have been analyzed in terms of a convergence in technology across the industries (see Chalm-Olmstead (1998), for example). The implication of such analysis is that there are industry specific conditions that favor conglomeration in an industry, i.e., conditions under which fewer of an industry's units are likely to be organized as stand alone firms rather than as conglomerate divisions. The question that arises then is whether conglomeration is a value-creating activity when such favorable industry conditions exist. Our empirical approach allows us to address this question. Our first goal in this paper is to identify industry factors that favor conglomeration from a shareholder value perspective. We do so by using existing theories that provide a valuecreating rationale for conglomeration, and then developing predictions regarding the impact of industry level factors on the degree of conglomeration in an industry. We focus on two primary factors that are predicted by such theories to influence the degree of conglomeration in an industry: the growth opportunities available to the firms in the industry, and the competitive environment in the industry, specifically, the degree concentration in the industry (i.e., the extent to which a few players dominate the industry). As we discuss in more detail later, value theories predict a lower degree of conglomeration in industries with higher growth opportunities and in those that are more concentrated. These implications are tested using panel

3 data for fifty of the largest industries in the U.S. over the twenty-year period 1978-1997. The median market-to-book value of stand-alone firms in an industry serves as our proxy for the industry's growth opportunities, while an asset-based Herfindahl index (using assets of both conglomerate divisions and stand-alone firms) is used to measure the extent to which an industry is concentrated. The extent of conglomeration in an industry is measured by the conglomeration divisions in an industry as a proportion of the total units (conglomerate divisions plus single-segment firms) in that industry. The empirical results indicate that the two industry factors are significant predictors of the level of conglomeration in the manner predicted by the value theories of conglomeration. Our second goal is to investigate whether conglomeration increases value when, on the basis of the industry factors considered, conglomeration is predicted to be value enhancing. This allows us to probe the long-standing question of whether conglomeration can enhance firm value in the right environment, in spite of recent evidence that conglomeration may not be beneficial on average. Addressing this issue empirically requires a measure of the extent to which conditions favor conglomeration from a shareholder value perspective. For this purpose we use the conglomeration levels predicted by the two industry factors (growth opportunities and industry concentration). If our interpretation of the role of these factors in value-driven conglomeration is correct, then when these industry factors predict high levels of conglomeration we should find that conglomerates operating in the industry are relatively more valuable. To measure the relative value of conglomerates operating in a particular industry, we introduce a metric called "industry excess value" that is derived from the excess values of individual conglomerate firm operating in the industry. (The excess value of an individual conglomerate is defined in the usual way, as the market value of the conglomerate relative to the sum of its divisions' values imputed from single-segment firms). We find that predicted conglomeration levels and industry excess values are strongly and positively related. Thus, conditions predicted by value theories to be favorable to conglomeration are also conditions that significantly enhance the relative benefits from conglomeration. It appears that there are indeed environments where conglomeration is value-increasing (or at least less valuedestroying), as predicted by value theories of conglomeration.

4 These results help to explain varying conglomeration levels across industries and over time. The results imply that conglomeration increases in industries that face a reduction in growth opportunities and concentration. More importantly, they imply that when conglomeration decisions are driven by changes in growth opportunities and concentration in the industry, they increase firm value on average. Collectively, these results prescribe some of the conditions under which conglomeration can add value (or destroy less value). While these results provide support for value theories of conglomeration, they do not refute the theories that claim that agency problems can result in value-destroying conglomeration. Our tests focus on how industry excess values change with the degree of conglomeration predicted by the industry factors of growth opportunities and concentration. They do not deal with the magnitude of excess value which can be positive or negative. Thus, it is very likely that agency motives and value-creation motives coexist, since the existing literature provides substantial evidence regarding the role of agency factors in conglomeration. Our findings, however, suggest that value maximization appears to play a role in at least some conglomeration decisions as well. The paper is organized as follows. Section II provides an overview of the various valuemaximization theories of conglomeration and develops predictions for what they imply for the relation between conglomeration levels and industry factors. Section III explains the construction of the data and provides definitions of the variables. Section IV discusses the results. Section V considers whether agency theories can explain the relation between the industry structure and the degree of conglomeration. It also explores the extent to which agency theories may explain industry excess values. Section VI concludes. II. Value theories of conglomeration A. Review of value theories In this section we discuss the prevailing value theories of conglomeration. There are three theories that fall under this rubric: market power theories, the resource hypothesis, and internal capital markets theories. The first two theories assume that managers are unconditional shareholder value maximizers and that there are no moral hazard problems between

5 shareholders and managers. These two theories depend on some form of market friction or failure that is not based on agency problems. By contrast, internal capital market theories regard the decision to diversify as a value-maximizing response to the conditions arising out of an agency problem. The agency problem is not specific to conglomerates, however, and the conglomeration decision is not divergent from shareholder interests. Therefore, we classify internal capital market theories as value theories, not agency theories. The key is that the conglomeration decision is made in order to maximize firm value and mitigate the negative effects of an agency problem. Market power theories argue that conglomerates exercise market power through several channels. For instance, they might employ predatory pricing tactics in one market using the profits from another market (i.e., the cross-subsidization or "deep pockets" argument).' Or, conglomerates that meet in multiple markets can tacitly collude by competing less vigorously with each other to create spheres of influence in specific markets (Bernheim and Whinston (1990)). The resource hypothesis argues that excess capacity in production factors leads to diversification if the production factors cannot be sold off at value (Penrose (1959) and Teece (1980, 1982)). For example, if a firm possesses indivisible physical resources beyond the optimal need for one product line, it can employ them in other product lines. In some cases, the excess capacity arises in human resources because managers or workers become more efficient through learning. In some other cases, organizational knowledge that is accumulated in the process of developing a product or process can be used effectively for other products. As Teece (1982) points out, in all these cases there must exist a market failure that prevents the transfer of these excess production resources to other parties. Models by Matsusaka (2001) and Maksimovic and Phillips (2002) are in the spirit of the resource hypothesis. Matsusaka (2001) argues that firms with broad organizational capabilities use diversification as part of a dynamic 1 See Bolton and Scharfstein (1990) for a model on how predatory pricing can drive a rival out.

6 value-maximizing strategy to seek matches for their capabilities. Maksimovic and Phillips (2002) suggest that firms optimally shift finite organizational capabilities to a more productive activity when the diminishing returns to scale reduce the productivity in their primary activity. Internal capital market theories argue that a conglomerate structure improves the allocation of investment capital. The common feature is that conglomeration provides management with an option to deploy capital to its best uses. This option is valuable because the external capital market, due to some market imperfection, is unable or unwilling to substitute for the internal market.2 Williamson (1975) argues that external capital markets can sometimes fail and that the internal capital market created by a conglomerate structure can increase allocational efficiency. In Stein (1997), for example, external markets impose a capital constraint on all projects (whether stand-alone or grouped as a conglomerate) because managers are reluctant to return excess cash to shareholders. Headquarters, having better information about project profitability, can allocate the limited capital more efficiently among the firm's projects if the projects are grouped together in a conglomerate structure. Fluck and Lynch (1999) suggest that conglomerates allow marginally profitable projects to obtain funding and survive a period of distress.3 B. Empirical evidence on value theories There is a large body of work that investigates whether or not conglomeration adds value. The initial evidence seems to support the notion that conglomerate segments are valued less than "equivalent" single-segment firms. Lang and Stulz (1994) and Berger and Ofek (1995) find that conglomerates trade at a discount of about 15% relative to a portfolio of 2 It is interesting to note that the internal capital markets theory can be viewed as a variant of the resource hypothesis, wherein the resource in excess is capital. Some of the limited capital available to managers might become excessive if external events make the investments in one segment unprofitable. Since the managers have no desire to return it to investors, they seek other avenues of investment to maximize shareholder value. 3 Aron (1988) argues that diversification can enhance shareholder value by mitigating moral hazard as multiple segments provide correlated signals of managerial effort. This theory also falls under this class since conglomeration is a value-enhancing response to a moral hazard problem prevalent in all firms.

7 median single-segment firms in the same industries. In addition, there is some evidence (see Lamont (1997) and Shin and Stulz (1998)) that one of the causes of the value differential might be the inefficiency of internal capital markets in conglomerates that misallocate capital among their segments. Rajan, Servaes, and Zingales (2000) and Scharfstein (1998) find evidence that is consistent with misallocation of capital within conglomerates.4 More recent work has questioned the evidence on both conglomerate discount as well as capital misallocation. Villalonga (2001) argues that the conglomerate discount disappears when conglomerate segments are evaluated against a more comparable benchmark using propensity scores, instead of a portfolio of median single-segment firms in the same industries. Maksimovic and Phillips (2002), using plant level data, argue that the evidence is consistent with efficient capital allocation in conglomerate. Whited (2001) claims that the observed capital misallocation is an artifact of measurement error arising from the use of Tobin's q to proxy for investment opportunities. Correcting for this error, she finds no evidence of capital misallocation in conglomerates. Lamont and Polk (2002), however, argue that they continue to find evidence consistent with inefficient investment even after considering the effects of measurement error. Several researchers argue that the diversification discount might reflect characteristics of firms that choose to conglomerate (see Fluck and Lynch (1999), Matsusaka (2001)). The general flavor of these theories is that firms that are inferior in some way choose to conglomerate, which explains their discount. Empirical support for this endogeneity bias is provided by Hyland (1999), Campa and Kedia (2002), Chevalier (2000), and Graham, Lemmon and Wolf (2002). Burch and Nanda (2002), however, study spin-off events and infer the excess 4 There are several papers that attempt to explain the diversification discount and/or the misallocation of capital. Scharfstein and Stein (2000) argue that capital allocation might be the least costly way to bribe divisional managers who are engaging in value decreasing activities. Rajan, Servaes, and Zingales (2000) provide a theory based on divisional managers' incentive to prefer investments that increase the market value of their human capital at the expense of shareholder value. Headquarters can only allocate resources, and it is unable to enforce optimal rules for sharing any divisional surplus. It turns out in their model that if divisions have very diverse resources, there will be suboptimal investment and headquarters tries to improve shareholder value by making resource allocation less diverse. Goel, Nanda, and Narayanan (2002) suggest that managers with career concerns will overallocate capital to the divisions that enhance their reputation the most.

8 value loss from a conglomerate structure (prior to spin-off) and its relation to divisional characteristics. The question of selection bias remains, however, since conglomerates that choose to spin-off are presumably ones for whom the spin-off creates value. In summary, the issue of whether the conglomerate decision creates value is still under debate. C. Value theories and the relation between industry structure and degree of conglomeration As stated in the introduction, the focus of the literature has been on the relative value and allocational efficiency of conglomerates at the individual firm level. Our goal in this paper is to use the value theories to motivate testable hypotheses about how the industry structure should relate to extent to which units in an industry are organized under the conglomeration form. In addition, we also wish to investigate whether there is evidence of beneficial conglomeration at the industry level, in spite of any agency problems that might distract managers from taking value-maximizing actions. We identify two determinants of the degree of conglomeration suggested by the value theories: industry growth opportunities and the degree of concentration. In this section we motivate the implications of the value theories (market power, resource hypothesis, and internal capital markets) for the relationship between each of these variables and the degree of conglomeration. The market power theory suggests that firms in high-growth industries, with sufficient value creation opportunities, have little incentive to operate under a conglomerate umbrella in order to use the deep pockets of another segment to fund predatory pricing tactics. When growth opportunities abound in an industry, rents are easier to obtain, and the need to engage in predatory pricing is diminished. Therefore, companies in high-growth industries are unlikely to acquire segments with deep pockets and, in general, there is little need to conglomerate. The market power theory's predicted relation between conglomeration and industry concentration is also negative, with a similar rationale. Firms that have already obtained market power have less of a need to obtain it through predatory pricing. The firms without market power, meanwhile,

9 are also unlikely to engage in predatory pricing since the firms with market power, with more resources and stronger market positions, are unlikely to be successfully driven out by such tactics. The resource hypothesis contends that companies expand into other segments in order to employ underutilized resources that cannot be sold. Since it is less likely that there are excess production resources in a high-growth industry, the resource hypothesis implies that the motivation for conglomeration is diminished in high-growth industries. The implication of the resource hypothesis regarding industry concentration is more subtle. One might argue that in concentrated industries, the productive resources are likely to be more fully deployed and, therefore, there is less need for diversification to deploy these assets. By itself, however, the resource hypothesis does not lead to a strong prediction regarding conglomeration and industry concentration. The internal capital market theories are based on the idea that conglomeration provides management an option to direct capital to more productive sources. Such an option is not as valuable when the industry has strong growth prospects. This is because growth industries have a greater need for capital and are less likely to face the need to reallocate capital. In a concentrated industry, the value of such an option will be lower as well, as the need to exercise the option arises less frequently. The reason is that there is less need, on average, to switch capital from segments operating in concentrated industries since the profitability of those segments, on average, is likely to be high (see Shepherd (1990) for evidence on the positive relation between concentration and profitability). Therefore, the internal capital market theories also imply a negative relation between concentration and conglomeration. In summary, the various value theories suggest that in industries with more growth or investment opportunities, there will be less conglomeration. To varying degrees, all of them also suggest that in more concentrated industries there will be less conglomeration. These predictions are stated in the following empirical hypothesis:

10 HI: Value theories imply that the conglomeration level in an industry will be negatively related to both its growth opportunities and its degree of concentration. D. Role of industry structure in value-creating conglomeration Thus far, any empirical support for Hypothesis (HI) provides only partial confirmation of the role of value theories in the conglomeration decision. This is because it is possible that agency and other non-value explanations have the same predictions regarding the relationship between the industry factors we examine and conglomeration levels. (We discuss the extent to which agency theories are likely to have the same predictions in a later section). To more convincingly conclude that value theories play a role in conglomeration, we need evidence that conglomeration actually results in value improvements when the value theories suggest it should do so. If Hypothesis (HI) is supported, then we can interpret predicted conglomeration levels (as predicted by growth opportunities and industry concentration) as proxying for the extent to which conditions suggest conglomeration should be value enhancing. This leads us to our second hypothesis: H2: Excess values of conglomerates operating in an industry should be positively related to the conglomeration level that is predicted by industry growth opportunities and industry concentration. It is worth noting that our discussion about the implications of the value theories has focused on the impact of various factors on the degree of conglomeration of a representative industry. Consider, for instance, the argument that value theories imply a negative relationship between growth opportunities and degree of conglomeration in an industry. If some firms in an industry with reduced growth opportunities choose to diversify, this increases the degree of conglomeration not only in the industry under consideration, but also in the industries into which these firms have diversified. One might, therefore, wonder if the degree of

11 conglomeration in an industry is influenced by not only conditions in that industry but also by conditions in other industries. We claim that this is likely to be a second-order effect. This is because when a firm from an industry diversifies because of fewer growth opportunities, it is more likely to merge with another firm from a different industry which is also interested in diversifying for any of the reasons for conglomeration we have detailed. Such a merger would lead to greater combined benefits from the merger. Hence, both industries are likely to have some of the conditions that promote conglomeration according to the value theories. III. Data sources and construction of variables A. Data sources and choice of industries We use the Compustat Industrial Segment (CIS) database for divisional data and the Compustat annual industrial database for single-segment firm data for the years 1978-1997. Both active and research data are used to avoid survivorship bias. Our intent is to test the hypothesis using the 50 largest U.S. industries, based on the number of market participants (single-segment firms plus divisions of conglomerates in an industry) in 1988.5 Following the usual practice in the literature, we eliminate financial services and regulated utility industries. We also eliminate all industries firms that have more than one missing necessary data item over the twenty-year period. We then select the top 50 industries in terms total number of market participants. Excluding the financial services and regulated utilities industries, this results in a sample consisting of 50 of the largest 79 industries (according to the total number of conglomerate divisions and stand-alone firms operating in the industry). In four of the 50 industries there is one missing data point. We relax our variable construction requirements for these four cases, and these adjustments are described in detail in the subsections that follow. 5 This year is chosen as a mid-way point during the time period we study. Results are robust to using alternative years to determine the 50 industries. Using many more than 50 industries results in missing data which is problematic for the panel data approach we employ.

12 In selecting the 50 industries, the number of market participants is determined as follows. We first count the total number of divisions and single-segment firms operating in each 3-digit SIC industry in 1988. To qualify as a single-segment firm in 1988, the firm must not have multiple divisions (as reported in the CIS database) during this year, and must have valid assets or valid sales. To qualify as a division in 1988, a candidate's parent firm must have multiple divisions reported in this year.6 Thus, all divisions of multi-divisional parent firms in our CIS database are included for counting purposes. B. Variable construction B]. Degree of conglomeration (Cong) This variable represents the extent to which units in an industry are under the conglomerate structure and is measured as the number of conglomerate divisions in an industry divided by the sum of the number of conglomerate divisions and the number of single-segment firms in the industry. Conglomerate divisions and single-segment firms are counted using the same methodology outlined above for industry selection. B2. Industry median market-to-book (IndMB) We measure growth opportunities by the median industry market-to-book ratio. The market-to-book ratio of a firm is defined as the market value of the firm (market value of common stock plus book value of long-term debt and current liabilities plus book value of preferred stock) divided by the book value of its assets. This ratio is calculated for the universe of single-segment firms in the Compustat annual database in the 50, three-digit SIC industries and the median is calculated for each industry in each year. We require that an industry have at 6 All of the divisions in our divisional data (for conglomerate firms) have positive 1988 sales, so we are not concerned that our sample includes invalid divisions.

13 least five valid ratios, except for one industry year (out of 1,000 total), where this restriction is relaxed so a median can be calculated. B3. Industry Herfindahl index (IndHerf) We measure industry concentration with an asset-based Herfindahl index. Because we wish to account for all players in the industry, both conglomerate divisions and single-segment firms in the industry are used. IndHerfis calculated as follows: IndHerf = -1 i ) i=l i ) where Ai is the book value of assets of the single-segment firm or the conglomerate division i operating in the industry, and n is the total number of single-segment firms and conglomerate divisions in the industry. B4. Industry excess values (Weighted-EV and OLS-EV) We introduce a metric called "industry excess value" to measure the relative values of conglomerates operating in an industry. To begin, we essentially follow Berger and Ofek (1995) in constructing conglomerate excess values (CEVs), defined as follows: CEV = In CM L DAi[INDi,(V /A)] i=l where, CMV= market value of common equity, plus book value of debt of the conglomerate, plus book value of preferred stock, DAi = asset size for Division i,

14 INDi(V/A) = median ratio of total capital (market value of common equity, plus book value of debt, plus book value of preferred stock) to assets for single-segment firms in the 3-digit SIC industry of Division I, and, n = number of divisions in the conglomerate firm. Unlike in Berger and Ofek, CMV includes preferred stock. Following Berger and Ofek, industry medians are taken from the narrowest SIC grouping that includes at least five singlesegment firms with sufficient data for computing the ratio. We also use their methodology in grossing-up divisional assets and the elimination of extreme excess values.7 Weighted-EV for an industry is calculated by taking a weighted average of CEVs for firms with segments operating in the industry. To illustrate, suppose there are only two conglomerate firms X and Y, with conglomerate excess values CEVx and CEVy, each with divisions operating in "industry 1." Suppose that conglomerate Xhas a total assets of 200, with 150 allocated in industry 1, and that conglomerate Yhas total assets of 1000, with 300 allocated to industry 2. The excess value for industry 1, Weighted-EVi, is calculated as follows: (150> (300> CEVx x ) + CEVy xI Xe y200) l1000) Weighted - EV1 =( (150 + ( 300 K200 1000) In the above measure, the weight given to the excess value of a conglomerate in calculating the excess value of an industry is directly related to what fraction of the conglomerate assets are devoted to that industry. The notion behind this construction is that a division that represents a larger fraction of a conglomerate has a greater impact on its excess 7 Berger and Ofek (1995) eliminate conglomerates where the sum of divisional assets deviates from parent firm aggregate assets by more than 25%. They then gross-up divisional assets so their sum equals the parent's aggregate assets. They also avoid extreme excess values by eliminating conglomerates where sum of divisional imputed values (the denominator in the CEV definition) is less than one-fourth or more than four times CMV.

15 value than a smaller division.8 We require that at least five divisions in an industry have valid parent CEV measures in order to compute Weighted-EV. In three of the 1000 industry-years, this condition is not met, and therefore, we ease the restriction on the number of divisions required. Using the same notion as above (that a division that represents a larger fraction of a conglomerate has a greater impact on its excess value), we construct another measure of industry excess value which we denote OLS-EV. This measure uses an ordinary least squares approach (one regression for each year), with the conglomerate excess value as the dependent variable and the relative asset weights of the divisions operating in various industries as the independent variables. The coefficients of each industry are then interpreted as the excess value of that industry. A detailed explanation of the construction of the OLS-EV variable is provided in the Appendix. IV. Results A. Intertemporal and inter-industry patterns in variables Tables 1 and 2 describe the data. In Table 1 we present the variables used for the industries in 1980 and 1995. These years are chosen because they are near the endpoints of our time period. We present Weighted-EV for the industry excess values, since OLS-EVhas fairly similar values and is used mainly as a robustness check on the results. As can be seen in 8 The measure we use takes into account the importance of the industry to a conglomerate and provides an indicator of the excess value of a typical (i.e., randomly-chosen) conglomerate firm with a division operating in a particular industry. Other measures of industry excess value, based on different weighting schemes, can also be computed. For example, we can construct a measure of excess value for an industry by assigning a weight to the excess value of each conglomerate equal to the asset size of each conglomerate's division in that industry relative to the total assets of all divisions (of all conglomerates) in that industry. In the construction of Weighted-EV1 above, the (150/200) weights would be replaced with (150/450) and the (300/1000) weights would be replaced with (300/450). Such a measure provides a value-weighted indicator of industry excess value regardless of the importance of the industry to a particular conglomerate. The results hold using this alternative measure of industry excess value.

16 Table 1, there is considerable variation across both industries and time. For example, in 1980 conglomeration levels range from a low of 17% (in the computer and office equipment industry and also the telephone communication industry) to a high of 87% (in the aircraft and parts industry and also the construction and related machinery industry). There are steep declines in the conglomeration levels of many industries, while others show a smaller decline or a slight increase. For example, the conglomeration level of the drugs industry declines from 68% to 18%, while that for the industrial organic chemicals industry experiences a slight increase from 80% to 83%. There is also considerable variation across industries and time for the other variables we report (IndMB, or industry market-to-book, IndHerf, or industry Herfindahl, and Weighted-EV, or the weighted industry excess value metric). B. Yearly summary statistics Table 2 presents summary statistics for the variables in our study. The trend toward focus during the 1980s and 1990s is quite apparent, as the mean (median) Cong steadily declines from a high of 70% (73%) in 1978 to a low of 44% (44%) in 1997. It is interesting to note that overall, industry concentration levels have not changed dramatically through time (as seen by the means and medians for IndHerf). There are two items of note in regards to industry excess values. First, there is considerable variation from year to year, but there is no noticeable overall pattern through time. Second, comparing Weighted-EV to OLS-EV confirms that the two measures have fairly similar values, at least at the yearly aggregate level as measured by the means and medians across industries. Thus, in spite of their quite different empirical constructions, the two measures seem consistent with each other. We note that the median Weighted-EV and OLS-EV across all years are -0. 13 and -0. 12, respectively. Not surprisingly given their construction, these are of the same order of magnitude as reported previous studies (e.g. Berger and Ofek (1995)).

17 C. Test of Hypothesis (HI) In Panel A of Table 3 we test Hypothesis (HI) and present regressions of conglomeration (Cong) on industry growth opportunities as measured by industry market-tobook values (IndMB), and industry concentration as measured by the industry Herfindahl index (IndHerf). We use a panel data approach to appropriately account for industry and time series effects. Specifically, a two-way fixed effects model is used (i.e., both industry and year dummies are included, although not reported) and we use error terms that are corrected for both autocorrelation and heteroskedasticity. As reported, IndMB and IndHerf are both negatively and significantly related to Cong, as Hypothesis (HI) predicts (see Model Al in Panel A of Table 3). For robustness purposes we also use a Fama-MacBeth approach (Fama and MacBeth (1973)) which involves averaging the coefficients of 20 yearly cross-sectional regressions. With this approach, which is not reported in the table, IndMB and IndHerf are negative and more strongly significant.9 We conclude that Hypothesis (HI) is supported by the data.'0 D. Test of Hypothesis (H2) We now address Hypothesis (H2), i.e., the issue of whether conglomeration in an industry creates value when growth opportunities and concentration in an industry are favorable to conglomeration. The results are reported in Panel B of Table 3, where industry excess value, as measured by Weighted-EV (Model Bl) or OLS-EV (Model B2), is regressed on the predicted conglomeration level, denoted ValueCongl. ValueCongl is obtained by using the model estimated in Panel A to construct a predicted value for each industry year. As can be seen, Models B] and B2 show that ValueCong is positively and significantly related to industry 9 We use Hansen-Hodrick-Newey-West autocorrelation and heteroskedasticity consistent t-values with 5 lags to establish the significance of the coefficients. The Fama-MacBeth approach is an inferior one for our purposes, since it does not control for the differences in industry-wide conglomeration levels (across all 20 years) due to factors we do not examine. The panel data regression approach, by contrast, allows us to control industry effects by explicitly including industry dummy variables via the fixed effects model. 10 We also test Hypothesis (HI) using the value-weighted industry excess value measure discussed in footnote 8. The two variables IndMB and IndHerfare significant and have the predicted negative sign.

18 excess value. In Model B] the coefficient is positive and the t-statistic is 8.33 while in Model B2 the coefficient is positive and the t-statistic is 8.73. We also repeat the analysis in Panel B using a Fama-MacBeth approach (using the 20 cross-sectional models from the Fama-MacBeth version of Panel A mentioned previously, and then averaging the slope coefficients of 20 yearly cross-sectional regressions for models B] and B2). Although not reported in the table, results are qualitatively similar. For example, in Model B] the slope coefficient is positive with a tvalue of 5.42. We conclude that Hypothesis (H2) is strongly supported. E. Robustness of results One potential concern has to do with the common components of our industry excess values and IndMB, leading to a potential "hardwiring" of the results. When single-segment firms in an industry have high market values, conglomerate excess values in that industry (and hence industry excess values) would be lower and IndMB would be higher, holding all else constant. Since ValueCongi is a negatively related to IndMB (recall it equals the predicted conglomeration level from Model Al), it could simply be serving as a noisy proxy for IndMB in the Panel B models. Thus the positive relation between ValueCongl and the measures of industry excess value we find in Panel B of Table 3 could be a result of the relation between IndMB and our excess value measures. We address this concern in three ways and conclude that the main results of the paper are not due to hardwiring. Table 4 reports the robustness tests. First, we use one year lagged values of IndMB and IndHerf(Lag(IndMB) and Lag(IndHerf), respectively) when estimating the model in Panel A to obtain the predicted conglomeration levels (ValueCong2). This also has the intuitive appeal of allowing some delay between changing industry factors and the firms becoming more or less conglomerated in response. The results of this estimation are reported as Model A2 in Table 4. The predicted conglomeration based on Model A2 is denoted by ValueCong2, and is used in a regression to

19 predict Weighted-EV(see Model B3)."1 The results remain highly significant. The t-value for ValueCong2 is slightly lower than that for ValueCongl in Table 3, but it still implies very strong significance with a value of 2.46). This result lends support to our contention that hardwiring is not driving the result. Second, we repeat the analysis by removing IndMB altogether (relying on only IndHerj) as a predictor of the degree of conglomeration (see Model A3 in Table 4). The predicted conglomeration based on Model A3 is denoted by ValueCong3 and is used as a regressor in Model B4. Again we find significant results (the t-value for ValueCong3 is 3.46). This result provides further evidence that hardwiring does not drive our findings. Finally, we repeat the estimation of Model B] of Table 3 (which uses ValueCongl from that table), but we include IndMB directly as a regressor as well. This completely removes the concern that the significance of ValueCong is being caused by its proxying for IndMB. As Model B5 of Table 4 shows, we find that IndMB is negatively and highly significant (t = - 9.92). This is not surprising, since industry excess values (constructed from conglomerate excess values) and single-segment firm market-to-book ratios (whose medians form IndMB) should be inversely related. The key result is that ValueCongl continues to be positive and significant, with a t-value of 3.2 1. Although all of our steps to address the hardwiring concern have an effect on the t-values, the various versions of ValueCong remain highly significant. Thus, we are not concerned that ValueCongl's significance in Table 3 is being driven by the extent to which it proxies for IndMB. We conclude that Hypothesis (H2) is strongly supported by the evidence. V. Alternative explanations: agency models The basis for Hypothesis (H2) is the premise that changes in conglomeration as predicted by the industry factors growth opportunities and concentration will, on average, be " We report only the results with Weighted-EVas the dependent variable. Results with OLS-EVas the dependent variable are similar.

20 value-increasing. The premise is based on Hypothesis (HI), where we have argued that value theories predict that these industry factors are negatively related to the degree of conglomeration in an industry. One might reasonably ask whether our conclusion that the conglomeration decision is partly motivated by value creation is in any way affected if other theories of conglomeration also predict the same negative relationship between the two industry factors and the degree of conglomeration. Because the empirical evidence in Table 3 indicates that industry excess values are positively related to predicted conglomeration, we can conclude that value motivation plays an important role in conglomeration decisions, even if other theories posit the same relationship between industry factors and the degree of conglomeration. Thus we can argue that our conclusions are unaffected even if other theories predict the relationship in hypothesis (HI). Nevertheless, it is reassuring to demonstrate that other theories of conglomeration do not predict a similar relationship between these industry factors and the degree of conglomeration. Alternative explanations of conglomeration (that is, explanations that are not based on value creation) can be grouped under the rubric of agency. In this section, we briefly discuss the prevailing agency theories of conglomeration and examine the extent to which these theories may explain the negative relations between the industry factors and observed level of conglomeration. Agency theories rely on some sort of friction or contracting problem that allows managers to engage in activities not entirely consistent with shareholder value maximization. There are three prevailing agency theories we consider that have possible links to conglomeration activity: the free cash flow theory, the managerial entrenchment hypothesis and, the managerial risk-aversion hypothesis. The existing literature presents varying degrees of support for the ability of these explanations to explain conglomeration activity.12 We briefly 12 The empirical literature on testing the agency motive for conglomeration has focused on the relationship between managerial ownership and diversification (see Aggarwal and Samwick (2001), Amihud and Lev (1981), Denis, Denis, and Sarin (1997), and May (1995)). The results are mixed.

21 describe the theories and investigate their implications for the relationship between the degree of conglomeration and the two industry variables we use, growth opportunities and concentration. A. Free cashflow theory A classic example of an agency theory is the free cash flow theory proposed by Jensen (1986). This theory argues that managers prefer to not return free cash flow to shareholders, and will instead spend it to control more resources to obtain the higher pay and prestige associated with managing larger firms. Because acquisitions are one way to spend free cash flow, the theory implies that that managers of firms with excess free cash flow will engage in "low-benefit or even value-destroying mergers." Jensen notes that diversification programs generally fall into this mold. Further, the arguments in the free cash flow theory suggest that firms in industries with low growth opportunities may be more inclined to make acquisitions since they have fewer profitable projects in which to invest. Thus, the negative relation between conglomeration and growth opportunities is consistent with Jensen's free cash flow theory. It is not clear, however, if the negative relation between conglomeration and industry concentration is consistent with the free cash flow theory. According to the free cash flow hypothesis, companies in concentrated industries are likely to be more diversified since they are likely to have more free cash flow (i.e., the conglomeration is predicted to be positively related to industry concentration). When both predictions are considered, it is hard to argue that the free cash flow theory explains the results presented in Table 3. B. Managerial entrenchment hypothesis Shleifer and Vishny (1989) discuss a managerial entrenchment hypothesis in which a manager has the incentive to invest in assets whose value would be lower under the next best manager, even if such investments do not enhance shareholder value. Once made, if such investments are costly to reverse the manager will be particularly valuable to shareholders.

22 Such entrenchment is valuable to the managers since compensation is an increasing function of the incremental value the manger brings to the firm. According to this theory, diversification strategies can emerge as a way for a manager to increase the dependence of the firm on his or her skills. This hypothesis has no clear implication for the relationship between the degree of conglomeration and either growth opportunities or industry concentration and is not helpful in explaining why conglomeration is negatively related to these factors.'3 C. Managerial risk aversion hypothesis Amihud and Lev (1981) present a managerial risk-aversion hypothesis in which riskaverse managers wish to manage the risk associated with their human capital. Since the risk associated with managers' income is related to firm risk, ceteris paribus, managers have an incentive to engage in conglomerate mergers to reduce firm risk even if it is not in the interest of shareholders to do so. As with the managerial entrenchment hypothesis, the managerial riskaversion hypothesis does not provide clear predictions regarding growth opportunities and industry concentration - and, hence, is not helpful to understanding the empirical findings. D. Agency theories and industry excess values On the whole, the agency explanations are unlikely to account for the negative relations between conglomeration and industry growth opportunities and concentration. This lends credence to our interpretation that value theories explain these findings. It is not the case, however, that our results preclude the existence of agency-motivated conglomeration. We now present indirect evidence that is perfectly consistent with agency issues playing a role in conglomeration decisions. 13 A variation of this theme is offered by Gibbons and Murphy (1992). They argue that managers may diversify the firm to enhance their value in the in the outside labor market because the ability to manage more complex organizations may be perceived as a valuable skill.

23 First, we note from Table 1 that the excess values of most industries are negative in the years of 1980 and 1995. Second, from Table 2 we can see that the average excess value across all industries is negative in each year of the sample. These statistics point to the possibility that conglomerates destroy value and that agency motives dominate the diversification strategy of firms.14 It must be noted that the presence of agency motives is entirely consistent with our earlier results that support value theories. These two motives are not mutually exclusive. To explore the role of agency, we extend models B] and B2 from Table 3 to include the unexplained portion of conglomeration in Model Al as an additional independent variable. This additional variable, ResidCongi, equals Cong - ValueCongl and captures both agency-driven conglomeration and any residual conglomeration explained by value theories but not captured by the two industry factors we use (growth opportunities and concentration). These results are reported in Table 5 (see Model B6). Again, we only report results using Weighted-EV since those using OLS-EV are similar. As can be seen from the table, ResidCongI is negative and insignificant, while ValueCongl is not materially affected. We also include the corresponding residuals for the alternative predicted conglomeration levels (ValueCong2 and ValueCong3) and re-estimate Models B3 and B4 from Table 4. These results are reported in Models B7 and B8 of Table 5, and once again the predicted conglomeration levels remain positive and significant while the residuals are insignificant. One interpretation of the insignificance of the residuals is that both agency motives and unspecified value-based factors are present and offset one other to some extent. As a further test of the presence of agency motives, we regress industry excess value on Cong, the observed conglomeration level. The results are shown in Model B9 of Table 5. As shown, there is no significant relation between Weighted-EV and Cong, and the result is similar if we use OLS-EV (the R-squared value is due to the inclusion of the year and industry 14 As discussed earlier, however, value destruction is only one of the explanations for negative excess values. A selection bias, for example, has also been suggested as an explanation.

24 dummies used in the two-way fixed effect approach). If conglomeration is predominantly driven by value-maximization (agency) explanations, then we should expect to see industry excess value positively (negatively) related to conglomeration levels. Hence, one interpretation of the result in Model B9 is that conglomeration takes place due to both value and agency reasons. In other words, the insignificance of Cong is consistent with value and agencymotivated conglomeration playing offsetting roles. VI. Conclusion The literature has been largely negative in its view of conglomeration. Most empirical studies indicate that, on average, conglomerates trade at a substantial discount to singlesegment firms. This has been interpreted as consistent with agency theories explaining conglomeration decisions. In theory, however, conglomeration can be beneficial and driven by value-maximizing goals, and managers and the press often discuss conglomeration decisions in the context of industry conditions. In this paper we find evidence that the degree of conglomeration in an industry is negatively related to the industry factors of growth opportunities and concentration, as predicted by value theories. Furthermore, we find that when changes in these factors favor the placement of an increasing fraction of units in an industry under the conglomerate structure, the excess value of conglomeratones in that industry increases. These results suggest that value motives can at least partially explain the conglomeration decision. While our results provide some support for value-motivated conglomeration, they do not rule out agency as a motive for conglomeration.

25 References Aggarwal, R. and A. Samwick, 2002, "Why Do Managers Diversify Their Firms? Agency Reconsidered," forthcoming, Journal of Finance. Amihud, Y. and B. Lev, 1981, "Risk reduction as a managerial motive for conglomerate mergers," Bell Journal of Economics 12, 605-617. Aron, D., 1988, "Ability, moral hazard, firm size, and diversification," Rand Journal of Economics 19 (Spring), 72-87. Berger, P. and E. Ofek, 1995, "Diversification's effect on firm value," Journal of Financial Economics 37, 39-65. Bernheim D. and M. Whinston, 1990, "Multi-market Contact and collusive behavior," Rand Journal of Economics 21, 1-26. Bolton, P., and D. Scharfstein, 1990, "A theory of predation based on agency problems in financial contracting," American Economic Review 80, 94-106. Burch, T. and V. Nanda, 2002, "Divisional diversity and the conglomerate discount: The evidence from spinoffs," forthcoming, Journal of Financial Economics. Campa, J. and S. Kedia, 2002, "Explaining the diversification discount,", Journal of Finance 57, 1731-1762. Chalm-Olmstead, S, 1998, "Mergers, acquisitions, and convergence: The strategic alliances of broadcasting, cable television, and telephone services," Journal of Media Economics 11, 33-46. Chevalier, J., 2000, "Why do firms undertake diversifying mergers? An analysis of the investment policies of merging firms," Working paper, University of Chicago, Chicago, IL. Comment, R. and G. Jarrell, 1995, "Corporate focus and stock returns," Journal of Financial Economics 37, 67-87. Denis, D. D. Denis, and A. Sarin, 1997, "Agency problems, equity ownership, and corporate diversification," Journal of Finance 52, 135-160. Fama, E. and J. MacBeth, 1973, "Risk, return, and equilibrium: Empirical tests," Journal of Political Economy 71, 607-636.

26 Fluck, Z. and A. Lynch, 1999, "Why do firms merge and then divest? A theory of financial synergy," Journal of Business 72, 319-46. Gibbons, R. and K. Murphy, 1992, "Optimal incentive contracts in the presence of career concerns: Theory and evidence," Journal of Political Economy 100, 468-505. Graham, J., M. Lemmon, and J. Wolf, 2002, "Does corporate diversification destroy value?" Journal of Finance 57, 695-720.. Goel, A., V. Nanda, and M. P. Narayanan, 2002, "Career concerns and resource allocations in conglomerates," Working paper, University of Michigan, Ann Arbor, MI. Hyland, D., 1999, "Why firms diversify: An empirical examination," Working paper, University of Texas at Arlington, TX. Jensen, M., 1986, "Agency costs of free cash flow, corporate finance and takeovers," American Economic Review 76, 323-329. Lamont, 0., 1997, "Cash flow and investment: Internal capital markets evidence," Journal of Finance 52, 83-109. Lamont, 0. and C. Polk, 2002, "Does diversification destroy value? Evidence from industry shocks," Journal of Financial Economics 63, 51-77. Lang, L. and R. Stulz, 1994, "Tobin's q, corporate diversification and firm performance," Journal of Political Economy 102, 1248-1280. Maksimovic, V. and G. Phillips, 2002, "Do conglomerate firms allocate resources inefficiently across industries? Theory and evidence," Journal of Finance 57, 721-767. Matsusaka, J., 2001, "Corporate diversification, value maximization, and organizational capabilities," Journal of Business 74, 409-431. May, D., 1995, "Do managerial motives influence firm risk-reduction strategies," Journal of Finance 50, 1291-1308. Montgomery, C., 1994, "Corporate diversification," Journal of Economic Perspectives 8, 163 -178. Penrose, E. (1959). The Theory of the growth of the firm, Oxford, Basil Blackwell. Rajan, R., H. Servaes, and L. Zingales, 2000, "The cost of diversity: The diversification discount and inefficient investment," Journal of Finance, 55:1, 35-80.. Scharfstein, D., 1998, "The dark side of internal capital markets II: Evidence from diversified conglomerates," NBER Working paper.

27 Scharfstein, D., and J. Stein, 2000, "The dark side of internal capital markets: Divisional rentseeking and inefficient investment", forthcoming, Journal of Finance. Servaes, H., 1996, "The value of diversification during the conglomerate merger wave," Journal of Finance 51, 1201-1225. Shepherd, W., 1990, "The economics of industrial organization," Third edition, Prentice Hall, Englewood, N. Shin, H., and R. Stulz, 1998, "Are internal capital markets efficient", Quarterly Journal of Economics 113, 531-552. Shleifer, A. and R. Vishny, 1989, "Managerial entrenchment: The case of manger-specific investments," Journal of Financial Economics, 17, 293-309. Stein, J., 1997, "Internal capital markets and the competition for corporate resources," Journal of Finance 52, 111-133. Stein, J., 2001, "Agency, investment and corporate investment," forthcoming in Handbook of the economics offinance, edited by G. Constantinides, M. Harris and R. Stulz. Teece, D., (1980) "Economics of scope and the scope of the enterprise," Journal of Economic Behavior and Organization 1, 223-247. Teece. D., (1982), "Towards an economic theory of the multiproduct firm," Journal of Economic Behavior and Organization 3, 39-63. Villalonga, B., 2001, "Does diversification cause the 'diversification discount'?," Working paper, Harvard University, Boston, MA. Whited, T., 2001 "Is it inefficient investment that causes the diversification discount?" Journal of Finance 58, 1667-1691. Williamson, 0., 1975, Markets and hierarchies: Analysis and antitrust implications (The Free Press, New York, NY.)

28 Appendix Construction of the industry excess value measure OLS-EV This measure uses an ordinary least squares (OLS) regression approach, where a separate regression is performed for each year. To estimate the fifty industry values of OLS-EV for a particular year, only conglomerates with valid divisional assets in one of the 50 industries in that year are retained. Fifty industry variables (Indl-Ind50) are then coded for each division year, where each division's asset weight within the conglomerate is assigned to the corresponding industry variable, and the remaining industry variables are set to zero. The table below provides an illustration using the example listed in the text for Weighted-EV, assuming all of each firm's assets not allocated to industry 1 are instead allocated to industry 2 (i.e. X allocates 50 to industry 2, while Y allocates 700). Regressors Observ Conglo Dependent ation merate variable Indl Ind2 Ind3...Ind50 1 X CEVX (150/200) (50/200) 0... 0 2 Y CEVy (300/1000) (700/1000) 0... 0 When a conglomerate contains a division operating outside of the 50 industries we consider, the weights for the divisions that are among the 50 are rescaled so the sum of the Indl through Ind50 for a given firm will always equal 1. Results are robust, however, if we do not rescale the industry weights. OLS regressions are estimated for each year with no intercept term, and the resulting 50 coefficients are used as each year's 50 industry excess values. The total number of observations in a given year's regression equals the total number of conglomerates with at least one division in the 50 industries. Observation sizes for the 20

29 regressions (one for each year) range from 616 to 1,112, and adjusted R-squared values range from 0.033 to 0.207 (all but two regressions have adjusted R-squared values exceeding 0.06).

30 Figure 1 Conglomeration levels in four selected industries during 1978-1997 Industries are defined at the 3-digit SIC code level. The conglomeration level is the number of conglomerate divisions operating in the industry divided by the sum of the number of conglomerate divisions and stand-alone firms operating in the industry. 0.8 0.7 o 0.6 E 0.5^ - ' 0.4 0 ' 0.3 0.2 & 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 Year

31 Table 1 Industry characteristics in 1980 and 1985 This table reports variable identities for 50 industries in 1980 and 1995. Cong is the conglomeration level, defined as the number of divisions in the industry divided by the total number of divisions and stand-alone firms in the industry. IndMB is the median market-to-book ratio for stand-alone firms in the industry. IndHerf is an asset-based Herfindahl index for conglomerate divisions and stand-alone firms in the industry. Wtd-EVis the industry's excess value using an asset-weighted average method that weights conglomerate excess values by the division's asset weight within the parent firm. 1980 Cong IndMB IndHerf Wtd-EV 1995 Cong IndMB IndHerf Wtd-EV Industry name Aircraft and parts Beverages Blast furnace and basic steel products Commercial printing Communications equipment Computer and data processing services Computer and office equipment Construction and related machinery Crude petroleum and natural gas Drugs Eating and drinking places Electric lighting and wiring equipment Electrical goods Electrical industrial apparatus Electronic components and accessories Fabricated structural metal products General industrial machinery Gold and silver ores Groceries and related products Grocery stores Hotels and motels Industrial organic chemicals Machinery, equipment, and supplies Measuring and controlling devices Medical instruments and supplies Metalworking machinery Misc. amusement, recreation services Misc. electrical equipment & supplies Misc. fabricated metal products Misc. chemical products Misc. durable goods Misc. manufactures Motion picture production & services Motor vehicles and equipment Nonferrous rolling and drawing 0.87 1.41 0.69 0.64 0.82 0.67 0.72 0.64 0.69 1.40 0.47 1.71 0.17 1.54 0.87 0.78 0.58 4.16 0.68 2.65 0.59 0.93 0.80 0.75 0.66 0.62 0.76 1.29 0.62 1.47 0.82 0.79 0.78 0.96 0.46 4.24 0.70 0.60 0.53 0.63 0.67 0.90 0.80 1.81 0.85 1.14 0.57 1.38 0.48 2.11 0.79 0.98 0.54 1.37 0.63 1.92 0.79 1.00 0.79 2.03 0.63 0.68 0.73 0.80 0.64 0.98 0.70 0.70 0.77 1.15 0.06 -0.18 0.06 0.05 0.05 -0.26 0.14 0.02 0.08 -0.10 0.15 -0.28 0.13 -0.46 0.05 0.01 0.03 -0.30 0.04 -0.52 0.06 -0.09 0.12 -0.01 0.09 0.19 0.24 -0.24 0.04 -0.10 0.07 -0.06 0.04 -0.03 0.06 -0.87 0.06 0.07 0.06 -0.09 0.11 -0.01 0.07 -0.64 0.03 -0.21 0.04 -0.24 0.06 -0.31 0.05 -0.16 0.05 -0.22 0.10 -0.55 0.05 -0.18 0.16 -0.41 0.15 -0.34 0.10 -0.03 0.12 -0.09 0.12 -0.22 0.20 -0.44 0.70 0.45 0.51 0.38 0.30 0.16 0.20 0.78 0.50 0.18 0.24 0.61 0.49 0.66 0.31 0.59 0.60 0.21 0.38 0.19 0.44 0.83 0.54 0.30 0.24 0.67 0.35 0.48 0.71 0.59 0.55 0.48 0.46 0.54 0.57 0.71 0.07 0.17 1.50 0.07 0.12 0.82 0.06 0.02 1.26 0.15 -0.11 2.16 0.09 -0.39 2.70 0.14 -0.31 2.16 0.08 -0.57 1.17 0.07 -0.29 1.24 0.04 -0.05 3.46 0.04 -0.41 1.28 0.09 -0.33 1.14 0.13 -0.26 1.20 0.08 -0.16 1.48 0.13 -0.25 1.91 0.05 -0.32 1.21 0.04 -0.23 1.25 0.04 -0.12 1.76 0.04 0.09 1.00 0.23 0.20 0.87 0.04 -0.04 1.08 0.06 0.10 1.39 0.04 -0.06 0.93 0.07 0.05 1.73 0.05 -0.39 2.55 0.05 -0.07 1.07 0.13 -0.26 1.21 0.04 0.11 1.36 0.10 -0.01 0.97 0.07 -0.10 1.51 0.08 0.05 1.31 0.09 -0.03 0.95 0.24 -0.31 1.27 0.13 0.19 1.11 0.09 -0.12 1.09 0.18 -0.05

32 Table 1 (continued) 1980 Cong IndMB IndHerf Wtd-EV 1995 Cong IndMB IndHerf Wtd-EV Industry name Nonstore retailers Oil and gas field services Paper mills Petroleum refining Photographic equipment and supplies Professional & commercial equipment Radio and television broadcasting Refrigeration and service machinery Search and navigation equipment Soap, cleaners, and toilet goods Special industry machinery Telephone communication Toys and sporting goods Trucking & courier services, excl. air Variety stores 0.68 0.76 0.79 2.31 0.81 0.76 0.80 1.00 0.67 1.23 0.43 0.77 0.82 1.37 0.78 0.93 0.70 1.45 0.72 0.96 0.80 0.93 0.17 0.79 0.75 0.54 0.66 0.60 0.57 0.66 0.13 -0.05 0.05 -0.33 0.03 -0.07 0.06 -0.41 0.20 -0.04 0.20 -0.31 0.32 -0.09 0.38 -0.18 0.10 -0.15 0.07 -0.31 0.03 -0.12 0.16 -0.09 0.05 -0.06 0.06 -0.06 0.13 0.02 0.33 0.61 0.80 0.81 0.38 0.35 0.58 0.64 0.79 0.49 0.43 0.36 0.30 0.48 0.36 1.66 0.07 -0.32 1.42 0.07 0.08 1.33 0.04 -0.24 0.93 0.07 -0.15 1.21 0.13 0.07 1.05 0.05 -0.21 1.77 0.06 0.10 1.45 0.06 -0.05 0.96 0.16 -0.32 1.74 0.06 -0.21 1.98 0.05 -0.44 1.56 0.05 -0.07 1.34 0.11 -0.41 1.05 0.35 -0.21 0.60 0.16 -0.02

33 Table 2 Summary statistics for key variables Statistics are shown for 50 three-digit SIC industries in each year and overall across all years. Cong (conglomeration level) is the number of divisions in an industry divided by the total number of divisions and single-segment firms in the industry. IndMB is the median market-to-book ratio for single-segment firms in the division's industry. IndHerf is an asset-based Herfindahl index for conglomerate divisions and stand-alone firms in the industry. Weighted-EVis the industry's excess value using an asset-weighted weighted average method that weights conglomerate excess values by the division's asset weight within the parent firm. OLS-EV uses an ordinary least squares regression approach (one regression for each year) to construct industry excess values. For a given year, all conglomerates operating in at least one of the 50 industries are retained. For each conglomerate year, the dependent variable is the conglomerate's excess value. Fifty industry-specific regressor variables are then coded (if the conglomerate has a division in a relevant industry, the asset-weight of the division is assigned, and 0 is assigned for all remaining industry regressors for which the conglomerate has no division, and then weights are rescaled so they sum to 1). The OLS coefficients are then used as the industry excess values. Cong IndMB IndHerf Weighted-EV OLS-EV (Conglomeration level) (Industry Market-to-Book) (Industry Herfindahl index) (Weighted-Excess value) (OLS-Excess value) Year Mean Med Std Dev Mean Med Std Dev Mean Med Std Dev Mean Med Std Dev Mean Med Std Dev 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 All years 0.70 0.69 0.68 0.66 0.64 0.62 0.61 0.57 0.55 0.54 0.53 0.52 0.51 0.52 0.52 0.51 0.50 0.48 0.47 0.44 0.56 0.73 0.73 0.70 0.68 0.65 0.64 0.63 0.58 0.56 0.53 0.54 0.52 0.51 0.50 0.52 0.50 0.50 0.48 0.46 0.44 0.58 0.15 0.16 0.15 0.16 0.15 0.16 0.17 0.17 0.18 0.18 0.17 0.17 0.17 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.94 1.06 1.24 1.09 1.15 1.35 1.16 1.24 1.25 1.12 1.10 1.13 0.98 1.13 1.26 1.46 1.31 1.40 1.39 1.45 1.21 0.87 0.35 0.88 0.66 0.97 0.78 0.98 0.45 0.98 0.57 1.16 0.52 1.04 0.39 1.18 0.40 1.14 0.44 1.05 0.32 1.04 0.29 1.05 0.32 0.90 0.31 1.04 0.47 1.12 0.44 1.46 0.47 1.28 0.35 1.27 0.53 1.26 0.46 1.36 0.43 1.10 0.48 0.10 0.10 0.10 0.10 0.11 0.11 0.11 0.11 0.11 0.11 0.11 0.11 0.11 0.10 0.10 0.10 0.10 0.09 0.09 0.09 0.10 0.07 0.08 0.07 0.08 0.07 0.07 0.07 0.07 0.07 0.09 0.07 0.08 0.07 0.08 0.08 0.08 0.08 0.08 0.08 0.08 0.08 0.07 0.08 0.08 0.09 0.07 0.08 0.07 0.08 0.06 0.08 0.06 0.07 0.07 0.07 0.06 0.07 0.06 0.07 0.07 0.08 0.07 -0.11 -0.09 0.16 -0.15 -0.13 0.16 -0.19 -0.16 0.20 -0.19 -0.16 0.19 -0.16 -0.15 0.19 -0.21 -0.20 0.17 -0.18 -0.16 0.16 -0.17 -0.18 0.15 -0.18 -0.16 0.15 -0.11 -0.10 0.18 -0.08 -0.08 0.11 -0.10 -0.08 0.13 -0.11 -0.10 0.13 -0.10 -0.07 0.16 -0.16 -0.17 0.19 -0.16 -0.16 0.16 -0.12 -0.14 0.14 -0.13 -0.11 0.19 -0.11 -0.13 0.17 -0.09 -0.10 0.19 -0.14 -0.13 0.17 -0.10 -0.08 -0.14 -0.09 -0.18 -0.14 -0.18 -0.14 -0.14 -0.11 -0.20 -0.19 -0.17 -0.15 -0.16 -0.17 -0.17 -0.15 -0.10 -0.07 -0.07 -0.07 -0.09 -0.08 -0.08 -0.08 -0.09 -0.08 -0.13 -0.13 -0.14 -0.15 -0.12 -0.11 -0.13 -0.10 -0.10 -0.10 -0.07 -0.07 -0.13 -0.12 0.20 0.27 0.30 0.20 0.22 0.18 0.18 0.15 0.18 0.18 0.14 0.17 0.16 0.20 0.19 0.17 0.14 0.21 0.17 0.18 0.19

34 Table 3 Panel data regression results Panel A reports a panel data regression pren dicting industry conglomeration levels (dependent variable = Cong). An industry's conglomeration level is the number of divisions in the industry divided by the total number of divisions and single-segment firms in the industry. IndMB is the median market-to-book ratio for single-segment firms in the industry. IndHerfis an asset-based Herfindahl index for conglomerate divisions and single-segment firms in the industry. Panel B reports panel regressions predicting two measures of industry excess values. Weighted-EV is the industry's excess value using an asset-weighted average method that weights conglomerate excess values by the division's asset weight within the parent firm. OLS-EVuses an ordinary least squares regression approach (one regression for each year) to construct industry excess values. For a given year, all conglomerates operating in at least one of the 50 industries are retained. For each conglomerate year, the dependent variable is the conglomerate's excess value. Fifty industry-specific regressor variables are then coded (if the conglomerate has a division in a relevant industry, the asset-weight of the division is assigned, and 0 is assigned for all remaining industry regressors for which the conglomerate has no division). The OLS coefficients are then used as the industry excess values. ValueCongl is the value-related conglomeration level, for which we use the predicted conglomeration level from Model Al in Panel A. All models in both panels use a panel data approach with twoway fixed effects (i.e. industry and year dummies are included, although not reported below), and have 1000 observations (50 industries, 20 years). Heteroskedasticity and autocorrelation-consistent t-values appear in parentheses below the coefficients. Panel A: Regression predicting industry conglomeration (dependent variable = Cong) Model Al IndMB -0.009 (-1.97) IndHerf -0.116 (-2.68) Adj. R-squared 0.85 Panel B: Regressions predicting industry excess value (dependent variable = Weighted-EV or OLS-EV) Model B] B2 Dep. Var. Weighted-EV OLS-EV ValueCongl 8.01 10.07 (8.33) (8.73) Adj. R-squared 0.19 0.17

35 Table 4 Robustness of regression results Panel A reports panel data regressions predicting industry conglomeration levels (dependent variable = Cong). An industry's conglomeration level is the number of divisions in the industry divided b the total number of divisions and single-segment firms in the industry. IndMB is the median market-to-book ratio for single-segment firms in the industry. IndHerf is an asset-based Herfindahl index for conglomerate divisions and single-segment firms in the industry. Lag(IndMB) and Lag(IndHerf) are IndMB is and IndHerf lagged by one year. Panel B reports panel data regressions predicting Weighted-EV, which is the industry's excess value using an asset-weighted average method that weights conglomerate excess values by the division's asset weight within the parent firm. ValueCong2 and ValueCong3 are the predicted conglomeration levels from Models A2 and A3, respectively, from Panel A in this table. ValueCongl is the predicted conglomeration level from Model Al in Table 3. All models in both panels use a panel data approach with two-way fixed effects (i.e. industry and year dummies are included, although not reported below), and have 1000 observations (50 industries, 20 years). Heteroskedasticity and autocorrelationconsistent t-values appear in parentheses below the coefficients. Panel A: Regressions predicting industry conglomeration (dependent variable = Cong) Model A2 A3 Lag(IndMB) -0.003 (-0.55) Lag(IndHerf) -0.073 (-1.67) IndHerf- -0.113 (-2.61) Adj. R-squared 0.86 0.85 Panel B: Regressions predicting industry excess value (dependent variable = Weighted-EV) Model B3 B4 B5 ValueCong2 2.47 (2.46) ValueCong3 -5.62 (3.46) ValueCongl - 2.87 -~- ~ (3.21) IndMB - - -0.21 (-9.92) Adj. R-squared 0.19 0.21 0.35

36 Table 5 Regressions including residuals and actual conglomeration levels This table reports panel data regressions predicting two definitions of industry excess values. Weighted-EV is the industry's excess value using an asset-weighted average method that weights conglomerate excess values by the division's asset weight within the parent firm. OLS-EVuses an ordinary least squares regression approach (one regression for each year) to construct industry excess values. For a given year, all conglomerates operating in at least one of the 50 industries are retained. For each conglomerate year, the dependent variable is the conglomerate's excess value. Fifty industry-specific regressor variables are then coded (if the conglomerate has a division in a relevant industry, the asset-weight of the division is assigned, and 0 is assigned for all remaining industry regressors for which the conglomerate has no division). The OLS coefficients are then used as the industry excess values. ValueCongl is the value-related conglomeration level, for which we use the predicted conglomeration level from the Model Al reported in Panel A of Table 3. ResidCongi = Cong - ValueCongl is the residual conglomeration level, where Cong is the actual (observed) conglomeration level. ValueCong2 and ResidCong2 are analogously defined using Model A2 of Table 4, and ValueCong3 and ResidCong3 are analogously defined using Model A3 of Table 4. All models in all three panels use a panel data approach with two-way fixed effects (i.e. industry and year dummies are included, although not reported below), and have 1000 observations (50 industries, 20 years). Heteroskedasticity and autocorrelation-consistent t-values appear in parentheses below the coefficients. Regressions predicting industry excess value (dependent variable = Weighted-EV) Model B6 B7 B8 B9 ValueCongl 8.03 - - - (8.32) ResidCongI -0.04 (0.53) - ValueCong2 - 2.45 - (2.45) - ResidCong2 - 0.04 ~~- (0.50) - - ValueCong3 - - 5.55 - - (3.40) - ResidCong3 - - 0.06 - - (0.68) - Cong - - - 0.07 (0.78) Adj. R-squared 0.19 0.19 0.21 0.13